![978-1-098-12195-2
[LSI]
C# 10 in a Nutshell
by Joseph Albahari
Copyright © 2022 Joseph Albahari. All rights reserved.
Printed in the United States of America.
Published by O’Reilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA 95472.
O’Reilly books may be purchased for educational, business, or sales promotional use. Online
editions are also available for most titles (http://oreilly.com). For more information, contact
our corporate/institutional sales department: 800-998-9938 or corporate@oreilly.com.
Acquisitions Editor: Amanda Quinn
Development Editor: Corbin Collins
Production Editor: Kristen Brown
Copyeditor: Charles Roumeliotis
Proofreader: Piper Editorial Consulting, LLC
Indexer: WordCo Indexing Services, Inc.
Interior Designer: David Futato
Cover Designer: Karen Montgomery
Illustrator: Kate Dullea
February 2022: First Edition
Revision History for the First Edition
2022-02-08: First Release
See http://oreilly.com/catalog/errata.csp?isbn=9781098121952 for release details.
The O’Reilly logo is a registered trademark of O’Reilly Media, Inc. C# 10 in a Nutshell, the
cover image, and related trade dress are trademarks of O’Reilly Media, Inc.
The views expressed in this work are those of the author, and do not represent the publish‐
er’s views. While the publisher and the author have used good faith efforts to ensure that
the information and instructions contained in this work are accurate, the publisher and
the author disclaim all responsibility for errors or omissions, including without limitation
responsibility for damages resulting from the use of or reliance on this work. Use of the
information and instructions contained in this work is at your own risk. If any code samples
or other technology this work contains or describes is subject to open source licenses or the
intellectual property rights of others, it is your responsibility to ensure that your use thereof
complies with such licenses and/or rights.](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-4-320.jpg)
![Lambda expression enhancements
The syntax around lambda expressions has been enhanced in a number of ways.
First, implicit typing (var) is permitted:
var greeter = () => "Hello, world";
The implicit type for a lambda expression is an Action or Func delegate, so greeter,
in this case, is of type Func<string>. You must explicitly state any parameter types:
var square = (int x) => x * x;
Second, a lambda expression can specify a return type:
var sqr = int (int x) => x;
This is primarily to improve compiler performance with complex nested lambdas.
Third, you can pass a lambda expression into a method parameter of type object,
Delegate, or Expression:
M1 (() => "test"); // Implicitly typed to Func<string>
M2 (() => "test"); // Implicitly typed to Func<string>
M3 (() => "test"); // Implicitly typed to Expression<Func<string>>
void M1 (object x) {}
void M2 (Delegate x) {}
void M3 (Expression x) {}
Finally, you can apply attributes to a lambda expression’s compile-generated target
method (as well as its parameters and return value):
Action a = [Description("test")] () => { };
See “Applying Attributes to Lambda Expressions (C# 10)” on page 228 for more
detail.
Nested property patterns
The following simplified syntax is legal in C# 10 for nested property pattern match‐
ing (see “Property Patterns” on page 225):
var obj = new Uri ("https://www.linqpad.net");
if (obj is Uri { Scheme.Length: 5 }) ...
This is equivalent to:
if (obj is Uri { Scheme: { Length: 5 }}) ...
CallerArgumentExpression
A method parameter to which you apply the [CallerArgumentExpression]
attribute captures an argument expression from the call site:
A Brief History of C# | 9
Introducing
C#
and
.NET](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-27-320.jpg)
![Print (Math.PI * 2);
void Print (double number,
[CallerArgumentExpression("number")] string expr = null)
=> Console.WriteLine (expr);
// Output: Math.PI * 2
This feature is intended primarily for validation and assertion libraries (see “Caller‐
ArgumentExpression (C# 10)” on page 231).
Other new features
The #line directive has been enhanced in C# 10 to allow a column and range to be
specified.
Interpolated strings in C# 10 can be constants, as long as the interpolated values are
constants.
Records can seal the ToString() method in C# 10.
C#’s definite assignment analysis has been improved so that expressions such as the
following work:
if (foo?.TryParse ("123", out var number) ?? false)
Console.WriteLine (number);
(Prior to C# 10, the compiler would generate an error: “Use of unassigned local
variable ‘number’.”)
What’s New in C# 9.0
C# 9.0 shipped with Visual Studio 2019, and is used when you target .NET 5.
Top-level statements
With top-level statements (see “Top-Level Statements” on page 35), you can write a
program without the baggage of a Main method and Program class:
using System;
Console.WriteLine ("Hello, world");
Top-level statements can include methods (which act as local methods). You can
also access command-line arguments via the “magic” args variable and return a
value to the caller. Top-level statements can be followed by type and namespace
declarations.
Init-only setters
An init-only setter (see “Init-only setters” on page 108) in a property declaration
uses the init keyword instead of the set keyword:
class Foo { public int ID { get; init; } }
10 | Chapter 1: Introducing C# and .NET](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-28-320.jpg)
![• Apply attributes to local functions (see “Attributes” on page 227)
•
• Apply the static keyword to lambda expressions or local functions to ensure
•
that you don’t accidentally capture local or instance variables (see “Static lamb‐
das” on page 177)
• Make any type work with the foreach statement, by writing a GetEnumerator
•
extension method
• Define a module initializer method that executes once when an assembly is
•
first loaded, by applying the [ModuleInitializer] attribute to a (static void
parameterless) method
• Use a “discard” (underscore symbol) as a lambda expression argument
•
• Write extended partial methods that are mandatory to implement—enabling
•
scenarios such as Roslyn’s new source generators (see “Extended partial meth‐
ods” on page 114)
• Apply an attribute to methods, types, or modules to prevent local variables
•
from being initialized by the runtime (see “[SkipLocalsInit]” on page 248)
What’s New in C# 8.0
C# 8.0 first shipped with Visual Studio 2019, and is still used today when you
target .NET Core 3 or .NET Standard 2.1.
Indices and ranges
Indices and ranges simplify working with elements or portions of an array (or the
low-level types Span<T> and ReadOnlySpan<T>).
Indices let you refer to elements relative to the end of an array by using the ^
operator. ^1 refers to the last element, ^2 refers to the second-to-last element, and
so on:
char[] vowels = new char[] {'a','e','i','o','u'};
char lastElement = vowels [^1]; // 'u'
char secondToLast = vowels [^2]; // 'o'
Ranges let you “slice” an array by using the .. operator:
char[] firstTwo = vowels [..2]; // 'a', 'e'
char[] lastThree = vowels [2..]; // 'i', 'o', 'u'
char[] middleOne = vowels [2..3] // 'i'
char[] lastTwo = vowels [^2..]; // 'o', 'u'
C# implements indexes and ranges with the help of the Index and Range types:
Index last = ^1;
Range firstTwoRange = 0..2;
char[] firstTwo = vowels [firstTwoRange]; // 'a', 'e'
A Brief History of C# | 13
Introducing
C#
and
.NET](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-31-320.jpg)
![You can support indices and ranges in your own classes by defining an indexer with
a parameter type of Index or Range:
class Sentence
{
string[] words = "The quick brown fox".Split();
public string this [Index index] => words [index];
public string[] this [Range range] => words [range];
}
For more information, see “Indices” on page 56.
Null-coalescing assignment
The ??= operator assigns a variable only if it’s null. Instead of
if (s == null) s = "Hello, world";
you can now write this:
s ??= "Hello, world";
Using declarations
If you omit the brackets and statement block following a using statement, it
becomes a using declaration. The resource is then disposed when execution falls
outside the enclosing statement block:
if (File.Exists ("file.txt"))
{
using var reader = File.OpenText ("file.txt");
Console.WriteLine (reader.ReadLine());
...
}
In this case, reader will be disposed when execution falls outside the if statement
block.
Read-only members
C# 8 lets you apply the readonly modifier to a struct’s functions, ensuring that if the
function attempts to modify any field, a compile-time error is generated:
struct Point
{
public int X, Y;
public readonly void ResetX() => X = 0; // Error!
}
If a readonly function calls a non-readonly function, the compiler generates a
warning (and defensively copies the struct to avoid the possibility of a mutation).
14 | Chapter 1: Introducing C# and .NET](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-32-320.jpg)
![Asynchronous streams
Prior to C# 8, you could use yield return to write an iterator, or await to write
an asynchronous function. But you couldn’t do both and write an iterator that
awaits, yielding elements asynchronously. C# 8 fixes this through the introduction
of asynchronous streams:
async IAsyncEnumerable<int> RangeAsync (
int start, int count, int delay)
{
for (int i = start; i < start + count; i++)
{
await Task.Delay (delay);
yield return i;
}
}
The await foreach statement consumes an asynchronous stream:
await foreach (var number in RangeAsync (0, 10, 100))
Console.WriteLine (number);
For more information, see “Asynchronous Streams” on page 650.
What’s New in C# 7.x
C# 7.x was first shipped with Visual Studio 2017. C# 7.3 is still used today by Visual
Studio 2019 when you target .NET Core 2, .NET Framework 4.6 to 4.8, or .NET
Standard 2.0.
C# 7.3
C# 7.3 made minor improvements to existing features, such as enabling use of
the equality operators with tuples, improving overload resolution, and offering the
ability to apply attributes to the backing fields of automatic properties:
[field:NonSerialized]
public int MyProperty { get; set; }
C# 7.3 also built on C# 7.2’s advanced low-allocation programming features, with
the ability to reassign ref locals, no requirement to pin when indexing fixed fields,
and field initializer support with stackalloc:
int* pointer = stackalloc int[] {1, 2, 3};
Span<int> arr = stackalloc [] {1, 2, 3};
Notice that stack-allocated memory can be assigned directly to a Span<T>. We
describe spans—and why you would use them—in Chapter 23.
C# 7.2
C# 7.2 added a new private protected modifier (the intersection of internal
and protected), the ability to follow named arguments with positional ones when
calling methods, and readonly structs. A readonly struct enforces that all fields are
A Brief History of C# | 17
Introducing
C#
and
.NET](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-35-320.jpg)
![Console.WriteLine (pos.row); // 3
Console.WriteLine (pos.column); // 10
}
Tuples implicitly support the deconstruction pattern, so you can easily deconstruct
them into individual variables:
static void Main()
{
(int row, int column) = GetFilePosition(); // Creates 2 local variables
Console.WriteLine (row); // 3
Console.WriteLine (column); // 10
}
throw expressions
Prior to C# 7, throw was always a statement. Now it can also appear as an expression
in expression-bodied functions:
public string Foo() => throw new NotImplementedException();
A throw expression can also appear in a ternary conditional expression:
string Capitalize (string value) =>
value == null ? throw new ArgumentException ("value") :
value == "" ? "" :
char.ToUpper (value[0]) + value.Substring (1);
What’s New in C# 6.0
C# 6.0, which shipped with Visual Studio 2015, features a new-generation compiler,
completely written in C#. Known as project “Roslyn,” the new compiler exposes the
entire compilation pipeline via libraries, allowing you to perform code analysis on
arbitrary source code. The compiler itself is open source, and the source code is
available at github.com/dotnet/roslyn.
In addition, C# 6.0 features several minor but significant enhancements, aimed
primarily at reducing code clutter.
The null-conditional (“Elvis”) operator (see “Null Operators” on page 74) avoids
having to explicitly check for null before calling a method or accessing a type
member. In the following example, result evaluates to null instead of throwing a
NullReferenceException:
System.Text.StringBuilder sb = null;
string result = sb?.ToString(); // result is null
Expression-bodied functions (see “Methods” on page 98) allow methods, properties,
operators, and indexers that comprise a single expression to be written more tersely,
in the style of a lambda expression:
public int TimesTwo (int x) => x * 2;
public string SomeProperty => "Property value";
A Brief History of C# | 21
Introducing
C#
and
.NET](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-39-320.jpg)
![Property initializers (Chapter 3) let you assign an initial value to an automatic
property:
public DateTime TimeCreated { get; set; } = DateTime.Now;
Initialized properties can also be read-only:
public DateTime TimeCreated { get; } = DateTime.Now;
Read-only properties can also be set in the constructor, making it easier to create
immutable (read-only) types.
Index initializers (Chapter 4) allow single-step initialization of any type that exposes
an indexer:
var dict = new Dictionary<int,string>()
{
[3] = "three",
[10] = "ten"
};
String interpolation (see “String Type” on page 52) offers a succinct alternative to
string.Format:
string s = $"It is {DateTime.Now.DayOfWeek} today";
Exception filters (see “try Statements and Exceptions” on page 181) let you apply a
condition to a catch block:
string html;
try
{
html = await new HttpClient().GetStringAsync ("http://asef");
}
catch (WebException ex) when (ex.Status == WebExceptionStatus.Timeout)
{
...
}
The using static (see “Namespaces” on page 87) directive lets you import all the
static members of a type so that you can use those members unqualified:
using static System.Console;
...
WriteLine ("Hello, world"); // WriteLine instead of Console.WriteLine
The nameof (Chapter 3) operator returns the name of a variable, type, or other
symbol as a string. This avoids breaking code when you rename a symbol in Visual
Studio:
int capacity = 123;
string x = nameof (capacity); // x is "capacity"
string y = nameof (Uri.Host); // y is "Host"
And finally, you’re now allowed to await inside catch and finally blocks.
22 | Chapter 1: Introducing C# and .NET](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-40-320.jpg)
![{
...
}
}
In this example, we also imported the Animals namespace so that our top-level
statements could access its types without qualification. Without that import, we’d
need to do this:
Animals.Panda p = new Animals.Panda ("Pan Dee");
We cover namespaces in detail at the end of this chapter (see “Namespaces” on page
87).
Defining a Main method
All of our examples, so far, have used top-level statements (a feature introduced in
C# 9).
Without top-level statements, a simple console or Windows application looks like
this:
using System;
class Program
{
static void Main() // Program entry point
{
int x = 12 * 30;
Console.WriteLine (x);
}
}
In the absence of top-level statements, C# looks for a static method called Main,
which becomes the entry point. The Main method can be defined inside any class
(and only one Main method can exist). Should your Main method need to access
private members of a particular class, defining a Main method inside that class can
be simpler than using top-level statements.
The Main method can optionally return an integer (rather than void) in order
to return a value to the execution environment (where a nonzero value typically
indicates an error). The Main method can also optionally accept an array of strings
as a parameter (that will be populated with any arguments passed to the executable).
For example:
static int Main (string[] args) {...}
An array (such as string[]) represents a fixed number of
elements of a particular type. Arrays are specified by placing
square brackets after the element type. We describe them in
“Arrays” on page 54.
34 | Chapter 2: C# Language Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-52-320.jpg)
![(The Main method can also be declared async and return a Task or Task<int> in
support of asynchronous programming, which we cover in Chapter 14.)
Top-Level Statements
Top-level statements (introduced in C# 9) let you avoid the baggage of a static Main
method and a containing class. A file with top-level statements comprises three
parts, in this order:
1. (Optionally) using directives
1.
2. A series of statements, optionally mixed with method declarations
2.
3. (Optionally) Type and namespace declarations
3.
For example:
using System; // Part 1
Console.WriteLine ("Hello, world"); // Part 2
void SomeMethod1() { ... } // Part 2
Console.WriteLine ("Hello again!"); // Part 2
void SomeMethod2() { ... } // Part 2
class SomeClass { ... } // Part 3
namespace SomeNamespace { ... } // Part 3
Because the CLR doesn’t explicitly support top-level statements, the compiler trans‐
lates your code into something like this:
using System; // Part 1
static class Program$ // Special compiler-generated name
{
static void Main$ (string[] args) // Compiler-generated name
{
Console.WriteLine ("Hello, world"); // Part 2
void SomeMethod1() { ... } // Part 2
Console.WriteLine ("Hello again!"); // Part 2
void SomeMethod2() { ... } // Part 2
}
}
class SomeClass { ... } // Part 3
namespace SomeNamespace { ... } // Part 3
Notice that everything in Part 2 is wrapped inside the main method. This means
that SomeMethod1 and SomeMethod2 act as local methods. We discuss the full implica‐
tions in “Local methods” on page 98, the most important being that local methods
(unless declared as static) can access variables declared within the containing
method:
int x = 3;
LocalMethod();
void LocalMethod() { Console.WriteLine (x); } // We can access x
Type Basics | 35
C#
Language
Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-53-320.jpg)
![1 A minor caveat is that very large long values lose some precision when converted to double.
Another consequence is that top-level methods cannot be accessed from other
classes or types.
Top-level statements can optionally return an integer value to the caller and access
a “magic” variable of type string[] called args, corresponding to command-line
arguments passed by the caller.
As a program can have only one entry point, there can be at most one file with
top-level statements in a C# project.
Types and Conversions
C# can convert between instances of compatible types. A conversion always creates
a new value from an existing one. Conversions can be either implicit or explicit:
implicit conversions happen automatically, and explicit conversions require a cast.
In the following example, we implicitly convert an int to a long type (which has
twice the bit capacity of an int) and explicitly cast an int to a short type (which has
half the bit capacity of an int):
int x = 12345; // int is a 32-bit integer
long y = x; // Implicit conversion to 64-bit integer
short z = (short)x; // Explicit conversion to 16-bit integer
Implicit conversions are allowed when both of the following are true:
• The compiler can guarantee that they will always succeed.
•
• No information is lost in conversion.1
•
Conversely, explicit conversions are required when one of the following is true:
• The compiler cannot guarantee that they will always succeed.
•
• Information might be lost during conversion.
•
(If the compiler can determine that a conversion will always fail, both kinds of
conversion are prohibited. Conversions that involve generics can also fail in certain
conditions—see “Type Parameters and Conversions” on page 152.)
The numeric conversions that we just saw are built into the
language. C# also supports reference conversions and boxing
conversions (see Chapter 3) as well as custom conversions (see
“Operator Overloading” on page 239). The compiler doesn’t
enforce the aforementioned rules with custom conversions, so
it’s possible for badly designed types to behave otherwise.
36 | Chapter 2: C# Language Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-54-320.jpg)
![Constant interpolated strings (C# 10)
From C# 10, interpolated strings can be constants, as long as the interpolated values
are constants:
const string greeting = "Hello";
const string message = $"{greeting}, world";
Arrays
An array represents a fixed number of variables (called elements) of a particular
type. The elements in an array are always stored in a contiguous block of memory,
providing highly efficient access.
An array is denoted with square brackets after the element type:
char[] vowels = new char[5]; // Declare an array of 5 characters
Square brackets also index the array, accessing a particular element by position:
vowels[0] = 'a';
vowels[1] = 'e';
vowels[2] = 'i';
vowels[3] = 'o';
vowels[4] = 'u';
Console.WriteLine (vowels[1]); // e
This prints “e” because array indexes start at 0. You can use a for loop statement to
iterate through each element in the array. The for loop in this example cycles the
integer i from 0 to 4:
for (int i = 0; i < vowels.Length; i++)
Console.Write (vowels[i]); // aeiou
The Length property of an array returns the number of elements in the array.
After an array has been created, you cannot change its length. The System.Collec
tion namespace and subnamespaces provide higher-level data structures, such as
dynamically sized arrays and dictionaries.
An array initialization expression lets you declare and populate an array in a single
step:
char[] vowels = new char[] {'a','e','i','o','u'};
Or simply:
char[] vowels = {'a','e','i','o','u'};
All arrays inherit from the System.Array class, providing common services for all
arrays. These members include methods to get and set elements regardless of the
array type. We describe them in “The Array Class” on page 355.
54 | Chapter 2: C# Language Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-72-320.jpg)
![Default Element Initialization
Creating an array always preinitializes the elements with default values. The default
value for a type is the result of a bitwise zeroing of memory. For example, consider
creating an array of integers. Because int is a value type, this allocates 1,000 integers
in one contiguous block of memory. The default value for each element will be 0:
int[] a = new int[1000];
Console.Write (a[123]); // 0
Value types versus reference types
Whether an array element type is a value type or a reference type has important
performance implications. When the element type is a value type, each element
value is allocated as part of the array, as shown here:
Point[] a = new Point[1000];
int x = a[500].X; // 0
public struct Point { public int X, Y; }
Had Point been a class, creating the array would have merely allocated 1,000 null
references:
Point[] a = new Point[1000];
int x = a[500].X; // Runtime error, NullReferenceException
public class Point { public int X, Y; }
To avoid this error, we must explicitly instantiate 1,000 Points after instantiating the
array:
Point[] a = new Point[1000];
for (int i = 0; i < a.Length; i++) // Iterate i from 0 to 999
a[i] = new Point(); // Set array element i with new point
An array itself is always a reference type object, regardless of the element type. For
instance, the following is legal:
int[] a = null;
Indices and Ranges
Indices and ranges (introduced in C# 8) simplify working with elements or portions
of an array.
Indices and ranges also work with the CLR types Span<T> and
ReadOnlySpan<T> (see Chapter 23).
You can also make your own types work with indices and
ranges, by defining an indexer of type Index or Range (see
“Indexers” on page 109).
Arrays | 55
C#
Language
Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-73-320.jpg)
![Indices
Indices let you refer to elements relative to the end of an array, with the ^ operator.
^1 refers to the last element, ^2 refers to the second-to-last element, and so on:
char[] vowels = new char[] {'a','e','i','o','u'};
char lastElement = vowels [^1]; // 'u'
char secondToLast = vowels [^2]; // 'o'
(^0 equals the length of the array, so vowels[^0] generates an error.)
C# implements indices with the help of the Index type, so you can also do the
following:
Index first = 0;
Index last = ^1;
char firstElement = vowels [first]; // 'a'
char lastElement = vowels [last]; // 'u'
Ranges
Ranges let you “slice” an array by using the .. operator:
char[] firstTwo = vowels [..2]; // 'a', 'e'
char[] lastThree = vowels [2..]; // 'i', 'o', 'u'
char[] middleOne = vowels [2..3]; // 'i'
The second number in the range is exclusive, so ..2 returns the elements before
vowels[2].
You can also use the ^ symbol in ranges. The following returns the last two
characters:
char[] lastTwo = vowels [^2..]; // 'o', 'u'
C# implements ranges with the help of the Range type, so you can also do the
following:
Range firstTwoRange = 0..2;
char[] firstTwo = vowels [firstTwoRange]; // 'a', 'e'
Multidimensional Arrays
Multidimensional arrays come in two varieties: rectangular and jagged. Rectangular
arrays represent an n-dimensional block of memory, and jagged arrays are arrays of
arrays.
Rectangular arrays
Rectangular arrays are declared using commas to separate each dimension. The
following declares a rectangular two-dimensional array for which the dimensions
are 3 by 3:
int[,] matrix = new int[3,3];
56 | Chapter 2: C# Language Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-74-320.jpg)
![The GetLength method of an array returns the length for a given dimension (start‐
ing at 0):
for (int i = 0; i < matrix.GetLength(0); i++)
for (int j = 0; j < matrix.GetLength(1); j++)
matrix[i,j] = i * 3 + j;
You can initialize a rectangular array with explicit values. The following code creates
an array identical to the previous example:
int[,] matrix = new int[,]
{
{0,1,2},
{3,4,5},
{6,7,8}
};
Jagged arrays
Jagged arrays are declared using successive square brackets to represent each
dimension. Here is an example of declaring a jagged two-dimensional array for
which the outermost dimension is 3:
int[][] matrix = new int[3][];
Interestingly, this is new int[3][] and not new int[][3].
Eric Lippert has written an excellent article on why this is so.
The inner dimensions aren’t specified in the declaration because, unlike a rectangu‐
lar array, each inner array can be an arbitrary length. Each inner array is implicitly
initialized to null rather than an empty array. You must manually create each inner
array:
for (int i = 0; i < matrix.Length; i++)
{
matrix[i] = new int[3]; // Create inner array
for (int j = 0; j < matrix[i].Length; j++)
matrix[i][j] = i * 3 + j;
}
You can initialize a jagged array with explicit values. The following code creates an
array identical to the previous example with an additional element at the end:
int[][] matrix = new int[][]
{
new int[] {0,1,2},
new int[] {3,4,5},
new int[] {6,7,8,9}
};
Simplified Array Initialization Expressions
There are two ways to shorten array initialization expressions. The first is to omit
the new operator and type qualifications:
Arrays | 57
C#
Language
Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-75-320.jpg)
![char[] vowels = {'a','e','i','o','u'};
int[,] rectangularMatrix =
{
{0,1,2},
{3,4,5},
{6,7,8}
};
int[][] jaggedMatrix =
{
new int[] {0,1,2},
new int[] {3,4,5},
new int[] {6,7,8,9}
};
The second approach is to use the var keyword, which instructs the compiler to
implicitly type a local variable:
var i = 3; // i is implicitly of type int
var s = "sausage"; // s is implicitly of type string
// Therefore:
var rectMatrix = new int[,] // rectMatrix is implicitly of type int[,]
{
{0,1,2},
{3,4,5},
{6,7,8}
};
var jaggedMat = new int[][] // jaggedMat is implicitly of type int[][]
{
new int[] {0,1,2},
new int[] {3,4,5},
new int[] {6,7,8,9}
};
Implicit typing can be taken one stage further with arrays: you can omit the type
qualifier after the new keyword and have the compiler infer the array type:
var vowels = new[] {'a','e','i','o','u'}; // Compiler infers char[]
For this to work, the elements must all be implicitly convertible to a single type (and
at least one of the elements must be of that type, and there must be exactly one best
type), as in the following example:
var x = new[] {1,10000000000}; // all convertible to long
Bounds Checking
All array indexing is bounds checked by the runtime. An IndexOutOfRange
Exception is thrown if you use an invalid index:
int[] arr = new int[3];
arr[3] = 1; // IndexOutOfRangeException thrown
58 | Chapter 2: C# Language Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-76-320.jpg)
![int[] ints = new int[2];
Console.WriteLine (ints[0]); // 0
The following code outputs 0, because fields are implicitly assigned a default value
(whether instance or static):
Console.WriteLine (Test.X); // 0
class Test { public static int X; } // field
Default Values
All type instances have a default value. The default value for the predefined types is
the result of a bitwise zeroing of memory:
Type Default value
Reference types (and nullable value types) null
Numeric and enum types 0
char type '0'
bool type false
You can obtain the default value for any type via the default keyword:
Console.WriteLine (default (decimal)); // 0
You can optionally omit the type when it can be inferred:
decimal d = default;
The default value in a custom value type (i.e., struct) is the same as the default
value for each field defined by the custom type.
Parameters
A method may have a sequence of parameters. Parameters define the set of argu‐
ments that must be provided for that method. In the following example, the method
Foo has a single parameter named p, of type int:
Foo (8); // 8 is an argument
static void Foo (int p) {...} // p is a parameter
You can control how parameters are passed with the ref, in, and out modifiers:
Parameter modifier Passed by Variable must be definitely assigned
(None) Value Going in
ref Reference Going in
in Reference (read-only) Going in
out Reference Going out
Variables and Parameters | 61
C#
Language
Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-79-320.jpg)
![Overloading solely on the presence of in is permitted:
void Foo ( SomeBigStruct a) { ... }
void Foo (in SomeBigStruct a) { ... }
To call the second overload, the caller must use the in modifier:
SomeBigStruct x = ...;
Foo (x); // Calls the first overload
Foo (in x); // Calls the second overload
When there’s no ambiguity,
void Bar (in SomeBigStruct a) { ... }
use of the in modifier is optional for the caller:
Bar (x); // OK (calls the 'in' overload)
Bar (in x); // OK (calls the 'in' overload)
To make this example meaningful, SomeBigStruct would be defined as a struct (see
“Structs” on page 129).
The params modifier
The params modifier, if applied to the last parameter of a method, allows the
method to accept any number of arguments of a particular type. The parameter
type must be declared as an (single-dimensional) array, as shown in the following
example:
int total = Sum (1, 2, 3, 4);
Console.WriteLine (total); // 10
// The call to Sum above is equivalent to:
int total2 = Sum (new int[] { 1, 2, 3, 4 });
int Sum (params int[] ints)
{
int sum = 0;
for (int i = 0; i < ints.Length; i++)
sum += ints [i]; // Increase sum by ints[i]
return sum;
}
If there are zero arguments in the params position, a zero-length array is created.
You can also supply a params argument as an ordinary array. The first line in our
example is semantically equivalent to this:
int total = Sum (new int[] { 1, 2, 3, 4 } );
Optional parameters
Methods, constructors, and indexers (Chapter 3) can declare optional parameters. A
parameter is optional if it specifies a default value in its declaration:
Variables and Parameters | 65
C#
Language
Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-83-320.jpg)
![A subtle difference is that argument expressions are evalu‐
ated in the order in which they appear at the calling site.
In general, this makes a difference only with interdependent
side-effecting expressions such as the following, which writes
0, 1:
int a = 0;
Foo (y: ++a, x: --a); // ++a is evaluated first
Of course, you would almost certainly avoid writing such code
in practice!
You can mix named and positional arguments:
Foo (1, y:2);
However, there is a restriction: positional arguments must come before named
arguments unless they are used in the correct position. So, you could call Foo like
this:
Foo (x:1, 2); // OK. Arguments in the declared positions
But not like this:
Foo (y:2, 1); // Compile-time error. y isn't in the first position
Named arguments are particularly useful in conjunction with optional parameters.
For instance, consider the following method:
void Bar (int a = 0, int b = 0, int c = 0, int d = 0) { ... }
You can call this supplying only a value for d, as follows:
Bar (d:3);
This is particularly useful when calling COM APIs, which we discuss in detail in
Chapter 24.
Ref Locals
A somewhat esoteric feature of C# is that you can define a local variable that
references an element in an array or field in an object (from C# 7):
int[] numbers = { 0, 1, 2, 3, 4 };
ref int numRef = ref numbers [2];
In this example, numRef is a reference to numbers[2]. When we modify numRef, we
modify the array element:
numRef *= 10;
Console.WriteLine (numRef); // 20
Console.WriteLine (numbers [2]); // 20
The target for a ref local must be an array element, field, or local variable; it cannot
be a property (Chapter 3). Ref locals are intended for specialized micro-optimization
scenarios and are typically used in conjunction with ref returns.
Variables and Parameters | 67
C#
Language
Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-85-320.jpg)
![You can insert parentheses to change the actual order of evaluation:
8 / ( 4 / 2 ) // 4
Right-associative operators
The assignment operators as well as the lambda, null-coalescing, and conditional
operators are right-associative; in other words, they are evaluated from right to left.
Right associativity allows multiple assignments, such as the following, to compile:
x = y = 3;
This first assigns 3 to y and then assigns the result of that expression (3) to x.
Operator Table
Table 2-3 lists C#’s operators in order of precedence. Operators in the same category
have the same precedence. We explain user-overloadable operators in “Operator
Overloading” on page 239.
Table 2-3. C# operators (categories in order of precedence)
Category Operator
symbol
Operator name Example User-
overloadable
Primary . Member access x.y No
?. and ?[] Null-conditional x?.y or x?[0] No
! (postfix) Null-forgiving x!.y or x![0] No
-> (unsafe) Pointer to struct x->y No
() Function call x() No
[] Array/index a[x] Via indexer
++ Post-increment x++ Yes
−− Post-decrement x−− Yes
new Create instance new Foo() No
stackalloc Stack allocation stackalloc(10) No
typeof Get type from
identifier
typeof(int) No
nameof Get name of
identifier
nameof(x) No
checked Integral overflow
check on
checked(x) No
unchecked Integral overflow
check off
unchecked(x) No
default Default value default(char) No
Unary await Await await myTask No
72 | Chapter 2: C# Language Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-90-320.jpg)
![Category Operator
symbol
Operator name Example User-
overloadable
sizeof Get size of struct sizeof(int) No
+ Positive value of +x Yes
− Negative value of −x Yes
! Not !x Yes
~ Bitwise complement ~x Yes
++ Pre-increment ++x Yes
−− Pre-decrement −−x Yes
() Cast (int)x No
^ Index from end array[^1] No
* (unsafe) Value at address *x No
& (unsafe) Address of value &x No
Range ..
..^
Range of indices x..y
x..^y
No
Switch & with switch Switch expression num switch {
1 => true,
_ => false
}
No
with With expression rec with
{ X = 123 }
No
Multiplicative * Multiply x * y Yes
/ Divide x / y Yes
% Remainder x % y Yes
Additive + Add x + y Yes
− Subtract x − y Yes
Shift << Shift left x << 1 Yes
>> Shift right x >> 1 Yes
Relational < Less than x < y Yes
> Greater than x > y Yes
<= Less than or equal
to
x <= y Yes
>= Greater than or
equal to
x >= y Yes
is Type is or is subclass
of
x is y No
as Type conversion x as y No
Equality == Equals x == y Yes
Expressions and Operators | 73
C#
Language
Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-91-320.jpg)
![If the lefthand expression is non-null, the righthand expression is never evaluated.
The null-coalescing operator also works with nullable value types (see “Nullable
Value Types” on page 194).
Null-Coalescing Assignment Operator
The ??= operator (introduced in C# 8) is the null-coalescing assignment operator. It
says, “If the operand to the left is null, assign the right operand to the left operand.”
Consider the following:
myVariable ??= someDefault;
This is equivalent to:
if (myVariable == null) myVariable = someDefault;
The ??= operator is particularly useful in implementing lazily calculated properties.
We’ll cover this topic later, in “Calculated Fields and Lazy Evaluation” on page 217.
Null-Conditional Operator
The ?. operator is the null-conditional or “Elvis” operator (after the Elvis emoticon).
It allows you to call methods and access members just like the standard dot operator
except that if the operand on the left is null, the expression evaluates to null instead
of throwing a NullReferenceException:
System.Text.StringBuilder sb = null;
string s = sb?.ToString(); // No error; s instead evaluates to null
The last line is equivalent to the following:
string s = (sb == null ? null : sb.ToString());
Null-conditional expressions also work with indexers:
string foo = null;
char? c = foo?[1]; // c is null
Upon encountering a null, the Elvis operator short-circuits the remainder of the
expression. In the following example, s evaluates to null, even with a standard dot
operator between ToString() and ToUpper():
System.Text.StringBuilder sb = null;
string s = sb?.ToString().ToUpper(); // s evaluates to null without error
Repeated use of Elvis is necessary only if the operand immediately to its left might
be null. The following expression is robust to both x being null and x.y being null:
x?.y?.z
It is equivalent to the following (except that x.y is evaluated only once):
x == null ? null
: (x.y == null ? null : x.y.z)
The final expression must be capable of accepting a null. The following is illegal:
Null Operators | 75
C#
Language
Basics](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-93-320.jpg)
![void Foo (int x) {...}
float Foo (int x) {...} // Compile-time error
void Goo (int[] x) {...}
void Goo (params int[] x) {...} // Compile-time error
Whether a parameter is pass-by-value or pass-by-reference is also part of the signa‐
ture. For example, Foo(int) can coexist with either Foo(ref int) or Foo(out int).
However, Foo(ref int) and Foo(out int) cannot coexist:
void Foo (int x) {...}
void Foo (ref int x) {...} // OK so far
void Foo (out int x) {...} // Compile-time error
Instance Constructors
Constructors run initialization code on a class or struct. A constructor is defined
like a method, except that the method name and return type are reduced to the
name of the enclosing type:
Panda p = new Panda ("Petey"); // Call constructor
public class Panda
{
string name; // Define field
public Panda (string n) // Define constructor
{
name = n; // Initialization code (set up field)
}
}
Instance constructors allow the following modifiers:
Access modifiers public internal private protected
Unmanaged code modifiers unsafe extern
Single-statement constructors can also be written as expression-bodied members:
public Panda (string n) => name = n;
Overloading constructors
A class or struct may overload constructors. To avoid code duplication, one con‐
structor can call another, using the this keyword:
using System;
public class Wine
{
public decimal Price;
public int Year;
public Wine (decimal price) { Price = price; }
100 | Chapter 3: Creating Types in C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-118-320.jpg)
![Init-only properties have another significant advantage, which
is that they allow for nondestructive mutation when used in
conjunction with records (see “Records” on page 211).
As with ordinary set accessors, init-only accessors can provide an implementation:
public class Note
{
readonly int _pitch;
public int Pitch { get => _pitch; init => _pitch = value; }
...
Notice that the _pitch field is read-only: init-only setters are permitted to modify
readonly fields in their own class. (Without this feature, _pitch would need to be
writable, and the class would fail at being internally immutable.)
Changing a property’s accessor from init to set (or vice
versa) is a binary breaking change: anyone that references your
assembly will need to recompile their assembly.
This should not be an issue when creating wholly immutable
types, in that your type will never require properties with a
(writable) set accessor.
CLR property implementation
C# property accessors internally compile to methods called get_XXX and set_XXX:
public decimal get_CurrentPrice {...}
public void set_CurrentPrice (decimal value) {...}
An init accessor is processed like a set accessor, but with an extra flag encoded
into the set accessor’s “modreq” metadata (see “Init-only properties” on page 794).
Simple nonvirtual property accessors are inlined by the Just-in-Time (JIT) compiler,
eliminating any performance difference between accessing a property and a field.
Inlining is an optimization in which a method call is replaced with the body of that
method.
Indexers
Indexers provide a natural syntax for accessing elements in a class or struct that
encapsulate a list or dictionary of values. Indexers are similar to properties but are
accessed via an index argument rather than a property name. The string class has
an indexer that lets you access each of its char values via an int index:
string s = "hello";
Console.WriteLine (s[0]); // 'h'
Console.WriteLine (s[3]); // 'l'
The syntax for using indexers is like that for using arrays, except that the index
argument(s) can be of any type(s).
Classes | 109
Creating
Types
in
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-127-320.jpg)
![Indexers have the same modifiers as properties (see “Properties” on page 105) and
can be called null-conditionally by inserting a question mark before the square
bracket (see “Null Operators” on page 74):
string s = null;
Console.WriteLine (s?[0]); // Writes nothing; no error.
Implementing an indexer
To write an indexer, define a property called this, specifying the arguments in
square brackets:
class Sentence
{
string[] words = "The quick brown fox".Split();
public string this [int wordNum] // indexer
{
get { return words [wordNum]; }
set { words [wordNum] = value; }
}
}
Here’s how we could use this indexer:
Sentence s = new Sentence();
Console.WriteLine (s[3]); // fox
s[3] = "kangaroo";
Console.WriteLine (s[3]); // kangaroo
A type can declare multiple indexers, each with parameters of different types. An
indexer can also take more than one parameter:
public string this [int arg1, string arg2]
{
get { ... } set { ... }
}
If you omit the set accessor, an indexer becomes read-only, and you can use
expression-bodied syntax to shorten its definition:
public string this [int wordNum] => words [wordNum];
CLR indexer implementation
Indexers internally compile to methods called get_Item and set_Item, as follows:
public string get_Item (int wordNum) {...}
public void set_Item (int wordNum, string value) {...}
Using indices and ranges with indexers
You can support indices and ranges (see “Indices and ranges” on page 13) in your
own classes by defining an indexer with a parameter type of Index or Range.
110 | Chapter 3: Creating Types in C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-128-320.jpg)
![We could extend our previous example by adding the following indexers to the
Sentence class:
public string this [Index index] => words [index];
public string[] this [Range range] => words [range];
This then enables the following:
Sentence s = new Sentence();
Console.WriteLine (s [^1]); // fox
string[] firstTwoWords = s [..2]; // (The, quick)
Static Constructors
A static constructor executes once per type rather than once per instance. A type can
define only one static constructor, and it must be parameterless and have the same
name as the type:
class Test
{
static Test() { Console.WriteLine ("Type Initialized"); }
}
The runtime automatically invokes a static constructor just prior to the type being
used. Two things trigger this:
• Instantiating the type
•
• Accessing a static member in the type
•
The only modifiers allowed by static constructors are unsafe and extern.
If a static constructor throws an unhandled exception (Chap‐
ter 4), that type becomes unusable for the life of the
application.
From C# 9, you can also define module initializers, which
execute once per assembly (when the assembly is first loaded).
To define a module initializer, write a static void method
and then apply the [ModuleInitializer] attribute to that
method:
[System.Runtime.CompilerServices.ModuleInitializer]
internal static void InitAssembly()
{
...
}
Static constructors and field initialization order
Static field initializers run just before the static constructor is called. If a type has
no static constructor, static field initializers will execute just prior to the type being
used—or anytime earlier at the whim of the runtime.
Classes | 111
Creating
Types
in
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-129-320.jpg)
![Overloading and Resolution
Inheritance has an interesting impact on method overloading. Consider the follow‐
ing two overloads:
static void Foo (Asset a) { }
static void Foo (House h) { }
When an overload is called, the most specific type has precedence:
House h = new House (...);
Foo(h); // Calls Foo(House)
The particular overload to call is determined statically (at compile time) rather than
at runtime. The following code calls Foo(Asset), even though the runtime type of a
is House:
Asset a = new House (...);
Foo(a); // Calls Foo(Asset)
If you cast Asset to dynamic (Chapter 4), the decision as to
which overload to call is deferred until runtime and is then
based on the object’s actual type:
Asset a = new House (...);
Foo ((dynamic)a); // Calls Foo(House)
The object Type
object (System.Object) is the ultimate base class for all types. Any type can be
upcast to object.
To illustrate how this is useful, consider a general-purpose stack. A stack is a data
structure based on the principle of LIFO—“last in, first out.” A stack has two
operations: push an object on the stack, and pop an object off the stack. Here is a
simple implementation that can hold up to 10 objects:
public class Stack
{
int position;
object[] data = new object[10];
public void Push (object obj) { data[position++] = obj; }
public object Pop() { return data[--position]; }
}
Because Stack works with the object type, we can Push and Pop instances of any
type to and from the Stack:
Stack stack = new Stack();
stack.Push ("sausage");
string s = (string) stack.Pop(); // Downcast, so explicit cast is needed
Console.WriteLine (s); // sausage
The object Type | 125
Creating
Types
in
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-143-320.jpg)
![Boxing conversions are crucial in providing a unified type
system. The system is not perfect, however: we’ll see in
“Generics” on page 145 that variance with arrays and generics
supports only reference conversions and not boxing conversions:
object[] a1 = new string[3]; // Legal
object[] a2 = new int[3]; // Error
Copying semantics of boxing and unboxing
Boxing copies the value-type instance into the new object, and unboxing copies the
contents of the object back into a value-type instance. In the following example,
changing the value of i doesn’t change its previously boxed copy:
int i = 3;
object boxed = i;
i = 5;
Console.WriteLine (boxed); // 3
Static and Runtime Type Checking
C# programs are type-checked both statically (at compile time) and at runtime (by
the CLR).
Static type checking enables the compiler to verify the correctness of your program
without running it. The following code will fail because the compiler enforces static
typing:
int x = "5";
Runtime type checking is performed by the CLR when you downcast via a reference
conversion or unboxing:
object y = "5";
int z = (int) y; // Runtime error, downcast failed
Runtime type checking is possible because each object on the heap internally stores
a little type token. You can retrieve this token by calling the GetType method of
object.
The GetType Method and typeof Operator
All types in C# are represented at runtime with an instance of System.Type. There
are two basic ways to get a System.Type object:
• Call GetType on the instance.
•
• Use the typeof operator on a type name.
•
GetType is evaluated at runtime; typeof is evaluated statically at compile time
(when generic type parameters are involved, it’s resolved by the JIT compiler).
The object Type | 127
Creating
Types
in
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-145-320.jpg)
![struct Point
{
int x, y;
public Point (int x, int y) { this.x = x; this.y = y; } // OK
}
If we added the following constructor, the struct would not compile, because y
would remain unassigned:
public Point (int x) { this.x = x; } // Not OK
The default constructor
In addition to any constructors that you define, a struct always has an implicit
parameterless constructor that performs a bitwise-zeroing of its fields (setting them
to their default values):
Point p = new Point(); // p.x and p.y will be 0
struct Point { int x, y; }
Even when you define a parameterless constructor of your own, the implicit param‐
eterless constructor still exists and can be accessed via the default keyword:
Point p1 = new Point(); // p1.x and p1.y will be 1
Point p2 = default; // p2.x and p2.y will be 0
struct Point
{
int x = 1;
int y;
public Point() => y = 1;
}
In this example, we initialized x to 1 via a field initializer, and we initialized y to 1
via the parameterless constructor. And yet with the default keyword, we were still
able to create a Point that bypassed both initializations. The default constructor can
be accessed other ways, too, as the following example illustrates:
var points = new Point[10]; // Each point in the array will be (0,0)
var test = new Test(); // test.p will be (0,0)
class Test { Point p; }
Having what amounts to two parameterless constructors can
be a source of confusion and is arguably a good reason to
avoid defining field initializers and explicit parameterless con‐
structors in structs.
A good strategy with structs is to design them such that their default value is a
valid state, thereby making initialization redundant. For example, rather than this:
struct WebOptions { public string Protocol { get; set; } = "https"; }
consider the following:
130 | Chapter 3: Creating Types in C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-148-320.jpg)
![But if a value type appears as a field in a class, it will reside on the heap:
class MyClass
{
Point p; // Lives on heap, because MyClass instances live on the heap
}
Similarly, arrays of structs live on the heap, and boxing a struct sends it to the heap.
Adding the ref modifier to a struct’s declaration ensures that it can only ever reside
on the stack. Attempting to use a ref struct in such a way that it could reside on the
heap generates a compile-time error:
var points = new Point [100]; // Error: will not compile!
ref struct Point { public int X, Y; }
class MyClass { Point P; } // Error: will not compile!
Ref structs were introduced mainly for the benefit of the Span<T> and ReadOnly
Span<T> structs. Because Span<T> and ReadOnlySpan<T> instances can exist only on
the stack, it’s possible for them to safely wrap stack-allocated memory.
Ref structs cannot partake in any C# feature that directly or indirectly introduces
the possibility of existing on the heap. This includes a number of advanced C#
features that we describe in Chapter 4, namely lambda expressions, iterators, and
asynchronous functions (because, behind the scenes, these features all create hidden
classes with fields). Also, ref structs cannot appear inside non-ref structs, and they
cannot implement interfaces (because this could result in boxing).
Access Modifiers
To promote encapsulation, a type or type member can limit its accessibility to other
types and other assemblies by adding one of five access modifiers to the declaration:
public
Fully accessible. This is the implicit accessibility for members of an enum or
interface.
internal
Accessible only within the containing assembly or friend assemblies. This is the
default accessibility for non-nested types.
private
Accessible only within the containing type. This is the default accessibility for
members of a class or struct.
protected
Accessible only within the containing type or subclasses.
protected internal
The union of protected and internal accessibility. A member that is
protected internal is accessible in two ways.
132 | Chapter 3: Creating Types in C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-150-320.jpg)
![private protected (from C# 7.2)
The intersection of protected and internal accessibility. A member that is
private protected is accessible only within the containing type, or subclasses
that reside in the same assembly (making it less accessible than protected or
internal alone).
Examples
Class2 is accessible from outside its assembly; Class1 is not:
class Class1 {} // Class1 is internal (default)
public class Class2 {}
ClassB exposes field x to other types in the same assembly; ClassA does not:
class ClassA { int x; } // x is private (default)
class ClassB { internal int x; }
Functions within Subclass can call Bar but not Foo:
class BaseClass
{
void Foo() {} // Foo is private (default)
protected void Bar() {}
}
class Subclass : BaseClass
{
void Test1() { Foo(); } // Error - cannot access Foo
void Test2() { Bar(); } // OK
}
Friend Assemblies
You can expose internal members to other friend assemblies by adding
the System.Runtime.CompilerServices.InternalsVisibleTo assembly attribute,
specifying the name of the friend assembly as follows:
[assembly: InternalsVisibleTo ("Friend")]
If the friend assembly has a strong name (see Chapter 17), you must specify its full
160-byte public key:
[assembly: InternalsVisibleTo ("StrongFriend, PublicKey=0024f000048c...")]
You can extract the full public key from a strongly named assembly with a LINQ
query (we explain LINQ in detail in Chapter 8):
string key = string.Join ("",
Assembly.GetExecutingAssembly().GetName().GetPublicKey()
.Select (b => b.ToString ("x2")));
The companion sample in LINQPad invites you to browse to
an assembly and then copies the assembly’s full public key to
the clipboard.
Access Modifiers | 133
Creating
Types
in
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-151-320.jpg)
![Flags Enums
You can combine enum members. To prevent ambiguities, members of a combina‐
ble enum require explicitly assigned values, typically in powers of two:
[Flags]
enum BorderSides { None=0, Left=1, Right=2, Top=4, Bottom=8 }
or:
enum BorderSides { None=0, Left=1, Right=1<<1, Top=1<<2, Bottom=1<<3 }
To work with combined enum values, you use bitwise operators such as | and &.
These operate on the underlying integral values:
BorderSides leftRight = BorderSides.Left | BorderSides.Right;
if ((leftRight & BorderSides.Left) != 0)
Console.WriteLine ("Includes Left"); // Includes Left
string formatted = leftRight.ToString(); // "Left, Right"
BorderSides s = BorderSides.Left;
s |= BorderSides.Right;
Console.WriteLine (s == leftRight); // True
s ^= BorderSides.Right; // Toggles BorderSides.Right
Console.WriteLine (s); // Left
By convention, the Flags attribute should always be applied to an enum type
when its members are combinable. If you declare such an enum without the Flags
attribute, you can still combine members, but calling ToString on an enum instance
will emit a number rather than a series of names.
By convention, a combinable enum type is given a plural rather than singular name.
For convenience, you can include combination members within an enum declara‐
tion itself:
[Flags]
enum BorderSides
{
None=0,
Left=1, Right=1<<1, Top=1<<2, Bottom=1<<3,
LeftRight = Left | Right,
TopBottom = Top | Bottom,
All = LeftRight | TopBottom
}
Enum Operators
The operators that work with enums are:
= == != < > <= >= + - ^ & | ˜
+= -= ++ -- sizeof
142 | Chapter 3: Creating Types in C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-160-320.jpg)
![}
[Flags]
public enum BorderSides { Left=1, Right=2, Top=4, Bottom=8 }
Nested Types
A nested type is declared within the scope of another type:
public class TopLevel
{
public class Nested { } // Nested class
public enum Color { Red, Blue, Tan } // Nested enum
}
A nested type has the following features:
• It can access the enclosing type’s private members and everything else the
•
enclosing type can access.
• You can declare it with the full range of access modifiers rather than just
•
public and internal.
• The default accessibility for a nested type is private rather than internal.
•
• Accessing a nested type from outside the enclosing type requires qualification
•
with the enclosing type’s name (like when accessing static members).
For example, to access Color.Red from outside our TopLevel class, we’d need to do
this:
TopLevel.Color color = TopLevel.Color.Red;
All types (classes, structs, interfaces, delegates, and enums) can be nested within
either a class or a struct.
Here is an example of accessing a private member of a type from a nested type:
public class TopLevel
{
static int x;
class Nested
{
static void Foo() { Console.WriteLine (TopLevel.x); }
}
}
Here is an example of applying the protected access modifier to a nested type:
public class TopLevel
{
protected class Nested { }
}
public class SubTopLevel : TopLevel
{
144 | Chapter 3: Creating Types in C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-162-320.jpg)
![static void Foo() { new TopLevel.Nested(); }
}
Here is an example of referring to a nested type from outside the enclosing type:
public class TopLevel
{
public class Nested { }
}
class Test
{
TopLevel.Nested n;
}
Nested types are used heavily by the compiler itself when it generates private classes
that capture state for constructs such as iterators and anonymous methods.
If the sole reason for using a nested type is to avoid cluttering
a namespace with too many types, consider using a nested
namespace instead. A nested type should be used because of
its stronger access control restrictions, or when the nested
class must access private members of the containing class.
Generics
C# has two separate mechanisms for writing code that is reusable across different
types: inheritance and generics. Whereas inheritance expresses reusability with a
base type, generics express reusability with a “template” that contains “placeholder”
types. Generics, when compared to inheritance, can increase type safety and reduce
casting and boxing.
C# generics and C++ templates are similar concepts, but they
work differently. We explain this difference in “C# Generics
Versus C++ Templates” on page 157.
Generic Types
A generic type declares type parameters—placeholder types to be filled in by the
consumer of the generic type, which supplies the type arguments. Here is a generic
type Stack<T>, designed to stack instances of type T. Stack<T> declares a single type
parameter T:
public class Stack<T>
{
int position;
T[] data = new T[100];
public void Push (T obj) => data[position++] = obj;
public T Pop() => data[--position];
}
Generics | 145
Creating
Types
in
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-163-320.jpg)
![We can use Stack<T> as follows:
var stack = new Stack<int>();
stack.Push (5);
stack.Push (10);
int x = stack.Pop(); // x is 10
int y = stack.Pop(); // y is 5
Stack<int> fills in the type parameter T with the type argument int, implicitly cre‐
ating a type on the fly (the synthesis occurs at runtime). Attempting to push a string
onto our Stack<int> would, however, produce a compile-time error. Stack<int>
effectively has the following definition (substitutions appear in bold, with the class
name hashed out to avoid confusion):
public class ###
{
int position;
int[] data = new int[100];
public void Push (int obj) => data[position++] = obj;
public int Pop() => data[--position];
}
Technically, we say that Stack<T> is an open type, whereas Stack<int> is a closed
type. At runtime, all generic type instances are closed—with the placeholder types
filled in. This means that the following statement is illegal:
var stack = new Stack<T>(); // Illegal: What is T?
However, it’s legal if it’s within a class or method that itself defines T as a type
parameter:
public class Stack<T>
{
...
public Stack<T> Clone()
{
Stack<T> clone = new Stack<T>(); // Legal
...
}
}
Why Generics Exist
Generics exist to write code that is reusable across different types. Suppose that
we need a stack of integers, but we don’t have generic types. One solution would
be to hardcode a separate version of the class for every required element type
(e.g., IntStack, StringStack, etc.). Clearly, this would cause considerable code
duplication. Another solution would be to write a stack that is generalized by using
object as the element type:
public class ObjectStack
{
int position;
object[] data = new object[10];
146 | Chapter 3: Creating Types in C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-164-320.jpg)
![public void Push (object obj) => data[position++] = obj;
public object Pop() => data[--position];
}
An ObjectStack, however, wouldn’t work as well as a hardcoded IntStack for
specifically stacking integers. An ObjectStack would require boxing and downcast‐
ing that could not be checked at compile time:
// Suppose we just want to store integers here:
ObjectStack stack = new ObjectStack();
stack.Push ("s"); // Wrong type, but no error!
int i = (int)stack.Pop(); // Downcast - runtime error
What we need is both a general implementation of a stack that works for all element
types as well as a way to easily specialize that stack to a specific element type for
increased type safety and reduced casting and boxing. Generics give us precisely this
by allowing us to parameterize the element type. Stack<T> has the benefits of both
ObjectStack and IntStack. Like ObjectStack, Stack<T> is written once to work
generally across all types. Like IntStack, Stack<T> is specialized for a particular
type—the beauty is that this type is T, which we substitute on the fly.
ObjectStack is functionally equivalent to Stack<object>.
Generic Methods
A generic method declares type parameters within the signature of a method.
With generic methods, many fundamental algorithms can be implemented in a
general-purpose way. Here is a generic method that swaps the contents of two
variables of any type T:
static void Swap<T> (ref T a, ref T b)
{
T temp = a;
a = b;
b = temp;
}
Swap<T> is called as follows:
int x = 5;
int y = 10;
Swap (ref x, ref y);
Generally, there is no need to supply type arguments to a generic method, because
the compiler can implicitly infer the type. If there is ambiguity, generic methods can
be called with type arguments as follows:
Swap<int> (ref x, ref y);
Generics | 147
Creating
Types
in
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-165-320.jpg)
![Within a generic type, a method is not classed as generic unless it introduces type
parameters (with the angle bracket syntax). The Pop method in our generic stack
merely uses the type’s existing type parameter, T, and is not classed as a generic
method.
Methods and types are the only constructs that can introduce type parameters.
Properties, indexers, events, fields, constructors, operators, and so on cannot
declare type parameters, although they can partake in any type parameters already
declared by their enclosing type. In our generic stack example, for instance, we
could write an indexer that returns a generic item:
public T this [int index] => data [index];
Similarly, constructors can partake in existing type parameters but not introduce
them:
public Stack<T>() { } // Illegal
Declaring Type Parameters
Type parameters can be introduced in the declaration of classes, structs, interfaces,
delegates (covered in Chapter 4), and methods. Other constructs, such as proper‐
ties, cannot introduce a type parameter, but they can use one. For example, the
property Value uses T:
public struct Nullable<T>
{
public T Value { get; }
}
A generic type or method can have multiple parameters:
class Dictionary<TKey, TValue> {...}
To instantiate:
Dictionary<int,string> myDict = new Dictionary<int,string>();
Or:
var myDict = new Dictionary<int,string>();
Generic type names and method names can be overloaded as long as the number
of type parameters is different. For example, the following three type names do not
conflict:
class A {}
class A<T> {}
class A<T1,T2> {}
By convention, generic types and methods with a single type
parameter typically name their parameter T, as long as the
intent of the parameter is clear. When using multiple type
parameters, each parameter is prefixed with T but has a more
descriptive name.
148 | Chapter 3: Creating Types in C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-166-320.jpg)
![typeof and Unbound Generic Types
Open generic types do not exist at runtime: they are closed as part of compilation.
However, it is possible for an unbound generic type to exist at runtime—purely as
a Type object. The only way to specify an unbound generic type in C# is via the
typeof operator:
class A<T> {}
class A<T1,T2> {}
...
Type a1 = typeof (A<>); // Unbound type (notice no type arguments).
Type a2 = typeof (A<,>); // Use commas to indicate multiple type args.
Open generic types are used in conjunction with the Reflection API (Chapter 18).
You can also use the typeof operator to specify a closed type:
Type a3 = typeof (A<int,int>);
Or, you can specify an open type (which is closed at runtime):
class B<T> { void X() { Type t = typeof (T); } }
The default Generic Value
You can use the default keyword to get the default value for a generic type parame‐
ter. The default value for a reference type is null, and the default value for a value
type is the result of bitwise-zeroing the value type’s fields:
static void Zap<T> (T[] array)
{
for (int i = 0; i < array.Length; i++)
array[i] = default(T);
}
From C# 7.1, you can omit the type argument for cases in which the compiler is able
to infer it. We could replace the last line of code with this:
array[i] = default;
Generic Constraints
By default, you can substitute a type parameter with any type whatsoever. Con‐
straints can be applied to a type parameter to require more specific type arguments.
These are the possible constraints:
where T : base-class // Base-class constraint
where T : interface // Interface constraint
where T : class // Reference-type constraint
where T : class? // (see "Nullable Reference Types" in Chapter 1)
where T : struct // Value-type constraint (excludes Nullable types)
where T : unmanaged // Unmanaged constraint
where T : new() // Parameterless constructor constraint
where U : T // Naked type constraint
Generics | 149
Creating
Types
in
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-167-320.jpg)
![The unmanaged constraint (introduced in C# 7.3) is a stronger version of a struct
constraint: T must be a simple value type or a struct that is (recursively) free of any
reference types.
The parameterless constructor constraint requires T to have a public parameterless
constructor. If this constraint is defined, you can call new() on T:
static void Initialize<T> (T[] array) where T : new()
{
for (int i = 0; i < array.Length; i++)
array[i] = new T();
}
The naked type constraint requires one type parameter to derive from (or match)
another type parameter. In this example, the method FilteredStack returns
another Stack, containing only the subset of elements where the type parameter
U is of the type parameter T:
class Stack<T>
{
Stack<U> FilteredStack<U>() where U : T {...}
}
Subclassing Generic Types
A generic class can be subclassed just like a nongeneric class. The subclass can leave
the base class’s type parameters open, as in the following example:
class Stack<T> {...}
class SpecialStack<T> : Stack<T> {...}
Or, the subclass can close the generic type parameters with a concrete type:
class IntStack : Stack<int> {...}
A subtype can also introduce fresh type arguments:
class List<T> {...}
class KeyedList<T,TKey> : List<T> {...}
Technically, all type arguments on a subtype are fresh: you
could say that a subtype closes and then reopens the base
type arguments. This means that a subclass can give new (and
potentially more meaningful) names to the type arguments
that it reopens:
class List<T> {...}
class KeyedList<TElement,TKey> : List<TElement> {...}
Self-Referencing Generic Declarations
A type can name itself as the concrete type when closing a type argument:
public interface IEquatable<T> { bool Equals (T obj); }
public class Balloon : IEquatable<Balloon>
Generics | 151
Creating
Types
in
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-169-320.jpg)
![use the as operator, which is unambiguous because it cannot perform custom
conversions:
StringBuilder Foo<T> (T arg)
{
StringBuilder sb = arg as StringBuilder;
if (sb != null) return sb;
...
}
A more general solution is to first cast to object. This works because conversions
to/from object are assumed not to be custom conversions, but reference or box‐
ing/unboxing conversions. In this case, StringBuilder is a reference type, so it
must be a reference conversion:
return (StringBuilder) (object) arg;
Unboxing conversions can also introduce ambiguities. The following could be an
unboxing, numeric, or custom conversion:
int Foo<T> (T x) => (int) x; // Compile-time error
The solution, again, is to first cast to object and then to int (which then unambig‐
uously signals an unboxing conversion in this case):
int Foo<T> (T x) => (int) (object) x;
Covariance
Assuming A is convertible to B, X has a covariant type parameter if X<A> is converti‐
ble to X<B>.
With C#’s notion of covariance (and contravariance), “conver‐
tible” means convertible via an implicit reference conversion—
such as A subclassing B, or A implementing B. Numeric conver‐
sions, boxing conversions, and custom conversions are not
included.
For instance, type IFoo<T> has a covariant T if the following is legal:
IFoo<string> s = ...;
IFoo<object> b = s;
Interfaces permit covariant type parameters (as do delegates; see Chapter 4), but
classes do not. Arrays also allow covariance (A[] can be converted to B[] if A has an
implicit reference conversion to B) and are discussed here for comparison.
Covariance and contravariance (or simply “variance”) are
advanced concepts. The motivation behind introducing and
enhancing variance in C# was to allow generic interface and
generic types (in particular, those defined in .NET, such as
IEnumerable<T>) to work more as you’d expect. You can
benefit from this without understanding the details behind
covariance and contravariance.
Generics | 153
Creating
Types
in
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-171-320.jpg)
![Variance is not automatic
To ensure static type safety, type parameters are not automatically variant. Consider
the following:
class Animal {}
class Bear : Animal {}
class Camel : Animal {}
public class Stack<T> // A simple Stack implementation
{
int position;
T[] data = new T[100];
public void Push (T obj) => data[position++] = obj;
public T Pop() => data[--position];
}
The following fails to compile:
Stack<Bear> bears = new Stack<Bear>();
Stack<Animal> animals = bears; // Compile-time error
That restriction prevents the possibility of runtime failure with the following code:
animals.Push (new Camel()); // Trying to add Camel to bears
Lack of covariance, however, can hinder reusability. Suppose, for instance, that we
wanted to write a method to Wash a stack of animals:
public class ZooCleaner
{
public static void Wash (Stack<Animal> animals) {...}
}
Calling Wash with a stack of bears would generate a compile-time error. One work‐
around is to redefine the Wash method with a constraint:
class ZooCleaner
{
public static void Wash<T> (Stack<T> animals) where T : Animal { ... }
}
We can now call Wash as follows:
Stack<Bear> bears = new Stack<Bear>();
ZooCleaner.Wash (bears);
Another solution is to have Stack<T> implement an interface with a covariant type
parameter, as you’ll see shortly.
Arrays
For historical reasons, array types support covariance. This means that B[] can be
cast to A[] if B subclasses A (and both are reference types):
Bear[] bears = new Bear[3];
Animal[] animals = bears; // OK
154 | Chapter 3: Creating Types in C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-172-320.jpg)
![The downside of this reusability is that element assignments can fail at runtime:
animals[0] = new Camel(); // Runtime error
Declaring a covariant type parameter
Type parameters on interfaces and delegates can be declared covariant by marking
them with the out modifier. This modifier ensures that, unlike with arrays, cova‐
riant type parameters are fully type-safe.
We can illustrate this with our Stack<T> class by having it implement the following
interface:
public interface IPoppable<out T> { T Pop(); }
The out modifier on T indicates that T is used only in output positions (e.g., return
types for methods). The out modifier flags the type parameter as covariant and
allows us to do this:
var bears = new Stack<Bear>();
bears.Push (new Bear());
// Bears implements IPoppable<Bear>. We can convert to IPoppable<Animal>:
IPoppable<Animal> animals = bears; // Legal
Animal a = animals.Pop();
The conversion from bears to animals is permitted by the compiler—by virtue of
the type parameter being covariant. This is type-safe because the case the compiler
is trying to avoid—pushing a Camel onto the stack—can’t occur, because there’s no
way to feed a Camel into an interface where T can appear only in output positions.
Covariance (and contravariance) in interfaces is something
that you typically consume: it’s less common that you need to
write variant interfaces.
Curiously, method parameters marked as out are not eligible
for covariance, due to a limitation in the CLR.
We can leverage the ability to cast covariantly to solve the reusability problem
described earlier:
public class ZooCleaner
{
public static void Wash (IPoppable<Animal> animals) { ... }
}
The IEnumerator<T> and IEnumerable<T> interfaces
described in Chapter 7 have a covariant T. This allows you
to cast IEnumerable<string> to IEnumerable<object>, for
instance.
The compiler will generate an error if you use a covariant type parameter in an
input position (e.g., a parameter to a method or a writable property).
Generics | 155
Creating
Types
in
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-173-320.jpg)
![delegate int Transformer (int x); // Delegate type declaration
A delegate instance literally acts as a delegate for the caller: the caller invokes the
delegate, and then the delegate calls the target method. This indirection decouples
the caller from the target method.
The statement
Transformer t = Square;
is shorthand for:
Transformer t = new Transformer (Square);
Technically, we are specifying a method group when we refer
to Square without brackets or arguments. If the method is
overloaded, C# will pick the correct overload based on the
signature of the delegate to which it’s being assigned.
The expression
t(3)
is shorthand for:
t.Invoke(3)
A delegate is similar to a callback, a general term that captures
constructs such as C function pointers.
Writing Plug-In Methods with Delegates
A delegate variable is assigned a method at runtime. This is useful for writing
plug-in methods. In this example, we have a utility method named Transform that
applies a transform to each element in an integer array. The Transform method has
a delegate parameter, which you can use for specifying a plug-in transform:
int[] values = { 1, 2, 3 };
Transform (values, Square); // Hook in the Square method
foreach (int i in values)
Console.Write (i + " "); // 1 4 9
void Transform (int[] values, Transformer t)
{
for (int i = 0; i < values.Length; i++)
values[i] = t (values[i]);
}
int Square (int x) => x * x;
int Cube (int x) => x * x * x;
delegate int Transformer (int x);
160 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-178-320.jpg)
![public delegate void ProgressReporter (int percentComplete);
public class Util
{
public static void HardWork (ProgressReporter p)
{
for (int i = 0; i < 10; i++)
{
p (i * 10); // Invoke delegate
System.Threading.Thread.Sleep (100); // Simulate hard work
}
}
}
To monitor progress, we can create a multicast delegate instance p, such that pro‐
gress is monitored by two independent methods:
ProgressReporter p = WriteProgressToConsole;
p += WriteProgressToFile;
Util.HardWork (p);
void WriteProgressToConsole (int percentComplete)
=> Console.WriteLine (percentComplete);
void WriteProgressToFile (int percentComplete)
=> System.IO.File.WriteAllText ("progress.txt",
percentComplete.ToString());
Generic Delegate Types
A delegate type can contain generic type parameters:
public delegate T Transformer<T> (T arg);
With this definition, we can write a generalized Transform utility method that
works on any type:
int[] values = { 1, 2, 3 };
Util.Transform (values, Square); // Hook in Square
foreach (int i in values)
Console.Write (i + " "); // 1 4 9
int Square (int x) => x * x;
public class Util
{
public static void Transform<T> (T[] values, Transformer<T> t)
{
for (int i = 0; i < values.Length; i++)
values[i] = t (values[i]);
}
}
Delegates | 163
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-181-320.jpg)
![The Func and Action Delegates
With generic delegates, it becomes possible to write a small set of delegate types that
are so general they can work for methods of any return type and any (reasonable)
number of arguments. These delegates are the Func and Action delegates, defined
in the System namespace (the in and out annotations indicate variance, which we
cover in the context of delegates shortly):
delegate TResult Func <out TResult> ();
delegate TResult Func <in T, out TResult> (T arg);
delegate TResult Func <in T1, in T2, out TResult> (T1 arg1, T2 arg2);
... and so on, up to T16
delegate void Action ();
delegate void Action <in T> (T arg);
delegate void Action <in T1, in T2> (T1 arg1, T2 arg2);
... and so on, up to T16
These delegates are extremely general. The Transformer delegate in our previous
example can be replaced with a Func delegate that takes a single argument of type T
and returns a same-typed value:
public static void Transform<T> (T[] values, Func<T,T> transformer)
{
for (int i = 0; i < values.Length; i++)
values[i] = transformer (values[i]);
}
The only practical scenarios not covered by these delegates are ref/out and pointer
parameters.
When C# was first introduced, the Func and Action delegates
did not exist (because generics did not exist). It’s for this his‐
torical reason that much of .NET uses custom delegate types
rather than Func and Action.
Delegates Versus Interfaces
A problem that you can solve with a delegate can also be solved with an interface.
For instance, we can rewrite our original example with an interface called ITrans
former instead of a delegate:
int[] values = { 1, 2, 3 };
Util.TransformAll (values, new Squarer());
foreach (int i in values)
Console.WriteLine (i);
public interface ITransformer
{
int Transform (int x);
}
public class Util
{
164 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-182-320.jpg)
![public static void TransformAll (int[] values, ITransformer t)
{
for (int i = 0; i < values.Length; i++)
values[i] = t.Transform (values[i]);
}
}
class Squarer : ITransformer
{
public int Transform (int x) => x * x;
}
A delegate design might be a better choice than an interface design if one or more of
these conditions are true:
• The interface defines only a single method.
•
• Multicast capability is needed.
•
• The subscriber needs to implement the interface multiple times.
•
In the ITransformer example, we don’t need to multicast. However, the interface
defines only a single method. Furthermore, our subscriber might need to implement
ITransformer multiple times, to support different transforms, such as square or
cube. With interfaces, we’re forced into writing a separate type per transform
because a class can implement ITransformer only once. This is quite cumbersome:
int[] values = { 1, 2, 3 };
Util.TransformAll (values, new Cuber());
foreach (int i in values)
Console.WriteLine (i);
class Squarer : ITransformer
{
public int Transform (int x) => x * x;
}
class Cuber : ITransformer
{
public int Transform (int x) => x * x * x;
}
Delegate Compatibility
Type compatibility
Delegate types are all incompatible with one another, even if their signatures are the
same:
D1 d1 = Method1;
D2 d2 = d1; // Compile-time error
void Method1() { }
Delegates | 165
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-183-320.jpg)
![From C# 9, you can ensure that a lambda expression, local function, or anonymous
method doesn’t capture state by applying the static keyword. This can be useful
in micro-optimization scenarios to prevent unintentional memory allocations. For
example, we can apply the static modifier to a lambda expression as follows:
Func<int, int> multiplier = static n => n * 2;
If we later try to modify the lambda expression such that it captures a local variable,
the compiler will generate an error:
int factor = 2;
Func<int, int> multiplier = static n => n * factor; // will not compile
The lambda itself evaluates to a delegate instance, which
requires a memory allocation. However, if the lambda doesn’t
capture variables, the compiler will reuse a single cached
instance across the life of the application, so there will be no
cost in practice.
This feature can also be used with local methods. In the following example, the
Multiply method cannot access the factor variable:
void Foo()
{
int factor = 123;
static int Multiply (int x) => x * 2; // Local static method
}
Of course, the Multiply method could still explicitly allocate memory by calling
new. What this protects us from is a potential allocation by stealth. Applying static
here is also arguably useful as a documentation tool, indicating a reduced level of
coupling.
Static lambdas can still access static variables and constants (because these do not
require a closure).
The static keyword acts merely as a check; it has no effect
on the IL that the compiler produces. Without the static
keyword, the compiler does not generate a closure unless it
needs to (and even then, it has tricks to mitigate the cost).
Capturing iteration variables
When you capture the iteration variable of a for loop, C# treats that variable as
though it were declared outside the loop. This means that the same variable is
captured in each iteration. The following program writes 333 instead of 012:
Action[] actions = new Action[3];
for (int i = 0; i < 3; i++)
actions [i] = () => Console.Write (i);
foreach (Action a in actions) a(); // 333
178 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-196-320.jpg)
![Each closure (shown in boldface) captures the same variable, i. (This actually makes
sense when you consider that i is a variable whose value persists between loop
iterations; you can even explicitly change i within the loop body if you want.) The
consequence is that when the delegates are later invoked, each delegate sees i’s value
at the time of invocation—which is 3. We can illustrate this better by expanding the
for loop, as follows:
Action[] actions = new Action[3];
int i = 0;
actions[0] = () => Console.Write (i);
i = 1;
actions[1] = () => Console.Write (i);
i = 2;
actions[2] = () => Console.Write (i);
i = 3;
foreach (Action a in actions) a(); // 333
The solution, if we want to write 012, is to assign the iteration variable to a local
variable that’s scoped within the loop:
Action[] actions = new Action[3];
for (int i = 0; i < 3; i++)
{
int loopScopedi = i;
actions [i] = () => Console.Write (loopScopedi);
}
foreach (Action a in actions) a(); // 012
Because loopScopedi is freshly created on every iteration, each closure captures a
different variable.
Prior to C# 5.0, foreach loops worked in the same way. This
caused considerable confusion: unlike with a for loop, the
iteration variable in a foreach loop is immutable, and so you
would expect it to be treated as local to the loop body. The
good news is that it’s now fixed, and you can safely capture a
foreach loop’s iteration variable without surprises.
Lambda Expressions Versus Local Methods
The functionality of local methods (see “Local methods” on page 19) overlaps with
that of lambda expressions. Local methods have the following three advantages:
• They can be recursive (they can call themselves) without ugly hacks.
•
• They avoid the clutter of specifying a delegate type.
•
• They incur slightly less overhead.
•
Local methods are more efficient because they avoid the indirection of a delegate
(which costs some CPU cycles and a memory allocation). They can also access
local variables of the containing method without the compiler having to “hoist” the
captured variables into a hidden class.
Lambda Expressions | 179
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-197-320.jpg)
![class Test
{
static void Main (string[] args)
{
try
{
byte b = byte.Parse (args[0]);
Console.WriteLine (b);
}
catch (IndexOutOfRangeException)
{
Console.WriteLine ("Please provide at least one argument");
}
catch (FormatException)
{
Console.WriteLine ("That's not a number!");
}
catch (OverflowException)
{
Console.WriteLine ("You've given me more than a byte!");
}
}
}
Only one catch clause executes for a given exception. If you want to include a safety
net to catch more general exceptions (such as System.Exception), you must put the
more-specific handlers first.
An exception can be caught without specifying a variable, if you don’t need to access
its properties:
catch (OverflowException) // no variable
{
...
}
Furthermore, you can omit both the variable and the type (meaning that all excep‐
tions will be caught):
catch { ... }
Exception filters
You can specify an exception filter in a catch clause by adding a when clause:
catch (WebException ex) when (ex.Status == WebExceptionStatus.Timeout)
{
...
}
If a WebException is thrown in this example, the Boolean expression following the
when keyword is then evaluated. If the result is false, the catch block in question is
ignored and any subsequent catch clauses are considered. With exception filters, it
can be meaningful to catch the same exception type again:
try Statements and Exceptions | 183
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-201-320.jpg)
![}
void Display (string name)
{
if (name == null)
throw new ArgumentNullException (nameof (name));
Console.WriteLine (name);
}
Because null-checking an argument and throwing an
ArgumentNullException is such a common code path, there’s
actually a shortcut for it, from .NET 6:
void Display (string name)
{
ArgumentNullException.ThrowIfNull (name);
Console.WriteLine (name);
}
Notice that we didn’t need to specify the name of the parame‐
ter. We’ll explain why later, in “CallerArgumentExpression (C#
10)” on page 231.
throw expressions
throw can also appear as an expression in expression-bodied functions:
public string Foo() => throw new NotImplementedException();
A throw expression can also appear in a ternary conditional expression:
string ProperCase (string value) =>
value == null ? throw new ArgumentException ("value") :
value == "" ? "" :
char.ToUpper (value[0]) + value.Substring (1);
Rethrowing an exception
You can capture and rethrow an exception as follows:
try { ... }
catch (Exception ex)
{
// Log error
...
throw; // Rethrow same exception
}
If we replaced throw with throw ex, the example would still
work, but the StackTrace property of the newly propagated
exception would no longer reflect the original error.
Rethrowing in this manner lets you log an error without swallowing it. It also lets
you back out of handling an exception should circumstances turn out to be beyond
186 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-204-320.jpg)
![{ 10, "ten" }
};
Or, more succinctly:
var dict = new Dictionary<int, string>()
{
[3] = "three",
[10] = "ten"
};
The latter is valid not only with dictionaries but with any type for which an indexer
exists.
Iterators
Whereas a foreach statement is a consumer of an enumerator, an iterator is a
producer of an enumerator. In this example, we use an iterator to return a sequence
of Fibonacci numbers (where each number is the sum of the previous two):
using System;
using System.Collections.Generic;
foreach (int fib in Fibs(6))
Console.Write (fib + " ");
}
IEnumerable<int> Fibs (int fibCount)
{
for (int i = 0, prevFib = 1, curFib = 1; i < fibCount; i++)
{
yield return prevFib;
int newFib = prevFib+curFib;
prevFib = curFib;
curFib = newFib;
}
}
OUTPUT: 1 1 2 3 5 8
Whereas a return statement expresses, “Here’s the value you asked me to return
from this method,” a yield return statement expresses, “Here’s the next element
you asked me to yield from this enumerator.” On each yield statement, control is
returned to the caller, but the callee’s state is maintained so that the method can
continue executing as soon as the caller enumerates the next element. The lifetime
of this state is bound to the enumerator such that the state can be released when the
caller has finished enumerating.
Enumeration and Iterators | 191
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-209-320.jpg)
![EmailAddress column on a Customer table), there is no problem because string is
a reference type in the CLR, which can be null. However, most other SQL column
types map to CLR struct types, making nullable value types very useful when
mapping SQL to the CLR:
// Maps to a Customer table in a database
public class Customer
{
...
public decimal? AccountBalance;
}
A nullable type can also be used to represent the backing field of what’s sometimes
called an ambient property. An ambient property, if null, returns the value of its
parent:
public class Row
{
...
Grid parent;
Color? color;
public Color Color
{
get { return color ?? parent.Color; }
set { color = value == parent.Color ? (Color?)null : value; }
}
}
Alternatives to Nullable Value Types
Before nullable value types were part of the C# language (i.e., before C# 2.0), there
were many strategies to deal with them, examples of which still appear in the .NET
libraries for historical reasons. One of these strategies is to designate a particular
non-null value as the “null value”; an example is in the string and array classes.
String.IndexOf returns the magic value of −1 when the character is not found:
int i = "Pink".IndexOf ('b');
Console.WriteLine (i); // −1
However, Array.IndexOf returns −1 only if the index is 0-bounded. The more
general formula is that IndexOf returns one less than the lower bound of the array.
In the next example, IndexOf returns 0 when an element is not found:
// Create an array whose lower bound is 1 instead of 0:
Array a = Array.CreateInstance (typeof (string),
new int[] {2}, new int[] {1});
a.SetValue ("a", 1);
a.SetValue ("b", 2);
Console.WriteLine (Array.IndexOf (a, "c")); // 0
Nullable Value Types | 199
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-217-320.jpg)
![Because nullable reference types are compile-time constructs,
there’s no runtime difference between string and string?.
In contrast, nullable value types introduce something concrete
into the type system, namely the Nullable<T> struct.
The following also generates a warning because x is not initialized:
class Foo { string x; }
The warning disappears if you initialize x, either via a field initializer or via code in
the constructor.
The Null-Forgiving Operator
The compiler also warns you upon dereferencing a nullable reference type, if it
thinks a NullReferenceException might occur. In the following example, accessing
the string’s Length property generates a warning:
void Foo (string? s) => Console.Write (s.Length);
You can remove the warning with the null-forgiving operator (!):
void Foo (string? s) => Console.Write (s!.Length);
Our use of the null-forgiving operator in this example is dangerous in that we could
end up throwing the very NullReferenceException we were trying to avoid in the
first place. We could fix it as follows:
void Foo (string? s)
{
if (s != null) Console.Write (s.Length);
}
Notice now that we don’t need the null-forgiving operator. This is because the
compiler performs static flow analysis and is smart enough to infer—at least in
simple cases—when a dereference is safe and there’s no chance of a NullReference
Exception.
The compiler’s ability to detect and warn is not bulletproof, and there are also limits
to what’s possible in terms of coverage. For instance, it’s unable to know whether
an array’s elements have been populated, and so the following does not generate a
warning:
var strings = new string[10];
Console.WriteLine (strings[0].Length);
Separating the Annotation and Warning Contexts
Enabling nullable reference types via the #nullable enable directive (or the
<Nullable>enable</Nullable> project setting) does two things:
Nullable Reference Types | 201
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-219-320.jpg)
![if (string.IsNullOrEmpty(s)) return false;
return char.IsUpper (s[0]);
}
}
The IsCapitalized extension method can be called as though it were an instance
method on a string, as follows:
Console.WriteLine ("Perth".IsCapitalized());
An extension method call, when compiled, is translated back into an ordinary static
method call:
Console.WriteLine (StringHelper.IsCapitalized ("Perth"));
The translation works as follows:
arg0.Method (arg1, arg2, ...); // Extension method call
StaticClass.Method (arg0, arg1, arg2, ...); // Static method call
Interfaces can be extended, too:
public static T First<T> (this IEnumerable<T> sequence)
{
foreach (T element in sequence)
return element;
throw new InvalidOperationException ("No elements!");
}
...
Console.WriteLine ("Seattle".First()); // S
Extension Method Chaining
Extension methods, like instance methods, provide a tidy way to chain functions.
Consider the following two functions:
public static class StringHelper
{
public static string Pluralize (this string s) {...}
public static string Capitalize (this string s) {...}
}
x and y are equivalent, and both evaluate to "Sausages", but x uses extension
methods, whereas y uses static methods:
string x = "sausage".Pluralize().Capitalize();
string y = StringHelper.Capitalize (StringHelper.Pluralize ("sausage"));
Ambiguity and Resolution
Namespaces
An extension method cannot be accessed unless its class is in scope, typically by its
namespace being imported. Consider the extension method IsCapitalized in the
following example:
Extension Methods | 203
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-221-320.jpg)
![using System;
namespace Utils
{
public static class StringHelper
{
public static bool IsCapitalized (this string s)
{
if (string.IsNullOrEmpty(s)) return false;
return char.IsUpper (s[0]);
}
}
}
To use IsCapitalized, the following application must import Utils to avoid a
compile-time error:
namespace MyApp
{
using Utils;
class Test
{
static void Main() => Console.WriteLine ("Perth".IsCapitalized());
}
}
Extension methods versus instance methods
Any compatible instance method will always take precedence over an extension
method. In the following example, Test’s Foo method will always take precedence,
even when called with an argument x of type int:
class Test
{
public void Foo (object x) { } // This method always wins
}
static class Extensions
{
public static void Foo (this Test t, int x) { }
}
The only way to call the extension method in this case is via normal static syntax, in
other words, Extensions.Foo(...).
Extension methods versus extension methods
If two extension methods have the same signature, the extension method must
be called as an ordinary static method to disambiguate the method to call. If one
extension method has more specific arguments, however, the more specific method
takes precedence.
To illustrate, consider the following two classes:
204 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-222-320.jpg)
![}
public string Name { get { return name; } }
public int Age { get { return age; } }
// The Equals and GetHashCode methods are overridden (see Chapter 6).
// The ToString method is also overridden.
}
...
var dude = new AnonymousGeneratedTypeName ("Bob", 23);
You must use the var keyword to reference an anonymous type because it doesn’t
have a name.
The property name of an anonymous type can be inferred from an expression that
is itself an identifier (or ends with one); thus
int Age = 23;
var dude = new { Name = "Bob", Age, Age.ToString().Length };
is equivalent to the following:
var dude = new { Name = "Bob", Age = Age, Length = Age.ToString().Length };
Two anonymous type instances declared within the same assembly will have the
same underlying type if their elements are named and typed identically:
var a1 = new { X = 2, Y = 4 };
var a2 = new { X = 2, Y = 4 };
Console.WriteLine (a1.GetType() == a2.GetType()); // True
Additionally, the Equals method is overridden to perform structural equality com‐
parison (comparison of the data):
Console.WriteLine (a1.Equals (a2)); // True
whereas the equality operator (==) performs referential comparison:
Console.WriteLine (a1 == a2); // False
You can create arrays of anonymous types as follows:
var dudes = new[]
{
new { Name = "Bob", Age = 30 },
new { Name = "Tom", Age = 40 }
};
A method cannot (usefully) return an anonymously typed object, because it is illegal
to write a method whose return type is var:
var Foo() => new { Name = "Bob", Age = 30 }; // Not legal!
Instead, you must use object or dynamic, and then whoever calls Foo must rely on
dynamic binding, with loss of static type safety (and IntelliSense in Visual Studio):
dynamic Foo() => new { Name = "Bob", Age = 30 }; // No static type safety.
206 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-224-320.jpg)
![Attributes
You’re already familiar with the notion of attributing code elements of a program
with modifiers, such as virtual or ref. These constructs are built into the language.
Attributes are an extensible mechanism for adding custom information to code
elements (assemblies, types, members, return values, parameters, and generic type
parameters). This extensibility is useful for services that integrate deeply into the
type system, without requiring special keywords or constructs in the C# language.
A good scenario for attributes is serialization—the process of converting arbitrary
objects to and from a particular format for storage or transmission. In this scenario,
an attribute on a field can specify the translation between C#’s representation of the
field and the format’s representation of the field.
Attribute Classes
An attribute is defined by a class that inherits (directly or indirectly) from the
abstract class System.Attribute. To attach an attribute to a code element, specify
the attribute’s type name in square brackets, before the code element. For example,
the following attaches the ObsoleteAttribute to the Foo class:
[ObsoleteAttribute]
public class Foo {...}
This particular attribute is recognized by the compiler and will cause compiler
warnings if a type or member marked as obsolete is referenced. By convention, all
attribute types end in the word Attribute. C# recognizes this and allows you to omit
the suffix when attaching an attribute:
[Obsolete]
public class Foo {...}
ObsoleteAttribute is a type declared in the System namespace as follows (simpli‐
fied for brevity):
public sealed class ObsoleteAttribute : Attribute {...}
The .NET libraries include many predefined attributes. We describe how to write
your own attributes in Chapter 18.
Named and Positional Attribute Parameters
Attributes can have parameters. In the following example, we apply XmlTypeAttri
bute to a class. This attribute instructs the XML serializer (in System.Xml.Seriali
zation) as to how an object is represented in XML and accepts several attribute
parameters. The following attribute maps the CustomerEntity class to an XML
element named Customer, which belongs to the http://oreilly.com namespace:
[XmlType ("Customer", Namespace="http://oreilly.com")]
public class CustomerEntity { ... }
Attributes | 227
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-245-320.jpg)
![Attribute parameters fall into one of two categories: positional or named. In the
preceding example, the first argument is a positional parameter; the second is a
named parameter. Positional parameters correspond to parameters of the attribute
type’s public constructors. Named parameters correspond to public fields or public
properties on the attribute type.
When specifying an attribute, you must include positional parameters that corre‐
spond to one of the attribute’s constructors. Named parameters are optional.
In Chapter 18, we describe the valid parameter types and rules for their evaluation.
Applying Attributes to Assemblies and Backing Fields
Implicitly, the target of an attribute is the code element it immediately precedes,
which is typically a type or type member. You can also attach attributes, however,
to an assembly. This requires that you explicitly specify the attribute’s target. Here
is how you can use the AssemblyFileVersion attribute to attach a version to the
assembly:
[assembly: AssemblyFileVersion ("1.2.3.4")]
From C# 7.3, you can use the field: prefix to apply an attribute to the backing
fields of an automatic property. This can be useful in controlling serialization:
[field:NonSerialized]
public int MyProperty { get; set; }
Applying Attributes to Lambda Expressions (C# 10)
From C# 10, you can apply attributes to the method, parameters, and return value
of a lambda expression:
Action<int> a = [Description ("Method")]
[return: Description ("Return value")]
([Description ("Parameter")]int x) => Console.Write (x);
This is useful when working with frameworks—such as
ASP.NET—that rely on you placing attributes on methods that
you write. With this feature, you can avoid having to create
named methods for simple operations.
These attributes are applied to the compiler-generated method to which the delegate
points. In Chapter 18, we’ll describe how to reflect over attributes in code. For now,
here’s the extra code you need to resolve that indirection:
var methodAtt = a.GetMethodInfo().GetCustomAttributes();
var paramAtt = a.GetMethodInfo().GetParameters()[0].GetCustomAttributes();
var returnAtt = a.GetMethodInfo().ReturnParameter.GetCustomAttributes();
To avoid syntactical ambiguity when applying attributes to a parameter on a
lambda expression, parentheses are always required. Attributes are not permitted
on expression-tree lambdas.
228 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-246-320.jpg)
![Specifying Multiple Attributes
You can specify multiple attributes for a single code element. You can list each
attribute either within the same pair of square brackets (separated by a comma) or
in separate pairs of square brackets (or a combination of the two). The following
three examples are semantically identical:
[Serializable, Obsolete, CLSCompliant(false)]
public class Bar {...}
[Serializable] [Obsolete] [CLSCompliant(false)]
public class Bar {...}
[Serializable, Obsolete]
[CLSCompliant(false)]
public class Bar {...}
Caller Info Attributes
You can tag optional parameters with one of three caller info attributes, which
instruct the compiler to feed information obtained from the caller’s source code into
the parameter’s default value:
• [CallerMemberName] applies the caller’s member name.
•
• [CallerFilePath] applies the path to the caller’s source code file.
•
• [CallerLineNumber] applies the line number in the caller’s source code file.
•
The Foo method in the following program demonstrates all three:
using System;
using System.Runtime.CompilerServices;
class Program
{
static void Main() => Foo();
static void Foo (
[CallerMemberName] string memberName = null,
[CallerFilePath] string filePath = null,
[CallerLineNumber] int lineNumber = 0)
{
Console.WriteLine (memberName);
Console.WriteLine (filePath);
Console.WriteLine (lineNumber);
}
}
Assuming that our program resides in c:sourcetestProgram.cs, the output would
be:
Caller Info Attributes | 229
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-247-320.jpg)
![Main
c:sourcetestProgram.cs
6
As with standard optional parameters, the substitution is done at the calling site.
Hence, our Main method is syntactic sugar for this:
static void Main() => Foo ("Main", @"c:sourcetestProgram.cs", 6);
Caller info attributes are useful for logging—and for implementing patterns such
as firing a single change notification event whenever any property on an object
changes. In fact, there’s a standard interface for this in the System.ComponentModel
namespace, called INotifyPropertyChanged:
public interface INotifyPropertyChanged
{
event PropertyChangedEventHandler PropertyChanged;
}
public delegate void PropertyChangedEventHandler
(object sender, PropertyChangedEventArgs e);
public class PropertyChangedEventArgs : EventArgs
{
public PropertyChangedEventArgs (string propertyName);
public virtual string PropertyName { get; }
}
Notice that PropertyChangedEventArgs requires the name of the property that
changed. By applying the [CallerMemberName] attribute, however, we can imple‐
ment this interface and invoke the event without ever specifying property names:
public class Foo : INotifyPropertyChanged
{
public event PropertyChangedEventHandler PropertyChanged = delegate { };
void RaisePropertyChanged ([CallerMemberName] string propertyName = null)
=> PropertyChanged (this, new PropertyChangedEventArgs (propertyName));
string customerName;
public string CustomerName
{
get => customerName;
set
{
if (value == customerName) return;
customerName = value;
RaisePropertyChanged();
// The compiler converts the above line to:
// RaisePropertyChanged ("CustomerName");
}
}
}
230 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-248-320.jpg)
![CallerArgumentExpression (C# 10)
A method parameter to which you apply the [CallerArgumentExpression]
attribute captures an argument expression from the call site:
Print (Math.PI * 2);
void Print (double number,
[CallerArgumentExpression("number")] string expr = null)
=> Console.WriteLine (expr);
// Output: Math.PI * 2
The compiler feeds in the calling expression’s source code literally, including
comments:
Print (Math.PI /*(π)*/ * 2);
// Output: Math.PI /*(π)*/ * 2
The main application for this feature is when writing validation and assertion
libraries. In the following example, an exception is thrown, whose message includes
the text “2 + 2 == 5”. This aids in debugging:
Assert (2 + 2 == 5);
void Assert (bool condition,
[CallerArgumentExpression ("condition")] string message = null)
{
if (!condition) throw new Exception ("Assertion failed: " + message);
}
Another example is the static ThrowIfNull method on the ArgumentNullException
class. This method was introduced in .NET 6 and is defined as follows:
public static void ThrowIfNull (object argument,
[CallerArgumentExpression("argument")] string paramName = null)
{
if (argument == null)
throw new ArgumentNullException (paramName);
}
It is used as follows:
void Print (string message)
{
ArgumentNullException.ThrowIfNull (message);
...
}
You can use [CallerArgumentExpression] multiple times, to capture multiple
argument expressions.
Caller Info Attributes | 231
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-249-320.jpg)
![A dynamic type is like object—it’s equally nondescriptive about a type. The dif‐
ference is that it lets you use it in ways that aren’t known at compile time. A
dynamic object binds at runtime based on its runtime type, not its compile-time
type. When the compiler sees a dynamically bound expression (which in general is
an expression that contains any value of type dynamic), it merely packages up the
expression such that the binding can be done later at runtime.
At runtime, if a dynamic object implements IDynamicMetaObjectProvider, that
interface is used to perform the binding. If not, binding occurs in almost the same
way as it would have had the compiler known the dynamic object’s runtime type.
These two alternatives are called custom binding and language binding.
Custom Binding
Custom binding occurs when a dynamic object implements IDynamicMetaObject
Provider (IDMOP). Although you can implement IDMOP on types that you write
in C#, and that is useful to do, the more common case is that you have acquired
an IDMOP object from a dynamic language that is implemented in .NET on the
Dynamic Language Runtime (DLR), such as IronPython or IronRuby. Objects from
those languages implicitly implement IDMOP as a means by which to directly
control the meanings of operations performed on them.
We discuss custom binders in greater detail in Chapter 19, but for now, let’s write a
simple one to demonstrate the feature:
using System;
using System.Dynamic;
dynamic d = new Duck();
d.Quack(); // Quack method was called
d.Waddle(); // Waddle method was called
public class Duck : DynamicObject
{
public override bool TryInvokeMember (
InvokeMemberBinder binder, object[] args, out object result)
{
Console.WriteLine (binder.Name + " method was called");
result = null;
return true;
}
}
The Duck class doesn’t actually have a Quack method. Instead, it uses custom bind‐
ing to intercept and interpret all method calls.
Language Binding
Language binding occurs when a dynamic object does not implement IDynamic
MetaObjectProvider. It is useful when working around imperfectly designed types
or inherent limitations in the .NET type system (we explore more scenarios in
Dynamic Binding | 233
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-251-320.jpg)
![Chapter 19). A typical problem when using numeric types is that they have no
common interface. We have seen that we can bind methods dynamically; the same
is true for operators:
int x = 3, y = 4;
Console.WriteLine (Mean (x, y));
dynamic Mean (dynamic x, dynamic y) => (x + y) / 2;
The benefit is obvious—you don’t need to duplicate code for each numeric
type. However, you lose static type safety, risking runtime exceptions rather than
compile-time errors.
Dynamic binding circumvents static type safety but not run‐
time type safety. Unlike with reflection (Chapter 18), you can’t
circumvent member accessibility rules with dynamic binding.
By design, language runtime binding behaves as similarly as possible to static
binding, had the runtime types of the dynamic objects been known at compile time.
In our previous example, the behavior of our program would be identical if we
hardcoded Mean to work with the int type. The most notable exception in parity
between static and dynamic binding is for extension methods, which we discuss in
“Uncallable Functions” on page 238.
Dynamic binding also incurs a performance hit. Because of
the DLR’s caching mechanisms, however, repeated calls to the
same dynamic expression are optimized—allowing you to effi‐
ciently call dynamic expressions in a loop. This optimization
brings the typical overhead for a simple dynamic expression
on today’s hardware down to less than 100 ns.
RuntimeBinderException
If a member fails to bind, a RuntimeBinderException is thrown. You can think of
this like a compile-time error at runtime:
dynamic d = 5;
d.Hello(); // throws RuntimeBinderException
The exception is thrown because the int type has no Hello method.
Runtime Representation of Dynamic
There is a deep equivalence between the dynamic and object types. The runtime
treats the following expression as true:
typeof (dynamic) == typeof (object)
This principle extends to constructed types and array types:
typeof (List<dynamic>) == typeof (List<object>)
typeof (dynamic[]) == typeof (object[])
234 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-252-320.jpg)
![Like an object reference, a dynamic reference can point to an object of any type
(except pointer types):
dynamic x = "hello";
Console.WriteLine (x.GetType().Name); // String
x = 123; // No error (despite same variable)
Console.WriteLine (x.GetType().Name); // Int32
Structurally, there is no difference between an object reference and a dynamic
reference. A dynamic reference simply enables dynamic operations on the object
it points to. You can convert from object to dynamic to perform any dynamic
operation you want on an object:
object o = new System.Text.StringBuilder();
dynamic d = o;
d.Append ("hello");
Console.WriteLine (o); // hello
Reflecting on a type exposing (public) dynamic members
reveals that those members are represented as annotated
objects; for example
public class Test
{
public dynamic Foo;
}
is equivalent to:
public class Test
{
[System.Runtime.CompilerServices.DynamicAttribute]
public object Foo;
}
This allows consumers of that type to know that Foo should
be treated as dynamic while allowing languages that don’t
support dynamic binding to fall back to object.
Dynamic Conversions
The dynamic type has implicit conversions to and from all other types:
int i = 7;
dynamic d = i;
long j = d; // No cast required (implicit conversion)
For the conversion to succeed, the runtime type of the dynamic object must be
implicitly convertible to the target static type. The preceding example worked
because an int is implicitly convertible to a long.
The following example throws a RuntimeBinderException because an int is not
implicitly convertible to a short:
int i = 7;
dynamic d = i;
short j = d; // throws RuntimeBinderException
Dynamic Binding | 235
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-253-320.jpg)
![memory within that scope. Here is an example of using pointers to quickly process a
bitmap:
unsafe void BlueFilter (int[,] bitmap)
{
int length = bitmap.Length;
fixed (int* b = bitmap)
{
int* p = b;
for (int i = 0; i < length; i++)
*p++ &= 0xFF;
}
}
Unsafe code can run faster than a corresponding safe implementation. In this case,
the code would have required a nested loop with array indexing and bounds check‐
ing. An unsafe C# method can also be faster than calling an external C function
given that there is no overhead associated with leaving the managed execution
environment.
The fixed Statement
The fixed statement is required to pin a managed object, such as the bitmap in the
previous example. During the execution of a program, many objects are allocated
and deallocated from the heap. To avoid unnecessary waste or fragmentation of
memory, the garbage collector moves objects around. Pointing to an object is futile
if its address could change while referencing it, so the fixed statement tells the
garbage collector to “pin” the object and not move it around. This can have an
impact on the efficiency of the runtime, so you should use fixed blocks only briefly,
and you should avoid heap allocation within the fixed block.
Within a fixed statement, you can get a pointer to any value type, an array of value
types, or a string. In the case of arrays and strings, the pointer will actually point to
the first element, which is a value type.
Value types declared inline within reference types require the reference type to be
pinned, as follows:
Test test = new Test();
unsafe
{
fixed (int* p = &test.X) // Pins test
{
*p = 9;
}
Console.WriteLine (test.X);
}
class Test { public int X; }
We describe the fixed statement further in “Mapping a Struct to Unmanaged
Memory” on page 973.
244 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-262-320.jpg)
![The Pointer-to-Member Operator
In addition to the & and * operators, C# also provides the C++ style -> operator,
which you can use on structs:
Test test = new Test();
unsafe
{
Test* p = &test;
p->X = 9;
System.Console.WriteLine (test.X);
}
struct Test { public int X; }
The stackalloc Keyword
You can allocate memory in a block on the stack explicitly by using the stackalloc
keyword. Because it is allocated on the stack, its lifetime is limited to the execution
of the method, just as with any other local variable (whose life hasn’t been extended
by virtue of being captured by a lambda expression, iterator block, or asynchronous
function). The block can use the [] operator to index into memory:
int* a = stackalloc int [10];
for (int i = 0; i < 10; ++i)
Console.WriteLine (a[i]);
In Chapter 23, we describe how you can use Span<T> to manage stack-allocated
memory without using the unsafe keyword:
Span<int> a = stackalloc int [10];
for (int i = 0; i < 10; ++i)
Console.WriteLine (a[i]);
Fixed-Size Buffers
The fixed keyword has another use, which is to create fixed-size buffers within
structs (this can be useful when calling an unmanaged function; see Chapter 24):
new UnsafeClass ("Christian Troy");
unsafe struct UnsafeUnicodeString
{
public short Length;
public fixed byte Buffer[30]; // Allocate block of 30 bytes
}
unsafe class UnsafeClass
{
UnsafeUnicodeString uus;
public UnsafeClass (string s)
{
uus.Length = (short)s.Length;
Unsafe Code and Pointers | 245
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-263-320.jpg)
![fixed (byte* p = uus.Buffer)
for (int i = 0; i < s.Length; i++)
p[i] = (byte) s[i];
}
}
Fixed-size buffers are not arrays: if Buffer were an array, it would consist of a
reference to an object stored on the (managed) heap, rather than 30 bytes within the
struct itself.
The fixed keyword is also used in this example to pin the object on the heap
that contains the buffer (which will be the instance of UnsafeClass). Hence, fixed
means two different things: fixed in size and fixed in place. The two are often used
together, in that a fixed-size buffer must be fixed in place to be used.
void*
A void pointer (void*) makes no assumptions about the type of the underlying data
and is useful for functions that deal with raw memory. An implicit conversion exists
from any pointer type to void*. A void* cannot be dereferenced, and arithmetic
operations cannot be performed on void pointers. Here’s an example:
short[] a = { 1, 1, 2, 3, 5, 8, 13, 21, 34, 55 };
unsafe
{
fixed (short* p = a)
{
//sizeof returns size of value-type in bytes
Zap (p, a.Length * sizeof (short));
}
}
foreach (short x in a)
System.Console.WriteLine (x); // Prints all zeros
unsafe void Zap (void* memory, int byteCount)
{
byte* b = (byte*)memory;
for (int i = 0; i < byteCount; i++)
*b++ = 0;
}
Native-Sized Integers
The nint and nuint native-sized integer types (introduced in C# 9) are sized to
match the address space of the process at runtime (in practice, 32 or 64 bits).
Native-sized integers can improve efficiency, overflow safety, and convenience when
performing pointer arithmetic.
The gain in efficiency arises because when you subtract two pointers in C#, the
result is always a 64-bit integer (long), which is inefficient on 32-bit platforms. By
first casting the pointers to nint, the result of a subtraction is also nint (which will
be 32 bits on a 32-bit platform):
246 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-264-320.jpg)
![A good example of the real-world use of nint and nuint is
in the implementation of Buffer.MemoryCopy. You can see
this in the .NET source code for Buffer.cs on GitHub or by
decompiling the method in ILSpy. A simplified version has
also been included in the LINQPad samples for C# 10 in a
Nutshell.
Function Pointers
A function pointer (from C# 9) is like a delegate, but without the indirection of a
delegate instance; instead, it points directly to a method. A function pointer can
point only to static methods, lacks multicast capability, and requires an unsafe con‐
text (because it bypasses runtime type safety). Its main purpose is to simplify and
optimize interop with unmanaged APIs (see “Callbacks from Unmanaged Code” on
page 967).
A function pointer type is declared as follows (with the return type appearing last):
delegate*<int, char, string, void> // (void refers to the return type)
This matches a function with this signature:
void SomeFunction (int x, char y, string z)
The & operator creates a function pointer from a method group. Here’s a complete
example:
unsafe
{
delegate*<string, int> functionPointer = &GetLength;
int length = functionPointer ("Hello, world");
static int GetLength (string s) => s.Length;
}
In this example, functionPointer is not an object upon which you can call a
method such as Invoke (or with a reference to a Target object). Instead, it’s a
variable that points directly to the target method’s address in memory:
Console.WriteLine ((IntPtr)functionPointer);
Like any other pointer, it’s not subject to runtime type checking. The following
treats our function’s return value as a decimal (which, being longer than an int,
means that we incorporate some random memory into the output):
var pointer2 = (delegate*<string, decimal>) (IntPtr) functionPointer;
Console.WriteLine (pointer2 ("Hello, unsafe world"));
[SkipLocalsInit]
When C# compiles a method, it emits a flag that instructs the runtime to initial‐
ize the method’s local variables to their default values (by zeroing the memory).
From C# 9, you can ask the compiler not to emit this flag by applying the
248 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-266-320.jpg)
![[SkipLocalInit] attribute to a method (in the System.Runtime.CompilerServices
namespace):
[SkipLocalsInit]
void Foo() ...
You can also apply this attribute to a type—which is equivalent to applying it to all
of the type’s methods—or even an entire module (the container for an assembly):
[module: System.Runtime.CompilerServices.SkipLocalsInit]
In normal safe scenarios, [SkipLocalsInit] has little effect on functionality or
performance, because C#’s definite assignment policy requires that you explicitly
assign local variables before they can be read. This means that the JIT optimizer is
likely to emit the same machine code, whether or not the attribute is applied.
In an unsafe context, however, use of [SkipLocalsInit] can usefully save the
CLR from the overhead of initializing value-typed local variables, creating a small
performance gain with methods that make extensive use of the stack (through
a large stackalloc). The following example prints uninitialized memory when
[SkipLocalsInit] is applied (instead of all zeros):
[SkipLocalsInit]
unsafe void Foo()
{
int local;
int* ptr = &local;
Console.WriteLine (*ptr);
int* a = stackalloc int [100];
for (int i = 0; i < 100; ++i) Console.WriteLine (a [i]);
}
Interestingly, you can achieve the same result in a “safe” context through the use of
Span<T>:
[SkipLocalsInit]
void Foo()
{
Span<int> a = stackalloc int [100];
for (int i = 0; i < 100; ++i) Console.WriteLine (a [i]);
}
Consequently, use of [SkipLocalsInit] requires that you compile your assembly
with the unsafe option—even if none of your methods are marked as unsafe.
Preprocessor Directives
Preprocessor directives supply the compiler with additional information about
regions of code. The most common preprocessor directives are the conditional
directives, which provide a way to include or exclude regions of code from
compilation:
Preprocessor Directives | 249
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-267-320.jpg)
![#define DEBUG
class MyClass
{
int x;
void Foo()
{
#if DEBUG
Console.WriteLine ("Testing: x = {0}", x);
#endif
}
...
}
In this class, the statement in Foo is compiled as conditionally dependent upon the
presence of the DEBUG symbol. If we remove the DEBUG symbol, the statement is not
compiled. You can define preprocessor symbols within a source file (as we have
done) or at a project level in the .csproj file:
<PropertyGroup>
<DefineConstants>DEBUG;ANOTHERSYMBOL</DefineConstants>
</PropertyGroup>
With the #if and #elif directives, you can use the ||, &&, and ! operators to
perform or, and, and not operations on multiple symbols. The following directive
instructs the compiler to include the code that follows if the TESTMODE symbol is
defined and the DEBUG symbol is not defined:
#if TESTMODE && !DEBUG
...
Keep in mind, however, that you’re not building an ordinary C# expression, and the
symbols upon which you operate have absolutely no connection to variables—static
or otherwise.
The #error and #warning symbols prevent accidental misuse of conditional direc‐
tives by making the compiler generate a warning or error given an undesirable set of
compilation symbols. Table 4-1 lists the preprocessor directives.
Table 4-1. Preprocessor directives
Preprocessor directive Action
#define symbol Defines symbol
#undef symbol Undefines symbol
#if symbol
[operator symbol2]...
symbol to test
operators are ==, !=, &&, and ||, followed by #else,
#elif, and #endif
#else Executes code to subsequent #endif
#elif symbol
[operator symbol2]
Combines #else branch and #if test
250 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-268-320.jpg)
![Preprocessor directive Action
#endif Ends conditional directives
#warning text text of the warning to appear in compiler output
#error text text of the error to appear in compiler output
#error version Reports the compiler version and exits
#pragma warning
[disable | restore]
Disables/restores compiler warning(s)
#line [ number ["file"] |
hidden]
number specifies the line in source code (a column can also
be specified from C# 10); file is the filename to appear in
computer output; hidden instructs debuggers to skip over
code from this point until the next #line directive
#region name Marks the beginning of an outline
#endregion Ends an outline region
#nullable option See “Nullable reference types” on page 16
Conditional Attributes
An attribute decorated with the Conditional attribute will be compiled only if a
given preprocessor symbol is present:
// file1.cs
#define DEBUG
using System;
using System.Diagnostics;
[Conditional("DEBUG")]
public class TestAttribute : Attribute {}
// file2.cs
#define DEBUG
[Test]
class Foo
{
[Test]
string s;
}
The compiler will incorporate the [Test] attributes only if the DEBUG symbol is in
scope for file2.cs.
Pragma Warning
The compiler generates a warning when it spots something in your code that seems
unintentional. Unlike errors, warnings don’t ordinarily prevent your application
from compiling.
Compiler warnings can be extremely valuable in spotting bugs. Their usefulness,
however, is undermined when you get false warnings. In a large application,
Preprocessor Directives | 251
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-269-320.jpg)
![• If placed in the same folder as the compiled assembly, tools such as Visual
•
Studio and LINQPad automatically read the XML file and use the information
to provide IntelliSense member listings to consumers of the assembly of the
same name.
• Third-party tools (such as Sandcastle and NDoc) can transform the XML file
•
into an HTML help file.
Standard XML Documentation Tags
Here are the standard XML tags that Visual Studio and documentation generators
recognize:
<summary>
<summary>...</summary>
Indicates the tool tip that IntelliSense should display for the type or member,
typically a single phrase or sentence.
<remarks>
<remarks>...</remarks>
Additional text that describes the type or member. Documentation generators
pick this up and merge it into the bulk of a type’s or member’s description.
<param>
<param name="name">...</param>
Explains a parameter on a method.
<returns>
<returns>...</returns>
Explains the return value for a method.
<exception>
<exception [cref="type"]>...</exception>
Lists an exception that a method can throw (cref refers to the exception type).
<example>
<example>...</example>
Denotes an example (used by documentation generators). This usually contains
both description text and source code (source code is typically within a <c> or
<code> tag).
<c>
<c>...</c>
Indicates an inline code snippet. This tag is usually used within an <example>
block.
XML Documentation | 253
Advanced
C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-271-320.jpg)
![<code>
<code>...</code>
Indicates a multiline code sample. This tag is usually used within an <example>
block.
<see>
<see cref="member">...</see>
Inserts an inline cross-reference to another type or member. HTML documen‐
tation generators typically convert this to a hyperlink. The compiler emits a
warning if the type or member name is invalid. To refer to generic types, use
curly braces; for example, cref="Foo{T,U}".
<seealso>
<seealso cref="member">...</seealso>
Cross-references another type or member. Documentation generators typically
write this into a separate “See Also” section at the bottom of the page.
<paramref>
<paramref name="name"/>
References a parameter from within a <summary> or <remarks> tag.
<list>
<list type=[ bullet | number | table ]>
<listheader>
<term>...</term>
<description>...</description>
</listheader>
<item>
<term>...</term>
<description>...</description>
</item>
</list>
Instructs documentation generators to emit a bulleted, numbered, or table-
style list.
<para>
<para>...</para>
Instructs documentation generators to format the contents into a separate
paragraph.
<include>
<include file='filename' path='tagpath[@name="id"]'>...</include>
Merges an external XML file that contains documentation. The path attribute
denotes an XPath query to a specific element in that file.
254 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-272-320.jpg)
![void X() {...}
/// M:NS.MyClass.Y(System.Int32,System.Double@,System.Decimal@)
void Y(int p1, ref double p2, out decimal p3) {...}
/// M:NS.MyClass.Z(System.Char[ ],System.Single[0:,0:])
void Z(char[ ] p1, float[,] p2) {...}
/// M:NS.MyClass.op_Addition(NS.MyClass,NS.MyClass)
public static MyClass operator+(MyClass c1, MyClass c2) {...}
/// M:NS.MyClass.op_Implicit(NS.MyClass)˜System.Int32
public static implicit operator int(MyClass c) {...}
/// M:NS.MyClass.#ctor
MyClass() {...}
/// M:NS.MyClass.Finalize
˜MyClass() {...}
/// M:NS.MyClass.#cctor
static MyClass() {...}
}
}
256 | Chapter 4: Advanced C#](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-274-320.jpg)
![For more granular categorization, char provides a static method called GetUnicode
Category; this returns a UnicodeCategory enumeration whose members are shown
in the rightmost column of Table 6-1.
By explicitly casting from an integer, it’s possible to produce
a char outside the allocated Unicode set. To test a charac‐
ter’s validity, call char.GetUnicodeCategory: if the result is
UnicodeCategory.OtherNotAssigned, the character is invalid.
A char is 16 bits wide—enough to represent any Unicode character in the Basic
Multilingual Plane. To go beyond this, you must use surrogate pairs: we describe the
methods for doing this in “Text Encodings and Unicode” on page 281.
String
A C# string (== System.String) is an immutable (unchangeable) sequence of
characters. In Chapter 2, we described how to express string literals, perform equal‐
ity comparisons, and concatenate two strings. This section covers the remaining
functions for working with strings, exposed through the static and instance mem‐
bers of the System.String class.
Constructing strings
The simplest way to construct a string is to assign a literal, as we saw in Chapter 2:
string s1 = "Hello";
string s2 = "First LinernSecond Line";
string s3 = @"serverfilesharehelloworld.cs";
To create a repeating sequence of characters, you can use string’s constructor:
Console.Write (new string ('*', 10)); // **********
You can also construct a string from a char array. The ToCharArray method does
the reverse:
char[] ca = "Hello".ToCharArray();
string s = new string (ca); // s = "Hello"
string’s constructor is also overloaded to accept various (unsafe) pointer types, in
order to create strings from types such as char*.
Null and empty strings
An empty string has a length of zero. To create an empty string, you can use either
a literal or the static string.Empty field; to test for an empty string, you can either
perform an equality comparison or test its Length property:
string empty = "";
Console.WriteLine (empty == ""); // True
Console.WriteLine (empty == string.Empty); // True
Console.WriteLine (empty.Length == 0); // True
String and Text Handling | 273
.NET
Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-291-320.jpg)
![Because strings are reference types, they can also be null:
string nullString = null;
Console.WriteLine (nullString == null); // True
Console.WriteLine (nullString == ""); // False
Console.WriteLine (nullString.Length == 0); // NullReferenceException
The static string.IsNullOrEmpty method is a useful shortcut for testing whether a
given string is either null or empty.
Accessing characters within a string
A string’s indexer returns a single character at the given index. As with all functions
that operate on strings, this is zero-indexed:
string str = "abcde";
char letter = str[1]; // letter == 'b'
string also implements IEnumerable<char>, so you can foreach over its
characters:
foreach (char c in "123") Console.Write (c + ","); // 1,2,3,
Searching within strings
The simplest methods for searching within strings are StartsWith, EndsWith, and
Contains. These all return true or false:
Console.WriteLine ("quick brown fox".EndsWith ("fox")); // True
Console.WriteLine ("quick brown fox".Contains ("brown")); // True
StartsWith and EndsWith are overloaded to let you specify a StringComparison
enum or a CultureInfo object to control case and culture sensitivity (see “Ordinal
versus culture comparison” on page 278). The default is to perform a case-sensitive
match using rules applicable to the current (localized) culture. The following
instead performs a case-insensitive search using the invariant culture’s rules:
"abcdef".StartsWith ("aBc", StringComparison.InvariantCultureIgnoreCase)
The Contains method doesn’t offer the convenience of this overload, although you
can achieve the same result with the IndexOf method.
IndexOf is more powerful: it returns the first position of a given character or
substring (or −1 if the substring isn’t found):
Console.WriteLine ("abcde".IndexOf ("cd")); // 2
IndexOf is also overloaded to accept a startPosition (an index from which to
begin searching) as well as a StringComparison enum:
Console.WriteLine ("abcde abcde".IndexOf ("CD", 6,
StringComparison.CurrentCultureIgnoreCase)); // 8
274 | Chapter 6: .NET Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-292-320.jpg)
![LastIndexOf is like IndexOf, but it works backward through the string.
IndexOfAny returns the first matching position of any one of a set of characters:
Console.Write ("ab,cd ef".IndexOfAny (new char[] {' ', ','} )); // 2
Console.Write ("pas5w0rd".IndexOfAny ("0123456789".ToCharArray() )); // 3
LastIndexOfAny does the same in the reverse direction.
Manipulating strings
Because String is immutable, all the methods that “manipulate” a string return a
new one, leaving the original untouched (the same goes for when you reassign a
string variable).
Substring extracts a portion of a string:
string left3 = "12345".Substring (0, 3); // left3 = "123";
string mid3 = "12345".Substring (1, 3); // mid3 = "234";
If you omit the length, you get the remainder of the string:
string end3 = "12345".Substring (2); // end3 = "345";
Insert and Remove insert or remove characters at a specified position:
string s1 = "helloworld".Insert (5, ", "); // s1 = "hello, world"
string s2 = s1.Remove (5, 2); // s2 = "helloworld";
PadLeft and PadRight pad a string to a given length with a specified character (or a
space if unspecified):
Console.WriteLine ("12345".PadLeft (9, '*')); // ****12345
Console.WriteLine ("12345".PadLeft (9)); // 12345
If the input string is longer than the padding length, the original string is returned
unchanged.
TrimStart and TrimEnd remove specified characters from the beginning or end of
a string; Trim does both. By default, these functions remove whitespace characters
(including spaces, tabs, new lines, and Unicode variations of these):
Console.WriteLine (" abc trn ".Trim().Length); // 3
Replace replaces all (nonoverlapping) occurrences of a particular character or
substring:
Console.WriteLine ("to be done".Replace (" ", " | ") ); // to | be | done
Console.WriteLine ("to be done".Replace (" ", "") ); // tobedone
ToUpper and ToLower return uppercase and lowercase versions of the input string.
By default, they honor the user’s current language settings; ToUpperInvariant and
ToLowerInvariant always apply English alphabet rules.
String and Text Handling | 275
.NET
Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-293-320.jpg)
![Splitting and joining strings
Split divides a string into pieces:
string[] words = "The quick brown fox".Split();
foreach (string word in words)
Console.Write (word + "|"); // The|quick|brown|fox|
By default, Split uses whitespace characters as delimiters; it’s also overloaded to
accept a params array of char or string delimiters. Split also optionally accepts a
StringSplitOptions enum, which has an option to remove empty entries: this is
useful when words are separated by several delimiters in a row.
The static Join method does the reverse of Split. It requires a delimiter and string
array:
string[] words = "The quick brown fox".Split();
string together = string.Join (" ", words); // The quick brown fox
The static Concat method is similar to Join but accepts only a params string
array and applies no separator. Concat is exactly equivalent to the + operator (the
compiler, in fact, translates + to Concat):
string sentence = string.Concat ("The", " quick", " brown", " fox");
string sameSentence = "The" + " quick" + " brown" + " fox";
String.Format and composite format strings
The static Format method provides a convenient way to build strings that embed
variables. The embedded variables (or values) can be of any type; the Format simply
calls ToString on them.
The master string that includes the embedded variables is called a composite for‐
mat string. When calling string.Format, you provide a composite format string
followed by each of the embedded variables:
string composite = "It's {0} degrees in {1} on this {2} morning";
string s = string.Format (composite, 35, "Perth", DateTime.Now.DayOfWeek);
// s == "It's 35 degrees in Perth on this Friday morning"
(And that’s Celsius!)
We can use interpolated string literals to the same effect (see “String Type” on page
52). Just precede the string with the $ symbol and put the expressions in braces:
string s = $"It's hot this {DateTime.Now.DayOfWeek} morning";
Each number in curly braces is called a format item. The number corresponds to the
argument position and is optionally followed by:
276 | Chapter 6: .NET Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-294-320.jpg)
![• Whether to emit a byte-order mark (a prefix that indicates endianness).
•
Encoding for file and stream I/O
The most common application for an Encoding object is to control how text is read
and written to a file or stream. For example, the following writes “Testing…” to a file
called data.txt in UTF-16 encoding:
System.IO.File.WriteAllText ("data.txt", "Testing...", Encoding.Unicode);
If you omit the final argument, WriteAllText applies the ubiquitous UTF-8
encoding.
UTF-8 is the default text encoding for all file and stream I/O.
We resume this subject in Chapter 15, in “Stream Adapters” on page 687.
Encoding to byte arrays
You can also use an Encoding object to go to and from a byte array. The GetBytes
method converts from string to byte[] with the given encoding; GetString con‐
verts from byte[] to string:
byte[] utf8Bytes = System.Text.Encoding.UTF8.GetBytes ("0123456789");
byte[] utf16Bytes = System.Text.Encoding.Unicode.GetBytes ("0123456789");
byte[] utf32Bytes = System.Text.Encoding.UTF32.GetBytes ("0123456789");
Console.WriteLine (utf8Bytes.Length); // 10
Console.WriteLine (utf16Bytes.Length); // 20
Console.WriteLine (utf32Bytes.Length); // 40
string original1 = System.Text.Encoding.UTF8.GetString (utf8Bytes);
string original2 = System.Text.Encoding.Unicode.GetString (utf16Bytes);
string original3 = System.Text.Encoding.UTF32.GetString (utf32Bytes);
Console.WriteLine (original1); // 0123456789
Console.WriteLine (original2); // 0123456789
Console.WriteLine (original3); // 0123456789
UTF-16 and surrogate pairs
Recall that .NET stores characters and strings in UTF-16. Because UTF-16 requires
one or two 16-bit words per character, and a char is only 16 bits in length,
some Unicode characters require two chars to represent. This has a couple of
consequences:
• A string’s Length property can be greater than its real character count.
•
• A single char is not always enough to fully represent a Unicode character.
•
String and Text Handling | 283
.NET
Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-301-320.jpg)
![Console.WriteLine (" Start: " + FormatTransitionTime
(rule.DaylightTransitionStart, false));
Console.WriteLine (" End: " + FormatTransitionTime
(rule.DaylightTransitionEnd, true));
Console.WriteLine();
}
In FormatTransitionTime, we honor the rules just described:
static string FormatTransitionTime (TimeZoneInfo.TransitionTime tt,
bool endTime)
{
if (endTime && tt.IsFixedDateRule
&& tt.Day == 1 && tt.Month == 1
&& tt.TimeOfDay == DateTime.MinValue)
return "-";
string s;
if (tt.IsFixedDateRule)
s = tt.Day.ToString();
else
s = "The " +
"first second third fourth last".Split() [tt.Week - 1] +
" " + tt.DayOfWeek + " in";
return s + " " + DateTimeFormatInfo.CurrentInfo.MonthNames [tt.Month-1]
+ " at " + tt.TimeOfDay.TimeOfDay;
}
Daylight Saving Time and DateTime
If you use a DateTimeOffset or a UTC DateTime, equality comparisons are unimpe‐
ded by the effects of Daylight Saving Time. But with local DateTimes, daylight saving
can be problematic.
We can summarize the rules as follows:
• Daylight saving affects local time but not UTC time.
•
• When the clocks turn back, comparisons that rely on time moving forward will
•
break if (and only if) they use local DateTimes.
• You can always reliably round-trip between UTC and local times (on the same
•
computer)—even as the clocks turn back.
The IsDaylightSavingTime tells you whether a given local DateTime is subject to
Daylight Saving Time. UTC times always return false:
Console.Write (DateTime.Now.IsDaylightSavingTime()); // True or False
Console.Write (DateTime.UtcNow.IsDaylightSavingTime()); // Always False
296 | Chapter 6: .NET Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-314-320.jpg)
![string s = string.Format (CultureInfo.InvariantCulture, "{0}", someObject);
This is equivalent to the following:
string s;
if (someObject is IFormattable)
s = ((IFormattable)someObject).ToString (null,
CultureInfo.InvariantCulture);
else if (someObject == null)
s = "";
else
s = someObject.ToString();
Parsing with format providers
There’s no standard interface for parsing through a format provider. Instead, each
participating type overloads its static Parse (and TryParse) method to accept a
format provider and, optionally, a NumberStyles or DateTimeStyles enum.
NumberStyles and DateTimeStyles control how parsing works: they let you specify
such things as whether parentheses or a currency symbol can appear in the input
string (by default, the answer to both questions is no). For example:
int error = int.Parse ("(2)"); // Exception thrown
int minusTwo = int.Parse ("(2)", NumberStyles.Integer |
NumberStyles.AllowParentheses); // OK
decimal fivePointTwo = decimal.Parse ("£5.20", NumberStyles.Currency,
CultureInfo.GetCultureInfo ("en-GB"));
The next section lists all NumberStyles and DateTimeStyles members as well as the
default parsing rules for each type.
IFormatProvider and ICustomFormatter
All format providers implement IFormatProvider:
public interface IFormatProvider { object GetFormat (Type formatType); }
The purpose of this method is to provide indirection—this is what allows Culture
Info to defer to an appropriate NumberFormatInfo or DateTimeInfo object to do the
work.
By implementing IFormatProvider—along with ICustomFormatter—you can also
write your own format provider that works in conjunction with existing types.
ICustomFormatter defines a single method, as follows:
string Format (string format, object arg, IFormatProvider formatProvider);
The following custom format provider writes numbers as words:
public class WordyFormatProvider : IFormatProvider, ICustomFormatter
{
static readonly string[] _numberWords =
Formatting and Parsing | 301
.NET
Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-319-320.jpg)
!["zero one two three four five six seven eight nine minus point".Split();
IFormatProvider _parent; // Allows consumers to chain format providers
public WordyFormatProvider () : this (CultureInfo.CurrentCulture) { }
public WordyFormatProvider (IFormatProvider parent) => _parent = parent;
public object GetFormat (Type formatType)
{
if (formatType == typeof (ICustomFormatter)) return this;
return null;
}
public string Format (string format, object arg, IFormatProvider prov)
{
// If it's not our format string, defer to the parent provider:
if (arg == null || format != "W")
return string.Format (_parent, "{0:" + format + "}", arg);
StringBuilder result = new StringBuilder();
string digitList = string.Format (CultureInfo.InvariantCulture,
"{0}", arg);
foreach (char digit in digitList)
{
int i = "0123456789-.".IndexOf (digit),
StringComparison.InvariantCulture);
if (i == -1) continue;
if (result.Length > 0) result.Append (' ');
result.Append (_numberWords[i]);
}
return result.ToString();
}
}
Notice that in the Format method, we used string.Format—with Invariant
Culture—to convert the input number to a string. It would have been simpler just
to call ToString() on arg, but then CurrentCulture would have been used instead.
The reason for needing the invariant culture is evident a few lines later:
int i = "0123456789-.".IndexOf (digit),
StringComparison.InvariantCulture);
It’s critical here that the number string comprises only the characters 0123456789-.
and not any internationalized versions of these.
Here’s an example of using WordyFormatProvider:
double n = -123.45;
IFormatProvider fp = new WordyFormatProvider();
Console.WriteLine (string.Format (fp, "{0:C} in words is {0:W}", n));
// -$123.45 in words is minus one two three point four five
You can use custom format providers only in composite format strings.
302 | Chapter 6: .NET Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-320-320.jpg)
![To represent a bigger number, such as one googol (10100
), you can use one of
BigInteger’s static methods, such as Pow (raise to the power):
BigInteger googol = BigInteger.Pow (10, 100);
Alternatively, you can Parse a string:
BigInteger googol = BigInteger.Parse ("1".PadRight (101, '0'));
Calling ToString() on this prints every digit:
Console.WriteLine (googol.ToString()); // 10000000000000000000000000000
00000000000000000000000000000000000000000000000000000000000000000000000
You can perform potentially lossy conversions between BigInteger and the stan‐
dard numeric types by using the explicit cast operator:
double g2 = (double) googol; // Explicit cast
BigInteger g3 = (BigInteger) g2; // Explicit cast
Console.WriteLine (g3);
The output from this demonstrates the loss of precision:
9999999999999999673361688041166912...
BigInteger overloads all the arithmetic operators including remainder (%) as well
as the comparison and equality operators.
You can also construct a BigInteger from a byte array. The following code gener‐
ates a 32-byte random number suitable for cryptography and then assigns it to a
BigInteger:
// This uses the System.Security.Cryptography namespace:
RandomNumberGenerator rand = RandomNumberGenerator.Create();
byte[] bytes = new byte [32];
rand.GetBytes (bytes);
var bigRandomNumber = new BigInteger (bytes); // Convert to BigInteger
The advantage of storing such a number in a BigInteger over a byte array is that
you get value-type semantics. Calling ToByteArray converts a BigInteger back to a
byte array.
Half
The Half struct is a 16-bit floating point type and was introduced with .NET 5. Half
is intended mainly for interoperating with graphics card processors and does not
have native support in most CPUs.
You can convert between Half and float or double via an explicit cast:
Half h = (Half) 123.456;
Console.WriteLine (h); // 123.44 (note loss of precision)
There are no arithmetic operations defined for this type, so you must convert to
another type such as float or double in order to perform calculations.
Working with Numbers | 317
.NET
Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-335-320.jpg)
![Random
The Random class generates a pseudorandom sequence of random bytes, integers,
or doubles.
To use Random, you first instantiate it, optionally providing a seed to initiate the
random number series. Using the same seed guarantees the same series of numbers
(if run under the same CLR version), which is sometimes useful when you want
reproducibility:
Random r1 = new Random (1);
Random r2 = new Random (1);
Console.WriteLine (r1.Next (100) + ", " + r1.Next (100)); // 24, 11
Console.WriteLine (r2.Next (100) + ", " + r2.Next (100)); // 24, 11
If you don’t want reproducibility, you can construct Random with no seed; in that
case, it uses the current system time to make one up.
Because the system clock has limited granularity, two Random
instances created close together (typically within 10 ms) will
yield the same sequence of values. A common trap is to
instantiate a new Random object every time you need a random
number rather than reusing the same object.
A good pattern is to declare a single static Random instance.
In multithreaded scenarios, however, this can cause trouble
because Random objects are not thread-safe. We describe a
workaround in “Thread-Local Storage” on page 898.
Calling Next(n) generates a random integer between 0 and n−1. NextDouble gener‐
ates a random double between 0 and 1. NextBytes fills a byte array with random
values.
Random is not considered random enough for high-security applications such as
cryptography. For this, .NET provides a cryptographically strong random number
generator, in the System.Security.Cryptography namespace. Here’s how to use it:
var rand = System.Security.Cryptography.RandomNumberGenerator.Create();
byte[] bytes = new byte [32];
rand.GetBytes (bytes); // Fill the byte array with random numbers.
The downside is that it’s less flexible: filling a byte array is the only means of
obtaining random numbers. To obtain an integer, you must use BitConverter:
byte[] bytes = new byte [4];
rand.GetBytes (bytes);
int i = BitConverter.ToInt32 (bytes, 0);
Working with Numbers | 319
.NET
Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-337-320.jpg)
![Enum Conversions
There are three ways to represent an enum value:
• As an enum member
•
• As its underlying integral value
•
• As a string
•
In this section, we describe how to convert between each.
Enum to integral conversions
Recall that an explicit cast converts between an enum member and its integral value.
An explicit cast is the correct approach if you know the enum type at compile time:
[Flags]
public enum BorderSides { Left=1, Right=2, Top=4, Bottom=8 }
...
int i = (int) BorderSides.Top; // i == 4
BorderSides side = (BorderSides) i; // side == BorderSides.Top
You can cast a System.Enum instance to its integral type in the same way. The trick is
to first cast to an object and then the integral type:
static int GetIntegralValue (Enum anyEnum)
{
return (int) (object) anyEnum;
}
This relies on you knowing the integral type: the method we just wrote would crash
if passed an enum whose integral type was long. To write a method that works with
an enum of any integral type, you can take one of three approaches. The first is to call
Convert.ToDecimal:
static decimal GetAnyIntegralValue (Enum anyEnum)
{
return Convert.ToDecimal (anyEnum);
}
This works because every integral type (including ulong) can be converted to
decimal without loss of information. The second approach is to call Enum.GetUnder
lyingType in order to obtain the enum’s integral type, and then call Convert.Change
Type:
static object GetBoxedIntegralValue (Enum anyEnum)
{
Type integralType = Enum.GetUnderlyingType (anyEnum.GetType());
return Convert.ChangeType (anyEnum, integralType);
}
This preserves the original integral type, as the following example shows:
Enums | 321
.NET
Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-339-320.jpg)
![foreach (Enum value in Enum.GetValues (typeof (BorderSides)))
Console.WriteLine (value);
Composite members such as LeftRight = Left | Right are included, too.
Enum.GetNames performs the same function but returns an array of strings.
Internally, the CLR implements GetValues and GetNames by
reflecting over the fields in the enum’s type. The results are
cached for efficiency.
How Enums Work
The semantics of enums are enforced largely by the compiler. In the CLR, there’s
no runtime difference between an enum instance (when unboxed) and its underly‐
ing integral value. Further, an enum definition in the CLR is merely a subtype
of System.Enum with static integral-type fields for each member. This makes the
ordinary use of an enum highly efficient, with a runtime cost matching that of
integral constants.
The downside of this strategy is that enums can provide static but not strong type
safety. We saw an example of this in Chapter 3:
[Flags] public enum BorderSides { Left=1, Right=2, Top=4, Bottom=8 }
...
BorderSides b = BorderSides.Left;
b += 1234; // No error!
When the compiler is unable to perform validation (as in this example), there’s no
backup from the runtime to throw an exception.
What we said about there being no runtime difference between an enum instance
and its integral value might seem at odds with the following:
[Flags] public enum BorderSides { Left=1, Right=2, Top=4, Bottom=8 }
...
Console.WriteLine (BorderSides.Right.ToString()); // Right
Console.WriteLine (BorderSides.Right.GetType().Name); // BorderSides
Given the nature of an enum instance at runtime, you’d expect this to print 2 and
Int32! The reason for its behavior is down to some more compile-time trickery.
C# explicitly boxes an enum instance before calling its virtual methods—such as
ToString or GetType. And when an enum instance is boxed, it gains a runtime
wrapping that references its enum type.
The Guid Struct
The Guid struct represents a globally unique identifier: a 16-byte value that, when
generated, is almost certainly unique in the world. Guids are often used for keys
of various sorts, in applications and databases. There are 2128
or 3.4 × 1038
unique
Guids.
The Guid Struct | 323
.NET
Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-341-320.jpg)
![The static Guid.NewGuid method generates a unique Guid:
Guid g = Guid.NewGuid ();
Console.WriteLine (g.ToString()); // 0d57629c-7d6e-4847-97cb-9e2fc25083fe
To instantiate an existing value, you use one of the constructors. The two most
useful constructors are:
public Guid (byte[] b); // Accepts a 16-byte array
public Guid (string g); // Accepts a formatted string
When represented as a string, a Guid is formatted as a 32-digit hexadecimal number,
with optional hyphens after the 8th, 12th, 16th, and 20th digits. The whole string
can also be optionally wrapped in brackets or braces:
Guid g1 = new Guid ("{0d57629c-7d6e-4847-97cb-9e2fc25083fe}");
Guid g2 = new Guid ("0d57629c7d6e484797cb9e2fc25083fe");
Console.WriteLine (g1 == g2); // True
Being a struct, a Guid honors value-type semantics; hence, the equality operator
works in the preceding example.
The ToByteArray method converts a Guid to a byte array.
The static Guid.Empty property returns an empty Guid (all zeros). This is often used
in place of null.
Equality Comparison
Until now, we’ve assumed that the == and != operators are all there is to equality
comparison. The issue of equality, however, is more complex and subtler, some‐
times requiring the use of additional methods and interfaces. This section explores
the standard C# and .NET protocols for equality, focusing particularly on two
questions:
• When are == and != adequate—and inadequate—for equality comparison, and
•
what are the alternatives?
• How and when should you customize a type’s equality logic?
•
But before exploring the details of equality protocols and how to customize them,
we first must look at the preliminary concept of value versus referential equality.
Value Versus Referential Equality
There are two kinds of equality:
Value equality
Two values are equivalent in some sense.
Referential equality
Two references refer to exactly the same object.
324 | Chapter 6: .NET Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-342-320.jpg)
![Console.WriteLine (a1.Equals (a2)); // True
Console.WriteLine (a1 == a2); // True
Pluggable equality comparers
If you want a type to take on different equality semantics just for a specific scenario,
you can use a pluggable IEqualityComparer. This is particularly useful in conjunc‐
tion with the standard collection classes, and we describe it in the following chapter,
in “Plugging in Equality and Order” on page 388.
Order Comparison
As well as defining standard protocols for equality, C# and .NET define two stan‐
dard protocols for determining the order of one object relative to another:
• The IComparable interfaces (IComparable and IComparable<T>)
•
• The > and < operators
•
The IComparable interfaces are used by general-purpose sorting algorithms. In the
following example, the static Array.Sort method works because System.String
implements the IComparable interfaces:
string[] colors = { "Green", "Red", "Blue" };
Array.Sort (colors);
foreach (string c in colors) Console.Write (c + " "); // Blue Green Red
The < and > operators are more specialized, and they are intended mostly for
numeric types. Because they are statically resolved, they can translate to highly
efficient bytecode, suitable for computationally intensive algorithms.
.NET also provides pluggable ordering protocols, via the IComparer interfaces. We
describe these in the final section of Chapter 7.
IComparable
The IComparable interfaces are defined as follows:
public interface IComparable { int CompareTo (object other); }
public interface IComparable<in T> { int CompareTo (T other); }
The two interfaces represent the same functionality. With value types, the generic
type-safe interface is faster than the nongeneric interface. In both cases, the Compare
To method works as follows:
• If a comes after b, a.CompareTo(b) returns a positive number.
•
• If a is the same as b, a.CompareTo(b) returns 0.
•
• If a comes before b, a.CompareTo(b) returns a negative number.
•
For example:
Order Comparison | 335
.NET
Fundamentals](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-353-320.jpg)
![Thanks to the improved static type safety, calling the following method with an
array of characters will generate a compile-time error:
void Test (IEnumerable<int> numbers) { ... }
It’s a standard practice for collection classes to publicly expose IEnumerable<T>
while “hiding” the nongeneric IEnumerable through explicit interface implementa‐
tion. This is so that if you directly call GetEnumerator(), you get back the type-safe
generic IEnumerator<T>. There are times, though, when this rule is broken for
reasons of backward compatibility (generics did not exist prior to C# 2.0). A good
example is arrays—these must return the nongeneric (the nice way of putting it
is “classic”) IEnumerator to prevent breaking earlier code. To get a generic IEnumera
tor<T>, you must cast to expose the explicit interface:
int[] data = { 1, 2, 3 };
var rator = ((IEnumerable <int>)data).GetEnumerator();
Fortunately, you rarely need to write this sort of code, thanks to the foreach
statement.
IEnumerable<T> and IDisposable
IEnumerator<T> inherits from IDisposable. This allows enumerators to hold refer‐
ences to resources such as database connections—and ensure that those resources
are released when enumeration is complete (or abandoned partway through). The
foreach statement recognizes this detail and translates the following:
foreach (var element in somethingEnumerable) { ... }
into the logical equivalent of this:
using (var rator = somethingEnumerable.GetEnumerator())
while (rator.MoveNext())
{
var element = rator.Current;
...
}
When to Use the Nongeneric Interfaces
Given the extra type safety of the generic collection interfaces such as IEnumera
ble<T>, the question arises: do you ever need to use the nongeneric IEnumerable (or
ICollection or IList)?
In the case of IEnumerable, you must implement this interface in conjunction with
IEnumerable<T>—because the latter derives from the former. However, it’s very rare
that you actually implement these interfaces from scratch: in nearly all cases, you
can take the higher-level approach of using iterator methods, Collection<T>, and
LINQ.
346 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-364-320.jpg)
![You can also subclass an existing collection: Collection<T> is
designed just for this purpose (see “Customizable Collections
and Proxies” on page 378). Yet another approach is to use the
LINQ query operators, which we cover in Chapter 8.
Returning another collection’s enumerator is just a matter of calling GetEnumerator
on the inner collection. However, this is viable only in the simplest scenarios in
which the items in the inner collection are exactly what are required. A more
flexible approach is to write an iterator, using C#’s yield return statement. An
iterator is a C# language feature that assists in writing collections, in the same way
the foreach statement assists in consuming collections. An iterator automatically
handles the implementation of IEnumerable and IEnumerator—or their generic
versions. Here’s a simple example:
public class MyCollection : IEnumerable
{
int[] data = { 1, 2, 3 };
public IEnumerator GetEnumerator()
{
foreach (int i in data)
yield return i;
}
}
Notice the “black magic”: GetEnumerator doesn’t appear to return an enumerator at
all! Upon parsing the yield return statement, the compiler writes a hidden nested
enumerator class behind the scenes and then refactors GetEnumerator to instantiate
and return that class. Iterators are powerful and simple (and are used extensively in
the implementation of LINQ-to-Object’s standard query operators).
Keeping with this approach, we can also implement the generic interface IEnumera
ble<T>:
public class MyGenCollection : IEnumerable<int>
{
int[] data = { 1, 2, 3 };
public IEnumerator<int> GetEnumerator()
{
foreach (int i in data)
yield return i;
}
// Explicit implementation keeps it hidden:
IEnumerator IEnumerable.GetEnumerator() => GetEnumerator();
}
Because IEnumerable<T> inherits from IEnumerable, we must implement both
the generic and the nongeneric versions of GetEnumerator. In accordance with
standard practice, we’ve implemented the nongeneric version explicitly. It can
348 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-366-320.jpg)
![simply call the generic GetEnumerator because IEnumerator<T> inherits from
IEnumerator.
The class we’ve just written would be suitable as a basis from which to write a
more sophisticated collection. However, if we need nothing above a simple IEnumer
able<T> implementation, the yield return statement allows for an easier variation.
Rather than writing a class, you can move the iteration logic into a method return‐
ing a generic IEnumerable<T> and let the compiler take care of the rest. Here’s an
example:
public static IEnumerable <int> GetSomeIntegers()
{
yield return 1;
yield return 2;
yield return 3;
}
Here’s our method in use:
foreach (int i in Test.GetSomeIntegers())
Console.WriteLine (i);
The final approach in writing GetEnumerator is to write a class that implements
IEnumerator directly. This is exactly what the compiler does behind the scenes, in
resolving iterators. (Fortunately, it’s rare that you’ll need to go this far yourself.) The
following example defines a collection that’s hardcoded to contain the integers 1, 2,
and 3:
public class MyIntList : IEnumerable
{
int[] data = { 1, 2, 3 };
public IEnumerator GetEnumerator() => new Enumerator (this);
class Enumerator : IEnumerator // Define an inner class
{ // for the enumerator.
MyIntList collection;
int currentIndex = -1;
public Enumerator (MyIntList items) => this.collection = items;
public object Current
{
get
{
if (currentIndex == -1)
throw new InvalidOperationException ("Enumeration not started!");
if (currentIndex == collection.data.Length)
throw new InvalidOperationException ("Past end of list!");
return collection.data [currentIndex];
}
}
public bool MoveNext()
Enumeration | 349
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-367-320.jpg)
![{
if (currentIndex >= collection.data.Length - 1) return false;
return ++currentIndex < collection.data.Length;
}
public void Reset() => currentIndex = -1;
}
}
Implementing Reset is optional—you can instead throw a
NotSupportedException.
Note that the first call to MoveNext should move to the first (and not the second)
item in the list.
To get on par with an iterator in functionality, we must also implement IEnumera
tor<T>. Here’s an example with bounds checking omitted for brevity:
class MyIntList : IEnumerable<int>
{
int[] data = { 1, 2, 3 };
// The generic enumerator is compatible with both IEnumerable and
// IEnumerable<T>. We implement the nongeneric GetEnumerator method
// explicitly to avoid a naming conflict.
public IEnumerator<int> GetEnumerator() => new Enumerator(this);
IEnumerator IEnumerable.GetEnumerator() => new Enumerator(this);
class Enumerator : IEnumerator<int>
{
int currentIndex = -1;
MyIntList collection;
public Enumerator (MyIntList items) => this.items = items;
public int Current => collection.data [currentIndex];
object IEnumerator.Current => Current;
public bool MoveNext() => ++currentIndex < collection.data.Length;
public void Reset() => currentIndex = -1;
// Given we don't need a Dispose method, it's good practice to
// implement it explicitly, so it's hidden from the public interface.
void IDisposable.Dispose() {}
}
}
The example with generics is faster because IEnumerator<int>.Current doesn’t
require casting from int to object and so avoids the overhead of boxing.
350 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-368-320.jpg)
![There is no consistent rationale in the way the words col‐
lection and list are applied throughout the .NET libraries.
For instance, because IList<T> is a more functional version
of ICollection<T>, you might expect the class List<T> to
be correspondingly more functional than the class Collec
tion<T>. This is not the case. It’s best to consider the terms
collection and list as broadly synonymous, except when a spe‐
cific type is involved.
ICollection<T> and ICollection
ICollection<T> is the standard interface for countable collections of objects. It
provides the ability to determine the size of a collection (Count), determine whether
an item exists in the collection (Contains), copy the collection into an array
(ToArray), and determine whether the collection is read-only (IsReadOnly). For
writable collections, you can also Add, Remove, and Clear items from the collection.
And because it extends IEnumerable<T>, it can also be traversed via the foreach
statement:
public interface ICollection<T> : IEnumerable<T>, IEnumerable
{
int Count { get; }
bool Contains (T item);
void CopyTo (T[] array, int arrayIndex);
bool IsReadOnly { get; }
void Add(T item);
bool Remove (T item);
void Clear();
}
The nongeneric ICollection is similar in providing a countable collection, but
it doesn’t provide functionality for altering the list or checking for element
membership:
public interface ICollection : IEnumerable
{
int Count { get; }
bool IsSynchronized { get; }
object SyncRoot { get; }
void CopyTo (Array array, int index);
}
The nongeneric interface also defines properties to assist with synchronization
(Chapter 14)—these were dumped in the generic version because thread safety is no
longer considered intrinsic to the collection.
Both interfaces are fairly straightforward to implement. If implementing a read-only
ICollection<T>, the Add, Remove, and Clear methods should throw a NotSuppor
tedException.
352 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-370-320.jpg)
![These interfaces are usually implemented in conjunction with either the IList or
the IDictionary interface.
IList<T> and IList
IList<T> is the standard interface for collections indexable by position. In addi‐
tion to the functionality inherited from ICollection<T> and IEnumerable<T>, it
provides the ability to read or write an element by position (via an indexer) and
insert/remove by position:
public interface IList<T> : ICollection<T>, IEnumerable<T>, IEnumerable
{
T this [int index] { get; set; }
int IndexOf (T item);
void Insert (int index, T item);
void RemoveAt (int index);
}
The IndexOf methods perform a linear search on the list, returning −1 if the
specified item is not found.
The nongeneric version of IList has more members because it inherits less from
ICollection:
public interface IList : ICollection, IEnumerable
{
object this [int index] { get; set }
bool IsFixedSize { get; }
bool IsReadOnly { get; }
int Add (object value);
void Clear();
bool Contains (object value);
int IndexOf (object value);
void Insert (int index, object value);
void Remove (object value);
void RemoveAt (int index);
}
The Add method on the nongeneric IList interface returns an integer—this is the
index of the newly added item. In contrast, the Add method on ICollection<T> has
a void return type.
The general-purpose List<T> class is the quintessential implementation of both
IList<T> and IList. C# arrays also implement both the generic and nongeneric
ILists (although the methods that add or remove elements are hidden via explicit
interface implementation and throw a NotSupportedException if called).
The ICollection and IList Interfaces | 353
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-371-320.jpg)
![An ArgumentException is thrown if you try to access a mul‐
tidimensional array via IList’s indexer. This is a trap when
writing methods such as the following:
public object FirstOrNull (IList list)
{
if (list == null || list.Count == 0) return null;
return list[0];
}
This might appear bulletproof, but it will throw an exception
if called with a multidimensional array. You can test for a
multidimensional array at runtime with this expression (more
on this in Chapter 19):
list.GetType().IsArray && list.GetType().GetArrayRank()>1
IReadOnlyCollection<T> and IReadOnlyList<T>
.NET also defines collection and list interfaces that expose just the members
required for read-only operations:
public interface IReadOnlyCollection<out T> : IEnumerable<T>, IEnumerable
{
int Count { get; }
}
public interface IReadOnlyList<out T> : IReadOnlyCollection<T>,
IEnumerable<T>, IEnumerable
{
T this[int index] { get; }
}
Because the type parameter for these interfaces is used only in output positions, it’s
marked as covariant. This allows a list of cats, for instance, to be treated as a read-
only list of animals. In contrast, T is not marked as covariant with ICollection<T>
and IList<T>, because T is used in both input and output positions.
These interfaces represent a read-only view of a collection or
list; the underlying implementation might still be writable.
Most of the writable (mutable) collections implement both the
read-only and read/write interfaces.
In addition to letting you work with collections covariantly, the read-only interfaces
allow a class to publicly expose a read-only view of a private writable collection. We
demonstrate this—along with a better solution—in “ReadOnlyCollection<T>” on
page 383.
IReadOnlyList<T> maps to the Windows Runtime type IVectorView<T>.
354 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-372-320.jpg)
![The Array Class
The Array class is the implicit base class for all single and multidimensional arrays,
and it is one of the most fundamental types implementing the standard collection
interfaces. The Array class provides type unification, so a common set of methods is
available to all arrays, regardless of their declaration or underlying element type.
Because arrays are so fundamental, C# provides explicit syntax for their decla‐
ration and initialization, which we described in Chapters 2 and 3. When an
array is declared using C#’s syntax, the CLR implicitly subtypes the Array class
—synthesizing a pseudotype appropriate to the array’s dimensions and element
types. This pseudotype implements the typed generic collection interfaces, such as
IList<string>.
The CLR also treats array types specially upon construction, assigning them a
contiguous space in memory. This makes indexing into arrays highly efficient, but
prevents them from being resized later on.
Array implements the collection interfaces up to IList<T> in both their generic and
nongeneric forms. IList<T> itself is implemented explicitly, though, to keep Array’s
public interface clean of methods such as Add or Remove, which throw an exception
on fixed-length collections such as arrays. The Array class does actually offer a
static Resize method, although this works by creating a new array and then copying
over each element. As well as being inefficient, references to the array elsewhere in
the program will still point to the original version. A better solution for resizable
collections is to use the List<T> class (described in the following section).
An array can contain value-type or reference-type elements. Value-type elements
are stored in place in the array, so an array of three long integers (each 8 bytes) will
occupy 24 bytes of contiguous memory. A reference type element, however, occu‐
pies only as much space in the array as a reference (4 bytes in a 32-bit environment
or 8 bytes in a 64-bit environment). Figure 7-2 illustrates the effect, in memory, of
the following program:
StringBuilder[] builders = new StringBuilder [5];
builders [0] = new StringBuilder ("builder1");
builders [1] = new StringBuilder ("builder2");
builders [2] = new StringBuilder ("builder3");
long[] numbers = new long [3];
numbers [0] = 12345;
numbers [1] = 54321;
The Array Class | 355
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-373-320.jpg)
![Figure 7-2. Arrays in memory
Because Array is a class, arrays are always (themselves) reference types—regardless
of the array’s element type. This means that the statement arrayB = arrayA results
in two variables that reference the same array. Similarly, two distinct arrays will
always fail an equality test, unless you employ a structural equality comparer, which
compares every element of the array:
object[] a1 = { "string", 123, true };
object[] a2 = { "string", 123, true };
Console.WriteLine (a1 == a2); // False
Console.WriteLine (a1.Equals (a2)); // False
IStructuralEquatable se1 = a1;
Console.WriteLine (se1.Equals (a2,
StructuralComparisons.StructuralEqualityComparer)); // True
Arrays can be duplicated by calling the Clone method: arrayB = arrayA.Clone().
However, this results in a shallow clone, meaning that only the memory represented
by the array itself is copied. If the array contains value-type objects, the values them‐
selves are copied; if the array contains reference type objects, just the references
are copied (resulting in two arrays whose members reference the same objects).
Figure 7-3 demonstrates the effect of adding the following code to our example:
StringBuilder[] builders2 = builders;
StringBuilder[] shallowClone = (StringBuilder[]) builders.Clone();
356 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-374-320.jpg)
![Figure 7-3. Shallow-cloning an array
To create a deep copy—for which reference type subobjects are duplicated—you
must loop through the array and clone each element manually. The same rules apply
to other .NET collection types.
Although Array is designed primarily for use with 32-bit indexers, it also has
limited support for 64-bit indexers (allowing an array to theoretically address up
to 264
elements) via several methods that accept both Int32 and Int64 parameters.
These overloads are useless in practice because the CLR does not permit any
object—including arrays—to exceed two gigabytes in size (whether running on a
32- or 64-bit environment).
Many of the methods on the Array class that you expect to
be instance methods are in fact static methods. This is an odd
design decision, and means that you should check for both
static and instance methods when looking for a method on
Array.
Construction and Indexing
The easiest way to create and index arrays is through C#’s language constructs:
int[] myArray = { 1, 2, 3 };
int first = myArray [0];
int last = myArray [myArray.Length - 1];
Alternatively, you can instantiate an array dynamically by calling Array.Create
Instance. This allows you to specify element type and rank (number of dimen‐
sions) at runtime as well as allowing nonzero-based arrays through specifying a
lower bound. Nonzero-based arrays are not compatible with the .NET Common
The Array Class | 357
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-375-320.jpg)
![Language Specification (CLS) and should not be exposed as public members in a
library that might be consumed by a program written in F# or Visual Basic.
The GetValue and SetValue methods let you access elements in a dynamically
created array (they also work on ordinary arrays):
// Create a string array 2 elements in length:
Array a = Array.CreateInstance (typeof(string), 2);
a.SetValue ("hi", 0); // → a[0] = "hi";
a.SetValue ("there", 1); // → a[1] = "there";
string s = (string) a.GetValue (0); // → s = a[0];
// We can also cast to a C# array as follows:
string[] cSharpArray = (string[]) a;
string s2 = cSharpArray [0];
Zero-indexed arrays created dynamically can be cast to a C# array of a matching
or compatible type (compatible by standard array-variance rules). For example, if
Apple subclasses Fruit, Apple[] can be cast to Fruit[]. This leads to the issue
of why object[] was not used as the unifying array type rather than the Array
class. The answer is that object[] is incompatible with both multidimensional and
value-type arrays (and nonzero-based arrays). An int[] array cannot be cast to
object[]. Hence, we require the Array class for full type unification.
GetValue and SetValue also work on compiler-created arrays, and they are useful
when writing methods that can deal with an array of any type and rank. For
multidimensional arrays, they accept an array of indexers:
public object GetValue (params int[] indices)
public void SetValue (object value, params int[] indices)
The following method prints the first element of any array, regardless of rank:
void WriteFirstValue (Array a)
{
Console.Write (a.Rank + "-dimensional; ");
// The indexers array will automatically initialize to all zeros, so
// passing it into GetValue or SetValue will get/set the zero-based
// (i.e., first) element in the array.
int[] indexers = new int[a.Rank];
Console.WriteLine ("First value is " + a.GetValue (indexers));
}
void Demo()
{
int[] oneD = { 1, 2, 3 };
int[,] twoD = { {5,6}, {8,9} };
WriteFirstValue (oneD); // 1-dimensional; first value is 1
WriteFirstValue (twoD); // 2-dimensional; first value is 5
}
358 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-376-320.jpg)
![For working with arrays of unknown type but known rank,
generics provide an easier and more efficient solution:
void WriteFirstValue<T> (T[] array)
{
Console.WriteLine (array[0]);
}
SetValue throws an exception if the element is of an incompatible type for the
array.
When an array is instantiated, whether via language syntax or Array.CreateIn
stance, its elements are automatically initialized. For arrays with reference type
elements, this means writing nulls; for arrays with value-type elements, this means
calling the value-type’s default constructor (effectively “zeroing” the members). The
Array class also provides this functionality on demand via the Clear method:
public static void Clear (Array array, int index, int length);
This method doesn’t affect the size of the array. This is in contrast to the usual use of
Clear (such as in ICollection<T>.Clear) whereby the collection is reduced to zero
elements.
Enumeration
Arrays are easily enumerated with a foreach statement:
int[] myArray = { 1, 2, 3};
foreach (int val in myArray)
Console.WriteLine (val);
You can also enumerate using the static Array.ForEach method, defined as follows:
public static void ForEach<T> (T[] array, Action<T> action);
This uses an Action delegate, with this signature:
public delegate void Action<T> (T obj);
Here’s the first example rewritten with Array.ForEach:
Array.ForEach (new[] { 1, 2, 3 }, Console.WriteLine);
Length and Rank
Array provides the following methods and properties for querying length and rank:
public int GetLength (int dimension);
public long GetLongLength (int dimension);
public int Length { get; }
public long LongLength { get; }
public int GetLowerBound (int dimension);
public int GetUpperBound (int dimension);
public int Rank { get; } // Returns number of dimensions in array
The Array Class | 359
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-377-320.jpg)
![GetLength and GetLongLength return the length for a given dimension (0 for a
single-dimensional array), and Length and LongLength return the total number of
elements in the array—all dimensions included.
GetLowerBound and GetUpperBound are useful with nonzero indexed arrays. Get
UpperBound returns the same result as adding GetLowerBound to GetLength for any
given dimension.
Searching
The Array class offers a range of methods for finding elements within a one-
dimensional array:
BinarySearch methods
For rapidly searching a sorted array for a particular item
IndexOf/LastIndex methods
For searching unsorted arrays for a particular item
Find/FindLast/FindIndex/FindLastIndex/FindAll/Exists/TrueForAll
For searching unsorted arrays for item(s) that satisfy a given Predicate<T>
None of the array searching methods throws an exception if the specified value is
not found. Instead, if an item is not found, methods returning an integer return −1
(assuming a zero-indexed array), and methods returning a generic type return the
type’s default value (e.g., 0 for an int, or null for a string).
The binary search methods are fast, but they work only on sorted arrays and require
that the elements be compared for order rather than simply equality. To this effect,
the binary search methods can accept an IComparer or IComparer<T> object to
arbitrate on ordering decisions (see “Plugging in Equality and Order” on page 388).
This must be consistent with any comparer used in originally sorting the array. If no
comparer is provided, the type’s default ordering algorithm will be applied based on
its implementation of IComparable / IComparable<T>.
The IndexOf and LastIndexOf methods perform a simple enumeration over the
array, returning the position of the first (or last) element that matches the given
value.
The predicate-based searching methods allow a method delegate or lambda expres‐
sion to arbitrate on whether a given element is a “match.” A predicate is simply a
delegate accepting an object and returning true or false:
public delegate bool Predicate<T> (T object);
In the following example, we search an array of strings for a name containing the
letter “a”:
string[] names = { "Rodney", "Jack", "Jill" };
string match = Array.Find (names, ContainsA);
Console.WriteLine (match); // Jack
360 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-378-320.jpg)
![ContainsA (string name) { return name.Contains ("a"); }
Here’s the same code shortened with a lambda expression:
string[] names = { "Rodney", "Jack", "Jill" };
string match = Array.Find (names, n => n.Contains ("a")); // Jack
FindAll returns an array of all items satisfying the predicate. In fact, it’s equivalent
to Enumerable.Where in the System.Linq namespace, except that FindAll returns
an array of matching items rather than an IEnumerable<T> of the same.
Exists returns true if any array member satisfies the given predicate, and is equiva‐
lent to Any in System.Linq.Enumerable.
TrueForAll returns true if all items satisfy the predicate, and is equivalent to All in
System.Linq.Enumerable.
Sorting
Array has the following built-in sorting methods:
// For sorting a single array:
public static void Sort<T> (T[] array);
public static void Sort (Array array);
// For sorting a pair of arrays:
public static void Sort<TKey,TValue> (TKey[] keys, TValue[] items);
public static void Sort (Array keys, Array items);
Each of these methods is additionally overloaded to also accept the following:
int index // Starting index at which to begin sorting
int length // Number of elements to sort
IComparer<T> comparer // Object making ordering decisions
Comparison<T> comparison // Delegate making ordering decisions
The following illustrates the simplest use of Sort:
int[] numbers = { 3, 2, 1 };
Array.Sort (numbers); // Array is now { 1, 2, 3 }
The methods accepting a pair of arrays work by rearranging the items of each array
in tandem, basing the ordering decisions on the first array. In the next example,
both the numbers and their corresponding words are sorted into numerical order:
int[] numbers = { 3, 2, 1 };
string[] words = { "three", "two", "one" };
Array.Sort (numbers, words);
// numbers array is now { 1, 2, 3 }
// words array is now { "one", "two", "three" }
The Array Class | 361
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-379-320.jpg)
![Array.Sort requires that the elements in the array implement IComparable (see
“Order Comparison” on page 335). This means that most built-in C# types (such as
integers, as in the preceding example) can be sorted. If the elements are not intrins‐
ically comparable or you want to override the default ordering, you must provide
Sort with a custom comparison provider that reports on the relative position of two
elements. There are ways to do this:
• Via a helper object that implements IComparer /IComparer<T> (see “Plugging
•
in Equality and Order” on page 388)
• Via a Comparison delegate:
•
public delegate int Comparison<T> (T x, T y);
The Comparison delegate follows the same semantics as IComparer<T>.CompareTo:
if x comes before y, a negative integer is returned; if x comes after y, a positive
integer is returned; if x and y have the same sorting position, 0 is returned.
In the following example, we sort an array of integers such that the odd numbers
come first:
int[] numbers = { 1, 2, 3, 4, 5 };
Array.Sort (numbers, (x, y) => x % 2 == y % 2 ? 0 : x % 2 == 1 ? -1 : 1);
// numbers array is now { 1, 3, 5, 2, 4 }
As an alternative to calling Sort, you can use LINQ’s OrderBy
and ThenBy operators. Unlike Array.Sort, the LINQ opera‐
tors don’t alter the original array, instead emitting the sorted
result in a fresh IEnumerable<T> sequence.
Reversing Elements
The following Array methods reverse the order of all—or a portion of—elements in
an array:
public static void Reverse (Array array);
public static void Reverse (Array array, int index, int length);
Copying
Array provides four methods to perform shallow copying: Clone, CopyTo, Copy, and
ConstrainedCopy. The former two are instance methods; the latter two are static
methods.
The Clone method returns a whole new (shallow-copied) array. The CopyTo and
Copy methods copy a contiguous subset of the array. Copying a multidimensional
rectangular array requires you to map the multidimensional index to a linear index.
For example, the middle square (position[1,1]) in a 3 × 3 array is represented
with the index 4, from the calculation: 1 * 3 + 1. The source and destination ranges
can overlap without causing a problem.
362 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-380-320.jpg)
![ConstrainedCopy performs an atomic operation: if all of the requested elements
cannot be successfully copied (due to a type error, for instance), the operation is
rolled back.
Array also provides an AsReadOnly method that returns a wrapper that prevents
elements from being reassigned.
Converting and Resizing
Array.ConvertAll creates and returns a new array of element type TOutput, calling
the supplied Converter delegate to copy over the elements. Converter is defined as
follows:
public delegate TOutput Converter<TInput,TOutput> (TInput input)
The following converts an array of floats to an array of integers:
float[] reals = { 1.3f, 1.5f, 1.8f };
int[] wholes = Array.ConvertAll (reals, r => Convert.ToInt32 (r));
// wholes array is { 1, 2, 2 }
The Resize method works by creating a new array and copying over the elements,
returning the new array via the reference parameter. However, any references to the
original array in other objects will remain unchanged.
The System.Linq namespace offers an additional buffet of
extension methods suitable for array conversion. These meth‐
ods return an IEnumerable<T>, which you can convert back to
an array via Enumerable’s ToArray method.
Lists, Queues, Stacks, and Sets
.NET provides a basic set of concrete collection classes that implement the inter‐
faces described in this chapter. This section concentrates on the list-like collections
(versus the dictionary-like collections, which we cover in “Dictionaries” on page
372). As with the interfaces we discussed previously, you usually have a choice of
generic or nongeneric versions of each type. In terms of flexibility and performance,
the generic classes win, making their nongeneric counterparts redundant except for
backward compatibility. This differs from the situation with collection interfaces,
for which the nongeneric versions are still occasionally useful.
Of the classes described in this section, the generic List class is the most commonly
used.
List<T> and ArrayList
The generic List and nongeneric ArrayList classes provide a dynamically sized
array of objects and are among the most commonly used of the collection
classes. ArrayList implements IList, whereas List<T> implements both IList
and IList<T> (and the read-only version, IReadOnlyList<T>). Unlike with arrays,
Lists, Queues, Stacks, and Sets | 363
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-381-320.jpg)
![all interfaces are implemented publicly, and methods such as Add and Remove are
exposed and work as you would expect.
Internally, List<T> and ArrayList work by maintaining an internal array of objects,
replaced with a larger array upon reaching capacity. Appending elements is efficient
(because there is usually a free slot at the end), but inserting elements can be slow
(because all elements after the insertion point must be shifted to make a free slot),
as can removing elements (especially near the start). As with arrays, searching is
efficient if the BinarySearch method is used on a list that has been sorted, but it is
otherwise inefficient because each item must be individually checked.
List<T> is up to several times faster than ArrayList if T is
a value type, because List<T> avoids the overhead of boxing
and unboxing elements.
List<T> and ArrayList provide constructors that accept an existing collection of
elements: these copy each element from the existing collection into the new List<T>
or ArrayList:
public class List<T> : IList<T>, IReadOnlyList<T>
{
public List ();
public List (IEnumerable<T> collection);
public List (int capacity);
// Add+Insert
public void Add (T item);
public void AddRange (IEnumerable<T> collection);
public void Insert (int index, T item);
public void InsertRange (int index, IEnumerable<T> collection);
// Remove
public bool Remove (T item);
public void RemoveAt (int index);
public void RemoveRange (int index, int count);
public int RemoveAll (Predicate<T> match);
// Indexing
public T this [int index] { get; set; }
public List<T> GetRange (int index, int count);
public Enumerator<T> GetEnumerator();
// Exporting, copying and converting:
public T[] ToArray();
public void CopyTo (T[] array);
public void CopyTo (T[] array, int arrayIndex);
public void CopyTo (int index, T[] array, int arrayIndex, int count);
public ReadOnlyCollection<T> AsReadOnly();
public List<TOutput> ConvertAll<TOutput> (Converter <T,TOutput>
converter);
// Other:
public void Reverse(); // Reverses order of elements in list.
public int Capacity { get;set; } // Forces expansion of internal array.
364 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-382-320.jpg)
![public void TrimExcess(); // Trims internal array back to size.
public void Clear(); // Removes all elements, so Count=0.
}
public delegate TOutput Converter <TInput, TOutput> (TInput input);
In addition to these members, List<T> provides instance versions of all of Array’s
searching and sorting methods.
The following code demonstrates List’s properties and methods (for examples on
searching and sorting, see “The Array Class” on page 355):
var words = new List<string>(); // New string-typed list
words.Add ("melon");
words.Add ("avocado");
words.AddRange (new[] { "banana", "plum" } );
words.Insert (0, "lemon"); // Insert at start
words.InsertRange (0, new[] { "peach", "nashi" }); // Insert at start
words.Remove ("melon");
words.RemoveAt (3); // Remove the 4th element
words.RemoveRange (0, 2); // Remove first 2 elements
// Remove all strings starting in 'n':
words.RemoveAll (s => s.StartsWith ("n"));
Console.WriteLine (words [0]); // first word
Console.WriteLine (words [words.Count - 1]); // last word
foreach (string s in words) Console.WriteLine (s); // all words
List<string> subset = words.GetRange (1, 2); // 2nd->3rd words
string[] wordsArray = words.ToArray(); // Creates a new typed array
// Copy first two elements to the end of an existing array:
string[] existing = new string [1000];
words.CopyTo (0, existing, 998, 2);
List<string> upperCaseWords = words.ConvertAll (s => s.ToUpper());
List<int> lengths = words.ConvertAll (s => s.Length);
The nongeneric ArrayList class requires clumsy casts—as the following example
demonstrates:
ArrayList al = new ArrayList();
al.Add ("hello");
string first = (string) al [0];
string[] strArr = (string[]) al.ToArray (typeof (string));
Such casts cannot be verified by the compiler; the following compiles successfully
but then fails at runtime:
int first = (int) al [0]; // Runtime exception
Lists, Queues, Stacks, and Sets | 365
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-383-320.jpg)
![An ArrayList is functionally similar to List<object>. Both
are useful when you need a list of mixed-type elements that
share no common base type (other than object). A possi‐
ble advantage of choosing an ArrayList, in this case, would
be if you need to deal with the list using reflection (Chap‐
ter 19). Reflection is easier with a nongeneric ArrayList than
a List<object>.
If you import the System.Linq namespace, you can convert an ArrayList to a
generic List by calling Cast and then ToList:
ArrayList al = new ArrayList();
al.AddRange (new[] { 1, 5, 9 } );
List<int> list = al.Cast<int>().ToList();
Cast and ToList are extension methods in the System.Linq.Enumerable class.
LinkedList<T>
LinkedList<T> is a generic doubly linked list (see Figure 7-4). A doubly linked list
is a chain of nodes in which each references the node before, the node after, and the
actual element. Its main benefit is that an element can always be inserted efficiently
anywhere in the list because it just involves creating a new node and updating a few
references. However, finding where to insert the node in the first place can be slow
because there’s no intrinsic mechanism to index directly into a linked list; each node
must be traversed, and binary-chop searches are not possible.
Figure 7-4. LinkedList<T>
LinkedList<T> implements IEnumerable<T> and ICollection<T> (and their non‐
generic versions), but not IList<T> because access by index is not supported. List
nodes are implemented via the following class:
366 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-384-320.jpg)
![public sealed class LinkedListNode<T>
{
public LinkedList<T> List { get; }
public LinkedListNode<T> Next { get; }
public LinkedListNode<T> Previous { get; }
public T Value { get; set; }
}
When adding a node, you can specify its position either relative to another node or
at the start/end of the list. LinkedList<T> provides the following methods for this:
public void AddFirst(LinkedListNode<T> node);
public LinkedListNode<T> AddFirst (T value);
public void AddLast (LinkedListNode<T> node);
public LinkedListNode<T> AddLast (T value);
public void AddAfter (LinkedListNode<T> node, LinkedListNode<T> newNode);
public LinkedListNode<T> AddAfter (LinkedListNode<T> node, T value);
public void AddBefore (LinkedListNode<T> node, LinkedListNode<T> newNode);
public LinkedListNode<T> AddBefore (LinkedListNode<T> node, T value);
Similar methods are provided to remove elements:
public void Clear();
public void RemoveFirst();
public void RemoveLast();
public bool Remove (T value);
public void Remove (LinkedListNode<T> node);
LinkedList<T> has internal fields to keep track of the number of elements in the
list as well as the head and tail of the list. These are exposed in the following public
properties:
public int Count { get; } // Fast
public LinkedListNode<T> First { get; } // Fast
public LinkedListNode<T> Last { get; } // Fast
LinkedList<T> also supports the following searching methods (each requiring that
the list be internally enumerated):
public bool Contains (T value);
public LinkedListNode<T> Find (T value);
public LinkedListNode<T> FindLast (T value);
Finally, LinkedList<T> supports copying to an array for indexed processing and
obtaining an enumerator to support the foreach statement:
public void CopyTo (T[] array, int index);
public Enumerator<T> GetEnumerator();
Lists, Queues, Stacks, and Sets | 367
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-385-320.jpg)
![Here’s a demonstration on the use of LinkedList<string>:
var tune = new LinkedList<string>();
tune.AddFirst ("do"); // do
tune.AddLast ("so"); // do - so
tune.AddAfter (tune.First, "re"); // do - re- so
tune.AddAfter (tune.First.Next, "mi"); // do - re - mi- so
tune.AddBefore (tune.Last, "fa"); // do - re - mi - fa- so
tune.RemoveFirst(); // re - mi - fa - so
tune.RemoveLast(); // re - mi - fa
LinkedListNode<string> miNode = tune.Find ("mi");
tune.Remove (miNode); // re - fa
tune.AddFirst (miNode); // mi- re - fa
foreach (string s in tune) Console.WriteLine (s);
Queue<T> and Queue
Queue<T> and Queue are first-in, first-out (FIFO) data structures, providing meth‐
ods to Enqueue (add an item to the tail of the queue) and Dequeue (retrieve and
remove the item at the head of the queue). A Peek method is also provided to
return the element at the head of the queue without removing it, as well as a Count
property (useful in checking that elements are present before dequeuing).
Although queues are enumerable, they do not implement IList<T>/IList, because
members cannot be accessed directly by index. A ToArray method is provided,
however, for copying the elements to an array from which they can be randomly
accessed:
public class Queue<T> : IEnumerable<T>, ICollection, IEnumerable
{
public Queue();
public Queue (IEnumerable<T> collection); // Copies existing elements
public Queue (int capacity); // To lessen auto-resizing
public void Clear();
public bool Contains (T item);
public void CopyTo (T[] array, int arrayIndex);
public int Count { get; }
public T Dequeue();
public void Enqueue (T item);
public Enumerator<T> GetEnumerator(); // To support foreach
public T Peek();
public T[] ToArray();
public void TrimExcess();
}
The following is an example of using Queue<int>:
var q = new Queue<int>();
q.Enqueue (10);
q.Enqueue (20);
368 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-386-320.jpg)
![int[] data = q.ToArray(); // Exports to an array
Console.WriteLine (q.Count); // "2"
Console.WriteLine (q.Peek()); // "10"
Console.WriteLine (q.Dequeue()); // "10"
Console.WriteLine (q.Dequeue()); // "20"
Console.WriteLine (q.Dequeue()); // throws an exception (queue empty)
Queues are implemented internally using an array that’s resized as required—much
like the generic List class. The queue maintains indexes that point directly to the
head and tail elements; therefore, enqueuing and dequeuing are extremely quick
operations (except when an internal resize is required).
Stack<T> and Stack
Stack<T> and Stack are last-in, first-out (LIFO) data structures, providing methods
to Push (add an item to the top of the stack) and Pop (retrieve and remove an
element from the top of the stack). A nondestructive Peek method is also provided,
as is a Count property and a ToArray method for exporting the data for random
access:
public class Stack<T> : IEnumerable<T>, ICollection, IEnumerable
{
public Stack();
public Stack (IEnumerable<T> collection); // Copies existing elements
public Stack (int capacity); // Lessens auto-resizing
public void Clear();
public bool Contains (T item);
public void CopyTo (T[] array, int arrayIndex);
public int Count { get; }
public Enumerator<T> GetEnumerator(); // To support foreach
public T Peek();
public T Pop();
public void Push (T item);
public T[] ToArray();
public void TrimExcess();
}
The following example demonstrates Stack<int>:
var s = new Stack<int>();
s.Push (1); // Stack = 1
s.Push (2); // Stack = 1,2
s.Push (3); // Stack = 1,2,3
Console.WriteLine (s.Count); // Prints 3
Console.WriteLine (s.Peek()); // Prints 3, Stack = 1,2,3
Console.WriteLine (s.Pop()); // Prints 3, Stack = 1,2
Console.WriteLine (s.Pop()); // Prints 2, Stack = 1
Console.WriteLine (s.Pop()); // Prints 1, Stack = <empty>
Console.WriteLine (s.Pop()); // throws exception
Stacks are implemented internally with an array that’s resized as required, as with
Queue<T> and List<T>.
Lists, Queues, Stacks, and Sets | 369
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-387-320.jpg)
![BitArray
A BitArray is a dynamically sized collection of compacted bool values. It is more
memory efficient than both a simple array of bool and a generic List of bool
because it uses only one bit for each value, whereas the bool type otherwise occu‐
pies one byte for each value.
BitArray’s indexer reads and writes individual bits:
var bits = new BitArray(2);
bits[1] = true;
There are four bitwise operator methods (And, Or, Xor, and Not). All but the last
accept another BitArray:
bits.Xor (bits); // Bitwise exclusive-OR bits with itself
Console.WriteLine (bits[1]); // False
HashSet<T> and SortedSet<T>
HashSet<T> and SortedSet<T> have the following distinguishing features:
• Their Contains methods execute quickly using a hash-based lookup.
•
• They do not store duplicate elements and silently ignore requests to add
•
duplicates.
• You cannot access an element by position.
•
SortedSet<T> keeps elements in order, whereas HashSet<T> does not.
The commonality of the HashSet<T> and SortedSet<T> types is captured by the
interface ISet<T>. From .NET 5, these classes also implement an interface called
IReadOnlySet<T>, which is also implemented by the immutable set types (see
“Immutable Collections” on page 384).
HashSet<T> is implemented with a hashtable that stores just keys; SortedSet<T> is
implemented with a red/black tree.
Both collections implement ICollection<T> and offer methods that you would
expect, such as Contains, Add, and Remove. In addition, there’s a predicate-based
removal method called RemoveWhere.
The following constructs a HashSet<char> from an existing collection, tests for
membership, and then enumerates the collection (notice the absence of duplicates):
var letters = new HashSet<char> ("the quick brown fox");
Console.WriteLine (letters.Contains ('t')); // true
Console.WriteLine (letters.Contains ('j')); // false
foreach (char c in letters) Console.Write (c); // the quickbrownfx
370 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-388-320.jpg)
![Type Internal
structure
Retrieve
by
index?
Memory
overhead
(avg.
bytes per
item)
Speed:
random
insertion
Speed:
sequential
insertion
Speed:
retrieval
by key
Sorted
SortedDictionary <K,V> Red/black
tree
No 20 130 100 120
SortedList <K,V> 2xArray Yes 2 3,300 30 40
SortedList 2xArray Yes 27 4,500 100 180
In Big-O notation, retrieval time by key is as follows:
• O(1) for Hashtable, Dictionary, and OrderedDictionary
•
• O(log n) for SortedDictionary and SortedList
•
• O(n) for ListDictionary (and nondictionary types such as List<T>)
•
n is the number of elements in the collection.
IDictionary<TKey,TValue>
IDictionary<TKey,TValue> defines the standard protocol for all key/value-based
collections. It extends ICollection<T> by adding methods and properties to access
elements based on a key of arbitrary type:
public interface IDictionary <TKey, TValue> :
ICollection <KeyValuePair <TKey, TValue>>, IEnumerable
{
bool ContainsKey (TKey key);
bool TryGetValue (TKey key, out TValue value);
void Add (TKey key, TValue value);
bool Remove (TKey key);
TValue this [TKey key] { get; set; } // Main indexer - by key
ICollection <TKey> Keys { get; } // Returns just keys
ICollection <TValue> Values { get; } // Returns just values
}
There’s also an interface called IReadOnlyDiction
ary<TKey,TValue>, which defines the read-only subset of dic‐
tionary members.
To add an item to a dictionary, you either call Add or use the index’s set accessor—
the latter adds an item to the dictionary if the key is not already present (or
updates the item if it is present). Duplicate keys are forbidden in all dictionary
implementations, so calling Add twice with the same key throws an exception.
Dictionaries | 373
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-391-320.jpg)
![Dictionary implements both the generic and nongeneric IDictionary interfaces,
the generic IDictionary being exposed publicly. Dictionary is, in fact, a “textbook”
implementation of the generic IDictionary.
Here’s how to use it:
var d = new Dictionary<string, int>();
d.Add("One", 1);
d["Two"] = 2; // adds to dictionary because "two" not already present
d["Two"] = 22; // updates dictionary because "two" is now present
d["Three"] = 3;
Console.WriteLine (d["Two"]); // Prints "22"
Console.WriteLine (d.ContainsKey ("One")); // true (fast operation)
Console.WriteLine (d.ContainsValue (3)); // true (slow operation)
int val = 0;
if (!d.TryGetValue ("onE", out val))
Console.WriteLine ("No val"); // "No val" (case sensitive)
// Three different ways to enumerate the dictionary:
foreach (KeyValuePair<string, int> kv in d) // One; 1
Console.WriteLine (kv.Key + "; " + kv.Value); // Two; 22
// Three; 3
foreach (string s in d.Keys) Console.Write (s); // OneTwoThree
Console.WriteLine();
foreach (int i in d.Values) Console.Write (i); // 1223
Its underlying hashtable works by converting each element’s key into an integer
hashcode—a pseudo-unique value—and then applying an algorithm to convert the
hashcode into a hash key. This hash key is used internally to determine which
“bucket” an entry belongs to. If the bucket contains more than one value, a linear
search is performed on the bucket. A good hash function does not strive to return
strictly unique hashcodes (which would usually be impossible); it strives to return
hashcodes that are evenly distributed across the 32-bit integer space. This prevents
the scenario of ending up with a few very large (and inefficient) buckets.
A dictionary can work with keys of any type, providing it’s able to determine
equality between keys and obtain hashcodes. By default, equality is determined via
the key’s object.Equals method, and the pseudo-unique hashcode is obtained via
the key’s GetHashCode method. You can change this behavior either by overriding
these methods or by providing an IEqualityComparer object when constructing the
dictionary. A common application of this is to specify a case-insensitive equality
comparer when using string keys:
var d = new Dictionary<string, int> (StringComparer.OrdinalIgnoreCase);
We discuss this further in “Plugging in Equality and Order” on page 388.
Dictionaries | 375
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-393-320.jpg)
![1 There’s also a functionally identical nongeneric version of this called SortedList.
heavy or slow in either scenario—you wouldn’t suffer unreasonably by using a
Dictionary to begin with.
Both classes come only in nongeneric form.
Sorted Dictionaries
The .NET BCL provides two dictionary classes internally structured such that their
content is always sorted by key:
• SortedDictionary<TKey,TValue>
•
• SortedList<TKey,TValue>1
•
(In this section, we abbreviate <TKey,TValue> to <,>.)
SortedDictionary<,> uses a red/black tree: a data structure designed to perform
consistently well in any insertion or retrieval scenario.
SortedList<,> is implemented internally with an ordered array pair, providing
fast retrieval (via a binary-chop search) but poor insertion performance (because
existing values need to be shifted to make room for a new entry).
SortedDictionary<,> is much faster than SortedList<,> at inserting elements in
a random sequence (particularly with large lists). SortedList<,>, however, has an
extra ability: to access items by index as well as by key. With a sorted list, you
can go directly to the nth element in the sorting sequence (via the indexer on
the Keys/Values properties). To do the same with a SortedDictionary<,>, you
must manually enumerate over n items. (Alternatively, you could write a class that
combines a sorted dictionary with a list class.)
None of the three collections allows duplicate keys (as is the case with all
dictionaries).
The following example uses reflection to load all of the methods defined in
System.Object into a sorted list keyed by name and then enumerates their keys
and values:
// MethodInfo is in the System.Reflection namespace
var sorted = new SortedList <string, MethodInfo>();
foreach (MethodInfo m in typeof (object).GetMethods())
sorted [m.Name] = m;
foreach (string name in sorted.Keys)
Console.WriteLine (name);
Dictionaries | 377
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-395-320.jpg)
![foreach (MethodInfo m in sorted.Values)
Console.WriteLine (m.Name + " returns a " + m.ReturnType);
Here’s the result of the first enumeration:
Equals
GetHashCode
GetType
ReferenceEquals
ToString
Here’s the result of the second enumeration:
Equals returns a System.Boolean
GetHashCode returns a System.Int32
GetType returns a System.Type
ReferenceEquals returns a System.Boolean
ToString returns a System.String
Notice that we populated the dictionary through its indexer. If we instead used the
Add method, it would throw an exception because the object class upon which
we’re reflecting overloads the Equals method, and you can’t add the same key twice
to a dictionary. By using the indexer, the later entry overwrites the earlier entry,
preventing this error.
You can store multiple members of the same key by making
each value element a list:
SortedList <string, List<MethodInfo>>
Extending our example, the following retrieves the MethodInfo whose key is
"GetHashCode", just as with an ordinary dictionary:
Console.WriteLine (sorted ["GetHashCode"]); // Int32 GetHashCode()
So far, everything we’ve done would also work with a SortedDictionary<,>. The
following two lines, however, which retrieve the last key and value, work only with a
sorted list:
Console.WriteLine (sorted.Keys [sorted.Count - 1]); // ToString
Console.WriteLine (sorted.Values[sorted.Count - 1].IsVirtual); // True
Customizable Collections and Proxies
The collection classes discussed in previous sections are convenient in that you can
directly instantiate them, but they don’t allow you to control what happens when an
item is added to or removed from the collection. With strongly typed collections in
an application, you sometimes need this control; for instance:
• To fire an event when an item is added or removed
•
• To update properties because of the added or removed item
•
378 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-396-320.jpg)
![{
Name = name; Popularity = popularity;
}
}
public class AnimalCollection : Collection <Animal>
{
// AnimalCollection is already a fully functioning list of animals.
// No extra code is required.
}
public class Zoo // The class that will expose AnimalCollection.
{ // This would typically have additional members.
public readonly AnimalCollection Animals = new AnimalCollection();
}
As it stands, AnimalCollection is no more functional than a simple List<Animal>;
its role is to provide a base for future extension. To illustrate, let’s now add a Zoo
property to Animal so that it can reference the Zoo in which it lives and override
each of the virtual methods in Collection<Animal> to maintain that property
automatically:
public class Animal
{
public string Name;
public int Popularity;
public Zoo Zoo { get; internal set; }
public Animal(string name, int popularity)
{
Name = name; Popularity = popularity;
}
}
public class AnimalCollection : Collection <Animal>
{
Zoo zoo;
public AnimalCollection (Zoo zoo) { this.zoo = zoo; }
protected override void InsertItem (int index, Animal item)
{
base.InsertItem (index, item);
item.Zoo = zoo;
}
protected override void SetItem (int index, Animal item)
{
base.SetItem (index, item);
item.Zoo = zoo;
}
protected override void RemoveItem (int index)
{
this [index].Zoo = null;
base.RemoveItem (index);
}
protected override void ClearItems()
380 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-398-320.jpg)
![// ...
protected abstract TKey GetKeyForItem(TItem item);
protected void ChangeItemKey(TItem item, TKey newKey);
// Fast lookup by key - this is in addition to lookup by index.
public TItem this[TKey key] { get; }
protected IDictionary<TKey, TItem> Dictionary { get; }
}
GetKeyForItem is what the implementer overrides to obtain an item’s key from
the underlying object. The ChangeItemKey method must be called if the item’s
key property changes, in order to update the internal dictionary. The Dictionary
property returns the internal dictionary used to implement the lookup, which is
created when the first item is added. This behavior can be changed by specifying
a creation threshold in the constructor, delaying the internal dictionary from being
created until the threshold is reached (in the interim, a linear search is performed
if an item is requested by key). A good reason not to specify a creation threshold
is that having a valid dictionary can be useful in obtaining an ICollection<> of
keys, via the Dictionary’s Keys property. This collection can then be passed on to a
public property.
The most common use for KeyedCollection<,> is in providing a collection of
items accessible both by index and by name. To demonstrate this, let’s revisit
the zoo, this time implementing AnimalCollection as a KeyedCollection<string,
Animal>:
public class Animal
{
string name;
public string Name
{
get { return name; }
set {
if (Zoo != null) Zoo.Animals.NotifyNameChange (this, value);
name = value;
}
}
public int Popularity;
public Zoo Zoo { get; internal set; }
public Animal (string name, int popularity)
{
Name = name; Popularity = popularity;
}
}
public class AnimalCollection : KeyedCollection <string, Animal>
{
Zoo zoo;
public AnimalCollection (Zoo zoo) { this.zoo = zoo; }
382 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-400-320.jpg)
![internal void NotifyNameChange (Animal a, string newName) =>
this.ChangeItemKey (a, newName);
protected override string GetKeyForItem (Animal item) => item.Name;
// The following methods would be implemented as in the previous example
protected override void InsertItem (int index, Animal item)...
protected override void SetItem (int index, Animal item)...
protected override void RemoveItem (int index)...
protected override void ClearItems()...
}
public class Zoo
{
public readonly AnimalCollection Animals;
public Zoo() { Animals = new AnimalCollection (this); }
}
The following code demonstrates its use:
Zoo zoo = new Zoo();
zoo.Animals.Add (new Animal ("Kangaroo", 10));
zoo.Animals.Add (new Animal ("Mr Sea Lion", 20));
Console.WriteLine (zoo.Animals [0].Popularity); // 10
Console.WriteLine (zoo.Animals ["Mr Sea Lion"].Popularity); // 20
zoo.Animals ["Kangaroo"].Name = "Mr Roo";
Console.WriteLine (zoo.Animals ["Mr Roo"].Popularity); // 10
DictionaryBase
The nongeneric version of KeyedCollection is called DictionaryBase. This legacy
class takes a very different approach in that it implements IDictionary and uses
clumsy hook methods like CollectionBase: OnInsert, OnInsertComplete, OnSet,
OnSetComplete, OnRemove, OnRemoveComplete, OnClear, and OnClearComplete (and
additionally, OnGet). The primary advantage of implementing IDictionary over
taking the KeyedCollection approach is that you don’t need to subclass it in order
to obtain keys. But since the very purpose of DictionaryBase is to be subclassed,
it’s no advantage at all. The improved model in KeyedCollection is almost certainly
due to the fact that it was written some years later, with the benefit of hindsight.
DictionaryBase is best considered useful for backward compatibility.
ReadOnlyCollection<T>
ReadOnlyCollection<T> is a wrapper, or proxy, that provides a read-only view of a
collection. This is useful in allowing a class to publicly expose read-only access to a
collection that the class can still update internally.
A read-only collection accepts the input collection in its constructor, to which it
maintains a permanent reference. It doesn’t take a static copy of the input collection,
so subsequent changes to the input collection are visible through the read-only
wrapper.
Customizable Collections and Proxies | 383
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-401-320.jpg)
![The disadvantage of immutability is that when you need to make a change, you
must create a whole new object. This incurs a performance hit, although there are
mitigating strategies that we discuss in this section, including the ability to reuse
portions of the original structure.
The immutable collections are part of .NET (in .NET Framework, they are available
via the System.Collections.Immutable NuGet package). All collections are defined in
the System.Collections.Immutable namespace:
Type Internal structure
ImmutableArray<T> Array
ImmutableList<T> AVL tree
ImmutableDictionary<K,V> AVL tree
ImmutableHashSet<T> AVL tree
ImmutableSortedDictionary<K,V> AVL tree
ImmutableSortedSet<T> AVL tree
ImmutableStack<T> Linked list
ImmutableQueue<T> Linked list
The ImmutableArray<T> and ImmutableList<T> types are both immutable versions
of List<T>. Both do the same job but with different performance characteristics
that we discuss in “Immutable Collections and Performance” on page 387.
The immutable collections expose a public interface similar to their mutable coun‐
terparts. The key difference is that the methods that appear to alter the collection
(such as Add or Remove) don’t alter the original collection; instead they return a new
collection with the requested item added or removed.
Immutable collections prevent the adding and removing of
items; they don’t prevent the items themselves from being
mutated. To get the full benefits of immutability, you need
to ensure that only immutable items end up in an immutable
collection.
Creating Immutable Collections
Each immutable collection type offers a Create<T>() method, which accepts
optional initial values and returns an initialized immutable collection:
ImmutableArray<int> array = ImmutableArray.Create<int> (1, 2, 3);
Each collection also offers a CreateRange<T> method, which does the same job as
Create<T>; the difference is that its parameter type is IEnumerable<T> instead of
params T[].
Immutable Collections | 385
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-403-320.jpg)
![You can also create an immutable collection from an existing IEnumerable<T>,
using appropriate extension methods (ToImmutableArray, ToImmutableList, ToIm
mutableDictionary, and so on):
var list = new[] { 1, 2, 3 }.ToImmutableList();
Manipulating Immutable Collections
The Add method returns a new collection containing the existing elements plus the
new one:
var oldList = ImmutableList.Create<int> (1, 2, 3);
ImmutableList<int> newList = oldList.Add (4);
Console.WriteLine (oldList.Count); // 3 (unaltered)
Console.WriteLine (newList.Count); // 4
The Remove method operates in the same fashion, returning a new collection with
the item removed.
Repeatedly adding or removing elements in this manner is inefficient, because a
new immutable collection is created for each add or remove operation. A better
solution is to call AddRange (or RemoveRange), which accepts an IEnumerable<T> of
items, which are all added or removed in one go:
var anotherList = oldList.AddRange (new[] { 4, 5, 6 });
The immutable list and array also defines Insert and InsertRange methods to
insert elements at a particular index, a RemoveAt method to remove at an index, and
RemoveAll, which removes based on a predicate.
Builders
For more complex initialization needs, each immutable collection class defines a
builder counterpart. Builders are classes that are functionally equivalent to a muta‐
ble collection, with similar performance characteristics. After the data is initialized,
calling .ToImmutable() on a builder returns an immutable collection:
ImmutableArray<int>.Builder builder = ImmutableArray.CreateBuilder<int>();
builder.Add (1);
builder.Add (2);
builder.Add (3);
builder.RemoveAt (0);
ImmutableArray<int> myImmutable = builder.ToImmutable();
You also can use builders to batch multiple updates to an existing immutable
collection:
var builder2 = myImmutable.ToBuilder();
builder2.Add (4); // Efficient
builder2.Remove (2); // Efficient
... // More changes to builder...
386 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-404-320.jpg)
![The semantics for Equals and GetHashCode follow the same rules for object
.Equals and object.GetHashCode, described in Chapter 6. In the following exam‐
ple, we define a Customer class with two fields and then write an equality comparer
that matches both the first and last names:
public class Customer
{
public string LastName;
public string FirstName;
public Customer (string last, string first)
{
LastName = last;
FirstName = first;
}
}
public class LastFirstEqComparer : EqualityComparer <Customer>
{
public override bool Equals (Customer x, Customer y)
=> x.LastName == y.LastName && x.FirstName == y.FirstName;
public override int GetHashCode (Customer obj)
=> (obj.LastName + ";" + obj.FirstName).GetHashCode();
}
To illustrate how this works, let’s create two customers:
Customer c1 = new Customer ("Bloggs", "Joe");
Customer c2 = new Customer ("Bloggs", "Joe");
Because we’ve not overridden object.Equals, normal reference type equality
semantics apply:
Console.WriteLine (c1 == c2); // False
Console.WriteLine (c1.Equals (c2)); // False
The same default equality semantics apply when using these customers in a
Dictionary without specifying an equality comparer:
var d = new Dictionary<Customer, string>();
d [c1] = "Joe";
Console.WriteLine (d.ContainsKey (c2)); // False
Now, with the custom equality comparer:
var eqComparer = new LastFirstEqComparer();
var d = new Dictionary<Customer, string> (eqComparer);
d [c1] = "Joe";
Console.WriteLine (d.ContainsKey (c2)); // True
In this example, we would have to be careful not to change the customer’s First
Name or LastName while it was in use in the dictionary; otherwise, its hashcode
would change and the Dictionary would break.
390 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-408-320.jpg)
![Here’s SurnameComparer in use in a sorted dictionary:
var dic = new SortedDictionary<string,string> (new SurnameComparer());
dic.Add ("MacPhail", "second!");
dic.Add ("MacWilliam", "third!");
dic.Add ("McDonald", "first!");
foreach (string s in dic.Values)
Console.Write (s + " "); // first! second! third!
StringComparer
StringComparer is a predefined plug-in class for equating and comparing strings,
allowing you to specify language and case sensitivity. StringComparer implements
both IEqualityComparer and IComparer (and their generic versions), so you can
use it with any type of dictionary or sorted collection.
Because StringComparer is abstract, you obtain instances via its static properties.
StringComparer.Ordinal mirrors the default behavior for string equality compari‐
son and StringComparer.CurrentCulture for order comparison. Here are all of its
static members:
public static StringComparer CurrentCulture { get; }
public static StringComparer CurrentCultureIgnoreCase { get; }
public static StringComparer InvariantCulture { get; }
public static StringComparer InvariantCultureIgnoreCase { get; }
public static StringComparer Ordinal { get; }
public static StringComparer OrdinalIgnoreCase { get; }
public static StringComparer Create (CultureInfo culture,
bool ignoreCase);
In the following example, an ordinal case-insensitive dictionary is created such that
dict["Joe"] and dict["JOE"] mean the same thing:
var dict = new Dictionary<string, int> (StringComparer.OrdinalIgnoreCase);
In the next example, an array of names is sorted, using Australian English:
string[] names = { "Tom", "HARRY", "sheila" };
CultureInfo ci = new CultureInfo ("en-AU");
Array.Sort<string> (names, StringComparer.Create (ci, false));
The final example is a culture-aware version of the SurnameComparer we wrote in
the previous section (to compare names suitable for a phonebook listing):
class SurnameComparer : Comparer<string>
{
StringComparer strCmp;
public SurnameComparer (CultureInfo ci)
{
// Create a case-sensitive, culture-sensitive string comparer
strCmp = StringComparer.Create (ci, false);
}
Plugging in Equality and Order | 393
Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-411-320.jpg)
![string Normalize (string s)
{
s = s.Trim();
if (s.ToUpper().StartsWith ("MC")) s = "MAC" + s.Substring (2);
return s;
}
public override int Compare (string x, string y)
{
// Directly call Compare on our culture-aware StringComparer
return strCmp.Compare (Normalize (x), Normalize (y));
}
}
IStructuralEquatable and IStructuralComparable
As we discussed in Chapter 6, structs implement structural comparison by default:
two structs are equal if all of their fields are equal. Sometimes, however, structural
equality and order comparison are useful as plug-in options on other types, as
well—such as arrays. The following interfaces help with this:
public interface IStructuralEquatable
{
bool Equals (object other, IEqualityComparer comparer);
int GetHashCode (IEqualityComparer comparer);
}
public interface IStructuralComparable
{
int CompareTo (object other, IComparer comparer);
}
The IEqualityComparer/IComparer that you pass in are applied to each individual
element in the composite object. We can demonstrate this by using arrays. In
the following example, we compare two arrays for equality, first using the default
Equals method, then using IStructuralEquatable’s version:
int[] a1 = { 1, 2, 3 };
int[] a2 = { 1, 2, 3 };
IStructuralEquatable se1 = a1;
Console.Write (a1.Equals (a2)); // False
Console.Write (se1.Equals (a2, EqualityComparer<int>.Default)); // True
Here’s another example:
string[] a1 = "the quick brown fox".Split();
string[] a2 = "THE QUICK BROWN FOX".Split();
IStructuralEquatable se1 = a1;
bool isTrue = se1.Equals (a2, StringComparer.InvariantCultureIgnoreCase);
394 | Chapter 7: Collections](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-412-320.jpg)
![8
LINQ Queries
LINQ, or Language Integrated Query, is a set of language and runtime features for
writing structured type-safe queries over local object collections and remote data
sources.
LINQ enables you to query any collection implementing IEnumerable<T>, whether
an array, list, or XML Document Object Model (DOM), as well as remote data
sources, such as tables in an SQL Server database. LINQ offers the benefits of both
compile-time type checking and dynamic query composition.
This chapter describes the LINQ architecture and the fundamentals of writing quer‐
ies. All core types are defined in the System.Linq and System.Linq.Expressions
namespaces.
The examples in this and the following two chapters are pre‐
loaded into an interactive querying tool called LINQPad. You
can download LINQPad from http://www.linqpad.net.
Getting Started
The basic units of data in LINQ are sequences and elements. A sequence is any object
that implements IEnumerable<T>, and an element is each item in the sequence. In
the following example, names is a sequence, and "Tom", "Dick", and "Harry" are
elements:
string[] names = { "Tom", "Dick", "Harry" };
We call this a local sequence because it represents a local collection of objects in
memory.
A query operator is a method that transforms a sequence. A typical query operator
accepts an input sequence and emits a transformed output sequence. In the Enumera
ble class in System.Linq, there are around 40 query operators—all implemented as
static extension methods. These are called standard query operators.
395](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-413-320.jpg)
![Queries that operate over local sequences are called local quer‐
ies or LINQ-to-objects queries.
LINQ also supports sequences that can be dynamically fed
from a remote data source such as an SQL Server database.
These sequences additionally implement the IQueryable<T>
interface and are supported through a matching set of stan‐
dard query operators in the Queryable class. We discuss this
further in “Interpreted Queries” on page 424.
A query is an expression that, when enumerated, transforms sequences with query
operators. The simplest query comprises one input sequence and one operator. For
instance, we can apply the Where operator on a simple array to extract those strings
whose length is at least four characters, as follows:
string[] names = { "Tom", "Dick", "Harry" };
IEnumerable<string> filteredNames = System.Linq.Enumerable.Where
(names, n => n.Length >= 4);
foreach (string n in filteredNames)
Console.WriteLine (n);
Dick
Harry
Because the standard query operators are implemented as extension methods, we
can call Where directly on names, as though it were an instance method:
IEnumerable<string> filteredNames = names.Where (n => n.Length >= 4);
For this to compile, you must import the System.Linq namespace. Here’s a com‐
plete example:
using System;
using System.Collections.Generic;
using System.Linq;
string[] names = { "Tom", "Dick", "Harry" };
IEnumerable<string> filteredNames = names.Where (n => n.Length >= 4);
foreach (string name in filteredNames) Console.WriteLine (name);
Dick
Harry
We could further shorten our code by implicitly typing filter
edNames:
var filteredNames = names.Where (n => n.Length >= 4);
This can hinder readability, however, outside of an IDE, where
there are no tool tips to help. For this reason, we make less use
of implicit typing in this chapter than you might in your own
projects.
396 | Chapter 8: LINQ Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-414-320.jpg)
![using System;
using System.Collections.Generic;
using System.Linq;
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
IEnumerable<string> query = names
.Where (n => n.Contains ("a"))
.OrderBy (n => n.Length)
.Select (n => n.ToUpper());
foreach (string name in query) Console.WriteLine (name);
JAY
MARY
HARRY
The variable, n, in our example, is privately scoped to each of
the lambda expressions. We can reuse the identifier n for the
same reason that we can reuse the identifier c in the following
method:
void Test()
{
foreach (char c in "string1") Console.Write (c);
foreach (char c in "string2") Console.Write (c);
foreach (char c in "string3") Console.Write (c);
}
Where, OrderBy, and Select are standard query operators that resolve to extension
methods in the Enumerable class (if you import the System.Linq namespace).
We already introduced the Where operator, which emits a filtered version of the
input sequence. The OrderBy operator emits a sorted version of its input sequence;
the Select method emits a sequence in which each input element is transformed
or projected with a given lambda expression (n.ToUpper(), in this case). Data flows
from left to right through the chain of operators, so the data is first filtered, then
sorted, and then projected.
A query operator never alters the input sequence; instead, it
returns a new sequence. This is consistent with the functional
programming paradigm from which LINQ was inspired.
Here are the signatures of each of these extension methods (with the OrderBy
signature slightly simplified):
public static IEnumerable<TSource> Where<TSource>
(this IEnumerable<TSource> source, Func<TSource,bool> predicate)
public static IEnumerable<TSource> OrderBy<TSource,TKey>
(this IEnumerable<TSource> source, Func<TSource,TKey> keySelector)
public static IEnumerable<TResult> Select<TSource,TResult>
(this IEnumerable<TSource> source, Func<TSource,TResult> selector)
398 | Chapter 8: LINQ Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-416-320.jpg)
![Lambda expressions and Func signatures
The standard query operators utilize generic Func delegates. Func is a family of
general-purpose generic delegates in the System namespace, defined with the fol‐
lowing intent:
The type arguments in Func appear in the same order they do in lambda
expressions.
Hence, Func<TSource,bool> matches a TSource=>bool lambda expression: one that
accepts a TSource argument and returns a bool value.
Similarly, Func<TSource,TResult> matches a TSource=>TResult lambda
expression.
The Func delegates are listed in “Lambda Expressions” on page 174.
Lambda expressions and element typing
The standard query operators use the following type parameter names:
Generic type letter Meaning
TSource Element type for the input sequence
TResult Element type for the output sequence (if different from TSource)
TKey Element type for the key used in sorting, grouping, or joining
TSource is determined by the input sequence. TResult and TKey are typically
inferred from your lambda expression.
For example, consider the signature of the Select query operator:
public static IEnumerable<TResult> Select<TSource,TResult>
(this IEnumerable<TSource> source, Func<TSource,TResult> selector)
Func<TSource,TResult> matches a TSource=>TResult lambda expression: one that
maps an input element to an output element. TSource and TResult can be different
types, so the lambda expression can change the type of each element. Further, the
lambda expression determines the output sequence type. The following query uses
Select to transform string type elements to integer type elements:
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
IEnumerable<int> query = names.Select (n => n.Length);
foreach (int length in query)
Console.Write (length + "|"); // 3|4|5|4|3|
The compiler can infer the type of TResult from the return value of the lambda
expression. In this case, n.Length returns an int value, so TResult is inferred to be
int.
Fluent Syntax | 401
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-419-320.jpg)
![The Where query operator is simpler and requires no type inference for the output
because input and output elements are of the same type. This makes sense because
the operator merely filters elements; it does not transform them:
public static IEnumerable<TSource> Where<TSource>
(this IEnumerable<TSource> source, Func<TSource,bool> predicate)
Finally, consider the signature of the OrderBy operator:
// Slightly simplified:
public static IEnumerable<TSource> OrderBy<TSource,TKey>
(this IEnumerable<TSource> source, Func<TSource,TKey> keySelector)
Func<TSource,TKey> maps an input element to a sorting key. TKey is inferred
from your lambda expression and is separate from the input and output element
types. For instance, we could choose to sort a list of names by length (int key) or
alphabetically (string key):
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
IEnumerable<string> sortedByLength, sortedAlphabetically;
sortedByLength = names.OrderBy (n => n.Length); // int key
sortedAlphabetically = names.OrderBy (n => n); // string key
You can call the query operators in Enumerable with tra‐
ditional delegates that refer to methods instead of lambda
expressions. This approach is effective in simplifying certain
kinds of local queries—particularly with LINQ to XML—and
is demonstrated in Chapter 10. It doesn’t work with IQuerya
ble<T>-based sequences, however (e.g., when querying a data‐
base), because the operators in Queryable require lambda
expressions in order to emit expression trees. We discuss this
later in “Interpreted Queries” on page 424.
Natural Ordering
The original ordering of elements within an input sequence is significant in LINQ.
Some query operators rely on this ordering, such as Take, Skip, and Reverse.
The Take operator outputs the first x elements, discarding the rest:
int[] numbers = { 10, 9, 8, 7, 6 };
IEnumerable<int> firstThree = numbers.Take (3); // { 10, 9, 8 }
The Skip operator ignores the first x elements and outputs the rest:
IEnumerable<int> lastTwo = numbers.Skip (3); // { 7, 6 }
Reverse does exactly as it says:
IEnumerable<int> reversed = numbers.Reverse(); // { 6, 7, 8, 9, 10 }
With local queries (LINQ-to-objects), operators such as Where and Select preserve
the original ordering of the input sequence (as do all other query operators, except
for those that specifically change the ordering).
402 | Chapter 8: LINQ Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-420-320.jpg)
![Other Operators
Not all query operators return a sequence. The element operators extract one ele‐
ment from the input sequence; examples are First, Last, and ElementAt:
int[] numbers = { 10, 9, 8, 7, 6 };
int firstNumber = numbers.First(); // 10
int lastNumber = numbers.Last(); // 6
int secondNumber = numbers.ElementAt(1); // 9
int secondLowest = numbers.OrderBy(n=>n).Skip(1).First(); // 7
Because these operators return a single element, you don’t usually call further query
operators on their result unless that element itself is a collection.
The aggregation operators return a scalar value, usually of numeric type:
int count = numbers.Count(); // 5;
int min = numbers.Min(); // 6;
The quantifiers return a bool value:
bool hasTheNumberNine = numbers.Contains (9); // true
bool hasMoreThanZeroElements = numbers.Any(); // true
bool hasAnOddElement = numbers.Any (n => n % 2 != 0); // true
Some query operators accept two input sequences. Examples are Concat, which
appends one sequence to another, and Union, which does the same but with dupli‐
cates removed:
int[] seq1 = { 1, 2, 3 };
int[] seq2 = { 3, 4, 5 };
IEnumerable<int> concat = seq1.Concat (seq2); // { 1, 2, 3, 3, 4, 5 }
IEnumerable<int> union = seq1.Union (seq2); // { 1, 2, 3, 4, 5 }
The joining operators also fall into this category. Chapter 9 covers all of the query
operators in detail.
Query Expressions
C# provides a syntactic shortcut for writing LINQ queries, called query expressions.
Contrary to popular belief, a query expression is not a means of embedding SQL
into C#. In fact, the design of query expressions was inspired primarily by list
comprehensions from functional programming languages such as LISP and Haskell,
although SQL had a cosmetic influence.
In this book, we refer to query expression syntax simply as
query syntax.
Query Expressions | 403
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-421-320.jpg)
![In the preceding section, we wrote a fluent-syntax query to extract strings contain‐
ing the letter “a,” sorted by length and converted to uppercase. Here’s the same thing
in query syntax:
using System;
using System.Collections.Generic;
using System.Linq;
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
IEnumerable<string> query =
from n in names
where n.Contains ("a") // Filter elements
orderby n.Length // Sort elements
select n.ToUpper(); // Translate each element (project)
foreach (string name in query) Console.WriteLine (name);
JAY
MARY
HARRY
Query expressions always start with a from clause and end with either a select
or group clause. The from clause declares a range variable (in this case, n), which
you can think of as traversing the input sequence—rather like foreach. Figure 8-2
illustrates the complete syntax as a railroad diagram.
To read this diagram, start at the left and then proceed along
the track as if you were a train. For instance, after the manda‐
tory from clause, you can optionally include an orderby,
where, let, or join clause. After that, you can either continue
with a select or group clause, or go back and include another
from, orderby, where, let, or join clause.
404 | Chapter 8: LINQ Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-422-320.jpg)
![example, with LINQ, you cannot use a variable before you declare it. In SQL, you
can reference a table alias in the SELECT clause before defining it in a FROM clause.
A subquery in LINQ is just another C# expression and so requires no special syntax.
Subqueries in SQL are subject to special rules.
With LINQ, data logically flows from left to right through the query. With SQL, the
order is less well structured with regard to data flow.
A LINQ query comprises a conveyor belt or pipeline of operators that accept and
emit sequences whose element order can matter. An SQL query comprises a network
of clauses that work mostly with unordered sets.
Query Syntax Versus Fluent Syntax
Query and fluent syntax each have advantages.
Query syntax is simpler for queries that involve any of the following:
• A let clause for introducing a new variable alongside the range variable
•
• SelectMany, Join, or GroupJoin, followed by an outer range variable reference
•
(We describe the let clause in “Composition Strategies” on page 418; we describe
SelectMany, Join, and GroupJoin in Chapter 9.)
The middle ground is queries that involve the simple use of Where, OrderBy, and
Select. Either syntax works well; the choice here is largely personal.
For queries that comprise a single operator, fluent syntax is shorter and less
cluttered.
Finally, there are many operators that have no keyword in query syntax. These
require that you use fluent syntax—at least in part. This means any operator outside
of the following:
Where, Select, SelectMany
OrderBy, ThenBy, OrderByDescending, ThenByDescending
GroupBy, Join, GroupJoin
Mixed-Syntax Queries
If a query operator has no query-syntax support, you can mix query syntax and
fluent syntax. The only restriction is that each query-syntax component must be
complete (i.e., start with a from clause and end with a select or group clause).
Assuming this array declaration
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
the following example counts the number of names containing the letter “a”:
int matches = (from n in names where n.Contains ("a") select n).Count();
// 3
Query Expressions | 407
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-425-320.jpg)
![Captured Variables
If your query’s lambda expressions capture outer variables, the query will honor the
value of those variables at the time the query runs:
int[] numbers = { 1, 2 };
int factor = 10;
IEnumerable<int> query = numbers.Select (n => n * factor);
factor = 20;
foreach (int n in query) Console.Write (n + "|"); // 20|40|
This can be a trap when building up a query within a for loop. For example,
suppose that we want to remove all vowels from a string. The following, although
inefficient, gives the correct result:
IEnumerable<char> query = "Not what you might expect";
query = query.Where (c => c != 'a');
query = query.Where (c => c != 'e');
query = query.Where (c => c != 'i');
query = query.Where (c => c != 'o');
query = query.Where (c => c != 'u');
foreach (char c in query) Console.Write (c); // Nt wht y mght xpct
Now watch what happens when we refactor this with a for loop:
IEnumerable<char> query = "Not what you might expect";
string vowels = "aeiou";
for (int i = 0; i < vowels.Length; i++)
query = query.Where (c => c != vowels[i]);
foreach (char c in query) Console.Write (c);
An IndexOutOfRangeException is thrown upon enumerating the query because, as
we saw in Chapter 4 (see “Capturing Outer Variables” on page 176), the compiler
scopes the iteration variable in the for loop as if it were declared outside the loop.
Hence, each closure captures the same variable (i) whose value is 5 when the query
is actually enumerated. To solve this, you must assign the loop variable to another
variable declared inside the statement block:
for (int i = 0; i < vowels.Length; i++)
{
char vowel = vowels[i];
query = query.Where (c => c != vowel);
}
This forces a fresh local variable to be captured on each loop iteration.
Another way to solve the problem is to replace the for loop
with a foreach loop:
foreach (char vowel in vowels)
query = query.Where (c => c != vowel);
410 | Chapter 8: LINQ Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-428-320.jpg)
![How Deferred Execution Works
Query operators provide deferred execution by returning decorator sequences.
Unlike a traditional collection class such as an array or linked list, a decorator
sequence (in general) has no backing structure of its own to store elements. Instead,
it wraps another sequence that you supply at runtime, to which it maintains a
permanent dependency. Whenever you request data from a decorator, it in turn
must request data from the wrapped input sequence.
The query operator’s transformation constitutes the “decora‐
tion.” If the output sequence performed no transformation, it
would be a proxy rather than a decorator.
Calling Where merely constructs the decorator wrapper sequence, which holds a
reference to the input sequence, the lambda expression, and any other arguments
supplied. The input sequence is enumerated only when the decorator is enumerated.
Figure 8-3 illustrates the composition of the following query:
IEnumerable<int> lessThanTen = new int[] { 5, 12, 3 }.Where (n => n < 10);
Figure 8-3. Decorator sequence
When you enumerate lessThanTen, you are, in effect, querying the array through
the Where decorator.
The good news—should you ever want to write your own query operator—is that
implementing a decorator sequence is easy with a C# iterator. Here’s how you can
write your own Select method:
public static IEnumerable<TResult> MySelect<TSource,TResult>
(this IEnumerable<TSource> source, Func<TSource,TResult> selector)
{
foreach (TSource element in source)
yield return selector (element);
}
Deferred Execution | 411
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-429-320.jpg)
![This method is an iterator by virtue of the yield return statement. Functionally,
it’s a shortcut for the following:
public static IEnumerable<TResult> MySelect<TSource,TResult>
(this IEnumerable<TSource> source, Func<TSource,TResult> selector)
{
return new SelectSequence (source, selector);
}
where SelectSequence is a (compiler-written) class whose enumerator encapsulates
the logic in the iterator method.
Hence, when you call an operator such as Select or Where, you’re doing nothing
more than instantiating an enumerable class that decorates the input sequence.
Chaining Decorators
Chaining query operators creates a layering of decorators. Consider the following
query:
IEnumerable<int> query = new int[] { 5, 12, 3 }.Where (n => n < 10)
.OrderBy (n => n)
.Select (n => n * 10);
Each query operator instantiates a new decorator that wraps the previous sequence
(rather like a Russian nesting doll). Figure 8-4 illustrates the object model of this
query. Note that this object model is fully constructed prior to any enumeration.
Figure 8-4. Layered decorator sequences
When you enumerate query, you’re querying the original array, transformed
through a layering or chain of decorators.
412 | Chapter 8: LINQ Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-430-320.jpg)
![Adding ToList onto the end of this query would cause the
preceding operators to execute immediately, collapsing the
whole object model into a single list.
Figure 8-5 shows the same object composition in Unified Modeling Language
(UML) syntax. Select’s decorator references the OrderBy decorator, which refer‐
ences Where’s decorator, which references the array. A feature of deferred execution
is that you build the identical object model if you compose the query progressively:
IEnumerable<int>
source = new int[] { 5, 12, 3 },
filtered = source .Where (n => n < 10),
sorted = filtered .OrderBy (n => n),
query = sorted .Select (n => n * 10);
Figure 8-5. UML decorator composition
How Queries Are Executed
Here are the results of enumerating the preceding query:
foreach (int n in query) Console.WriteLine (n);
30
50
Behind the scenes, the foreach calls GetEnumerator on Select’s decorator (the
last or outermost operator), which kicks off everything. The result is a chain of
enumerators that structurally mirrors the chain of decorator sequences. Figure 8-6
illustrates the flow of execution as enumeration proceeds.
Deferred Execution | 413
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-431-320.jpg)
![Figure 8-6. Execution of a local query
In the first section of this chapter, we depicted a query as a production line of
conveyor belts. Extending this analogy, we can say a LINQ query is a lazy produc‐
tion line, where the conveyor belts roll elements only upon demand. Constructing
a query constructs a production line—with everything in place—but with nothing
rolling. Then, when the consumer requests an element (enumerates over the query),
the rightmost conveyor belt activates; this in turn triggers the others to roll—as
and when input sequence elements are needed. LINQ follows a demand-driven pull
model, rather than a supply-driven push model. This is important—as you’ll see
later—in allowing LINQ to scale to querying SQL databases.
Subqueries
A subquery is a query contained within another query’s lambda expression. The
following example uses a subquery to sort musicians by their last name:
string[] musos =
{ "David Gilmour", "Roger Waters", "Rick Wright", "Nick Mason" };
IEnumerable<string> query = musos.OrderBy (m => m.Split().Last());
m.Split converts each string into a collection of words, upon which we then call the
Last query operator. m.Split().Last is the subquery; query references the outer
query.
414 | Chapter 8: LINQ Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-432-320.jpg)
![Subqueries are permitted because you can put any valid C# expression on the right‐
hand side of a lambda. A subquery is simply another C# expression. This means
that the rules for subqueries are a consequence of the rules for lambda expressions
(and the behavior of query operators in general).
The term subquery, in the general sense, has a broader mean‐
ing. For the purpose of describing LINQ, we use the term only
for a query referenced from within the lambda expression of
another query. In a query expression, a subquery amounts to a
query referenced from an expression in any clause except the
from clause.
A subquery is privately scoped to the enclosing expression and can reference
parameters in the outer lambda expression (or range variables in a query
expression).
m.Split().Last is a very simple subquery. The next query retrieves all strings in an
array whose length matches that of the shortest string:
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
IEnumerable<string> outerQuery = names
.Where (n => n.Length == names.OrderBy (n2 => n2.Length)
.Select (n2 => n2.Length).First());
// Tom, Jay
Here’s the same thing as a query expression:
IEnumerable<string> outerQuery =
from n in names
where n.Length ==
(from n2 in names orderby n2.Length select n2.Length).First()
select n;
Because the outer range variable (n) is in scope for a subquery, we cannot reuse n as
the subquery’s range variable.
A subquery is executed whenever the enclosing lambda expression is evaluated.
This means that a subquery is executed upon demand, at the discretion of the outer
query. You could say that execution proceeds from the outside in. Local queries
follow this model literally; interpreted queries (e.g., database queries) follow this
model conceptually.
Subqueries | 415
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-433-320.jpg)
![Rather than calling string’s Replace method five times, we
could remove vowels from a string more efficiently with a
regular expression:
n => Regex.Replace (n, "[aeiou]", "")
string’s Replace method has the advantage, though, of also
working in database queries.
Translating this directly into a query expression is troublesome because the select
clause must come after the where and orderby clauses. And if we rearrange the
query so as to project last, the result would be different:
IEnumerable<string> query =
from n in names
where n.Length > 2
orderby n
select n.Replace ("a", "").Replace ("e", "").Replace ("i", "")
.Replace ("o", "").Replace ("u", "");
// Dck
// Hrry
// Jy
// Mry
// Tm
Fortunately, there are a number of ways to get the original result in query syntax.
The first is by querying progressively:
IEnumerable<string> query =
from n in names
select n.Replace ("a", "").Replace ("e", "").Replace ("i", "")
.Replace ("o", "").Replace ("u", "");
query = from n in query where n.Length > 2 orderby n select n;
// Dck
// Hrry
// Mry
The into Keyword
The into keyword is interpreted in two very different ways by
query expressions, depending on context. The meaning we’re
describing now is for signaling query continuation (the other is
for signaling a GroupJoin).
The into keyword lets you “continue” a query after a projection and is a shortcut for
progressively querying. With into, we can rewrite the preceding query as follows:
IEnumerable<string> query =
from n in names
select n.Replace ("a", "").Replace ("e", "").Replace ("i", "")
.Replace ("o", "").Replace ("u", "")
into noVowel
where noVowel.Length > 2 orderby noVowel select noVowel;
Composition Strategies | 419
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-437-320.jpg)
![We then can project into it with object initializers:
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
IEnumerable<TempProjectionItem> temp =
from n in names
select new TempProjectionItem
{
Original = n,
Vowelless = n.Replace ("a", "").Replace ("e", "").Replace ("i", "")
.Replace ("o", "").Replace ("u", "")
};
The result is of type IEnumerable<TempProjectionItem>, which we can subse‐
quently query:
IEnumerable<string> query = from item in temp
where item.Vowelless.Length > 2
select item.Original;
// Dick
// Harry
// Mary
Anonymous Types
Anonymous types allow you to structure your intermediate results without writing
special classes. We can eliminate the TempProjectionItem class in our previous
example with anonymous types:
var intermediate = from n in names
select new
{
Original = n,
Vowelless = n.Replace ("a", "").Replace ("e", "").Replace ("i", "")
.Replace ("o", "").Replace ("u", "")
};
IEnumerable<string> query = from item in intermediate
where item.Vowelless.Length > 2
select item.Original;
This gives the same result as the previous example, but without needing to write
a one-off class. The compiler does the job instead, generating a temporary class
with fields that match the structure of our projection. This means, however, that the
intermediate query has the following type:
IEnumerable <random-compiler-generated-name>
The only way we can declare a variable of this type is with the var keyword. In this
case, var is more than just a clutter reduction device; it’s a necessity.
We can write the entire query more succinctly with the into keyword:
422 | Chapter 8: LINQ Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-440-320.jpg)
![var query = from n in names
select new
{
Original = n,
Vowelless = n.Replace ("a", "").Replace ("e", "").Replace ("i", "")
.Replace ("o", "").Replace ("u", "")
}
into temp
where temp.Vowelless.Length > 2
select temp.Original;
Query expressions provide a shortcut for writing this kind of query: the let
keyword.
The let Keyword
The let keyword introduces a new variable alongside the range variable.
With let, we can write a query extracting strings whose length, excluding vowels,
exceeds two characters, as follows:
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
IEnumerable<string> query =
from n in names
let vowelless = n.Replace ("a", "").Replace ("e", "").Replace ("i", "")
.Replace ("o", "").Replace ("u", "")
where vowelless.Length > 2
orderby vowelless
select n; // Thanks to let, n is still in scope.
The compiler resolves a let clause by projecting into a temporary anonymous type
that contains both the range variable and the new expression variable. In other
words, the compiler translates this query into the preceding example.
let accomplishes two things:
• It projects new elements alongside existing elements.
•
• It allows an expression to be used repeatedly in a query without being
•
rewritten.
The let approach is particularly advantageous in this example because it allows the
select clause to project either the original name (n) or its vowel-removed version
(vowelless).
You can have any number of let statements before or after a where statement
(see Figure 8-2). A let statement can reference variables introduced in earlier let
statements (subject to the boundaries imposed by an into clause). let reprojects all
existing variables transparently.
A let expression need not evaluate to a scalar type: sometimes it’s useful to have it
evaluate to a subsequence, for instance.
Projection Strategies | 423
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-441-320.jpg)
![With this table in place, we can write an interpreted LINQ query in C# that uses EF
Core to retrieve customers whose name contains the letter “a,” as follows:
using System;
using System.Linq;
using Microsoft.EntityFrameworkCore;
using var dbContext = new NutshellContext();
IQueryable<string> query = from c in dbContext.Customers
where c.Name.Contains ("a")
orderby c.Name.Length
select c.Name.ToUpper();
foreach (string name in query) Console.WriteLine (name);
public class Customer
{
public int ID { get; set; }
public string Name { get; set; }
}
// We’ll explain the following class in more detail in the next section.
public class NutshellContext : DbContext
{
public virtual DbSet<Customer> Customers { get; set; }
protected override void OnConfiguring (DbContextOptionsBuilder builder)
=> builder.UseSqlServer ("...connection string...");
protected override void OnModelCreating (ModelBuilder modelBuilder)
=> modelBuilder.Entity<Customer>().ToTable ("Customer")
.HasKey (c => c.ID);
}
EF Core translates this query into the following SQL:
SELECT UPPER([c].[Name])
FROM [Customers] AS [c]
WHERE CHARINDEX(N'a', [c].[Name]) > 0
ORDER BY CAST(LEN([c].[Name]) AS int)
Here’s the end result:
// JAY
// MARY
// HARRY
Interpreted Queries | 425
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-443-320.jpg)
![.OrderBy (n => n)
.Pair() // Local from this point on.
.Select ((n, i) => "Pair " + i.ToString() + " = " + n);
foreach (string element in q) Console.WriteLine (element);
// Pair 0 = DICK, HARRY
// Pair 1 = JAY, MARY
Because dbContext.Customers is of a type implementing IQueryable<T>, the
Select operator resolves to Queryable.Select. This returns an output sequence
also of type IQueryable<T>, so the OrderBy operator similarly resolves to
Queryable.OrderBy. But the next query operator, Pair, has no overload accepting
IQueryable<T>—only the less specific IEnumerable<T>. So, it resolves to our local
Pair method—wrapping the interpreted query in a local query. Pair also returns
IEnumerable, so the Select that follows resolves to another local operator.
On the EF Core side, the resulting SQL statement is equivalent to this:
SELECT UPPER([c].[Name]) FROM [Customers] AS [c] ORDER BY UPPER([c].[Name])
The remaining work is done locally. In effect, we end up with a local query (on the
outside) whose source is an interpreted query (the inside).
AsEnumerable
Enumerable.AsEnumerable is the simplest of all query operators. Here’s its complete
definition:
public static IEnumerable<TSource> AsEnumerable<TSource>
(this IEnumerable<TSource> source)
{
return source;
}
Its purpose is to cast an IQueryable<T> sequence to IEnumerable<T>, forcing sub‐
sequent query operators to bind to Enumerable operators instead of Queryable
operators. This causes the remainder of the query to execute locally.
To illustrate, suppose that we had a MedicalArticles table in SQL Server and
wanted to use EF Core to retrieve all articles on influenza whose abstract contained
fewer than 100 words. For the latter predicate, we need a regular expression:
Regex wordCounter = new Regex (@"b(w|[-'])+b");
using var dbContext = new NutshellContext ();
var query = dbContext.MedicalArticles
.Where (article => article.Topic == "influenza" &&
wordCounter.Matches (article.Abstract).Count < 100);
The problem is that SQL Server doesn’t support regular expressions, so EF Core
will throw an exception, complaining that the query cannot be translated to SQL.
We can solve this by querying in two steps: first retrieving all articles on influenza
Interpreted Queries | 429
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-447-320.jpg)
![through an EF Core query, and then filtering locally for abstracts of fewer than 100
words:
Regex wordCounter = new Regex (@"b(w|[-'])+b");
using var dbContext = new NutshellContext ();
IEnumerable<MedicalArticle> efQuery = dbContext.MedicalArticles
.Where (article => article.Topic == "influenza");
IEnumerable<MedicalArticle> localQuery = efQuery
.Where (article => wordCounter.Matches (article.Abstract).Count < 100);
Because efQuery is of type IEnumerable<MedicalArticle>, the second query binds
to the local query operators, forcing that part of the filtering to run on the client.
With AsEnumerable, we can do the same in a single query:
Regex wordCounter = new Regex (@"b(w|[-'])+b");
using var dbContext = new NutshellContext ();
var query = dbContext.MedicalArticles
.Where (article => article.Topic == "influenza")
.AsEnumerable()
.Where (article => wordCounter.Matches (article.Abstract).Count < 100);
An alternative to calling AsEnumerable is to call ToArray or ToList. The advantage
of AsEnumerable is that it doesn’t force immediate query execution, nor does it
create any storage structure.
Moving query processing from the database server to the cli‐
ent can hurt performance, especially if it means retrieving
more rows. A more efficient (though more complex) way
to solve our example would be to use SQL CLR integration
to expose a function on the database that implemented the
regular expression.
We further demonstrate combined interpreted and local queries in Chapter 10.
EF Core
Throughout this and Chapter 9, we use EF Core to demonstrate interpreted queries.
Let’s now examine the key features of this technology.
EF Core Entity Classes
EF Core lets you use any class to represent data, as long as it contains a public
property for each column that you want to query.
For instance, we could define the following entity class to query and update a
Customers table in the database:
430 | Chapter 8: LINQ Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-448-320.jpg)
![Install-Package Microsoft.EntityFrameworkCore.Tools
Add-Migration InitialCreate
Update-Database
The first command installs tools to manage EF Core from within Visual Studio.
The second command generates a special C# class known as a code migration that
contains instructions to create the database. The final command runs those instruc‐
tions against the database connection string specified in the project’s application
configuration file.
Using DbContext
After you’ve defined Entity classes and subclassed DbContext, you can instantiate
your DbContext and query the database, as follows:
using var dbContext = new NutshellContext();
Console.WriteLine (dbContext.Customers.Count());
// Executes "SELECT COUNT(*) FROM [Customer] AS [c]"
You can also use your DbContext instance to write to the database. The following
code inserts a row into the Customer table:
using var dbContext = new NutshellContext();
Customer cust = new Customer()
{
Name = "Sara Wells"
};
dbContext.Customers.Add (cust);
dbContext.SaveChanges(); // Writes changes back to database
The following queries the database for the customer that was just inserted:
using var dbContext = new NutshellContext();
Customer cust = dbContext.Customers
.Single (c => c.Name == "Sara Wells")
The following updates that customer’s name and writes the change to the database:
cust.Name = "Dr. Sara Wells";
dbContext.SaveChanges();
The Single operator is ideal for retrieving a row by primary
key. Unlike First, it throws an exception if more than one
element is returned.
Object Tracking
A DbContext instance keeps track of all the entities it instantiates, so it can feed the
same ones back to you whenever you request the same rows in a table. In other
words, a context in its lifetime will never emit two separate entities that refer to the
same row in a table (where a row is identified by primary key). This capability is
called object tracking.
EF Core | 435
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-453-320.jpg)
![When you assign a lambda expression to an intermediate variable, however, you
must be explicit about whether to resolve to a delegate (i.e., Func<>) or an expres‐
sion tree (i.e., Expression<Func<>>). In the following example, predicate1 and
predicate2 are not interchangeable:
Func <Product, bool> predicate1 = p => !p.Discontinued;
IEnumerable<Product> q1 = localProducts.Where (predicate1);
Expression <Func <Product, bool>> predicate2 = p => !p.Discontinued;
IQueryable<Product> q2 = sqlProducts.Where (predicate2);
Compiling expression trees
You can convert an expression tree to a delegate by calling Compile. This is of par‐
ticular value when writing methods that return reusable expressions. To illustrate,
let’s add a static method to the Product class that returns a predicate evaluating to
true if a product is not discontinued and has sold in the past 30 days:
public class Product
{
public static Expression<Func<Product, bool>> IsSelling()
{
return p => !p.Discontinued && p.LastSale > DateTime.Now.AddDays (-30);
}
}
The method just written can be used both in interpreted and local queries, as
follows:
void Test()
{
var dbContext = new NutshellContext();
Product[] localProducts = dbContext.Products.ToArray();
IQueryable<Product> sqlQuery =
dbContext.Products.Where (Product.IsSelling());
IEnumerable<Product> localQuery =
localProducts.Where (Product.IsSelling().Compile());
}
.NET does not provide an API to convert in the reverse
direction, from a delegate to an expression tree. This makes
expression trees more versatile.
AsQueryable
The AsQueryable operator lets you write whole queries that can run over either local
or remote sequences:
IQueryable<Product> FilterSortProducts (IQueryable<Product> input)
{
return from p in input
where ...
Building Query Expressions | 443
LINQ
Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-461-320.jpg)
![orderby ...
select p;
}
void Test()
{
var dbContext = new NutshellContext();
Product[] localProducts = dbContext.Products.ToArray();
var sqlQuery = FilterSortProducts (dbContext.Products);
var localQuery = FilterSortProducts (localProducts.AsQueryable());
...
}
AsQueryable wraps IQueryable<T> clothing around a local sequence so that subse‐
quent query operators resolve to expression trees. When you later enumerate over
the result, the expression trees are implicitly compiled (at a small performance cost),
and the local sequence enumerates as it would ordinarily.
Expression Trees
We said previously that an implicit conversion from a lambda expression to Expres
sion<TDelegate> causes the C# compiler to emit code that builds an expression
tree. With some programming effort, you can do the same thing manually at
runtime—in other words, dynamically build an expression tree from scratch. The
result can be cast to an Expression<TDelegate> and used in EF Core queries or
compiled into an ordinary delegate by calling Compile.
The Expression DOM
An expression tree is a miniature code DOM. Each node in the tree is represented
by a type in the System.Linq.Expressions namespace. Figure 8-10 illustrates these
types.
Figure 8-10. Expression types
444 | Chapter 8: LINQ Queries](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-462-320.jpg)
![9
LINQ Operators
This chapter describes each of the LINQ query operators. As well as serving as a
reference, two of the sections, “Projecting” on page 449 and “Joining” on page 449,
cover a number of conceptual areas:
• Projecting object hierarchies
•
• Joining with Select, SelectMany, Join, and GroupJoin
•
• Query expressions with multiple range variables
•
All of the examples in this chapter assume that a names array is defined as follows:
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
Examples that query a database assume that a variable called dbContext is instanti‐
ated as
var dbContext = new NutshellContext();
where NutshellContext is defined as follows:
public class NutshellContext : DbContext
{
public DbSet<Customer> Customers { get; set; }
public DbSet<Purchase> Purchases { get; set; }
protected override void OnModelCreating(ModelBuilder modelBuilder)
{
modelBuilder.Entity<Customer>(entity =>
{
entity.ToTable("Customer");
entity.Property(e => e.Name).IsRequired(); // Column is not nullable
});
modelBuilder.Entity<Purchase>(entity =>
{
entity.ToTable("Purchase");
entity.Property(e => e.Date).IsRequired();
447](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-465-320.jpg)
![element pair. Rather than naming the type arguments TOuter and TInner, the Zip
operator names them TFirst and TSecond:
IEnumerable<TFirst>, IEnumerable<TSecond>→IEnumerable<TResult>
Join, GroupJoin, Zip
Ordering
IEnumerable<TSource>→IOrderedEnumerable<TSource>
Returns a reordering of a sequence.
OrderBy, OrderByDescending, ThenBy, ThenByDescending, Reverse
Grouping
IEnumerable<TSource>→IEnumerable<IGrouping<TKey,TElement>>
IEnumerable<TSource>→IEnumerable<TElement[]>
Groups a sequence into subsequences.
GroupBy, Chunk
Set operators
IEnumerable<TSource>, IEnumerable<TSource>→IEnumerable<TSource>
Takes two same-typed sequences and returns their commonality, sum, or difference.
Concat, Union, UnionBy, Intersect, IntersectBy, Except, ExceptBy
Conversion methods: Import
IEnumerable→IEnumerable<TResult>
OfType, Cast
Conversion methods: Export
IEnumerable<TSource>→An array, list, dictionary, lookup, or sequence
ToArray, ToList, ToDictionary, ToLookup, AsEnumerable, AsQueryable
Sequence→Element or Value
The following query operators accept an input sequence and emit a single element
or value.
Element operators
IEnumerable<TSource>→TSource
Picks a single element from a sequence.
450 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-468-320.jpg)
![Overview
Where returns the elements from the input sequence that satisfy the given predicate.
For instance:
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
IEnumerable<string> query = names.Where (name => name.EndsWith ("y"));
// Harry
// Mary
// Jay
In query syntax:
IEnumerable<string> query = from n in names
where n.EndsWith ("y")
select n;
A where clause can appear more than once in a query and be interspersed with let,
orderby, and join clauses:
from n in names
where n.Length > 3
let u = n.ToUpper()
where u.EndsWith ("Y")
select u;
// HARRY
// MARY
Standard C# scoping rules apply to such queries. In other words, you cannot refer to
a variable prior to declaring it with a range variable or a let clause.
Indexed filtering
Where’s predicate optionally accepts a second argument, of type int. This is fed with
the position of each element within the input sequence, allowing the predicate to
use this information in its filtering decision. For example, the following skips every
second element:
IEnumerable<string> query = names.Where ((n, i) => i % 2 == 0);
// Tom
// Harry
// Jay
An exception is thrown if you use indexed filtering in EF Core.
SQL LIKE comparisons in EF Core
The following methods on string translate to SQL’s LIKE operator:
Contains, StartsWith, EndsWith
Filtering | 453
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-471-320.jpg)
![For instance, c.Name.Contains ("abc") translates to customer.Name LIKE
'%abc%' (or more accurately, a parameterized version of this). Contains lets you
compare only against a locally evaluated expression; to compare against another
column, you must use the EF.Functions.Like method:
... where EF.Functions.Like (c.Description, "%" + c.Name + "%")
EF.Functions.Like also lets you perform more complex comparisons (e.g., LIKE
'abc%def%').
< and > string comparisons in EF Core
You can perform order comparison on strings with string’s CompareTo method; this
maps to SQL’s < and > operators:
dbContext.Purchases.Where (p => p.Description.CompareTo ("C") < 0)
WHERE x IN (…, …, …) in EF Core
With EF Core, you can apply the Contains operator to a local collection within a
filter predicate. For instance:
string[] chosenOnes = { "Tom", "Jay" };
from c in dbContext.Customers
where chosenOnes.Contains (c.Name)
...
This maps to SQL’s IN operator. In other words:
WHERE customer.Name IN ("Tom", "Jay")
If the local collection is an array of entities or nonscalar types, EF Core might
instead emit an EXISTS clause.
Take, TakeLast, Skip, and SkipLast
Argument Type
Source sequence IEnumerable<TSource>
Number of elements to take or skip int
Take emits the first n elements and discards the rest; Skip discards the first n ele‐
ments and emits the rest. The two methods are useful together when implementing
a web page allowing a user to navigate through a large set of matching records. For
instance, suppose that a user searches a book database for the term “mercury”, and
there are 100 matches. The following returns the first 20:
IQueryable<Book> query = dbContext.Books
.Where (b => b.Title.Contains ("mercury"))
.OrderBy (b => b.Title)
.Take (20);
454 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-472-320.jpg)
![The next query returns books 21 to 40:
IQueryable<Book> query = dbContext.Books
.Where (b => b.Title.Contains ("mercury"))
.OrderBy (b => b.Title)
.Skip (20).Take (20);
EF Core translates Take and Skip to the ROW_NUMBER function in SQL Server 2005,
or a TOP n subquery in earlier versions of SQL Server.
The TakeLast and SkipLast methods take or skip the last n elements.
From .NET 6, the Take method is overloaded to accept a Range variable. This
overload can subsume the functionality of all four methods; for instance, Take(5..)
is equivalent to Skip(5), and Take(..^5) is equivalent to SkipLast(5).
TakeWhile and SkipWhile
Argument Type
Source sequence IEnumerable<TSource>
Predicate TSource => bool or (TSource,int) => bool
TakeWhile enumerates the input sequence, emitting each item until the given predi‐
cate is false. It then ignores the remaining elements:
int[] numbers = { 3, 5, 2, 234, 4, 1 };
var takeWhileSmall = numbers.TakeWhile (n => n < 100); // { 3, 5, 2 }
SkipWhile enumerates the input sequence, ignoring each item until the given predi‐
cate is false. It then emits the remaining elements:
int[] numbers = { 3, 5, 2, 234, 4, 1 };
var skipWhileSmall = numbers.SkipWhile (n => n < 100); // { 234, 4, 1 }
TakeWhile and SkipWhile have no translation to SQL and throw an exception if
used in an EF Core query.
Distinct and DistinctBy
Distinct returns the input sequence, stripped of duplicates. You can optionally pass
in a custom equality comparer. The following returns distinct letters in a string:
char[] distinctLetters = "HelloWorld".Distinct().ToArray();
string s = new string (distinctLetters); // HeloWrd
We can call LINQ methods directly on a string because string implements
IEnumerable<char>.
The DistinctBy method was introduced in .NET 6 and lets you specify a key
selector to be applied before performing equality comparison. The result of the
following expression is {1,2,3}:
Filtering | 455
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-473-320.jpg)
![new[] { 1.0, 1.1, 2.0, 2.1, 3.0, 3.1 }.DistinctBy (n => Math.Round (n, 0))
Projecting
IEnumerable<TSource>→ IEnumerable<TResult>
Method Description SQL equivalents
Select Transforms each input element with the given lambda expression SELECT
SelectMany Transforms each input element, and then flattens and concatenates
the resultant subsequences
INNER JOIN,
LEFT OUTER JOIN,
CROSS JOIN
When querying a database, Select and SelectMany are the
most versatile joining constructs; for local queries, Join and
GroupJoin are the most efficient joining constructs.
Select
Argument Type
Source sequence IEnumerable<TSource>
Result selector TSource => TResult or (TSource,int) => TResulta
a Prohibited with EF Core
Query syntax
select projection-expression
Enumerable implementation
public static IEnumerable<TResult> Select<TSource,TResult>
(this IEnumerable<TSource> source, Func<TSource,TResult> selector)
{
foreach (TSource element in source)
yield return selector (element);
}
Overview
With Select, you always get the same number of elements that you started with.
Each element, however, can be transformed in any manner by the lambda function.
The following selects the names of all fonts installed on the computer (from
System.Drawing):
IEnumerable<string> query = from f in FontFamily.Families
select f.Name;
456 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-474-320.jpg)
![foreach (string name in query) Console.WriteLine (name);
In this example, the select clause converts a FontFamily object to its name. Here’s
the lambda equivalent:
IEnumerable<string> query = FontFamily.Families.Select (f => f.Name);
Select statements are often used to project into anonymous types:
var query =
from f in FontFamily.Families
select new { f.Name, LineSpacing = f.GetLineSpacing (FontStyle.Bold) };
A projection with no transformation is sometimes used with query syntax to satisfy
the requirement that the query end in a select or group clause. The following
selects fonts supporting strikeout:
IEnumerable<FontFamily> query =
from f in FontFamily.Families
where f.IsStyleAvailable (FontStyle.Strikeout)
select f;
foreach (FontFamily ff in query) Console.WriteLine (ff.Name);
In such cases, the compiler omits the projection when translating to fluent syntax.
Indexed projection
The selector expression can optionally accept an integer argument, which acts as
an indexer, providing the expression with the position of each input in the input
sequence. This works only with local queries:
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
IEnumerable<string> query = names
.Select ((s,i) => i + "=" + s); // { "0=Tom", "1=Dick", ... }
Select subqueries and object hierarchies
You can nest a subquery in a select clause to build an object hierarchy. The
following example returns a collection describing each directory under Path.Get
TempPath(), with a subcollection of files under each directory:
string tempPath = Path.GetTempPath();
DirectoryInfo[] dirs = new DirectoryInfo (tempPath).GetDirectories();
var query =
from d in dirs
where (d.Attributes & FileAttributes.System) == 0
select new
{
DirectoryName = d.FullName,
Created = d.CreationTime,
Projecting | 457
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-475-320.jpg)
![You can use SelectMany to expand child sequences, flatten nested collections, and
join two collections into a flat output sequence. Using the conveyor belt analogy,
SelectMany funnels fresh material onto a conveyor belt. With SelectMany, each
input element is the trigger for the introduction of fresh material. The fresh material
is emitted by the selector lambda expression and must be a sequence. In other
words, the lambda expression must emit a child sequence per input element. The
final result is a concatenation of the child sequences emitted for each input element.
Starting with a simple example, suppose that we have the following array of names,
string[] fullNames = { "Anne Williams", "John Fred Smith", "Sue Green" };
that we want to convert to a single flat collection of words—in other words:
"Anne", "Williams", "John", "Fred", "Smith", "Sue", Green"
SelectMany is ideal for this task, because we’re mapping each input element to a
variable number of output elements. All we must do is come up with a selector
expression that converts each input element to a child sequence. string.Split does
the job nicely: it takes a string and splits it into words, emitting the result as an
array:
string testInputElement = "Anne Williams";
string[] childSequence = testInputElement.Split();
// childSequence is { "Anne", "Williams" };
So, here’s our SelectMany query and the result:
IEnumerable<string> query = fullNames.SelectMany (name => name.Split());
foreach (string name in query)
Console.Write (name + "|"); // Anne|Williams|John|Fred|Smith|Sue|Green|
If you replace SelectMany with Select, you get the same
results in hierarchical form. The following emits a sequence
of string arrays, requiring nested foreach statements to
enumerate:
IEnumerable<string[]> query =
fullNames.Select (name => name.Split());
foreach (string[] stringArray in query)
foreach (string name in stringArray)
Console.Write (name + "|");
The benefit of SelectMany is that it yields a single flat result
sequence.
SelectMany is supported in query syntax and is invoked by having an additional
generator—in other words, an extra from clause in the query. The from keyword
has two meanings in query syntax. At the start of a query, it introduces the original
range variable and input sequence. Anywhere else in the query, it translates to
SelectMany. Here’s our query in query syntax:
462 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-480-320.jpg)
![IEnumerable<string> query = fullNames
.SelectMany (fName => fName.Split()
.Select (name => new { name, fName } ))
.OrderBy (x => x.fName)
.ThenBy (x => x.name)
.Select (x => x.name + " came from " + x.fName);
Thinking in query syntax
As we just demonstrated, there are good reasons to use query syntax if you need
multiple range variables. In such cases, it helps to not only use query syntax but also
to think directly in its terms.
There are two basic patterns when writing additional generators. The first is
expanding and flattening subsequences. To do this, you call a property or method on
an existing range variable in your additional generator. We did this in the previous
example:
from fullName in fullNames
from name in fullName.Split()
Here, we’ve expanded from enumerating full names to enumerating words. An
analogous EF Core query is when you expand collection navigation properties. The
following query lists all customers along with their purchases:
IEnumerable<string> query = from c in dbContext.Customers
from p in c.Purchases
select c.Name + " bought a " + p.Description;
Tom bought a Bike
Tom bought a Holiday
Dick bought a Phone
Harry bought a Car
...
Here, we’ve expanded each customer into a subsequence of purchases.
The second pattern is performing a cartesian product, or cross join, in which every
element of one sequence is matched with every element of another. To do this,
introduce a generator whose selector expression returns a sequence unrelated to a
range variable:
int[] numbers = { 1, 2, 3 }; string[] letters = { "a", "b" };
IEnumerable<string> query = from n in numbers
from l in letters
select n.ToString() + l;
// RESULT: { "1a", "1b", "2a", "2b", "3a", "3b" }
This style of query is the basis of SelectMany-style joins.
464 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-482-320.jpg)
![Joining with SelectMany
You can use SelectMany to join two sequences simply by filtering the results of a
cross product. For instance, suppose that we want to match players for a game. We
could start as follows:
string[] players = { "Tom", "Jay", "Mary" };
IEnumerable<string> query = from name1 in players
from name2 in players
select name1 + " vs " + name2;
//RESULT: { "Tom vs Tom", "Tom vs Jay", "Tom vs Mary",
// "Jay vs Tom", "Jay vs Jay", "Jay vs Mary",
// "Mary vs Tom", "Mary vs "Jay", "Mary vs Mary" }
The query reads “For every player, reiterate every player, selecting player 1 versus
player 2.” Although we got what we asked for (a cross join), the results are not useful
until we add a filter:
IEnumerable<string> query = from name1 in players
from name2 in players
where name1.CompareTo (name2) < 0
orderby name1, name2
select name1 + " vs " + name2;
//RESULT: { "Jay vs Mary", "Jay vs Tom", "Mary vs Tom" }
The filter predicate constitutes the join condition. Our query can be called a non-
equi join because the join condition doesn’t use an equality operator.
SelectMany in EF Core
SelectMany in EF Core can perform cross joins, non-equi joins, inner joins, and left
outer joins. You can use SelectMany with both predefined associations and ad hoc
relationships—just as with Select. The difference is that SelectMany returns a flat
rather than a hierarchical result set.
An EF Core cross join is written just as in the preceding section. The following
query matches every customer to every purchase (a cross join):
var query = from c in dbContext.Customers
from p in dbContext.Purchases
select c.Name + " might have bought a " + p.Description;
More typically, though, you’d want to match customers to only their own purchases.
You achieve this by adding a where clause with a joining predicate. This results in a
standard SQL-style equi-join:
var query = from c in dbContext.Customers
from p in dbContext.Purchases
where c.ID == p.CustomerID
select c.Name + " bought a " + p.Description;
Projecting | 465
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-483-320.jpg)
![GroupJoin arguments
Argument Type
Outer sequence IEnumerable<TOuter>
Inner sequence IEnumerable<TInner>
Outer key selector TOuter => TKey
Inner key selector TInner => TKey
Result selector (TOuter,IEnumerable<TInner>) => TResult
Query syntax
from outer-var in outer-enumerable
join inner-var in inner-enumerable on outer-key-expr equals inner-key-expr
[ into identifier ]
Overview
Join and GroupJoin mesh two input sequences into a single output sequence. Join
emits flat output; GroupJoin emits hierarchical output.
Join and GroupJoin provide an alternative strategy to Select and SelectMany.
The advantage of Join and GroupJoin is that they execute efficiently over local
in-memory collections because they first load the inner sequence into a keyed
lookup, avoiding the need to repeatedly enumerate over every inner element. The
disadvantage is that they offer the equivalent of inner and left outer joins only; cross
joins and non-equi joins must still be done using Select/SelectMany. With EF Core
queries, Join and GroupJoin offer no real benefits over Select and SelectMany.
Table 9-1 summarizes the differences between each of the joining strategies.
Table 9-1. Joining strategies
Strategy Result
shape
Local query
efficiency
Inner
joins
Left
outer
joins
Cross
joins
Non-
equi
joins
Select + SelectMany Flat Bad Yes Yes Yes Yes
Select + Select Nested Bad Yes Yes Yes Yes
Join Flat Good Yes — — —
GroupJoin Nested Good Yes Yes — —
GroupJoin + SelectMany Flat Good Yes Yes — —
Joining | 469
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-487-320.jpg)
![Join
The Join operator performs an inner join, emitting a flat output sequence.
The following query lists all customers alongside their purchases without using a
navigation property:
IQueryable<string> query =
from c in dbContext.Customers
join p in dbContext.Purchases on c.ID equals p.CustomerID
select c.Name + " bought a " + p.Description;
The results match what we would get from a SelectMany-style query:
Tom bought a Bike
Tom bought a Holiday
Dick bought a Phone
Harry bought a Car
To see the benefit of Join over SelectMany, we must convert this to a local query.
We can demonstrate this by first copying all customers and purchases to arrays and
then querying the arrays:
Customer[] customers = dbContext.Customers.ToArray();
Purchase[] purchases = dbContext.Purchases.ToArray();
var slowQuery = from c in customers
from p in purchases where c.ID == p.CustomerID
select c.Name + " bought a " + p.Description;
var fastQuery = from c in customers
join p in purchases on c.ID equals p.CustomerID
select c.Name + " bought a " + p.Description;
Although both queries yield the same results, the Join query is considerably
faster because its implementation in Enumerable preloads the inner collection
(purchases) into a keyed lookup.
The query syntax for join can be written in general terms, as follows:
join inner-var in inner-sequence on outer-key-expr equals inner-key-expr
Join operators in LINQ differentiate between the outer sequence and inner sequence.
Syntactically:
• The outer sequence is the input sequence (customers, in this case).
•
• The inner sequence is the new collection you introduce (purchases, in this
•
case).
Join performs inner joins, meaning customers without purchases are excluded
from the output. With inner joins, you can swap the inner and outer sequences in
the query and still get the same results:
from p in purchases // p is now outer
join c in customers on p.CustomerID equals c.ID // c is now inner
...
470 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-488-320.jpg)
![The result selector expression at the end creates each element in the output
sequence. If you have additional clauses prior to projecting, such as orderby in
this example:
from c in customers
join p in purchases on c.ID equals p.CustomerID
orderby p.Price
select c.Name + " bought a " + p.Description;
you must manufacture a temporary anonymous type in the result selector in fluent
syntax. This keeps both c and p in scope following the join:
customers.Join ( // outer collection
purchases, // inner collection
c => c.ID, // outer key selector
p => p.CustomerID, // inner key selector
(c, p) => new { c, p } ) // result selector
.OrderBy (x => x.p.Price)
.Select (x => x.c.Name + " bought a " + x.p.Description);
Query syntax is usually preferable when joining; it’s less fiddly.
GroupJoin
GroupJoin does the same work as Join, but instead of yielding a flat result, it yields
a hierarchical result, grouped by each outer element. It also allows left outer joins.
GroupJoin is not currently supported in EF Core.
The query syntax for GroupJoin is the same as for Join, but is followed by the into
keyword.
Here’s the most basic example, using a local query:
Customer[] customers = dbContext.Customers.ToArray();
Purchase[] purchases = dbContext.Purchases.ToArray();
IEnumerable<IEnumerable<Purchase>> query =
from c in customers
join p in purchases on c.ID equals p.CustomerID
into custPurchases
select custPurchases; // custPurchases is a sequence
An into clause translates to GroupJoin only when it appears
directly after a join clause. After a select or group clause,
it means query continuation. The two uses of the into key‐
word are quite different, although they have one feature in
common: they both introduce a new range variable.
The result is a sequence of sequences, which we could enumerate as follows:
foreach (IEnumerable<Purchase> purchaseSequence in query)
foreach (Purchase p in purchaseSequence)
Console.WriteLine (p.Description);
472 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-490-320.jpg)
![expands and flattens all the purchase subsequences, concatenating them into a single
sequence of purchase elements.
Joining with lookups
The Join and GroupJoin methods in Enumerable work in two steps. First, they
load the inner sequence into a lookup. Second, they query the outer sequence in
combination with the lookup.
A lookup is a sequence of groupings that can be accessed directly by key. Another
way to think of it is as a dictionary of sequences—a dictionary that can accept
many elements under each key (sometimes called a multidictionary). Lookups are
read-only and defined by the following interface:
public interface ILookup<TKey,TElement> :
IEnumerable<IGrouping<TKey,TElement>>, IEnumerable
{
int Count { get; }
bool Contains (TKey key);
IEnumerable<TElement> this [TKey key] { get; }
}
The joining operators—like other sequence-emitting opera‐
tors—honor deferred or lazy execution semantics. This means
the lookup is not built until you begin enumerating the output
sequence (and then the entire lookup is built right then).
You can create and query lookups manually as an alternative strategy to using the
joining operators when dealing with local collections. There are a couple of benefits
to doing so:
• You can reuse the same lookup over multiple queries—as well as in ordinary
•
imperative code.
• Querying a lookup is an excellent way of understanding how Join and Group
•
Join work.
The ToLookup extension method creates a lookup. The following loads all purchases
into a lookup—keyed by their CustomerID:
ILookup<int?,Purchase> purchLookup =
purchases.ToLookup (p => p.CustomerID, p => p);
The first argument selects the key; the second argument selects the objects that are
to be loaded as values into the lookup.
Reading a lookup is rather like reading a dictionary except that the indexer returns
a sequence of matching items rather than a single matching item. The following
enumerates all purchases made by the customer whose ID is 1:
foreach (Purchase p in purchLookup [1])
Console.WriteLine (p.Description);
474 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-492-320.jpg)
![With a lookup in place, you can write SelectMany/Select queries that execute as
efficiently as Join/GroupJoin queries. Join is equivalent to using SelectMany on a
lookup:
from c in customers
from p in purchLookup [c.ID]
select new { c.Name, p.Description, p.Price };
Tom Bike 500
Tom Holiday 2000
Dick Bike 600
Dick Phone 300
...
Adding a call to DefaultIfEmpty makes this into an outer join:
from c in customers
from p in purchLookup [c.ID].DefaultIfEmpty()
select new {
c.Name,
Descript = p == null ? null : p.Description,
Price = p == null ? (decimal?) null : p.Price
};
GroupJoin is equivalent to reading the lookup inside a projection:
from c in customers
select new {
CustName = c.Name,
CustPurchases = purchLookup [c.ID]
};
Enumerable implementations
Here’s the simplest valid implementation of Enumerable.Join, null checking aside:
public static IEnumerable <TResult> Join
<TOuter,TInner,TKey,TResult> (
this IEnumerable <TOuter> outer,
IEnumerable <TInner> inner,
Func <TOuter,TKey> outerKeySelector,
Func <TInner,TKey> innerKeySelector,
Func <TOuter,TInner,TResult> resultSelector)
{
ILookup <TKey, TInner> lookup = inner.ToLookup (innerKeySelector);
return
from outerItem in outer
from innerItem in lookup [outerKeySelector (outerItem)]
select resultSelector (outerItem, innerItem);
}
Joining | 475
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-493-320.jpg)
![GroupJoin’s implementation is like that of Join but simpler:
public static IEnumerable <TResult> GroupJoin
<TOuter,TInner,TKey,TResult> (
this IEnumerable <TOuter> outer,
IEnumerable <TInner> inner,
Func <TOuter,TKey> outerKeySelector,
Func <TInner,TKey> innerKeySelector,
Func <TOuter,IEnumerable<TInner>,TResult> resultSelector)
{
ILookup <TKey, TInner> lookup = inner.ToLookup (innerKeySelector);
return
from outerItem in outer
select resultSelector
(outerItem, lookup [outerKeySelector (outerItem)]);
}
The Zip Operator
IEnumerable<TFirst>, IEnumerable<TSecond>→IEnumerable<TResult>
The Zip operator enumerates two sequences in step (like a zipper), returning a
sequence based on applying a function over each element pair. For instance, the
following:
int[] numbers = { 3, 5, 7 };
string[] words = { "three", "five", "seven", "ignored" };
IEnumerable<string> zip = numbers.Zip (words, (n, w) => n + "=" + w);
produces a sequence with the following elements:
3=three
5=five
7=seven
Extra elements in either input sequence are ignored. Zip is not supported by EF
Core.
Ordering
IEnumerable<TSource>→IOrderedEnumerable<TSource>
Method Description SQL equivalents
OrderBy, ThenBy Sorts a sequence in ascending order ORDER BY ...
OrderByDescending,
ThenByDescending
Sorts a sequence in descending order ORDER BY ... DESC
Reverse Returns a sequence in reverse order Exception thrown
Ordering operators return the same elements in a different order.
476 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-494-320.jpg)
![OrderBy, OrderByDescending, ThenBy, and ThenByDescending
OrderBy and OrderByDescending arguments
Argument Type
Input sequence IEnumerable<TSource>
Key selector TSource => TKey
Return type = IOrderedEnumerable<TSource>
ThenBy and ThenByDescending arguments
Argument Type
Input sequence IOrderedEnumerable<TSource>
Key selector TSource => TKey
Query syntax
orderby expression1 [descending] [, expression2 [descending] ... ]
Overview
OrderBy returns a sorted version of the input sequence, using the keySelector
expression to make comparisons. The following query emits a sequence of names in
alphabetical order:
IEnumerable<string> query = names.OrderBy (s => s);
The following sorts names by length:
IEnumerable<string> query = names.OrderBy (s => s.Length);
// Result: { "Jay", "Tom", "Mary", "Dick", "Harry" };
The relative order of elements with the same sorting key (in this case, Jay/Tom and
Mary/Dick) is indeterminate—unless you append a ThenBy operator:
IEnumerable<string> query = names.OrderBy (s => s.Length).ThenBy (s => s);
// Result: { "Jay", "Tom", "Dick", "Mary", "Harry" };
ThenBy reorders only elements that had the same sorting key in the preceding sort.
You can chain any number of ThenBy operators. The following sorts first by length,
then by the second character, and finally by the first character:
names.OrderBy (s => s.Length).ThenBy (s => s[1]).ThenBy (s => s[0]);
Ordering | 477
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-495-320.jpg)
![Here’s the equivalent in query syntax:
from s in names
orderby s.Length, s[1], s[0]
select s;
The following variation is incorrect—it will actually order first
by s[1] and then by s.Length (or in the case of a database
query, it will order only by s[1] and discard the former
ordering):
from s in names
orderby s.Length
orderby s[1]
...
LINQ also provides OrderByDescending and ThenByDescending operators, which
do the same things, emitting the results in reverse order. The following EF Core
query retrieves purchases in descending order of price, with those of the same price
listed alphabetically:
dbContext.Purchases.OrderByDescending (p => p.Price)
.ThenBy (p => p.Description);
In query syntax:
from p in dbContext.Purchases
orderby p.Price descending, p.Description
select p;
Comparers and collations
In a local query, the key selector objects themselves determine the ordering algo‐
rithm via their default IComparable implementation (see Chapter 7). You can
override the sorting algorithm by passing in an IComparer object. The following
performs a case-insensitive sort:
names.OrderBy (n => n, StringComparer.CurrentCultureIgnoreCase);
Passing in a comparer is not supported in query syntax or in any way by EF
Core. When querying a database, the comparison algorithm is determined by the
participating column’s collation. If the collation is case sensitive, you can request a
case-insensitive sort by calling ToUpper in the key selector:
from p in dbContext.Purchases
orderby p.Description.ToUpper()
select p;
IOrderedEnumerable and IOrderedQueryable
The ordering operators return special subtypes of IEnumerable<T>. Those in
Enumerable return IOrderedEnumerable<TSource>; those in Queryable return
IOrderedQueryable<TSource>. These subtypes allow a subsequent ThenBy operator
to refine rather than replace the existing ordering.
478 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-496-320.jpg)
![Query syntax
group element-expression by key-expression
Overview
GroupBy organizes a flat input sequence into sequences of groups. For example, the
following organizes all of the files in Path.GetTempPath() by extension:
string[] files = Directory.GetFiles (Path.GetTempPath());
IEnumerable<IGrouping<string,string>> query =
files.GroupBy (file => Path.GetExtension (file));
Or, with implicit typing:
var query = files.GroupBy (file => Path.GetExtension (file));
Here’s how to enumerate the result:
foreach (IGrouping<string,string> grouping in query)
{
Console.WriteLine ("Extension: " + grouping.Key);
foreach (string filename in grouping)
Console.WriteLine (" - " + filename);
}
Extension: .pdf
-- chapter03.pdf
-- chapter04.pdf
Extension: .doc
-- todo.doc
-- menu.doc
-- Copy of menu.doc
...
Enumerable.GroupBy works by reading the input elements into a temporary dictio‐
nary of lists so that all elements with the same key end up in the same sublist. It then
emits a sequence of groupings. A grouping is a sequence with a Key property:
public interface IGrouping <TKey,TElement> : IEnumerable<TElement>,
IEnumerable
{
TKey Key { get; } // Key applies to the subsequence as a whole
}
By default, the elements in each grouping are untransformed input elements unless
you specify an elementSelector argument. The following projects each input ele‐
ment to uppercase:
files.GroupBy (file => Path.GetExtension (file), file => file.ToUpper());
An elementSelector is independent of the keySelector. In our case, this means
that the Key on each grouping is still in its original case:
480 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-498-320.jpg)
![Extension: .pdf
-- CHAPTER03.PDF
-- CHAPTER04.PDF
Extension: .doc
-- TODO.DOC
Note that the subcollections are not emitted in alphabetical order of key. GroupBy
merely groups; it does not sort. In fact, it preserves the original ordering. To sort,
you must add an OrderBy operator:
files.GroupBy (file => Path.GetExtension (file), file => file.ToUpper())
.OrderBy (grouping => grouping.Key);
GroupBy has a simple and direct translation in query syntax:
group element-expr by key-expr
Here’s our example in query syntax:
from file in files
group file.ToUpper() by Path.GetExtension (file);
As with select, group “ends” a query—unless you add a query continuation clause:
from file in files
group file.ToUpper() by Path.GetExtension (file) into grouping
orderby grouping.Key
select grouping;
Query continuations are often useful in a group by query. The next query filters out
groups that have fewer than five files in them:
from file in files
group file.ToUpper() by Path.GetExtension (file) into grouping
where grouping.Count() >= 5
select grouping;
A where after a group by is equivalent to HAVING in SQL.
It applies to each subsequence or grouping as a whole rather
than the individual elements.
Sometimes, you’re interested purely in the result of an aggregation on a grouping
and so can abandon the subsequences:
string[] votes = { "Dogs", "Cats", "Cats", "Dogs", "Dogs" };
IEnumerable<string> query = from vote in votes
group vote by vote into g
orderby g.Count() descending
select g.Key;
string winner = query.First(); // Dogs
Grouping | 481
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-499-320.jpg)
![GroupBy in EF Core
Grouping works in the same way when querying a database. If you have navigation
properties set up, you’ll find, however, that the need to group arises less frequently
than with standard SQL. For instance, to select customers with at least two purcha‐
ses, you don’t need to group; the following query does the job nicely:
from c in dbContext.Customers
where c.Purchases.Count >= 2
select c.Name + " has made " + c.Purchases.Count + " purchases";
An example of when you might use grouping is to list total sales by year:
from p in dbContext.Purchases
group p.Price by p.Date.Year into salesByYear
select new {
Year = salesByYear.Key,
TotalValue = salesByYear.Sum()
};
LINQ’s grouping is more powerful than SQL’s GROUP BY in that you can fetch all
detail rows without any aggregation:
from p in dbContext.Purchases
group p by p.Date.Year
Date.Year
However, this doesn’t work in EF Core. An easy workaround is to call .AsEnumera
ble() just before grouping so that the grouping happens on the client. This is no
less efficient as long as you perform any filtering before grouping so that you only
fetch the data you need from the server.
Another departure from traditional SQL comes in there being no obligation to
project the variables or expressions used in grouping or sorting.
Grouping by multiple keys
You can group by a composite key, using an anonymous type:
from n in names
group n by new { FirstLetter = n[0], Length = n.Length };
Custom equality comparers
You can pass a custom equality comparer into GroupBy, in a local query, to change
the algorithm for key comparison. Rarely is this required, though, because changing
the key selector expression is usually sufficient. For instance, the following creates a
case-insensitive grouping:
group n by n.ToUpper()
482 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-500-320.jpg)
![Chunk
IEnumerable<TSource>→IEnumerable<TElement[]>
Argument Type
Input sequence IEnumerable<TSource>
size int
Introduced in .NET 6, Chunk groups a sequence into chunks of a given size (or
fewer, if there aren’t enough elements):
foreach (int[] chunk in new[] { 1, 2, 3, 4, 5, 6, 7, 8 }.Chunk (3))
Console.WriteLine (string.Join (", ", chunk));
Output:
1, 2, 3
4, 5, 6
7, 8
Set Operators
IEnumerable<TSource>, IEnumerable<TSource>→IEnumerable<TSource>
Method Description SQL equivalents
Concat Returns a concatenation of elements in each of
the two sequences
UNION ALL
Union, UnionBy Returns a concatenation of elements in each of
the two sequences, excluding duplicates
UNION
Intersect,
IntersectBy
Returns elements present in both sequences WHERE ... IN (...)
Except,
ExceptBy
Returns elements present in the first but not
the second sequence
EXCEPT or
WHERE ... NOT IN (...)
Concat, Union, and UnionBy
Concat returns all the elements of the first sequence, followed by all the elements of
the second. Union does the same but removes any duplicates:
int[] seq1 = { 1, 2, 3 }, seq2 = { 3, 4, 5 };
IEnumerable<int>
concat = seq1.Concat (seq2), // { 1, 2, 3, 3, 4, 5 }
union = seq1.Union (seq2); // { 1, 2, 3, 4, 5 }
Specifying the type argument explicitly is useful when the sequences are differently
typed but the elements have a common base type. For instance, with the reflection
API (Chapter 18), methods and properties are represented with MethodInfo and
Set Operators | 483
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-501-320.jpg)
![PropertyInfo classes, which have a common base class called MemberInfo. We can
concatenate methods and properties by stating that base class explicitly when calling
Concat:
MethodInfo[] methods = typeof (string).GetMethods();
PropertyInfo[] props = typeof (string).GetProperties();
IEnumerable<MemberInfo> both = methods.Concat<MemberInfo> (props);
In the next example, we filter the methods before concatenating:
var methods = typeof (string).GetMethods().Where (m => !m.IsSpecialName);
var props = typeof (string).GetProperties();
var both = methods.Concat<MemberInfo> (props);
This example relies on interface type parameter variance: methods is of type
IEnumerable<MethodInfo>, which requires a covariant conversion to IEnumerable
<MemberInfo>. It’s a good illustration of how variance makes things work more like
you’d expect.
UnionBy (introduced in .NET 6) takes a keySelector, which is used in determin‐
ing whether an element is a duplicate. In the following example, we perform a
case-insensitive union:
string[] seq1 = { "A", "b", "C" };
string[] seq2 = { "a", "B", "c" };
var union = seq1.UnionBy (seq2, x => x.ToUpperInvariant());
// union is { "A", "b", "C" }
In this case, the same thing can be accomplished with Union, if we supply an
equality comparer:
var union = seq1.Union (seq2, StringComparer.InvariantCultureIgnoreCase);
Intersect, Intersect By, Except, and ExceptBy
Intersect returns the elements that two sequences have in common. Except
returns the elements in the first input sequence that are not present in the second:
int[] seq1 = { 1, 2, 3 }, seq2 = { 3, 4, 5 };
IEnumerable<int>
commonality = seq1.Intersect (seq2), // { 3 }
difference1 = seq1.Except (seq2), // { 1, 2 }
difference2 = seq2.Except (seq1); // { 4, 5 }
Enumerable.Except works internally by loading all of the elements in the first
collection into a dictionary and then removing from the dictionary all elements
present in the second sequence. The equivalent in SQL is a NOT EXISTS or NOT IN
subquery:
SELECT number FROM numbers1Table
WHERE number NOT IN (SELECT number FROM numbers2Table)
484 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-502-320.jpg)
![The IntersectBy and ExceptBy methods (from .NET 6) let you specify a key
selector that’s applied before performing equality comparison (see the discussion on
UnionBy in the preceding section).
Conversion Methods
LINQ deals primarily in sequences; in other words, collections of type IEnumera
ble<T>. The conversion methods convert to and from other types of collections:
Method Description
OfType Converts IEnumerable to IEnumerable<T>, discarding wrongly typed elements
Cast Converts IEnumerable to IEnumerable<T>, throwing an exception if there are any
wrongly typed elements
ToArray Converts IEnumerable<T> to T[]
ToList Converts IEnumerable<T> to List<T>
ToDictionary Converts IEnumerable<T> to Dictionary<TKey,TValue>
ToLookup Converts IEnumerable<T> to ILookup<TKey,TElement>
AsEnumerable Upcasts to IEnumerable<T>
AsQueryable Casts or converts to IQueryable<T>
OfType and Cast
OfType and Cast accept a nongeneric IEnumerable collection and emit a generic
IEnumerable<T> sequence that you can subsequently query:
ArrayList classicList = new ArrayList(); // in System.Collections
classicList.AddRange ( new int[] { 3, 4, 5 } );
IEnumerable<int> sequence1 = classicList.Cast<int>();
Cast and OfType differ in their behavior when encountering an input element that’s
of an incompatible type. Cast throws an exception; OfType ignores the incompatible
element. Continuing the preceding example:
DateTime offender = DateTime.Now;
classicList.Add (offender);
IEnumerable<int>
sequence2 = classicList.OfType<int>(), // OK - ignores offending DateTime
sequence3 = classicList.Cast<int>(); // Throws exception
The rules for element compatibility exactly follow those of C#’s is operator, and
therefore consider only reference conversions and unboxing conversions. We can
see this by examining the internal implementation of OfType:
public static IEnumerable<TSource> OfType <TSource> (IEnumerable source)
{
foreach (object element in source)
if (element is TSource)
Conversion Methods | 485
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-503-320.jpg)
![yield return (TSource)element;
}
Cast has an identical implementation, except that it omits the type compatibility
test:
public static IEnumerable<TSource> Cast <TSource> (IEnumerable source)
{
foreach (object element in source)
yield return (TSource)element;
}
A consequence of these implementations is that you cannot use Cast to perform
numeric or custom conversions (for these, you must perform a Select operation
instead). In other words, Cast is not as flexible as C#’s cast operator:
int i = 3;
long l = i; // Implicit numeric conversion int->long
int i2 = (int) l; // Explicit numeric conversion long->int
We can demonstrate this by attempting to use OfType or Cast to convert a sequence
of ints to a sequence of longs:
int[] integers = { 1, 2, 3 };
IEnumerable<long> test1 = integers.OfType<long>();
IEnumerable<long> test2 = integers.Cast<long>();
When enumerated, test1 emits zero elements and test2 throws an exception.
Examining OfType’s implementation, it’s fairly clear why. After substituting TSource,
we get the following expression:
(element is long)
This returns false for an int element, due to the lack of an inheritance
relationship.
The reason that test2 throws an exception when enumerated
is more subtle. Notice in Cast’s implementation that element
is of type object. When TSource is a value type, the CLR
assumes this is an unboxing conversion and synthesizes a
method that reproduces the scenario described in the section
“Boxing and Unboxing” on page 126:
int value = 123;
object element = value;
long result = (long) element; // exception
Because the element variable is declared of type object, an
object-to-long cast is performed (an unboxing) rather than
an int-to-long numeric conversion. Unboxing operations
require an exact type match, so the object-to-long unbox
fails when given an int.
486 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-504-320.jpg)
![Element Operators
IEnumerable<TSource>→ TSource
Method Description SQL equivalents
First, FirstOrDefault Returns the first element in the sequence, optionally
satisfying a predicate
SELECT TOP 1 ...
ORDER BY ...
Last,
LastOrDefault
Returns the last element in the sequence, optionally
satisfying a predicate
SELECT TOP 1 ...
ORDER BY ... DESC
Single,
SingleOrDefault
Equivalent to First/FirstOrDefault, but
throws an exception if there is more than one match
ElementAt,
ElementAtOrDefault
Returns the element at the specified position Exception thrown
MinBy, MaxBy Returns the element with the smallest or largest
value
Exception thrown
DefaultIfEmpty Returns a single-element sequence whose value
is default(TSource) if the sequence has no
elements
OUTER JOIN
Methods ending in “OrDefault” return default(TSource) rather than throwing an
exception if the input sequence is empty or if no elements match the supplied
predicate.
default(TSource) is null for reference type elements, false for the bool type, and
zero for numeric types.
First, Last, and Single
Argument Type
Source sequence IEnumerable<TSource>
Predicate (optional) TSource => bool
The following example demonstrates First and Last:
int[] numbers = { 1, 2, 3, 4, 5 };
int first = numbers.First(); // 1
int last = numbers.Last(); // 5
int firstEven = numbers.First (n => n % 2 == 0); // 2
int lastEven = numbers.Last (n => n % 2 == 0); // 4
The following demonstrates First versus FirstOrDefault:
int firstBigError = numbers.First (n => n > 10); // Exception
int firstBigNumber = numbers.FirstOrDefault (n => n > 10); // 0
488 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-506-320.jpg)
![To prevent an exception, Single requires exactly one matching element; SingleOr
Default requires one or zero matching elements:
int onlyDivBy3 = numbers.Single (n => n % 3 == 0); // 3
int divBy2Err = numbers.Single (n => n % 2 == 0); // Error: 2 & 4 match
int singleError = numbers.Single (n => n > 10); // Error
int noMatches = numbers.SingleOrDefault (n => n > 10); // 0
int divBy2Error = numbers.SingleOrDefault (n => n % 2 == 0); // Error
Single is the “fussiest” in this family of element operators. FirstOrDefault and
LastOrDefault are the most tolerant.
In EF Core, Single is often used to retrieve a row from a table by primary key:
Customer cust = dataContext.Customers.Single (c => c.ID == 3);
ElementAt
Argument Type
Source sequence IEnumerable<TSource>
Index of element to return int
ElementAt picks the nth element from the sequence:
int[] numbers = { 1, 2, 3, 4, 5 };
int third = numbers.ElementAt (2); // 3
int tenthError = numbers.ElementAt (9); // Exception
int tenth = numbers.ElementAtOrDefault (9); // 0
Enumerable.ElementAt is written such that if the input sequence happens to imple‐
ment IList<T>, it calls IList<T>’s indexer. Otherwise, it enumerates n times and
then returns the next element. ElementAt is not supported in EF Core.
MinBy and MaxBy
MinBy and MaxBy (introduced in .NET 6) return the element with the smallest or
largest value, as determined by a keySelector:
string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" };
Console.WriteLine (names.MaxBy (n => n.Length)); // Harry
In contrast, Min and Max (which we will cover in the following section) return the
smallest or largest value itself:
Console.WriteLine (names.Max (n => n.Length)); // 5
If two or more elements share a minimum/maximum value, MinBy/MaxBy returns
the first:
Console.WriteLine (names.MinBy (n => n.Length)); // Tom
Element Operators | 489
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-507-320.jpg)
![If the input sequence is empty, MinBy and MaxBy return null if the element type is
nullable (or throw an exception if the element type is not nullable).
DefaultIfEmpty
DefaultIfEmpty returns a sequence containing a single element whose value is
default(TSource) if the input sequence has no elements; otherwise, it returns the
input sequence unchanged. You use this in writing flat outer joins: see “Outer joins
with SelectMany” on page 467 and “Flat outer joins” on page 473.
Aggregation Methods
IEnumerable<TSource>→scalar
Method Description SQL equivalents
Count,
LongCount
Returns the number of elements in the input sequence,
optionally satisfying a predicate
COUNT (...)
Min, Max Returns the smallest or largest element in the sequence MIN (...),
MAX (...)
Sum, Average Calculates a numeric sum or average over elements in the
sequence
SUM (...),
AVG (...)
Aggregate Performs a custom aggregation Exception thrown
Count and LongCount
Argument Type
Source sequence IEnumerable<TSource>
Predicate (optional) TSource => bool
Count simply enumerates over a sequence, returning the number of items:
int fullCount = new int[] { 5, 6, 7 }.Count(); // 3
The internal implementation of Enumerable.Count tests the input sequence to see
whether it happens to implement ICollection<T>. If it does, it simply calls ICollec
tion<T>.Count; otherwise, it enumerates over every item, incrementing a counter.
You can optionally supply a predicate:
int digitCount = "pa55w0rd".Count (c => char.IsDigit (c)); // 3
LongCount does the same job as Count but returns a 64-bit integer, allowing for
sequences of greater than two billion elements.
490 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-508-320.jpg)
![Min and Max
Argument Type
Source sequence IEnumerable<TSource>
Result selector (optional) TSource => TResult
Min and Max return the smallest or largest element from a sequence:
int[] numbers = { 28, 32, 14 };
int smallest = numbers.Min(); // 14;
int largest = numbers.Max(); // 32;
If you include a selector expression, each element is first projected:
int smallest = numbers.Max (n => n % 10); // 8;
A selector expression is mandatory if the items themselves are not intrinsically
comparable—in other words, if they do not implement IComparable<T>:
Purchase runtimeError = dbContext.Purchases.Min (); // Error
decimal? lowestPrice = dbContext.Purchases.Min (p => p.Price); // OK
A selector expression determines not only how elements are compared, but also
the final result. In the preceding example, the final result is a decimal value, not a
purchase object. To get the cheapest purchase, you need a subquery:
Purchase cheapest = dbContext.Purchases
.Where (p => p.Price == dbContext.Purchases.Min (p2 => p2.Price))
.FirstOrDefault();
In this case, you could also formulate the query without an aggregation by using an
OrderBy followed by FirstOrDefault.
Sum and Average
Argument Type
Source sequence IEnumerable<TSource>
Result selector (optional) TSource => TResult
Sum and Average are aggregation operators that are used in a similar manner to Min
and Max:
decimal[] numbers = { 3, 4, 8 };
decimal sumTotal = numbers.Sum(); // 15
decimal average = numbers.Average(); // 5 (mean value)
The following returns the total length of each of the strings in the names array:
int combinedLength = names.Sum (s => s.Length); // 19
Aggregation Methods | 491
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-509-320.jpg)
![Sum and Average are fairly restrictive in their typing. Their definitions are hard‐
wired to each of the numeric types (int, long, float, double, decimal, and their
nullable versions). In contrast, Min and Max can operate directly on anything that
implements IComparable<T>—such as a string, for instance.
Further, Average always returns either decimal, float, or double, according to the
following table:
Selector type Result type
decimal decimal
float float
int, long, double double
This means that the following does not compile (“cannot convert double to int”):
int avg = new int[] { 3, 4 }.Average();
But this will compile:
double avg = new int[] { 3, 4 }.Average(); // 3.5
Average implicitly upscales the input values to prevent loss of precision. In this
example, we averaged integers and got 3.5 without needing to resort to an input
element cast:
double avg = numbers.Average (n => (double) n);
When querying a database, Sum and Average translate to the standard SQL aggrega‐
tions. The following query returns customers whose average purchase was more
than $500:
from c in dbContext.Customers
where c.Purchases.Average (p => p.Price) > 500
select c.Name;
Aggregate
Aggregate allows you to specify a custom accumulation algorithm for implement‐
ing unusual aggregations. Aggregate is not supported in EF Core and is somewhat
specialized in its use cases. The following demonstrates how Aggregate can do the
work of Sum:
int[] numbers = { 1, 2, 3 };
int sum = numbers.Aggregate (0, (total, n) => total + n); // 6
The first argument to Aggregate is the seed, from which accumulation starts. The
second argument is an expression to update the accumulated value, given a fresh
element. You can optionally supply a third argument to project the final result value
from the accumulated value.
492 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-510-320.jpg)
![Most problems for which Aggregate has been designed can be
solved as easily with a foreach loop—and with more familiar
syntax. The advantage of using Aggregate is that with large
or complex aggregations, you can automatically parallelize the
operation with PLINQ (see Chapter 22).
Unseeded aggregations
You can omit the seed value when calling Aggregate, in which case the first element
becomes the implicit seed, and aggregation proceeds from the second element.
Here’s the preceding example, unseeded:
int[] numbers = { 1, 2, 3 };
int sum = numbers.Aggregate ((total, n) => total + n); // 6
This gives the same result as before, but we’re actually doing a different calculation.
Before, we were calculating 0 + 1 + 2 + 3; now we’re calculating 1 + 2 + 3. We can
better illustrate the difference by multiplying instead of adding:
int[] numbers = { 1, 2, 3 };
int x = numbers.Aggregate (0, (prod, n) => prod * n); // 0*1*2*3 = 0
int y = numbers.Aggregate ( (prod, n) => prod * n); // 1*2*3 = 6
As you’ll see in Chapter 22, unseeded aggregations have the advantage of being
parallelizable without requiring the use of special overloads. However, there are
some traps with unseeded aggregations.
Traps with unseeded aggregations
The unseeded aggregation methods are intended for use with delegates that are
commutative and associative. If used otherwise, the result is either unintuitive (with
ordinary queries) or nondeterministic (in the case that you parallelize the query with
PLINQ). For example, consider the following function:
(total, n) => total + n * n
This is neither commutative nor associative. (For example, 1 + 2 * 2 != 2 + 1 * 1.)
Let’s see what happens when we use it to sum the square of the numbers 2, 3, and 4:
int[] numbers = { 2, 3, 4 };
int sum = numbers.Aggregate ((total, n) => total + n * n); // 27
Instead of calculating
2*2 + 3*3 + 4*4 // 29
it calculates:
2 + 3*3 + 4*4 // 27
We can fix this in a number of ways. First, we could include 0 as the first element:
int[] numbers = { 0, 2, 3, 4 };
Aggregation Methods | 493
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-511-320.jpg)
![Quantifiers
IEnumerable<TSource>→bool
Method Description SQL equivalents
Contains Returns true if the input sequence contains the given element WHERE ... IN (...)
Any Returns true if any elements satisfy the given predicate WHERE ... IN (...)
All Returns true if all elements satisfy the given predicate WHERE (...)
SequenceEqual Returns true if the second sequence has identical elements to
the input sequence
Contains and Any
The Contains method accepts an argument of type TSource; Any accepts an
optional predicate.
Contains returns true if the given element is present:
bool hasAThree = new int[] { 2, 3, 4 }.Contains (3); // true;
Any returns true if the given expression is true for at least one element. We can
rewrite the preceding query with Any as follows:
bool hasAThree = new int[] { 2, 3, 4 }.Any (n => n == 3); // true;
Any can do everything that Contains can do, and more:
bool hasABigNumber = new int[] { 2, 3, 4 }.Any (n => n > 10); // false;
Calling Any without a predicate returns true if the sequence has one or more
elements. Here’s another way to write the preceding query:
bool hasABigNumber = new int[] { 2, 3, 4 }.Where (n => n > 10).Any();
Any is particularly useful in subqueries and is used often when querying databases;
for example:
from c in dbContext.Customers
where c.Purchases.Any (p => p.Price > 1000)
select c
All and SequenceEqual
All returns true if all elements satisfy a predicate. The following returns customers
whose purchases are less than $100:
dbContext.Customers.Where (c => c.Purchases.All (p => p.Price < 100));
SequenceEqual compares two sequences. To return true, each sequence must have
identical elements, in the identical order. You can optionally provide an equality
comparer; the default is EqualityComparer<T>.Default.
Quantifiers | 495
LINQ
Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-513-320.jpg)
![Generation Methods
void→IEnumerable<TResult>
Method Description
Empty Creates an empty sequence
Repeat Creates a sequence of repeating elements
Range Creates a sequence of integers
Empty, Repeat, and Range are static (nonextension) methods that manufacture
simple local sequences.
Empty
Empty manufactures an empty sequence and requires just a type argument:
foreach (string s in Enumerable.Empty<string>())
Console.Write (s); // <nothing>
In conjunction with the ?? operator, Empty does the reverse of DefaultIfEmpty. For
example, suppose that we have a jagged array of integers and we want to get all the
integers into a single flat list. The following SelectMany query fails if any of the
inner arrays is null:
int[][] numbers =
{
new int[] { 1, 2, 3 },
new int[] { 4, 5, 6 },
null // this null makes the query below fail.
};
IEnumerable<int> flat = numbers.SelectMany (innerArray => innerArray);
Empty in conjunction with ?? fixes the problem:
IEnumerable<int> flat = numbers
.SelectMany (innerArray => innerArray ?? Enumerable.Empty <int>());
foreach (int i in flat)
Console.Write (i + " "); // 1 2 3 4 5 6
Range and Repeat
Range accepts a starting index and count (both integers):
foreach (int i in Enumerable.Range (5, 3))
Console.Write (i + " "); // 5 6 7
Repeat accepts an element to repeat, and the number of repetitions:
foreach (bool x in Enumerable.Repeat (true, 3))
Console.Write (x + " "); // True True True
496 | Chapter 9: LINQ Operators](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-514-320.jpg)
![1 The X-DOM actually optimizes this step internally by storing simple text content in a string. The
XTEXT node is not actually created until you call Nodes( ) on the XContainer.
This has two benefits. First, the code resembles the shape of the XML. Second, it can
be incorporated into the select clause of a LINQ query. For example, the following
query projects from an EF Core entity class into an X-DOM:
XElement query =
new XElement ("customers",
from c in dbContext.Customers.AsEnumerable()
select
new XElement ("customer", new XAttribute ("id", c.ID),
new XElement ("firstname", c.FirstName),
new XElement ("lastname", c.LastName,
new XComment ("nice name")
)
)
);
We examine this further later in this chapter in “Projecting into an X-DOM” on
page 525.
Specifying Content
Functional construction is possible because the constructors for XElement (and
XDocument) are overloaded to accept a params object array:
public XElement (XName name, params object[] content)
The same holds true for the Add method in XContainer:
public void Add (params object[] content)
Hence, you can specify any number of child objects of any type when building or
appending an X-DOM. This works because anything counts as legal content. To see
how, we need to examine how each content object is processed internally. Here are
the decisions made by XContainer, in order:
1. If the object is null, it’s ignored.
1.
2. If the object is based on XNode or XStreamingElement, it’s added as is to the
2.
Nodes collection.
3. If the object is an XAttribute, it’s added to the Attributes collection.
3.
4. If the object is a string, it’s wrapped in an XText node and added to Nodes.1
4.
5. If the object implements IEnumerable, it’s enumerated, and the same rules are
5.
applied to each element.
6. Otherwise, the object is converted to a string, wrapped in an XText node, and
6.
then added to Nodes.2
Instantiating an X-DOM | 503
LINQ
to
XML](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-521-320.jpg)
![Updating an X-DOM
You can update elements and attributes in the following ways:
• Call SetValue or reassign the Value property.
•
• Call SetElementValue or SetAttributeValue.
•
• Call one of the RemoveXXX methods.
•
• Call one of the AddXXX or ReplaceXXX methods, specifying fresh content.
•
You can also reassign the Name property on XElement objects.
Simple Value Updates
Members Works on
SetValue (object value) XElement, XAttribute
Value { get; set } XElement, XAttribute
The SetValue method replaces an element or attribute’s content with a simple value.
Setting the Value property does the same but accepts string data only. We describe
both of these functions in detail later in “Working with Values” on page 513. An
effect of calling SetValue (or reassigning Value) is that it replaces all child nodes:
XElement settings = new XElement ("settings",
new XElement ("timeout", 30)
);
settings.SetValue ("blah");
Console.WriteLine (settings.ToString()); // <settings>blah</settings>
Updating Child Nodes and Attributes
Category Members Works on
Add Add (params object[] content) XContainer
AddFirst (params object[] content) XContainer
Remove RemoveNodes() XContainer
RemoveAttributes() XElement
RemoveAll() XElement
Update ReplaceNodes (params object[] content) XContainer
ReplaceAttributes (params object[] content) XElement
ReplaceAll (params object[] content XElement
SetElementValue (XName name, object value) XElement
SetAttributeValue (XName name, object value) XElement
510 | Chapter 10: LINQ to XML](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-528-320.jpg)
![The most convenient methods in this group are the last two: SetElementValue
and SetAttributeValue. They serve as shortcuts for instantiating an XElement
or XAttribute and then Adding it to a parent, replacing any existing element or
attribute of that name:
XElement settings = new XElement ("settings");
settings.SetElementValue ("timeout", 30); // Adds child node
settings.SetElementValue ("timeout", 60); // Update it to 60
Add appends a child node to an element or document. AddFirst does the same thing
but inserts at the beginning of the collection rather than the end.
You can remove all child nodes or attributes in one hit with RemoveNodes or
RemoveAttributes. RemoveAll is equivalent to calling both methods.
The ReplaceXXX methods are equivalent to Removing and then Adding. They take a
snapshot of the input, so e.ReplaceNodes(e.Nodes()) works as expected.
Updating Through the Parent
Members Works on
AddBeforeSelf (params object[] content) XNode
AddAfterSelf (params object[] content) XNode
Remove() XNode, XAttribute
ReplaceWith (params object[] content) XNode
The methods AddBeforeSelf, AddAfterSelf, Remove, and ReplaceWith don’t oper‐
ate on the node’s children. Instead, they operate on the collection in which the
node itself is in. This requires that the node have a parent element—otherwise, an
exception is thrown. AddBeforeSelf and AddAfterSelf are useful for inserting a
node into an arbitrary position:
XElement items = new XElement ("items",
new XElement ("one"),
new XElement ("three")
);
items.FirstNode.AddAfterSelf (new XElement ("two"));
Here’s the result:
<items><one /><two /><three /></items>
Inserting into an arbitrary position within a long sequence of elements is efficient
because nodes are stored internally in a linked list.
The Remove method removes the current node from its parent. ReplaceWith does
the same and then inserts some other content at the same position:
XElement items = XElement.Parse ("<items><one/><two/><three/></items>");
items.FirstNode.ReplaceWith (new XComment ("One was here"));
Updating an X-DOM | 511
LINQ
to
XML](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-529-320.jpg)
![• A single XDocumentType object (to reference a document type definition
•
[DTD])
• Any number of XProcessingInstruction objects
•
• Any number of XComment objects
•
Of these, only the root XElement is mandatory in order to
have a valid XDocument. The XDeclaration is optional—if
omitted, default settings are applied during serialization.
The simplest valid XDocument has just a root element:
var doc = new XDocument (
new XElement ("test", "data")
);
Notice that we didn’t include an XDeclaration object. The file generated by calling
doc.Save would still contain an XML declaration, however, because one is gener‐
ated by default.
The next example produces a simple but correct XHTML file, illustrating all the
constructs that an XDocument can accept:
var styleInstruction = new XProcessingInstruction (
"xml-stylesheet", "href='styles.css' type='text/css'");
var docType = new XDocumentType ("html",
"-//W3C//DTD XHTML 1.0 Strict//EN",
"http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd", null);
XNamespace ns = "http://www.w3.org/1999/xhtml";
var root =
new XElement (ns + "html",
new XElement (ns + "head",
new XElement (ns + "title", "An XHTML page")),
new XElement (ns + "body",
new XElement (ns + "p", "This is the content"))
);
var doc =
new XDocument (
new XDeclaration ("1.0", "utf-8", "no"),
new XComment ("Reference a stylesheet"),
styleInstruction,
docType,
root);
doc.Save ("test.html");
The resultant test.html reads as follows:
<?xml version="1.0" encoding="utf-8" standalone="no"?>
<!--Reference a stylesheet-->
<?xml-stylesheet href='styles.css' type='text/css'?>
516 | Chapter 10: LINQ to XML](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-534-320.jpg)
![NextSibling moves the cursor to the start of the first sibling node with the specified
name/namespace.
There are also two legacy methods: ReadString and ReadElementString behave
like ReadContentAsString and ReadElementContentAsString, except that they
throw an exception if there’s more than a single text node within the element. You
should avoid these methods because they throw an exception if an element contains
a comment.
Reading Attributes
XmlReader provides an indexer giving you direct (random) access to an ele‐
ment’s attributes—by name or position. Using the indexer is equivalent to calling
GetAttribute.
Given the XML fragment
<customer id="123" status="archived"/>
we could read its attributes, as follows:
Console.WriteLine (reader ["id"]); // 123
Console.WriteLine (reader ["status"]); // archived
Console.WriteLine (reader ["bogus"] == null); // True
The XmlReader must be positioned on a start element in
order to read attributes. After calling ReadStartElement, the
attributes are gone forever!
Although attribute order is semantically irrelevant, you can access attributes by
their ordinal position. We could rewrite the preceding example as follows:
Console.WriteLine (reader [0]); // 123
Console.WriteLine (reader [1]); // archived
The indexer also lets you specify the attribute’s namespace—if it has one.
AttributeCount returns the number of attributes for the current node.
Attribute nodes
To explicitly traverse attribute nodes, you must make a special diversion from the
normal path of just calling Read. A good reason to do so is if you want to parse
attribute values into other types, via the ReadContentAsXXX methods.
The diversion must begin from a start element. To make the job easier, the forward-
only rule is relaxed during attribute traversal: you can jump to any attribute (for‐
ward or backward) by calling MoveToAttribute.
MoveToElement returns you to the start element from any‐
place within the attribute node diversion.
XmlReader | 535
Other
XML
and
JSON
Technologies](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-553-320.jpg)
![Consider the following JSON file named people.json:
{
"FirstName":"Sara",
"LastName":"Wells",
"Age":35,
"Friends":["Dylan","Ian"]
}
The curly braces indicate a JSON object (which contains properties such as "First
Name" and "LastName"), whereas the square brackets indicate a JSON array (which
contains repeating elements). In this case, the repeating elements are strings, but
they could be objects (or other arrays).
The following code parses the file by enumerating its JSON tokens. A token is the
beginning or end of an object, the beginning or end of an array, the name of a
property, or an array or property value (string, number, true, false, or null):
byte[] data = File.ReadAllBytes ("people.json");
Utf8JsonReader reader = new Utf8JsonReader (data);
while (reader.Read())
{
switch (reader.TokenType)
{
case JsonTokenType.StartObject:
Console.WriteLine ($"Start of object");
break;
case JsonTokenType.EndObject:
Console.WriteLine ($"End of object");
break;
case JsonTokenType.StartArray:
Console.WriteLine();
Console.WriteLine ($"Start of array");
break;
case JsonTokenType.EndArray:
Console.WriteLine ($"End of array");
break;
case JsonTokenType.PropertyName:
Console.Write ($"Property: {reader.GetString()}");
break;
case JsonTokenType.String:
Console.WriteLine ($" Value: {reader.GetString()}");
break;
case JsonTokenType.Number:
Console.WriteLine ($" Value: {reader.GetInt32()}");
break;
default:
Console.WriteLine ($"No support for {reader.TokenType}");
break;
}
}
Working with JSON | 545
Other
XML
and
JSON
Technologies](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-563-320.jpg)
![Here’s the output:
Start of object
Property: FirstName Value: Sara
Property: LastName Value: Wells
Property: Age Value: 35
Property: Friends
Start of array
Value: Dylan
Value: Ian
End of array
End of object
Because Utf8JsonReader works directly with UTF-8, it steps through the tokens
without first having to convert the input into UTF-16 (the format of .NET strings).
Conversion to UTF-16 takes place only when you call a method such as Get
String().
Interestingly, Utf8JsonReader’s constructor does not accept a byte array, but rather
a ReadOnlySpan<byte> (for this reason, Utf8JsonReader is defined as a ref struct).
You can pass in a byte array because there’s an implicit conversion from T[] to
ReadOnlySpan<T>. In Chapter 23, we describe how spans work and how you can use
them to improve performance by minimizing memory allocations.
JsonReaderOptions
By default, Utf8JsonReader requires that the JSON conform strictly to the JSON
RFC 8259 standard. You can instruct the reader to be more tolerant by passing an
instance of JsonReaderOptions to the Utf8JsonReader constructor. The options
allow the following:
C-Style comments
By default, comments in JSON cause a JsonException to be thrown. Setting
the CommentHandling property to JsonCommentHandling.Skip causes com‐
ments to be ignored, whereas JsonCommentHandling.Allow causes the reader
to recognize them and emit JsonTokenType.Comment tokens when they are
encountered. Comments cannot appear in the middle of other tokens.
Trailing commas
Per the standard, the last property of an object and the last element of an array
must not have a trailing comma. Setting the AllowTrailingCommas property to
e relaxes this restriction.
Control over the maximum nesting depth
By default, objects and arrays can nest to 64 levels. Setting the MaxDepth to a
different number overrides this setting.
546 | Chapter 11: Other XML and JSON Technologies](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-564-320.jpg)
![using JsonDocument document = JsonDocument.Parse (@"[1, 2, 3, 4, 5]");
int length = document.RootElement.GetArrayLength(); // 5
int value = document.RootElement[3].GetInt32(); // 4
Reading JSON objects
If the element represents a JSON object, you can call the following methods:
EnumerateObject()
Enumerates all of the object’s property names and values.
GetProperty (string propertyName)
Gets a property by name (returning another JsonElement). Throws an excep‐
tion if the name isn’t present.
TryGetProperty (string propertyName, out JsonElement value)
Returns an object’s property if present.
For example:
using JsonDocument document = JsonDocument.Parse (@"{ ""Age"": 32}");
JsonElement root = document.RootElement;
int age = root.GetProperty ("Age").GetInt32();
Here’s how we could “discover” the Age property:
JsonProperty ageProp = root.EnumerateObject().First();
string name = ageProp.Name; // Age
JsonElement value = ageProp.Value;
Console.WriteLine (value.ValueKind); // Number
Console.WriteLine (value.GetInt32()); // 32
JsonDocument and LINQ
JsonDocument lends itself well to LINQ. Given the following JSON file:
[
{
"FirstName":"Sara",
"LastName":"Wells",
"Age":35,
"Friends":["Ian"]
},
{
"FirstName":"Ian",
"LastName":"Weems",
"Age":42,
"Friends":["Joe","Eric","Li"]
},
{
"FirstName":"Dylan",
"LastName":"Lockwood",
"Age":46,
"Friends":["Sara","Ian"]
550 | Chapter 11: Other XML and JSON Technologies](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-568-320.jpg)
![}
]
we can use JsonDocument to query this with LINQ, as follows:
using var stream = File.OpenRead (jsonPath);
using JsonDocument document = JsonDocument.Parse (json);
var query =
from person in document.RootElement.EnumerateArray()
select new
{
FirstName = person.GetProperty ("FirstName").GetString(),
Age = person.GetProperty ("Age").GetInt32(),
Friends =
from friend in person.GetProperty ("Friends").EnumerateArray()
select friend.GetString()
};
Because LINQ queries are lazily evaluated, it’s important to enumerate the query
before the document goes out of scope and JsonDocument is implicitly disposed of
by virtue of the using statement.
Making updates with a JSON writer
Although JsonDocument is read-only, you can send the content of a JsonElement
to a Utf8JsonWriter with the WriteTo method. This provides a mechanism for
emitting a modified version of the JSON. Here’s how we can take the JSON from the
preceding example and write it to a new JSON file that includes only people with
two or more friends:
using var json = File.OpenRead (jsonPath);
using JsonDocument document = JsonDocument.Parse (json);
var options = new JsonWriterOptions { Indented = true };
using (var outputStream = File.Create ("NewFile.json"))
using (var writer = new Utf8JsonWriter (outputStream, options))
{
writer.WriteStartArray();
foreach (var person in document.RootElement.EnumerateArray())
{
int friendCount = person.GetProperty ("Friends").GetArrayLength();
if (friendCount >= 2)
person.WriteTo (writer);
}
}
If you need the ability to update the DOM, however, JsonNode is a better solution.
JsonNode
JsonNode (in System.Text.Json.Nodes) was introduced in .NET 6, primarily to
satisfy the demand for a writable DOM. However, it’s also suitable in read-only
Working with JSON | 551
Other
XML
and
JSON
Technologies](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-569-320.jpg)
![Nodes that have been parsed from JSON text are internally
backed by a JsonElement (part of the JsonDocument read-only
JSON API). You can extract the underlying JsonElement as
follows:
JsonElement je = node.GetValue<JsonElement>();
However, this doesn’t work when the node is instantiated
explicitly (as will be the case when we update the DOM).
Such nodes are backed not by a JsonElement but by the actual
parsed value (see “Making updates with JsonNode” on page
554).
Reading JSON arrays
A JsonNode that represents a JSON array will be of type JsonArray.
JsonArray implements IList<JsonNode>, so you can enumerate over it and access
the elements like you would an array or list:
var node = JsonNode.Parse (@"[1, 2, 3, 4, 5]");
Console.WriteLine (node.AsArray().Count); // 5
foreach (JsonNode child in node.AsArray())
{ ... }
As a shortcut, you can access the indexer directly from the JsonNode class:
Console.WriteLine ((int)node[0]); // 1
Reading JSON objects
A JsonNode that represents a JSON object will be of type JsonObject.
JsonObject implements IDictionary<string,JsonNode>, so you can access a
member via the indexer, as well as enumerating over the dictionary’s key/value
pairs.
And as with JsonArray, you can access the indexer directly from the JsonNode class:
var node = JsonNode.Parse (@"{ ""Name"":""Alice"", ""Age"": 32}");
string name = (string) node ["Name"]; // Alice
int age = (int) node ["Age"]; // 32
Here’s how we could “discover” the Name and Age properties:
// Enumerate over the dictionary’s key/value pairs:
foreach (KeyValuePair<string,JsonNode> keyValuePair in node.AsObject())
{
string propertyName = keyValuePair.Key; // "Name" (then "Age")
JsonNode value = keyValuePair.Value;
}
If you’re not sure whether a property has been defined, the following pattern also
works:
Working with JSON | 553
Other
XML
and
JSON
Technologies](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-571-320.jpg)
![if (node.AsObject().TryGetPropertyValue ("Name", out JsonNode nameNode))
{ ... }
Fluent traversal and LINQ
You can reach deep into a hierarchy just with indexers. For example, given the
following JSON file:
[
{
"FirstName":"Sara",
"LastName":"Wells",
"Age":35,
"Friends":["Ian"]
},
{
"FirstName":"Ian",
"LastName":"Weems",
"Age":42,
"Friends":["Joe","Eric","Li"]
},
{
"FirstName":"Dylan",
"LastName":"Lockwood",
"Age":46,
"Friends":["Sara","Ian"]
}
]
we can extract the second person’s third friend as follows:
string li = (string) node[1]["Friends"][2];
Such a file is also easy to query via LINQ:
JsonNode node = JsonNode.Parse (File.ReadAllText (jsonPath));
var query =
from person in node.AsArray()
select new
{
FirstName = (string) person ["FirstName"],
Age = (int) person ["Age"],
Friends =
from friend in person ["Friends"].AsArray()
select (string) friend
};
Unlike JsonDocument, JsonNode is not disposable, so we don’t have to worry about
the potential for disposal during lazy enumeration.
Making updates with JsonNode
JsonObject and JsonArray are mutable, so you can update their content.
554 | Chapter 11: Other XML and JSON Technologies](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-572-320.jpg)
![The easiest way to replace or add properties to a JsonObject is via the indexer. In
the following example, we change the Color property’s value from “Red” to “White”
and add a new property called “Valid”:
var node = JsonNode.Parse ("{ "Color": "Red" }");
node ["Color"] = "White";
node ["Valid"] = true;
Console.WriteLine (node.ToJsonString()); // {"Color":"White","Valid":true}
The second line in that example is a shortcut for the following:
node ["Color"] = JsonValue.Create ("White");
Rather than assigning the property a simple value, you can assign it a JsonArray
or JsonObject. (We will demonstrate how to construct JsonArray and JsonObject
instances in the following section.)
To remove a property, first cast to JsonObject (or call AsObject) and then call the
Remove method:
node.AsObject().Remove ("Valid");
(JsonObject also exposes an Add method, which throws an exception if the property
already exists.)
JsonArray also lets you use the indexer to replace items:
var node = JsonNode.Parse ("[1, 2, 3]");
node[0] = 10;
Calling AsArray exposes the Add/Insert/Remove/RemoveAt methods. In the follow‐
ing example, we remove the first element in the array and add one to the end:
var arrayNode = JsonNode.Parse ("[1, 2, 3]");
arrayNode.AsArray().RemoveAt(0);
arrayNode.AsArray().Add (4);
Console.WriteLine (arrayNode.ToJsonString()); // [2,3,4]
Constructing a JsonNode DOM programmatically
JsonArray and JsonObject have constructors that support object initialization
syntax, which allows you to build an entire JsonNode DOM in one expression:
var node = new JsonArray
{
new JsonObject {
["Name"] = "Tracy",
["Age"] = 30,
["Friends"] = new JsonArray ("Lisa", "Joe")
},
new JsonObject {
["Name"] = "Jordyn",
["Age"] = 25,
["Friends"] = new JsonArray ("Tracy", "Li")
}
};
Working with JSON | 555
Other
XML
and
JSON
Technologies](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-573-320.jpg)
![This evaluates to the following JSON:
[
{
"Name": "Tracy",
"Age": 30,
"Friends": ["Lisa", "Joe"]
},
{
"Name": "Jordyn",
"Age": 25,
"Friends": ["Tracy","Li"]
}
]
556 | Chapter 11: Other XML and JSON Technologies](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-574-320.jpg)
![we can reduce our SuspendEvents method to the following:
public IDisposable SuspendEvents()
{
_suspendCount++;
return Disposable.Create (() => _suspendCount--);
}
Automatic Garbage Collection
Regardless of whether an object requires a Dispose method for custom tear-down
logic, at some point the memory it occupies on the heap must be freed. The
CLR handles this side of it entirely automatically via an automatic GC. You never
deallocate managed memory yourself. For example, consider the following method:
public void Test()
{
byte[] myArray = new byte[1000];
...
}
When Test executes, an array to hold 1,000 bytes is allocated on the memory heap.
The array is referenced by the variable myArray, stored on the local variable stack.
When the method exits, this local variable myArray pops out of scope, meaning that
nothing is left to reference the array on the memory heap. The orphaned array then
becomes eligible to be reclaimed in garbage collection.
In debug mode with optimizations disabled, the lifetime of an
object referenced by a local variable extends to the end of the
code block to ease debugging. Otherwise, it becomes eligible
for collection at the earliest point at which it’s no longer used.
Garbage collection does not happen immediately after an object is orphaned. Rather
like garbage collection on the street, it happens periodically, although (unlike
garbage collection on the street) not to a fixed schedule. The CLR bases its deci‐
sion on when to collect upon a number of factors, such as the available memory,
the amount of memory allocation, and the time since the last collection (the GC
self-tunes to optimize for an application’s specific memory access patterns). This
means that there’s an indeterminate delay between an object being orphaned and
being released from memory. This delay can range from nanoseconds to days.
The GC doesn’t collect all garbage with every collection.
Instead, the memory manager divides objects into genera‐
tions, and the GC collects new generations (recently alloca‐
ted objects) more frequently than old generations (long-lived
objects). We discuss this in more detail in “How the GC
Works” on page 569.
Automatic Garbage Collection | 563
Disposal
&
Garbage
Collection](https://image.slidesharecdn.com/c10inanutshellthedefinitivereference-oreillymedia2022-241015222114-477944cc/85/C-10-in-a-Nutshell_-The-Definitive-Reference-O-Reilly-Media-2022-pdf-581-320.jpg)
C# 10 in a Nutshell_ The Definitive Reference-O'Reilly Media (2022).pdf
- 1. Joseph Albahari C# 10 in a Nutshell The Definitive Reference
- 3. C# 10 IN A NUTSHELL THE DEFINITIVE REFERENCE Joseph Albahari
- 4. 978-1-098-12195-2 [LSI] C# 10 in a Nutshell by Joseph Albahari Copyright © 2022 Joseph Albahari. All rights reserved. Printed in the United States of America. Published by O’Reilly Media, Inc., 1005 Gravenstein Highway North, Sebastopol, CA 95472. O’Reilly books may be purchased for educational, business, or sales promotional use. Online editions are also available for most titles (http://oreilly.com). For more information, contact our corporate/institutional sales department: 800-998-9938 or corporate@oreilly.com. Acquisitions Editor: Amanda Quinn Development Editor: Corbin Collins Production Editor: Kristen Brown Copyeditor: Charles Roumeliotis Proofreader: Piper Editorial Consulting, LLC Indexer: WordCo Indexing Services, Inc. Interior Designer: David Futato Cover Designer: Karen Montgomery Illustrator: Kate Dullea February 2022: First Edition Revision History for the First Edition 2022-02-08: First Release See http://oreilly.com/catalog/errata.csp?isbn=9781098121952 for release details. The O’Reilly logo is a registered trademark of O’Reilly Media, Inc. C# 10 in a Nutshell, the cover image, and related trade dress are trademarks of O’Reilly Media, Inc. The views expressed in this work are those of the author, and do not represent the publish‐ er’s views. While the publisher and the author have used good faith efforts to ensure that the information and instructions contained in this work are accurate, the publisher and the author disclaim all responsibility for errors or omissions, including without limitation responsibility for damages resulting from the use of or reliance on this work. Use of the information and instructions contained in this work is at your own risk. If any code samples or other technology this work contains or describes is subject to open source licenses or the intellectual property rights of others, it is your responsibility to ensure that your use thereof complies with such licenses and/or rights.
- 5. Table of Contents Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi 1. Introducing C# and .NET. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Object Orientation 1 Type Safety 2 Memory Management 3 Platform Support 3 CLRs, BCLs, and Runtimes 3 A Brief History of C# 7 2. C# Language Basics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 A First C# Program 25 Syntax 28 Type Basics 30 Numeric Types 41 Boolean Type and Operators 49 Strings and Characters 51 Arrays 54 Variables and Parameters 59 Expressions and Operators 70 Null Operators 74 Statements 76 Namespaces 87 3. Creating Types in C#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 Classes 95 Inheritance 115 The object Type 125 Structs 129 iii
- 6. Access Modifiers 132 Interfaces 134 Enums 140 Nested Types 144 Generics 145 4. Advanced C#. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159 Delegates 159 Events 167 Lambda Expressions 174 Anonymous Methods 180 try Statements and Exceptions 181 Enumeration and Iterators 189 Nullable Value Types 194 Nullable Reference Types 200 Extension Methods 202 Anonymous Types 205 Tuples 207 Records 211 Patterns 222 Attributes 227 Caller Info Attributes 229 Dynamic Binding 232 Operator Overloading 239 Unsafe Code and Pointers 243 Preprocessor Directives 249 XML Documentation 252 5. .NET Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257 .NET Standard 259 Runtime and C# Language Versions 261 Reference Assemblies 261 The CLR and BCL 262 Application Layers 266 6. .NET Fundamentals. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271 String and Text Handling 271 Dates and Times 284 DateOnly and TimeOnly 292 DateTime and Time Zones 292 iv | Table of Contents
- 7. Formatting and Parsing 297 Standard Format Strings and Parsing Flags 303 Other Conversion Mechanisms 310 Globalization 314 Working with Numbers 315 BitOperations 320 Enums 320 The Guid Struct 323 Equality Comparison 324 Order Comparison 335 Utility Classes 338 7. Collections. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343 Enumeration 343 The ICollection and IList Interfaces 351 The Array Class 355 Lists, Queues, Stacks, and Sets 363 Dictionaries 372 Customizable Collections and Proxies 378 Immutable Collections 384 Plugging in Equality and Order 388 8. LINQ Queries. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395 Getting Started 395 Fluent Syntax 397 Query Expressions 403 Deferred Execution 408 Subqueries 414 Composition Strategies 418 Projection Strategies 421 Interpreted Queries 424 EF Core 430 Building Query Expressions 442 9. LINQ Operators. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447 Overview 448 Filtering 451 Projecting 456 Joining 468 Ordering 476 Table of Contents | v
- 8. Grouping 479 Set Operators 483 Conversion Methods 485 Element Operators 488 Aggregation Methods 490 Quantifiers 495 Generation Methods 496 10. LINQ to XML. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 497 Architectural Overview 497 X-DOM Overview 498 Instantiating an X-DOM 502 Navigating and Querying 504 Updating an X-DOM 510 Working with Values 513 Documents and Declarations 515 Names and Namespaces 519 Annotations 524 Projecting into an X-DOM 525 11. Other XML and JSON Technologies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 529 XmlReader 529 XmlWriter 537 Patterns for Using XmlReader/XmlWriter 539 Working with JSON 544 12. Disposal and Garbage Collection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 557 IDisposable, Dispose, and Close 557 Automatic Garbage Collection 563 Finalizers 565 How the GC Works 569 Managed Memory Leaks 576 Weak References 579 13. Diagnostics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585 Conditional Compilation 585 Debug and Trace Classes 589 Debugger Integration 592 Processes and Process Threads 593 StackTrace and StackFrame 594 vi | Table of Contents
- 9. Windows Event Logs 596 Performance Counters 598 The Stopwatch Class 602 Cross-Platform Diagnostic Tools 603 14. Concurrency and Asynchrony. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 609 Introduction 609 Threading 610 Tasks 626 Principles of Asynchrony 634 Asynchronous Functions in C# 639 Asynchronous Patterns 659 Obsolete Patterns 667 15. Streams and I/O. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 671 Stream Architecture 671 Using Streams 673 Stream Adapters 687 Compression Streams 695 Working with ZIP Files 698 File and Directory Operations 699 OS Security 710 Memory-Mapped Files 712 16. Networking. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 717 Network Architecture 717 Addresses and Ports 719 URIs 720 HttpClient 722 Writing an HTTP Server 731 Using DNS 734 Sending Mail with SmtpClient 734 Using TCP 735 Receiving POP3 Mail with TCP 739 17. Assemblies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 741 What’s in an Assembly 741 Strong Names and Assembly Signing 746 Assembly Names 747 Authenticode Signing 749 Table of Contents | vii
- 10. Resources and Satellite Assemblies 752 Loading, Resolving, and Isolating Assemblies 759 18. Reflection and Metadata. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 781 Reflecting and Activating Types 782 Reflecting and Invoking Members 789 Reflecting Assemblies 802 Working with Attributes 803 Dynamic Code Generation 808 Emitting Assemblies and Types 815 Emitting Type Members 818 Emitting Generic Methods and Types 823 Awkward Emission Targets 825 Parsing IL 829 19. Dynamic Programming. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 835 The Dynamic Language Runtime 835 Numeric Type Unification 836 Dynamic Member Overload Resolution 838 Implementing Dynamic Objects 844 Interoperating with Dynamic Languages 847 20. Cryptography. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 851 Overview 851 Windows Data Protection 852 Hashing 853 Symmetric Encryption 855 Public-Key Encryption and Signing 860 21. Advanced Threading. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 865 Synchronization Overview 866 Exclusive Locking 866 Locking and Thread Safety 874 Nonexclusive Locking 880 Signaling with Event Wait Handles 887 The Barrier Class 894 Lazy Initialization 896 Thread-Local Storage 898 Timers 901 viii | Table of Contents
- 11. 22. Parallel Programming. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 907 Why PFX? 908 PLINQ 911 The Parallel Class 924 Task Parallelism 930 Working with AggregateException 940 Concurrent Collections 942 BlockingCollection<T> 945 23. Span<T> and Memory<T>. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 951 Spans and Slicing 952 Memory<T> 955 Forward-Only Enumerators 957 Working with Stack-Allocated and Unmanaged Memory 958 24. Native and COM Interoperability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 961 Calling into Native DLLs 961 Type and Parameter Marshaling 962 Callbacks from Unmanaged Code 967 Simulating a C Union 970 Shared Memory 971 Mapping a Struct to Unmanaged Memory 973 COM Interoperability 977 Calling a COM Component from C# 979 Embedding Interop Types 982 Exposing C# Objects to COM 983 25. Regular Expressions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 987 Regular Expression Basics 987 Quantifiers 992 Zero-Width Assertions 993 Groups 996 Replacing and Splitting Text 997 Cookbook Regular Expressions 999 Regular Expressions Language Reference 1003 Index. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1007 Table of Contents | ix
- 13. Preface C# 10 represents the ninth major update to Microsoft’s flagship programming language, positioning C# as a language with unusual flexibility and breadth. At one end, it offers high-level abstractions such as query expressions and asynchronous continuations, whereas at the other end, it allows low-level efficiency through con‐ structs such as custom value types and optional pointers. The price of this growth is that there’s more than ever to learn. Although tools such as Microsoft’s IntelliSense—and online references—are excellent in helping you on the job, they presume an existing map of conceptual knowledge. This book provides exactly that map of knowledge in a concise and unified style—free of clutter and long introductions. Like the past seven editions, C# 10 in a Nutshell is organized around concepts and use cases, making it friendly both to sequential reading and to random browsing. It also plumbs significant depths while assuming only basic background knowledge, making it accessible to intermediate as well as advanced readers. This book covers C#, the Common Language Runtime (CLR), and the .NET 6 Base Class Library (BCL). We’ve chosen this focus to allow space for difficult and advanced topics without compromising depth or readability. Features recently added to C# are flagged so that you can also use this book as a reference for C# 9, C# 8, and C# 7. Intended Audience This book targets intermediate to advanced audiences. No prior knowledge of C# is required, but some general programming experience is necessary. For the beginner, this book complements, rather than replaces, a tutorial-style introduction to programming. This book is an ideal companion to any of the vast array of books that focus on an applied technology such as ASP.NET Core or Windows Presentation Foundation xi
- 14. (WPF). C# 10 in a Nutshell covers the areas of the language and .NET that such books omit, and vice versa. If you’re looking for a book that skims every .NET technology, this is not for you. This book is also unsuitable if you want to learn about APIs specific to mobile device development. How This Book Is Organized Chapters 2 through 4 concentrate purely on C#, starting with the basics of syntax, types, and variables, and finishing with advanced topics such as unsafe code and preprocessor directives. If you’re new to the language, you should read these chap‐ ters sequentially. The remaining chapters focus on .NET 6’s Base Class Libraries (BCLs), covering such topics as Language-Integrated Query (LINQ), XML, collections, concurrency, I/O and networking, memory management, reflection, dynamic programming, attributes, cryptography, and native interoperability. You can read most of these chapters randomly, except for Chapters 5 and 6, which lay a foundation for subse‐ quent topics. You’re also best off reading the three chapters on LINQ in sequence, and some chapters assume some knowledge of concurrency, which we cover in Chapter 14. What You Need to Use This Book The examples in this book require .NET 6. You will also find Microsoft’s .NET documentation useful to look up individual types and members (which is available online). Although it’s possible to write source code in a simple text editor and build your program from the command line, you’ll be much more productive with a code scratchpad for instantly testing code snippets, plus an integrated development envi‐ ronment (IDE) for producing executables and libraries. For a Windows code scratchpad, download LINQPad 7 from www.linqpad.net (free). LINQPad fully supports C# 10 and is maintained by the author. For a Windows IDE, download Visual Studio 2022: any edition is suitable for what’s taught in this book. For a cross-platform IDE, download Visual Studio Code. All code listings for all chapters are available as interactive (editable) LINQPad samples. You can download the entire lot in a single click: at the bottom left, click the LINQPad’s Samples tab, click “Download more samples,” and then choose “C# 10 in a Nutshell.” xii | Preface
- 15. Conventions Used in This Book The book uses basic UML notation to illustrate relationships between types, as shown in Figure P-1. A slanted rectangle means an abstract class; a circle means an interface. A line with a hollow triangle denotes inheritance, with the triangle pointing to the base type. A line with an arrow denotes a one-way association; a line without an arrow denotes a two-way association. Figure P-1. Sample diagram The following typographical conventions are used in this book: Italic Indicates new terms, URIs, filenames, and directories Constant width Indicates C# code, keywords and identifiers, and program output Constant width bold Shows a highlighted section of code Constant width italic Shows text that should be replaced with user-supplied values Preface | xiii
- 16. Using Code Examples Supplemental material (code examples, exercises, etc.) is available for download at http://www.albahari.com/nutshell. This book is here to help you get your job done. In general, you may use the code in this book in your programs and documentation. You do not need to contact us for permission unless you’re reproducing a significant portion of the code. For example, writing a program that uses several chunks of code from this book does not require permission. Selling or distributing examples from O’Reilly books does require permission. Answering a question by citing this book and quoting example code does not require permission (although we appreciate attribution). Incorporating a significant amount of example code from this book into your product’s documentation does require permission. We appreciate, but generally do not require, attribution. An attribution usually includes the title, author, publisher, and ISBN. For example: “C# 10 in a Nutshell by Joseph Albahari (O’Reilly). Copyright 2022 Joseph Albahari, 978-1-098-12195-2.” If you feel your use of code examples falls outside fair use or the permission given here, feel free to contact us at permissions@oreilly.com. O’Reilly Online Learning For more than 40 years, O’Reilly Media has pro‐ vided technology and business training, knowl‐ edge, and insight to help companies succeed. Our unique network of experts and innovators share their knowledge and expertise through books, articles, and our online learning platform. O’Reilly’s online learning platform gives you on-demand access to live training courses, in-depth learning paths, interactive coding environments, and a vast collec‐ tion of text and video from O’Reilly and 200+ other publishers. For more informa‐ tion, visit http://oreilly.com. How to Contact Us Please address comments and questions concerning this book to the publisher: O’Reilly Media, Inc. 1005 Gravenstein Highway North Sebastopol, CA 95472 800-998-9938 (in the United States or Canada) 707-829-0515 (international or local) 707-829-0104 (fax) We have a web page for this book, where we list errata, examples, and any additional information. You can access this page at https://oreil.ly/c-sharp-nutshell-10. xiv | Preface
- 17. Email bookquestions@oreilly.com to comment or ask technical questions about this book. For news and information about our books and courses, visit http://oreilly.com. Find us on Facebook: http://facebook.com/oreilly Follow us on Twitter: http://twitter.com/oreillymedia Watch us on YouTube: http://youtube.com/oreillymedia Acknowledgments Since its first incarnation in 2007, this book has relied on input from some superb technical reviewers. For their input into recent editions, I’d like to extend particular thanks to Stephen Toub, Paulo Morgado, Fred Silberberg, Vitek Karas, Aaron Rob‐ inson, Jan Vorlicek, Sam Gentile, Rod Stephens, Jared Parsons, Matthew Groves, Dixin Yan, Lee Coward, Bonnie DeWitt, Wonseok Chae, Lori Lalonde, and James Montemagno. And for their input into earlier editions, I’m most grateful to Eric Lippert, Jon Skeet, Stephen Toub, Nicholas Paldino, Chris Burrows, Shawn Farkas, Brian Grunkemeyer, Maoni Stephens, David DeWinter, Mike Barnett, Melitta Andersen, Mitch Wheat, Brian Peek, Krzysztof Cwalina, Matt Warren, Joel Pobar, Glyn Griffiths, Ion Vasi‐ lian, Brad Abrams, and Adam Nathan. I appreciate that many of the technical reviewers are accomplished individuals at Microsoft, and I particularly thank you for taking the time to raise this book to the next quality bar. I want to thank Ben Albahari and Eric Johannsen, who contributed to previous editions, and the O’Reilly team—particularly my efficient and responsive editor Corbin Collins. Finally, my deepest thanks to my wonderful wife, Li Albahari, whose presence kept me happy throughout the project. Preface | xv
- 19. 1 Introducing C# and .NET C# is a general-purpose, type-safe, object-oriented programming language. The goal of the language is programmer productivity. To this end, C# balances simplicity, expressiveness, and performance. The chief architect of the language since its first version is Anders Hejlsberg (creator of Turbo Pascal and architect of Delphi). The C# language is platform neutral and works with a range of platform-specific runtimes. Object Orientation C# is a rich implementation of the object-orientation paradigm, which includes encapsulation, inheritance, and polymorphism. Encapsulation means creating a boundary around an object to separate its external (public) behavior from its inter‐ nal (private) implementation details. Following are the distinctive features of C# from an object-oriented perspective: Unified type system The fundamental building block in C# is an encapsulated unit of data and func‐ tions called a type. C# has a unified type system in which all types ultimately share a common base type. This means that all types, whether they represent business objects or are primitive types such as numbers, share the same basic functionality. For example, an instance of any type can be converted to a string by calling its ToString method. Classes and interfaces In a traditional object-oriented paradigm, the only kind of type is a class. In C#, there are several other kinds of types, one of which is an interface. An interface is like a class that cannot hold data. This means that it can define only behavior (and not state), which allows for multiple inheritance as well as a separation between specification and implementation. 1
- 20. Properties, methods, and events In the pure object-oriented paradigm, all functions are methods. In C#, meth‐ ods are only one kind of function member, which also includes properties and events (there are others, too). Properties are function members that encapsulate a piece of an object’s state, such as a button’s color or a label’s text. Events are function members that simplify acting on object state changes. Although C# is primarily an object-oriented language, it also borrows from the functional programming paradigm; specifically: Functions can be treated as values Using delegates, C# allows functions to be passed as values to and from other functions. C# supports patterns for purity Core to functional programming is avoiding the use of variables whose values change, in favor of declarative patterns. C# has key features to help with those patterns, including the ability to write unnamed functions on the fly that “capture” variables (lambda expressions) and the ability to perform list or reactive programming via query expressions. C# also provides records, which make it easy to write immutable (read-only) types. Type Safety C# is primarily a type-safe language, meaning that instances of types can interact only through protocols they define, thereby ensuring each type’s internal consis‐ tency. For instance, C# prevents you from interacting with a string type as though it were an integer type. More specifically, C# supports static typing, meaning that the language enforces type safety at compile time. This is in addition to type safety being enforced at runtime. Static typing eliminates a large class of errors before a program is even run. It shifts the burden away from runtime unit tests onto the compiler to verify that all the types in a program fit together correctly. This makes large programs much easier to manage, more predictable, and more robust. Furthermore, static typing allows tools such as IntelliSense in Visual Studio to help you write a program because it knows for a given variable what type it is, and hence what methods you can call on that variable. Such tools can also identify everywhere in your program that a variable, type, or method is used, allowing for reliable refactoring. C# also allows parts of your code to be dynamically typed via the dynamic keyword. However, C# remains a predominantly statically typed language. C# is also called a strongly typed language because its type rules are strictly enforced (whether statically or at runtime). For instance, you cannot call a function that’s designed to accept an integer with a floating-point number, unless you first explicitly convert the floating-point number to an integer. This helps prevent mistakes. 2 | Chapter 1: Introducing C# and .NET
- 21. Memory Management C# relies on the runtime to perform automatic memory management. The Com‐ mon Language Runtime has a garbage collector that executes as part of your program, reclaiming memory for objects that are no longer referenced. This frees programmers from explicitly deallocating the memory for an object, eliminating the problem of incorrect pointers encountered in languages such as C++. C# does not eliminate pointers: it merely makes them unnecessary for most pro‐ gramming tasks. For performance-critical hotspots and interoperability, pointers and explicit memory allocation is permitted in blocks that are marked unsafe. Platform Support C# has runtimes that support the following platforms: • Windows Desktop 7-11 (for rich-client, web, server, and command-line • applications) • macOS (for rich-client, web, and command-line applications) • • Linux and macOS (for web and command-line applications) • • Android and iOS (for mobile applications) • • Windows 10 devices (Xbox, Surface Hub, and HoloLens) • There is also a technology called Blazor that can compile C# to web assembly that runs in a browser. CLRs, BCLs, and Runtimes Runtime support for C# programs consists of a Common Language Runtime and a Base Class Library. A runtime can also include a higher-level application layer that contains libraries for developing rich-client, mobile, or web applications (see Figure 1-1). Different runtimes exist to allow for different kinds of applications, as well as different platforms. Figure 1-1. Runtime architecture Memory Management | 3 Introducing C# and .NET
- 22. Common Language Runtime A Common Language Runtime (CLR) provides essential runtime services such as automatic memory management and exception handling. (The word “common” refers to the fact that the same runtime can be shared by other managed program‐ ming languages, such as F#, Visual Basic, and Managed C++.) C# is called a managed language because it compiles source code into managed code, which is represented in Intermediate Language (IL). The CLR converts the IL into the native code of the machine, such as X64 or X86, usually just prior to execution. This is referred to as Just-in-Time (JIT) compilation. Ahead-of-time compilation is also available to improve startup time with large assemblies or resource-constrained devices (and to satisfy iOS app store rules when developing mobile apps). The container for managed code is called an assembly. An assembly contains not only IL but also type information (metadata). The presence of metadata allows assemblies to reference types in other assemblies without needing additional files. You can examine and disassemble the contents of an assembly with Microsoft’s ildasm tool. And with tools such as ILSpy or JetBrain’s dotPeek, you can go further and decompile the IL to C#. Because IL is higher level than native machine code, the decompiler can do quite a good job of reconstructing the original C#. A program can query its own metadata (reflection) and even generate new IL at runtime (reflection.emit). Base Class Library A CLR always ships with a set of assemblies called a Base Class Library (BCL). A BCL provides core functionality to programmers, such as collections, input/out‐ put, text processing, XML/JSON handling, networking, encryption, interop, con‐ currency, and parallel programming. A BCL also implements types that the C# language itself requires (for features such as enumeration, querying, and asynchrony) and lets you explicitly access features of the CLR, such as reflection and memory management. Runtimes A runtime (also called a framework) is a deployable unit that you download and install. A runtime consists of a CLR (with its BCL), plus an optional application layer specific to the kind of application that you’re writing—web, mobile, rich client, etc. (If you’re writing a command-line console application or a non-UI library, you don’t need an application layer.) When writing an application, you target a particular runtime, which means that your application uses and depends on the functionality that the runtime provides. 4 | Chapter 1: Introducing C# and .NET
- 23. Your choice of runtime also determines which platforms your application will support. The following table lists the major runtime options: Application layer CLR/BCL Program type Runs on... ASP.NET .NET 6 Web Windows, Linux, macOS Windows Desktop .NET 6 Windows Windows 7–10+ MAUI (early 2022) .NET 6 Mobile, desktop iOS, Android, macOS, Windows 10+ WinUI 3 (early 2022) .NET 6 Win10 Windows 10+ desktop UWP .NET Core 2.2 Win10 + Win10 devices Windows 10+ desktop & devices (Legacy) .NET Framework .NET Framework Web, Windows Windows 7–10+ Figure 1-2 shows this information graphically and also serves as a guide to what’s covered in the book. Figure 1-2. Runtimes for C# .NET 6 .NET 6 is Microsoft’s flagship open-source runtime. You can write web and console applications that run on Windows, Linux, and macOS; rich-client applications that run on Windows 7 through 11 and macOS; and mobile apps that run on iOS and Android. This book focuses on the .NET 6 CLR and BCL. CLRs, BCLs, and Runtimes | 5 Introducing C# and .NET
- 24. Unlike .NET Framework, .NET 6 is not preinstalled on Windows machines. If you try to run a .NET 6 application without the correct runtime being present, a mes‐ sage will appear directing you to a web page where you can download the runtime. You can avoid this by creating a self-contained deployment, which includes the parts of the runtime required by the application. .NET 6’s predecessor was .NET 5, whose predecessor was .NET Core 3. (Microsoft removed “Core” from the name and skipped version 4.) The reason for skipping a version was to avoid confusion with .NET Framework 4.x. This means that assemblies compiled under .NET Core ver‐ sions 1, 2, and 3 (and .NET 5) will, in most cases, run without modification under .NET 6. In contrast, assemblies compiled under (any version of) .NET Framework are usually incom‐ patible with .NET 6. The .NET 6 BCL and CLR are very similar to .NET 5 (and .NET Core 3), with their differences centering mostly on performance and deployment. MAUI MAUI (Multi-platform App UI, early 2022) is designed for creating mobile apps for iOS and Android, as well as cross-platform desktop apps for macOS and Windows. MAUI is an evolution of Xamarin, and allows a single project to target multiple platforms. UWP and WinUI 3 Universal Windows Platform (UWP) is designed for writing immersive touch-first applications that run on Windows 10+ desktop and devices (Xbox, Surface Hub, and HoloLens). UWP apps are sandboxed and ship via the Windows Store. UWP is preinstalled with Windows 10. It uses a version of the .NET Core 2.2 CLR/BCL, and it’s unlikely that this dependency will be updated. Instead, Microsoft has released a successor called WinUI 3, as part of the Windows App SDK. The Windows App SDK works with the latest .NET, integrates better with the .NET desktop APIs, and can run outside a sandbox. However, it does not yet support devices such as Xbox or HoloLens. .NET Framework .NET Framework is Microsoft’s original Windows-only runtime for writing web and rich-client applications that run (only) on Windows desktops/servers. No major new releases are planned, although Microsoft will continue to support and maintain the current 4.8 release due to the wealth of existing applications. With the .NET Framework, the CLR/BCL is integrated with the application layer. Applications written in .NET Framework can be recompiled under .NET 6, although they usually require some modification. Some features of .NET Frame‐ work are not present in .NET 6 (and vice versa). 6 | Chapter 1: Introducing C# and .NET
- 25. .NET Framework is preinstalled with Windows and is automatically patched via Windows Update. When you target .NET Framework 4.8, you can use the features of C# 7.3 and earlier. The word “.NET” has long been used as an umbrella term for any technology that includes the word “.NET” (.NET Frame‐ work, .NET Core, .NET Standard, and so on). This means that Microsoft’s renaming of .NET Core to .NET has created an unfortunate ambiguity. In this book, we’ll refer to the new .NET as .NET 5+. And to refer to .NET Core and its successors, we’ll use the phrase “.NET Core and .NET 5+.” To add to the confusion, .NET (5+) is a framework, yet it’s very different from the .NET Framework. Hence, we’ll use the term runtime in preference to framework, where possible. Niche Runtimes There are also the following niche runtimes: • The .NET Micro Framework is for running .NET code on highly resource- • constrained embedded devices (under one megabyte). • Unity is a game development platform that allows game logic to be scripted • with C#. It’s also possible to run managed code within SQL Server. With SQL Server CLR integration, you can write custom functions, stored procedures, and aggregations in C# and then call them from SQL. This works in conjunction with .NET Framework and a special “hosted” CLR that enforces a sandbox to protect the integrity of the SQL Server process. A Brief History of C# The following is a reverse chronology of the new features in each C# version, for the benefit of readers who are already familiar with an older version of the language. What’s New in C# 10 C# 10 ships with Visual Studio 2022, and is used when you target .NET 6. File-scoped namespaces In the common case that all types in a file are defined in a single namespace, a file-scoped namespace declaration in C# 10 reduces clutter and eliminates an unnecessary level of indentation: namespace MyNamespace; // Applies to everything that follows in the file. class Class1 {} // inside MyNamespace class Class2 {} // inside MyNamespace A Brief History of C# | 7 Introducing C# and .NET
- 26. The global using directive When you prefix a using directive with the global keyword, it applies the directive to all files in the project: global using System; global using System.Collection.Generic; This lets you avoid repeating the same directives in every file. global using direc‐ tives work with using static. Additionally, .NET 6 projects now support implicit global using directives: if the ImplicitUsings element is set to true in the project file, the most commonly used namespaces are automatically imported (based on the SDK project type). See “The global using Directive (C# 10)” on page 88 for more detail. Nondestructive mutation for anonymous types C# 9 introduced the with keyword, to perform nondestructive mutation on records. In C# 10, the with keyword also works with anonymous types: var a1 = new { A = 1, B = 2, C = 3, D = 4, E = 5 }; var a2 = a1 with { E = 10 }; Console.WriteLine (a2); // { A = 1, B = 2, C = 3, D = 4, E = 10 } New deconstruction syntax C# 7 introduced the deconstruction syntax for tuples (or any type with a Deconstruct method). C# 10 takes this syntax further, letting you mix assignment and declaration in the same deconstruction: var point = (3, 4); double x = 0; (x, double y) = point; Field initializers and parameterless constructors in structs From C# 10, you can include field initializers and parameterless constructors in structs (see “Structs” on page 129). These execute only when the constructor is called explicitly, and so can easily be bypassed—for instance, via the default key‐ word. This feature was introduced primarily for the benefit of struct records. Record structs Records were first introduced in C# 9, where they acted as a compiled-enhanced class. In C# 10, records can also be structs: record struct Point (int X, int Y); The rules are otherwise similar: record structs have much the same features as class structs (see “Records” on page 211). An exception is that the compiler-generated properties on record structs are writable, unless you prefix the record declaration with the readonly keyword. 8 | Chapter 1: Introducing C# and .NET
- 27. Lambda expression enhancements The syntax around lambda expressions has been enhanced in a number of ways. First, implicit typing (var) is permitted: var greeter = () => "Hello, world"; The implicit type for a lambda expression is an Action or Func delegate, so greeter, in this case, is of type Func<string>. You must explicitly state any parameter types: var square = (int x) => x * x; Second, a lambda expression can specify a return type: var sqr = int (int x) => x; This is primarily to improve compiler performance with complex nested lambdas. Third, you can pass a lambda expression into a method parameter of type object, Delegate, or Expression: M1 (() => "test"); // Implicitly typed to Func<string> M2 (() => "test"); // Implicitly typed to Func<string> M3 (() => "test"); // Implicitly typed to Expression<Func<string>> void M1 (object x) {} void M2 (Delegate x) {} void M3 (Expression x) {} Finally, you can apply attributes to a lambda expression’s compile-generated target method (as well as its parameters and return value): Action a = [Description("test")] () => { }; See “Applying Attributes to Lambda Expressions (C# 10)” on page 228 for more detail. Nested property patterns The following simplified syntax is legal in C# 10 for nested property pattern match‐ ing (see “Property Patterns” on page 225): var obj = new Uri ("https://www.linqpad.net"); if (obj is Uri { Scheme.Length: 5 }) ... This is equivalent to: if (obj is Uri { Scheme: { Length: 5 }}) ... CallerArgumentExpression A method parameter to which you apply the [CallerArgumentExpression] attribute captures an argument expression from the call site: A Brief History of C# | 9 Introducing C# and .NET
- 28. Print (Math.PI * 2); void Print (double number, [CallerArgumentExpression("number")] string expr = null) => Console.WriteLine (expr); // Output: Math.PI * 2 This feature is intended primarily for validation and assertion libraries (see “Caller‐ ArgumentExpression (C# 10)” on page 231). Other new features The #line directive has been enhanced in C# 10 to allow a column and range to be specified. Interpolated strings in C# 10 can be constants, as long as the interpolated values are constants. Records can seal the ToString() method in C# 10. C#’s definite assignment analysis has been improved so that expressions such as the following work: if (foo?.TryParse ("123", out var number) ?? false) Console.WriteLine (number); (Prior to C# 10, the compiler would generate an error: “Use of unassigned local variable ‘number’.”) What’s New in C# 9.0 C# 9.0 shipped with Visual Studio 2019, and is used when you target .NET 5. Top-level statements With top-level statements (see “Top-Level Statements” on page 35), you can write a program without the baggage of a Main method and Program class: using System; Console.WriteLine ("Hello, world"); Top-level statements can include methods (which act as local methods). You can also access command-line arguments via the “magic” args variable and return a value to the caller. Top-level statements can be followed by type and namespace declarations. Init-only setters An init-only setter (see “Init-only setters” on page 108) in a property declaration uses the init keyword instead of the set keyword: class Foo { public int ID { get; init; } } 10 | Chapter 1: Introducing C# and .NET
- 29. This behaves like a read-only property, except that it can also be set via an object initializer: var foo = new Foo { ID = 123 }; This makes it possible to create immutable (read-only) types that can be populated via an object initializer instead of a constructor, and helps to avoid the antipattern of constructors that accept a large number of optional parameters. Init-only setters also allow for nondestructive mutation when used in records. Records A record (see “Records” on page 211) is a special kind of class that’s designed to work well with immutable data. Its most special feature is that it supports nondes‐ tructive mutation via a new keyword (with): Point p1 = new Point (2, 3); Point p2 = p1 with { Y = 4 }; // p2 is a copy of p1, but with Y set to 4 Console.WriteLine (p2); // Point { X = 2, Y = 4 } record Point { public Point (double x, double y) => (X, Y) = (x, y); public double X { get; init; } public double Y { get; init; } } In simple cases, a record can also eliminate the boilerplate code of defining proper‐ ties and writing a constructor and deconstructor. We can replace our Point record definition with the following, without loss of functionality: record Point (double X, double Y); Like tuples, records exhibit structural equality by default. Records can subclass other records and can include the same constructs that classes can include. The compiler implements records as classes at runtime. Pattern-matching improvements The relational pattern (see “Patterns” on page 222) allows the <, >, <=, and >= operators to appear in patterns: string GetWeightCategory (decimal bmi) => bmi switch { < 18.5m => "underweight", < 25m => "normal", < 30m => "overweight", _ => "obese" }; With pattern combinators, you can combine patterns via three new keywords (and, or, and not): bool IsVowel (char c) => c is 'a' or 'e' or 'i' or 'o' or 'u'; A Brief History of C# | 11 Introducing C# and .NET
- 30. bool IsLetter (char c) => c is >= 'a' and <= 'z' or >= 'A' and <= 'Z'; As with the && and || operators, and has higher precedence than or. You can override this with parentheses. The not combinator can be used with the type pattern to test whether an object is (not) a type: if (obj is not string) ... Target-typed new expressions When constructing an object, C# 9 lets you omit the type name when the compiler can infer it unambiguously: System.Text.StringBuilder sb1 = new(); System.Text.StringBuilder sb2 = new ("Test"); This is particularly useful when the variable declaration and initialization are in different parts of your code: class Foo { System.Text.StringBuilder sb; public Foo (string initialValue) => sb = new (initialValue); } And in the following scenario: MyMethod (new ("test")); void MyMethod (System.Text.StringBuilder sb) { ... } see “Target-Typed new Expressions” on page 69 for more information. Interop improvements C# 9 introduced function pointers (see “Function Pointers” on page 248 and “Call‐ backs with Function Pointers” on page 967). Their main purpose is to allow unman‐ aged code to call static methods in C# without the overhead of a delegate instance, with the ability to bypass the P/Invoke layer when the arguments and return types are blittable (represented identically on each side). C# 9 also introduced the nint and nuint native-sized integer types (see “Native- Sized Integers” on page 246), which map at runtime to System.IntPtr and System.UIntPtr. At compile time, they behave like numeric types with support for arithmetic operations. Other new features Additionally, C# 9 now lets you do the following: • Override a method or read-only property such that it returns a more derived • type (see “Covariant return types” on page 120) 12 | Chapter 1: Introducing C# and .NET
- 31. • Apply attributes to local functions (see “Attributes” on page 227) • • Apply the static keyword to lambda expressions or local functions to ensure • that you don’t accidentally capture local or instance variables (see “Static lamb‐ das” on page 177) • Make any type work with the foreach statement, by writing a GetEnumerator • extension method • Define a module initializer method that executes once when an assembly is • first loaded, by applying the [ModuleInitializer] attribute to a (static void parameterless) method • Use a “discard” (underscore symbol) as a lambda expression argument • • Write extended partial methods that are mandatory to implement—enabling • scenarios such as Roslyn’s new source generators (see “Extended partial meth‐ ods” on page 114) • Apply an attribute to methods, types, or modules to prevent local variables • from being initialized by the runtime (see “[SkipLocalsInit]” on page 248) What’s New in C# 8.0 C# 8.0 first shipped with Visual Studio 2019, and is still used today when you target .NET Core 3 or .NET Standard 2.1. Indices and ranges Indices and ranges simplify working with elements or portions of an array (or the low-level types Span<T> and ReadOnlySpan<T>). Indices let you refer to elements relative to the end of an array by using the ^ operator. ^1 refers to the last element, ^2 refers to the second-to-last element, and so on: char[] vowels = new char[] {'a','e','i','o','u'}; char lastElement = vowels [^1]; // 'u' char secondToLast = vowels [^2]; // 'o' Ranges let you “slice” an array by using the .. operator: char[] firstTwo = vowels [..2]; // 'a', 'e' char[] lastThree = vowels [2..]; // 'i', 'o', 'u' char[] middleOne = vowels [2..3] // 'i' char[] lastTwo = vowels [^2..]; // 'o', 'u' C# implements indexes and ranges with the help of the Index and Range types: Index last = ^1; Range firstTwoRange = 0..2; char[] firstTwo = vowels [firstTwoRange]; // 'a', 'e' A Brief History of C# | 13 Introducing C# and .NET
- 32. You can support indices and ranges in your own classes by defining an indexer with a parameter type of Index or Range: class Sentence { string[] words = "The quick brown fox".Split(); public string this [Index index] => words [index]; public string[] this [Range range] => words [range]; } For more information, see “Indices” on page 56. Null-coalescing assignment The ??= operator assigns a variable only if it’s null. Instead of if (s == null) s = "Hello, world"; you can now write this: s ??= "Hello, world"; Using declarations If you omit the brackets and statement block following a using statement, it becomes a using declaration. The resource is then disposed when execution falls outside the enclosing statement block: if (File.Exists ("file.txt")) { using var reader = File.OpenText ("file.txt"); Console.WriteLine (reader.ReadLine()); ... } In this case, reader will be disposed when execution falls outside the if statement block. Read-only members C# 8 lets you apply the readonly modifier to a struct’s functions, ensuring that if the function attempts to modify any field, a compile-time error is generated: struct Point { public int X, Y; public readonly void ResetX() => X = 0; // Error! } If a readonly function calls a non-readonly function, the compiler generates a warning (and defensively copies the struct to avoid the possibility of a mutation). 14 | Chapter 1: Introducing C# and .NET
- 33. Static local methods Adding the static modifier to a local method prevents it from seeing the local variables and parameters of the enclosing method. This helps to reduce coupling and enables the local method to declare variables as it pleases, without risk of colliding with those in the containing method. Default interface members C# 8 lets you add a default implementation to an interface member, making it optional to implement: interface ILogger { void Log (string text) => Console.WriteLine (text); } This means that you can add a member to an interface without breaking implemen‐ tations. Default implementations must be called explicitly through the interface: ((ILogger)new Logger()).Log ("message"); Interfaces can also define static members (including fields), which can be accessed from code inside default implementations: interface ILogger { void Log (string text) => Console.WriteLine (Prefix + text); static string Prefix = ""; } Or from outside the interface unless restricted via an accessibility modifier on the static interface member (such as private, protected, or internal): ILogger.Prefix = "File log: "; Instance fields are prohibited. For more details, see “Default Interface Members” on page 139. Switch expressions From C# 8, you can use switch in the context of an expression: string cardName = cardNumber switch // assuming cardNumber is an int { 13 => "King", 12 => "Queen", 11 => "Jack", _ => "Pip card" // equivalent to 'default' }; For more examples, see “Switch expressions” on page 82. A Brief History of C# | 15 Introducing C# and .NET
- 34. Tuple, positional, and property patterns C# 8 supports three new patterns, mostly for the benefit of switch state‐ ments/expressions (see “Patterns” on page 222). Tuple patterns let you switch on multiple values: int cardNumber = 12; string suite = "spades"; string cardName = (cardNumber, suite) switch { (13, "spades") => "King of spades", (13, "clubs") => "King of clubs", ... }; Positional patterns allow a similar syntax for objects that expose a deconstructor, and property patterns let you match on an object’s properties. You can use all of the patterns both in switches and with the is operator. The following example uses a property pattern to test whether obj is a string with a length of 4: if (obj is string { Length:4 }) ... Nullable reference types Whereas nullable value types bring nullability to value types, nullable reference types do the opposite and bring (a degree of) non-nullability to reference types, with the purpose of helping to avoid NullReferenceExceptions. Nullable reference types introduce a level of safety that’s enforced purely by the compiler in the form of warnings or errors when it detects code that’s at risk of generating a NullReferenceException. Nullable reference types can be enabled either at the project level (via the Nullable element in the .csproj project file) or in code (via the #nullable directive). After it’s enabled, the compiler makes non-nullability the default: if you want a reference type to accept nulls, you must apply the ? suffix to indicate a nullable reference type: #nullable enable // Enable nullable reference types from this point on string s1 = null; // Generates a compiler warning! (s1 is non-nullable) string? s2 = null; // OK: s2 is nullable reference type Uninitialized fields also generate a warning (if the type is not marked as nulla‐ ble), as does dereferencing a nullable reference type, if the compiler thinks a NullReferenceException might occur: void Foo (string? s) => Console.Write (s.Length); // Warning (.Length) To remove the warning, you can use the null-forgiving operator (!): void Foo (string? s) => Console.Write (s!.Length); For a full discussion, see “Nullable Reference Types” on page 200. 16 | Chapter 1: Introducing C# and .NET
- 35. Asynchronous streams Prior to C# 8, you could use yield return to write an iterator, or await to write an asynchronous function. But you couldn’t do both and write an iterator that awaits, yielding elements asynchronously. C# 8 fixes this through the introduction of asynchronous streams: async IAsyncEnumerable<int> RangeAsync ( int start, int count, int delay) { for (int i = start; i < start + count; i++) { await Task.Delay (delay); yield return i; } } The await foreach statement consumes an asynchronous stream: await foreach (var number in RangeAsync (0, 10, 100)) Console.WriteLine (number); For more information, see “Asynchronous Streams” on page 650. What’s New in C# 7.x C# 7.x was first shipped with Visual Studio 2017. C# 7.3 is still used today by Visual Studio 2019 when you target .NET Core 2, .NET Framework 4.6 to 4.8, or .NET Standard 2.0. C# 7.3 C# 7.3 made minor improvements to existing features, such as enabling use of the equality operators with tuples, improving overload resolution, and offering the ability to apply attributes to the backing fields of automatic properties: [field:NonSerialized] public int MyProperty { get; set; } C# 7.3 also built on C# 7.2’s advanced low-allocation programming features, with the ability to reassign ref locals, no requirement to pin when indexing fixed fields, and field initializer support with stackalloc: int* pointer = stackalloc int[] {1, 2, 3}; Span<int> arr = stackalloc [] {1, 2, 3}; Notice that stack-allocated memory can be assigned directly to a Span<T>. We describe spans—and why you would use them—in Chapter 23. C# 7.2 C# 7.2 added a new private protected modifier (the intersection of internal and protected), the ability to follow named arguments with positional ones when calling methods, and readonly structs. A readonly struct enforces that all fields are A Brief History of C# | 17 Introducing C# and .NET
- 36. readonly, to aid in declaring intent and to allow the compiler more optimization freedom: readonly struct Point { public readonly int X, Y; // X and Y must be readonly } C# 7.2 also added specialized features to help with micro-optimization and low- allocation programming: see “The in modifier” on page 64, “Ref Locals” on page 67, “Ref Returns” on page 68, and “Ref Structs” on page 131). C# 7.1 From C# 7.1, you can omit the type when using the default keyword, if the type can be inferred: decimal number = default; // number is decimal C# 7.1 also relaxed the rules for switch statements (so that you can pattern-match on generic type parameters), allowed a program’s Main method to be asynchronous, and allowed tuple element names to be inferred: var now = DateTime.Now; var tuple = (now.Hour, now.Minute, now.Second); Numeric literal improvements Numeric literals in C# 7 can include underscores to improve readability. These are called digit separators and are ignored by the compiler: int million = 1_000_000; Binary literals can be specified with the 0b prefix: var b = 0b1010_1011_1100_1101_1110_1111; Out variables and discards C# 7 makes it easier to call methods that contain out parameters. First, you can now declare out variables on the fly (see “Out variables and discards” on page 64): bool successful = int.TryParse ("123", out int result); Console.WriteLine (result); And when calling a method with multiple out parameters, you can discard ones you’re uninterested in with the underscore character: SomeBigMethod (out _, out _, out _, out int x, out _, out _, out _); Console.WriteLine (x); Type patterns and pattern variables You can also introduce variables on the fly with the is operator. These are called pattern variables (see “Introducing a pattern variable” on page 118): 18 | Chapter 1: Introducing C# and .NET
- 37. void Foo (object x) { if (x is string s) Console.WriteLine (s.Length); } The switch statement also supports type patterns, so you can switch on type as well as constants (see “Switching on types” on page 81). You can specify conditions with a when clause and also switch on the null value: switch (x) { case int i: Console.WriteLine ("It's an int!"); break; case string s: Console.WriteLine (s.Length); // We can use the s variable break; case bool b when b == true: // Matches only when b is true Console.WriteLine ("True"); break; case null: Console.WriteLine ("Nothing"); break; } Local methods A local method is a method declared within another function (see “Local methods” on page 98): void WriteCubes() { Console.WriteLine (Cube (3)); Console.WriteLine (Cube (4)); Console.WriteLine (Cube (5)); int Cube (int value) => value * value * value; } Local methods are visible only to the containing function and can capture local variables in the same way that lambda expressions do. More expression-bodied members C# 6 introduced the expression-bodied “fat-arrow” syntax for methods, read-only properties, operators, and indexers. C# 7 extends this to constructors, read/write properties, and finalizers: public class Person { string name; public Person (string name) => Name = name; A Brief History of C# | 19 Introducing C# and .NET
- 38. public string Name { get => name; set => name = value ?? ""; } ~Person () => Console.WriteLine ("finalize"); } Deconstructors C# 7 introduces the deconstructor pattern (see “Deconstructors” on page 102). Whereas a constructor typically takes a set of values (as parameters) and assigns them to fields, a deconstructor does the reverse and assigns fields back to a set of variables. We could write a deconstructor for the Person class in the preceding example as follows (exception handling aside): public void Deconstruct (out string firstName, out string lastName) { int spacePos = name.IndexOf (' '); firstName = name.Substring (0, spacePos); lastName = name.Substring (spacePos + 1); } Deconstructors are called with the following special syntax: var joe = new Person ("Joe Bloggs"); var (first, last) = joe; // Deconstruction Console.WriteLine (first); // Joe Console.WriteLine (last); // Bloggs Tuples Perhaps the most notable improvement to C# 7 is explicit tuple support (see “Tuples” on page 207). Tuples provide a simple way to store a set of related values: var bob = ("Bob", 23); Console.WriteLine (bob.Item1); // Bob Console.WriteLine (bob.Item2); // 23 C#’s new tuples are syntactic sugar for using the System.ValueTuple<…> generic structs. But thanks to compiler magic, tuple elements can be named: var tuple = (name:"Bob", age:23); Console.WriteLine (tuple.name); // Bob Console.WriteLine (tuple.age); // 23 With tuples, functions can return multiple values without resorting to out parame‐ ters or extra type baggage: static (int row, int column) GetFilePosition() => (3, 10); static void Main() { var pos = GetFilePosition(); 20 | Chapter 1: Introducing C# and .NET
- 39. Console.WriteLine (pos.row); // 3 Console.WriteLine (pos.column); // 10 } Tuples implicitly support the deconstruction pattern, so you can easily deconstruct them into individual variables: static void Main() { (int row, int column) = GetFilePosition(); // Creates 2 local variables Console.WriteLine (row); // 3 Console.WriteLine (column); // 10 } throw expressions Prior to C# 7, throw was always a statement. Now it can also appear as an expression in expression-bodied functions: public string Foo() => throw new NotImplementedException(); A throw expression can also appear in a ternary conditional expression: string Capitalize (string value) => value == null ? throw new ArgumentException ("value") : value == "" ? "" : char.ToUpper (value[0]) + value.Substring (1); What’s New in C# 6.0 C# 6.0, which shipped with Visual Studio 2015, features a new-generation compiler, completely written in C#. Known as project “Roslyn,” the new compiler exposes the entire compilation pipeline via libraries, allowing you to perform code analysis on arbitrary source code. The compiler itself is open source, and the source code is available at github.com/dotnet/roslyn. In addition, C# 6.0 features several minor but significant enhancements, aimed primarily at reducing code clutter. The null-conditional (“Elvis”) operator (see “Null Operators” on page 74) avoids having to explicitly check for null before calling a method or accessing a type member. In the following example, result evaluates to null instead of throwing a NullReferenceException: System.Text.StringBuilder sb = null; string result = sb?.ToString(); // result is null Expression-bodied functions (see “Methods” on page 98) allow methods, properties, operators, and indexers that comprise a single expression to be written more tersely, in the style of a lambda expression: public int TimesTwo (int x) => x * 2; public string SomeProperty => "Property value"; A Brief History of C# | 21 Introducing C# and .NET
- 40. Property initializers (Chapter 3) let you assign an initial value to an automatic property: public DateTime TimeCreated { get; set; } = DateTime.Now; Initialized properties can also be read-only: public DateTime TimeCreated { get; } = DateTime.Now; Read-only properties can also be set in the constructor, making it easier to create immutable (read-only) types. Index initializers (Chapter 4) allow single-step initialization of any type that exposes an indexer: var dict = new Dictionary<int,string>() { [3] = "three", [10] = "ten" }; String interpolation (see “String Type” on page 52) offers a succinct alternative to string.Format: string s = $"It is {DateTime.Now.DayOfWeek} today"; Exception filters (see “try Statements and Exceptions” on page 181) let you apply a condition to a catch block: string html; try { html = await new HttpClient().GetStringAsync ("http://asef"); } catch (WebException ex) when (ex.Status == WebExceptionStatus.Timeout) { ... } The using static (see “Namespaces” on page 87) directive lets you import all the static members of a type so that you can use those members unqualified: using static System.Console; ... WriteLine ("Hello, world"); // WriteLine instead of Console.WriteLine The nameof (Chapter 3) operator returns the name of a variable, type, or other symbol as a string. This avoids breaking code when you rename a symbol in Visual Studio: int capacity = 123; string x = nameof (capacity); // x is "capacity" string y = nameof (Uri.Host); // y is "Host" And finally, you’re now allowed to await inside catch and finally blocks. 22 | Chapter 1: Introducing C# and .NET
- 41. What’s New in C# 5.0 C# 5.0’s big new feature was support for asynchronous functions via two new keywords, async and await. Asynchronous functions enable asynchronous continu‐ ations, which make it easier to write responsive and thread-safe rich-client applica‐ tions. They also make it easy to write highly concurrent and efficient I/O-bound applications that don’t tie up a thread resource per operation. We cover asynchro‐ nous functions in detail in Chapter 14. What’s New in C# 4.0 C# 4.0 introduced four major enhancements: Dynamic binding (Chapters 4 and 19) defers binding—the process of resolving types and members—from compile time to runtime and is useful in scenarios that would otherwise require complicated reflection code. Dynamic binding is also useful when interoperating with dynamic languages and COM components. Optional parameters (Chapter 2) allow functions to specify default parameter values so that callers can omit arguments, and named arguments allow a function caller to identify an argument by name rather than position. Type variance rules were relaxed in C# 4.0 (Chapters 3 and 4), such that type parameters in generic interfaces and generic delegates can be marked as covariant or contravariant, allowing more natural type conversions. COM interoperability (Chapter 24) was enhanced in C# 4.0 in three ways. First, arguments can be passed by reference without the ref keyword (particularly useful in conjunction with optional parameters). Second, assemblies that contain COM interop types can be linked rather than referenced. Linked interop types support type equivalence, avoiding the need for Primary Interop Assemblies and putting an end to versioning and deployment headaches. Third, functions that return COM Variant types from linked interop types are mapped to dynamic rather than object, eliminating the need for casting. What’s New in C# 3.0 The features added to C# 3.0 were mostly centered on Language-Integrated Query (LINQ) capabilities. LINQ enables queries to be written directly within a C# pro‐ gram and checked statically for correctness, and query both local collections (such as lists or XML documents) or remote data sources (such as a database). The C# 3.0 features added to support LINQ comprised implicitly typed local variables, anonymous types, object initializers, lambda expressions, extension methods, query expressions, and expression trees. Implicitly typed local variables (var keyword, Chapter 2) let you omit the variable type in a declaration statement, allowing the compiler to infer it. This reduces clutter as well as allows anonymous types (Chapter 4), which are simple classes created on the fly that are commonly used in the final output of LINQ queries. You can also implicitly type arrays (Chapter 2). A Brief History of C# | 23 Introducing C# and .NET
- 42. Object initializers (Chapter 3) simplify object construction by allowing you to set properties inline after the constructor call. Object initializers work with both named and anonymous types. Lambda expressions (Chapter 4) are miniature functions created by the compiler on the fly; they are particularly useful in “fluent” LINQ queries (Chapter 8). Extension methods (Chapter 4) extend an existing type with new methods (without altering the type’s definition), making static methods feel like instance methods. LINQ’s query operators are implemented as extension methods. Query expressions (Chapter 8) provide a higher-level syntax for writing LINQ quer‐ ies that can be substantially simpler when working with multiple sequences or range variables. Expression trees (Chapter 8) are miniature code Document Object Models (DOMs) that describe lambda expressions assigned to the special type Expression<TDele gate>. Expression trees make it possible for LINQ queries to execute remotely (e.g., on a database server) because they can be introspected and translated at runtime (e.g., into an SQL statement). C# 3.0 also added automatic properties and partial methods. Automatic properties (Chapter 3) cut the work in writing properties that simply get/set a private backing field by having the compiler do that work automatically. Partial methods (Chapter 3) let an autogenerated partial class provide customizable hooks for manual authoring that “melt away” if unused. What’s New in C# 2.0 The big new features in C# 2 were generics (Chapter 3), nullable value types (Chap‐ ter 4), iterators (Chapter 4), and anonymous methods (the predecessor to lambda expressions). These features paved the way for the introduction of LINQ in C# 3. C# 2 also added support for partial classes and static classes, and a host of minor and miscellaneous features such as the namespace alias qualifier, friend assemblies, and fixed-size buffers. The introduction of generics required a new CLR (CLR 2.0), because generics maintain full type fidelity at runtime. 24 | Chapter 1: Introducing C# and .NET
- 43. 2 C# Language Basics In this chapter, we introduce the basics of the C# language. Almost all of the code listings in this book are available as interactive samples in LINQPad. Working through these sam‐ ples in conjunction with the book accelerates learning in that you can edit the samples and instantly see the results without needing to set up projects and solutions in Visual Studio. To download the samples, in LINQPad, click the Samples tab, and then click “Download more samples.” LINQPad is free— go to http://www.linqpad.net. A First C# Program Following is a program that multiplies 12 by 30 and prints the result, 360, to the screen. The double forward slash indicates that the remainder of a line is a comment: int x = 12 * 30; // Statement 1 System.Console.WriteLine (x); // Statement 2 Our program consists of two statements. Statements in C# execute sequentially and are terminated by a semicolon. The first statement computes the expression 12 * 30 and stores the result in a variable, named x, whose type is a 32-bit integer (int). The second statement calls the WriteLine method on a class called Console, which is defined in a namespace called System. This prints the variable x to a text window on the screen. A method performs a function; a class groups function members and data members to form an object-oriented building block. The Console class groups members that handle command-line input/output (I/O) functionality, such as the WriteLine method. A class is a kind of type, which we examine in “Type Basics” on page 30. 25
- 44. At the outermost level, types are organized into namespaces. Many commonly used types—including the Console class—reside in the System namespace. The .NET libraries are organized into nested namespaces. For example, the System.Text namespace contains types for handling text, and System.IO contain types for input/ output. Qualifying the Console class with the System namespace on every use adds clutter. The using directive lets you avoid this clutter by importing a namespace: using System; // Import the System namespace int x = 12 * 30; Console.WriteLine (x); // No need to specify System. A basic form of code reuse is to write higher-level functions that call lower-level functions. We can refactor our program with a reusable method called FeetToInches that multiplies an integer by 12, as follows: using System; Console.WriteLine (FeetToInches (30)); // 360 Console.WriteLine (FeetToInches (100)); // 1200 int FeetToInches (int feet) { int inches = feet * 12; return inches; } Our method contains a series of statements surrounded by a pair of braces. This is called a statement block. A method can receive input data from the caller by specifying parameters and output data back to the caller by specifying a return type. Our FeetToInches method has a parameter for inputting feet, and a return type for outputting inches: int FeetToInches (int feet) ... The literals 30 and 100 are the arguments passed to the FeetToInches method. If a method doesn’t receive input, use empty parentheses. If it doesn’t return any‐ thing, use the void keyword: using System; SayHello(); void SayHello() { Console.WriteLine ("Hello, world"); } Methods are one of several kinds of functions in C#. Another kind of function we used in our example program was the * operator, which performs multiplication. There are also constructors, properties, events, indexers, and finalizers. 26 | Chapter 2: C# Language Basics
- 45. Compilation The C# compiler compiles source code (a set of files with the .cs extension) into an assembly. An assembly is the unit of packaging and deployment in .NET. An assembly can be either an application or a library. A normal console or Windows application has an entry point, whereas a library does not. The purpose of a library is to be called upon (referenced) by an application or by other libraries. .NET itself is a set of libraries (as well as a runtime environment). Each of the programs in the preceding section began directly with a series of state‐ ments (called top-level statements). The presence of top-level statements implicitly creates an entry point for a console or Windows application. (Without top-level statements, a Main method denotes an application’s entry point—see “Custom Types” on page 31.) Unlike .NET Framework, .NET 6 assemblies never have an .exe extension. The .exe that you see after building a .NET 6 application is a platform-specific native loader responsible for starting your application’s .dll assembly. .NET 6 also allows you to create a self-contained deployment that includes the loader, your assemblies, and the .NET run‐ time—all in a single .exe file. The dotnet tool (dotnet.exe on Windows) helps you to manage .NET source code and binaries from the command line. You can use it to both build and run your program, as an alternative to using an integrated development environment (IDE) such as Visual Studio or Visual Studio Code. You can obtain the dotnet tool either by installing the .NET 6 SDK or by instal‐ ling Visual Studio. Its default location is %ProgramFiles%dotnet on Windows or /usr/bin/dotnet on Ubuntu Linux. To compile an application, the dotnet tool requires a project file as well as one or more C# files. The following command scaffolds a new console project (creates its basic structure): dotnet new Console -n MyFirstProgram This creates a subfolder called MyFirstProgram containing a project file called MyFirstProgram.csproj and a C# file called Program.cs that prints “Hello world”. To build and run your program, run the following command from the MyFirstPro‐ gram folder: dotnet run MyFirstProgram Or, if you just want to build without running: dotnet build MyFirstProgram.csproj The output assembly will be written to a subdirectory under bindebug. We explain assemblies in detail in Chapter 17. A First C# Program | 27 C# Language Basics
- 46. Syntax C# syntax is inspired by C and C++ syntax. In this section, we describe C#’s elements of syntax, using the following program: using System; int x = 12 * 30; Console.WriteLine (x); Identifiers and Keywords Identifiers are names that programmers choose for their classes, methods, variables, and so on. Here are the identifiers in our example program, in the order in which they appear: System x Console WriteLine An identifier must be a whole word, essentially made up of Unicode characters starting with a letter or underscore. C# identifiers are case sensitive. By convention, parameters, local variables, and private fields should be in camel case (e.g., myVaria ble), and all other identifiers should be in Pascal case (e.g., MyMethod). Keywords are names that mean something special to the compiler. There are two keywords in our example program, using and int. Most keywords are reserved, which means that you can’t use them as identifiers. Here is the full list of C# reserved keywords: abstract as base bool break byte case catch char checked class const continue decimal default delegate do double else enum event explicit extern false finally fixed float for foreach goto if implicit in int interface internal is lock long namespace new null object operator out override params private protected public readonly record ref return sbyte sealed short sizeof stackalloc static string struct switch this throw true try typeof uint ulong unchecked unsafe ushort using virtual void volatile while 28 | Chapter 2: C# Language Basics
- 47. If you really want to use an identifier that clashes with a reserved keyword, you can do so by qualifying it with the @ prefix; for instance: int using = 123; // Illegal int @using = 123; // Legal The @ symbol doesn’t form part of the identifier itself. So, @myVariable is the same as myVariable. Contextual keywords Some keywords are contextual, meaning that you also can use them as identifiers— without an @ symbol: add alias and ascending async await by descending dynamic equals from get global group init into join let managed nameof nint not notnull nuint on or orderby partial remove select set unmanaged value var with when where yield With contextual keywords, ambiguity cannot arise within the context in which they are used. Literals, Punctuators, and Operators Literals are primitive pieces of data lexically embedded into the program. The literals we used in our example program are 12 and 30. Punctuators help demarcate the structure of the program. An example is the semi‐ colon, which terminates a statement. Statements can wrap multiple lines: Console.WriteLine (1 + 2 + 3 + 4 + 5 + 6 + 7 + 8 + 9 + 10); An operator transforms and combines expressions. Most operators in C# are deno‐ ted with a symbol, such as the multiplication operator, *. We discuss operators in more detail later in this chapter. These are the operators we used in our example program: = * . () A period denotes a member of something (or a decimal point with numeric literals). Parentheses are used when declaring or calling a method; empty parentheses are used when the method accepts no arguments. (Parentheses also have other purposes that you’ll see later in this chapter.) An equals sign performs assignment. (The double equals sign, ==, performs equality comparison, as you’ll see later.) Syntax | 29 C# Language Basics
- 48. Comments C# offers two different styles of source-code documentation: single-line comments and multiline comments. A single-line comment begins with a double forward slash and continues until the end of the line; for example: int x = 3; // Comment about assigning 3 to x A multiline comment begins with /* and ends with */; for example: int x = 3; /* This is a comment that spans two lines */ Comments can embed XML documentation tags, which we explain in “XML Docu‐ mentation” on page 252. Type Basics A type defines the blueprint for a value. In this example, we use two literals of type int with values 12 and 30. We also declare a variable of type int whose name is x: int x = 12 * 30; Console.WriteLine (x); Because most of the code listings in this book require types from the System namespace, we will omit “using System” from now on, unless we’re illustrating a concept relating to namespaces. A variable denotes a storage location that can contain different values over time. In contrast, a constant always represents the same value (more on this later): const int y = 360; All values in C# are instances of a type. The meaning of a value and the set of possible values a variable can have are determined by its type. Predefined Type Examples Predefined types are types that are specially supported by the compiler. The int type is a predefined type for representing the set of integers that fit into 32 bits of memory, from −231 to 231 − 1, and is the default type for numeric literals within this range. You can perform functions such as arithmetic with instances of the int type, as follows: int x = 12 * 30; Another predefined C# type is string. The string type represents a sequence of characters, such as “.NET” or “http://oreilly.com”. You can work with strings by calling functions on them, as follows: string message = "Hello world"; string upperMessage = message.ToUpper(); Console.WriteLine (upperMessage); // HELLO WORLD 30 | Chapter 2: C# Language Basics
- 49. int x = 2022; message = message + x.ToString(); Console.WriteLine (message); // Hello world2022 In this example, we called x.ToString() to obtain a string representation of the integer x. You can call ToString() on a variable of almost any type. The predefined bool type has exactly two possible values: true and false. The bool type is commonly used with an if statement to conditionally branch execution flow: bool simpleVar = false; if (simpleVar) Console.WriteLine ("This will not print"); int x = 5000; bool lessThanAMile = x < 5280; if (lessThanAMile) Console.WriteLine ("This will print"); In C#, predefined types (also referred to as built-in types) are recognized with a C# keyword. The System namespace in .NET contains many important types that are not prede‐ fined by C# (e.g., DateTime). Custom Types Just as we can write our own methods, we can write our own types. In this next example, we define a custom type named UnitConverter—a class that serves as a blueprint for unit conversions: UnitConverter feetToInchesConverter = new UnitConverter (12); UnitConverter milesToFeetConverter = new UnitConverter (5280); Console.WriteLine (feetToInchesConverter.Convert(30)); // 360 Console.WriteLine (feetToInchesConverter.Convert(100)); // 1200 Console.WriteLine (feetToInchesConverter.Convert( milesToFeetConverter.Convert(1))); // 63360 public class UnitConverter { int ratio; // Field public UnitConverter (int unitRatio) // Constructor { ratio = unitRatio; } public int Convert (int unit) // Method { return unit * ratio; Type Basics | 31 C# Language Basics
- 50. } } In this example, our class definition appears in the same file as our top-level statements. This is legal—as long as the top- level statements appear first—and is acceptable when writ‐ ing small test programs. With larger programs, the standard approach is to put the class definition in a separate file such as UnitConverter.cs. Members of a type A type contains data members and function members. The data member of UnitConverter is the field called ratio. The function members of UnitConverter are the Convert method and the UnitConverter’s constructor. Symmetry of predefined types and custom types A beautiful aspect of C# is that predefined types and custom types have few differences. The predefined int type serves as a blueprint for integers. It holds data—32 bits—and provides function members that use that data, such as ToString. Similarly, our custom UnitConverter type acts as a blueprint for unit conversions. It holds data—the ratio—and provides function members to use that data. Constructors and instantiation Data is created by instantiating a type. Predefined types can be instantiated simply by using a literal such as 12 or "Hello world". The new operator creates instances of a custom type. We created and declared an instance of the UnitConverter type with this statement: UnitConverter feetToInchesConverter = new UnitConverter (12); Immediately after the new operator instantiates an object, the object’s constructor is called to perform initialization. A constructor is defined like a method, except that the method name and return type are reduced to the name of the enclosing type: public UnitConverter (int unitRatio) { ratio = unitRatio; } Instance versus static members The data members and function members that operate on the instance of the type are called instance members. The UnitConverter’s Convert method and the int’s ToString method are examples of instance members. By default, members are instance members. Data members and function members that don’t operate on the instance of the type can be marked as static. To refer to a static member from outside its type, you specify its type name rather than an instance. An example is the WriteLine method of the Console class. Because this is static, we call Console.WriteLine() and not new Console().WriteLine(). 32 | Chapter 2: C# Language Basics
- 51. (The Console class is actually declared as a static class, which means that all of its members are static, and you can never create instances of a Console.) In the following code, the instance field Name pertains to an instance of a particular Panda, whereas Population pertains to the set of all Panda instances. We create two instances of the Panda, print their names, and then print the total population: Panda p1 = new Panda ("Pan Dee"); Panda p2 = new Panda ("Pan Dah"); Console.WriteLine (p1.Name); // Pan Dee Console.WriteLine (p2.Name); // Pan Dah Console.WriteLine (Panda.Population); // 2 public class Panda { public string Name; // Instance field public static int Population; // Static field public Panda (string n) // Constructor { Name = n; // Assign the instance field Population = Population + 1; // Increment the static Population field } } Attempting to evaluate p1.Population or Panda.Name will generate a compile-time error. The public keyword The public keyword exposes members to other classes. In this example, if the Name field in Panda was not marked as public, it would be private and could not be accessed from outside the class. Marking a member public is how a type communi‐ cates: “Here is what I want other types to see—everything else is my own private implementation details.” In object-oriented terms, we say that the public members encapsulate the private members of the class. Defining namespaces Particularly with larger programs, it makes sense to organize types into namespaces. Here’s how to define the Panda class inside a namespace called Animals: using System; using Animals; Panda p = new Panda ("Pan Dee"); Console.WriteLine (p.Name); namespace Animals { public class Panda Type Basics | 33 C# Language Basics
- 52. { ... } } In this example, we also imported the Animals namespace so that our top-level statements could access its types without qualification. Without that import, we’d need to do this: Animals.Panda p = new Animals.Panda ("Pan Dee"); We cover namespaces in detail at the end of this chapter (see “Namespaces” on page 87). Defining a Main method All of our examples, so far, have used top-level statements (a feature introduced in C# 9). Without top-level statements, a simple console or Windows application looks like this: using System; class Program { static void Main() // Program entry point { int x = 12 * 30; Console.WriteLine (x); } } In the absence of top-level statements, C# looks for a static method called Main, which becomes the entry point. The Main method can be defined inside any class (and only one Main method can exist). Should your Main method need to access private members of a particular class, defining a Main method inside that class can be simpler than using top-level statements. The Main method can optionally return an integer (rather than void) in order to return a value to the execution environment (where a nonzero value typically indicates an error). The Main method can also optionally accept an array of strings as a parameter (that will be populated with any arguments passed to the executable). For example: static int Main (string[] args) {...} An array (such as string[]) represents a fixed number of elements of a particular type. Arrays are specified by placing square brackets after the element type. We describe them in “Arrays” on page 54. 34 | Chapter 2: C# Language Basics
- 53. (The Main method can also be declared async and return a Task or Task<int> in support of asynchronous programming, which we cover in Chapter 14.) Top-Level Statements Top-level statements (introduced in C# 9) let you avoid the baggage of a static Main method and a containing class. A file with top-level statements comprises three parts, in this order: 1. (Optionally) using directives 1. 2. A series of statements, optionally mixed with method declarations 2. 3. (Optionally) Type and namespace declarations 3. For example: using System; // Part 1 Console.WriteLine ("Hello, world"); // Part 2 void SomeMethod1() { ... } // Part 2 Console.WriteLine ("Hello again!"); // Part 2 void SomeMethod2() { ... } // Part 2 class SomeClass { ... } // Part 3 namespace SomeNamespace { ... } // Part 3 Because the CLR doesn’t explicitly support top-level statements, the compiler trans‐ lates your code into something like this: using System; // Part 1 static class Program$ // Special compiler-generated name { static void Main$ (string[] args) // Compiler-generated name { Console.WriteLine ("Hello, world"); // Part 2 void SomeMethod1() { ... } // Part 2 Console.WriteLine ("Hello again!"); // Part 2 void SomeMethod2() { ... } // Part 2 } } class SomeClass { ... } // Part 3 namespace SomeNamespace { ... } // Part 3 Notice that everything in Part 2 is wrapped inside the main method. This means that SomeMethod1 and SomeMethod2 act as local methods. We discuss the full implica‐ tions in “Local methods” on page 98, the most important being that local methods (unless declared as static) can access variables declared within the containing method: int x = 3; LocalMethod(); void LocalMethod() { Console.WriteLine (x); } // We can access x Type Basics | 35 C# Language Basics
- 54. 1 A minor caveat is that very large long values lose some precision when converted to double. Another consequence is that top-level methods cannot be accessed from other classes or types. Top-level statements can optionally return an integer value to the caller and access a “magic” variable of type string[] called args, corresponding to command-line arguments passed by the caller. As a program can have only one entry point, there can be at most one file with top-level statements in a C# project. Types and Conversions C# can convert between instances of compatible types. A conversion always creates a new value from an existing one. Conversions can be either implicit or explicit: implicit conversions happen automatically, and explicit conversions require a cast. In the following example, we implicitly convert an int to a long type (which has twice the bit capacity of an int) and explicitly cast an int to a short type (which has half the bit capacity of an int): int x = 12345; // int is a 32-bit integer long y = x; // Implicit conversion to 64-bit integer short z = (short)x; // Explicit conversion to 16-bit integer Implicit conversions are allowed when both of the following are true: • The compiler can guarantee that they will always succeed. • • No information is lost in conversion.1 • Conversely, explicit conversions are required when one of the following is true: • The compiler cannot guarantee that they will always succeed. • • Information might be lost during conversion. • (If the compiler can determine that a conversion will always fail, both kinds of conversion are prohibited. Conversions that involve generics can also fail in certain conditions—see “Type Parameters and Conversions” on page 152.) The numeric conversions that we just saw are built into the language. C# also supports reference conversions and boxing conversions (see Chapter 3) as well as custom conversions (see “Operator Overloading” on page 239). The compiler doesn’t enforce the aforementioned rules with custom conversions, so it’s possible for badly designed types to behave otherwise. 36 | Chapter 2: C# Language Basics
- 55. Value Types Versus Reference Types All C# types fall into the following categories: • Value types • • Reference types • • Generic type parameters • • Pointer types • In this section, we describe value types and reference types. We cover generic type parameters in “Generics” on page 145 and pointer types in “Unsafe Code and Pointers” on page 243. Value types comprise most built-in types (specifically, all numeric types, the char type, and the bool type) as well as custom struct and enum types. Reference types comprise all class, array, delegate, and interface types. (This includes the predefined string type.) The fundamental difference between value types and reference types is how they are handled in memory. Value types The content of a value type variable or constant is simply a value. For example, the content of the built-in value type, int, is 32 bits of data. You can define a custom value type with the struct keyword (see Figure 2-1): public struct Point { public int X; public int Y; } Or more tersely: public struct Point { public int X, Y; } Figure 2-1. A value-type instance in memory The assignment of a value-type instance always copies the instance; for example: Point p1 = new Point(); p1.X = 7; Point p2 = p1; // Assignment causes copy Console.WriteLine (p1.X); // 7 Console.WriteLine (p2.X); // 7 Type Basics | 37 C# Language Basics
- 56. p1.X = 9; // Change p1.X Console.WriteLine (p1.X); // 9 Console.WriteLine (p2.X); // 7 Figure 2-2 shows that p1 and p2 have independent storage. Figure 2-2. Assignment copies a value-type instance Reference types A reference type is more complex than a value type, having two parts: an object and the reference to that object. The content of a reference-type variable or constant is a reference to an object that contains the value. Here is the Point type from our previous example rewritten as a class rather than a struct (shown in Figure 2-3): public class Point { public int X, Y; } Figure 2-3. A reference-type instance in memory Assigning a reference-type variable copies the reference, not the object instance. This allows multiple variables to refer to the same object—something not ordinarily possible with value types. If we repeat the previous example, but with Point now a class, an operation to p1 affects p2: Point p1 = new Point(); p1.X = 7; Point p2 = p1; // Copies p1 reference Console.WriteLine (p1.X); // 7 Console.WriteLine (p2.X); // 7 p1.X = 9; // Change p1.X Console.WriteLine (p1.X); // 9 Console.WriteLine (p2.X); // 9 38 | Chapter 2: C# Language Basics
- 57. Figure 2-4 shows that p1 and p2 are two references that point to the same object. Figure 2-4. Assignment copies a reference Null A reference can be assigned the literal null, indicating that the reference points to no object: Point p = null; Console.WriteLine (p == null); // True // The following line generates a runtime error // (a NullReferenceException is thrown): Console.WriteLine (p.X); class Point {...} In “Nullable Reference Types” on page 200, we describe a feature of C# that helps to reduce accidental NullReference Exception errors. In contrast, a value type cannot ordinarily have a null value: Point p = null; // Compile-time error int x = null; // Compile-time error struct Point {...} C# also has a construct called nullable value types for repre‐ senting value-type nulls. For more information, see “Nullable Value Types” on page 194. Storage overhead Value-type instances occupy precisely the memory required to store their fields. In this example, Point takes 8 bytes of memory: struct Point { int x; // 4 bytes int y; // 4 bytes } Type Basics | 39 C# Language Basics
- 58. Technically, the CLR positions fields within the type at an address that’s a multiple of the fields’ size (up to a maximum of 8 bytes). Thus, the following actually consumes 16 bytes of memory (with the 7 bytes following the first field “wasted”): struct A { byte b; long l; } You can override this behavior by applying the StructLayout attribute (see “Mapping a Struct to Unmanaged Memory” on page 973). Reference types require separate allocations of memory for the reference and object. The object consumes as many bytes as its fields, plus additional administrative overhead. The precise overhead is intrinsically private to the implementation of the .NET runtime, but at minimum, the overhead is 8 bytes, used to store a key to the object’s type as well as temporary information such as its lock state for multithreading and a flag to indicate whether it has been fixed from movement by the garbage collector. Each reference to an object requires an extra 4 or 8 bytes, depending on whether the .NET runtime is running on a 32- or 64-bit platform. Predefined Type Taxonomy The predefined types in C# are as follows: Value types • Numeric • — Signed integer (sbyte, short, int, long) — — Unsigned integer (byte, ushort, uint, ulong) — — Real number (float, double, decimal) — • Logical (bool) • • Character (char) • Reference types • String (string) • • Object (object) • Predefined types in C# alias .NET types in the System namespace. There is only a syntactic difference between these two statements: int i = 5; System.Int32 i = 5; The set of predefined value types excluding decimal are known as primitive types in the CLR. Primitive types are so called because they are supported directly via instructions in compiled code, and this usually translates to direct support on the underlying processor; for example: 40 | Chapter 2: C# Language Basics
- 59. // Underlying hexadecimal representation int i = 7; // 0x7 bool b = true; // 0x1 char c = 'A'; // 0x41 float f = 0.5f; // uses IEEE floating-point encoding The System.IntPtr and System.UIntPtr types are also primitive (see Chapter 24). Numeric Types C# has the predefined numeric types shown in Table 2-1. Table 2-1. Predefined numeric types in C# C# type System type Suffix Size Range Integral—signed sbyte SByte 8 bits –27 to 27–1 short Int16 16 bits –215 to 215–1 int Int32 32 bits –231 to 231–1 long Int64 L 64 bits –263 to 263–1 nint IntPtr 32/64 bits Integral—unsigned byte Byte 8 bits 0 to 28–1 ushort UInt16 16 bits 0 to 216–1 uint UInt32 U 32 bits 0 to 232–1 ulong UInt64 UL 64 bits 0 to 264–1 unint UIntPtr 32/64 bits Real float Single F 32 bits ± (~10–45 to 1038) double Double D 64 bits ± (~10–324 to 10308) decimal Decimal M 128 bits ± (~10–28 to 1028) Of the integral types, int and long are first-class citizens and are favored by both C# and the runtime. The other integral types are typically used for interoperability or when space efficiency is paramount. The nint and nuint native-sized integer types (introduced in C# 9) are most useful in helping with pointer arithmetic, so we will describe these in a later chapter (see “Native-Sized Integers” on page 246). Numeric Types | 41 C# Language Basics
- 60. 2 Technically, decimal is a floating-point type, too, although it’s not referred to as such in the C# language specification. Of the real number types, float and double are called floating-point types2 and are typically used for scientific and graphical calculations. The decimal type is typically used for financial calculations, for which base-10-accurate arithmetic and high precision are required. From .NET 5, there is a 16-bit floating point type called Half. This is intended mainly for interoperating with graphics card processors and does not have native support in most CPUs. Half is not a primitive CLR type and does not have special language support in C#. Numeric Literals Integral-type literals can use decimal or hexadecimal notation; hexadecimal is deno‐ ted with the 0x prefix. For example: int x = 127; long y = 0x7F; You can insert an underscore anywhere within a numeric literal to make it more readable: int million = 1_000_000; You can specify numbers in binary with the 0b prefix: var b = 0b1010_1011_1100_1101_1110_1111; Real literals can use decimal and/or exponential notation: double d = 1.5; double million = 1E06; Numeric literal type inference By default, the compiler infers a numeric literal to be either double or an integral type: • If the literal contains a decimal point or the exponential symbol (E), it is a • double. • Otherwise, the literal’s type is the first type in this list that can fit the literal’s • value: int, uint, long, and ulong. For example: Console.WriteLine ( 1.0.GetType()); // Double (double) Console.WriteLine ( 1E06.GetType()); // Double (double) Console.WriteLine ( 1.GetType()); // Int32 (int) 42 | Chapter 2: C# Language Basics
- 61. Console.WriteLine ( 0xF0000000.GetType()); // UInt32 (uint) Console.WriteLine (0x100000000.GetType()); // Int64 (long) Numeric suffixes Numeric suffixes explicitly define the type of a literal. Suffixes can be either lower‐ case or uppercase, and are as follows: Category C# type Example F float float f = 1.0F; D double double d = 1D; M decimal decimal d = 1.0M; U uint uint i = 1U; L long long i = 1L; UL ulong ulong i = 1UL; The suffixes U and L are rarely necessary because the uint, long, and ulong types can nearly always be either inferred or implicitly converted from int: long i = 5; // Implicit lossless conversion from int literal to long The D suffix is technically redundant in that all literals with a decimal point are inferred to be double. And you can always add a decimal point to a numeric literal: double x = 4.0; The F and M suffixes are the most useful and should always be applied when specifying float or decimal literals. Without the F suffix, the following line would not compile, because 4.5 would be inferred to be of type double, which has no implicit conversion to float: float f = 4.5F; The same principle is true for a decimal literal: decimal d = -1.23M; // Will not compile without the M suffix. We describe the semantics of numeric conversions in detail in the following section. Numeric Conversions Converting between integral types Integral type conversions are implicit when the destination type can represent every possible value of the source type. Otherwise, an explicit conversion is required; for example: int x = 12345; // int is a 32-bit integer long y = x; // Implicit conversion to 64-bit integral type short z = (short)x; // Explicit conversion to 16-bit integral type Numeric Types | 43 C# Language Basics
- 62. Converting between floating-point types A float can be implicitly converted to a double given that a double can represent every possible value of a float. The reverse conversion must be explicit. Converting between floating-point and integral types All integral types can be implicitly converted to all floating-point types: int i = 1; float f = i; The reverse conversion must be explicit: int i2 = (int)f; When you cast from a floating-point number to an integral type, any fractional portion is truncated; no rounding is per‐ formed. The static class System.Convert provides methods that round while converting between various numeric types (see Chapter 6). Implicitly converting a large integral type to a floating-point type preserves magni‐ tude but can occasionally lose precision. This is because floating-point types always have more magnitude than integral types but can have less precision. Rewriting our example with a larger number demonstrates this: int i1 = 100000001; float f = i1; // Magnitude preserved, precision lost int i2 = (int)f; // 100000000 Decimal conversions All integral types can be implicitly converted to the decimal type given that a decimal can represent every possible C# integral-type value. All other numeric conversions to and from a decimal type must be explicit because they introduce the possibility of either a value being out of range or precision being lost. Arithmetic Operators The arithmetic operators (+, -, *, /, %) are defined for all numeric types except the 8- and 16-bit integral types: + Addition - Subtraction * Multiplication / Division % Remainder after division 44 | Chapter 2: C# Language Basics
- 63. Increment and Decrement Operators The increment and decrement operators (++, --, respectively) increment and decre‐ ment numeric types by 1. The operator can either follow or precede the variable, depending on whether you want its value before or after the increment/decrement; for example: int x = 0, y = 0; Console.WriteLine (x++); // Outputs 0; x is now 1 Console.WriteLine (++y); // Outputs 1; y is now 1 Specialized Operations on Integral Types The integral types are int, uint, long, ulong, short, ushort, byte, and sbyte. Division Division operations on integral types always eliminate the remainder (round toward zero). Dividing by a variable whose value is zero generates a runtime error (a DivideByZeroException): int a = 2 / 3; // 0 int b = 0; int c = 5 / b; // throws DivideByZeroException Dividing by the literal or constant 0 generates a compile-time error. Overflow At runtime, arithmetic operations on integral types can overflow. By default, this happens silently—no exception is thrown, and the result exhibits “wraparound” behavior, as though the computation were done on a larger integer type and the extra significant bits discarded. For example, decrementing the minimum possible int value results in the maximum possible int value: int a = int.MinValue; a--; Console.WriteLine (a == int.MaxValue); // True Overflow check operators The checked operator instructs the runtime to generate an OverflowException rather than overflowing silently when an integral-type expression or statement exceeds the arithmetic limits of that type. The checked operator affects expressions with the ++, −−, +, − (binary and unary), *, /, and explicit conversion operators between integral types. Overflow checking incurs a small performance cost. The checked operator has no effect on the double and float types (which overflow to special “infinite” values, as you’ll see soon) and no effect on the decimal type (which is always checked). Numeric Types | 45 C# Language Basics
- 64. You can use checked around either an expression or a statement block: int a = 1000000; int b = 1000000; int c = checked (a * b); // Checks just the expression. checked // Checks all expressions { // in statement block. ... c = a * b; ... } You can make arithmetic overflow checking the default for all expressions in a program by selecting the “checked” option at the project level (in Visual Studio, go to Advanced Build Settings). If you then need to disable overflow checking just for specific expressions or statements, you can do so with the unchecked operator. For example, the following code will not throw exceptions—even if the project’s “checked” option is selected: int x = int.MaxValue; int y = unchecked (x + 1); unchecked { int z = x + 1; } Overflow checking for constant expressions Regardless of the “checked” project setting, expressions evaluated at compile time are always overflow-checked—unless you apply the unchecked operator: int x = int.MaxValue + 1; // Compile-time error int y = unchecked (int.MaxValue + 1); // No errors Bitwise operators C# supports the following bitwise operators: Operator Meaning Sample expression Result ~ Complement ~0xfU 0xfffffff0U & And 0xf0 & 0x33 0x30 | Or 0xf0 | 0x33 0xf3 ^ Exclusive Or 0xff00 ^ 0x0ff0 0xf0f0 << Shift left 0x20 << 2 0x80 >> Shift right 0x20 >> 1 0x10 From .NET 6, additional bitwise operations are exposed via a new class called BitOperations in the System.Numerics namespace (see “BitOperations” on page 320). 46 | Chapter 2: C# Language Basics
- 65. 8- and 16-Bit Integral Types The 8- and 16-bit integral types are byte, sbyte, short, and ushort. These types lack their own arithmetic operators, so C# implicitly converts them to larger types as required. This can cause a compile-time error when trying to assign the result back to a small integral type: short x = 1, y = 1; short z = x + y; // Compile-time error In this case, x and y are implicitly converted to int so that the addition can be performed. This means that the result is also an int, which cannot be implicitly cast back to a short (because it could cause loss of data). To make this compile, you must add an explicit cast: short z = (short) (x + y); // OK Special Float and Double Values Unlike integral types, floating-point types have values that certain operations treat specially. These special values are NaN (Not a Number), +∞, −∞, and −0. The float and double classes have constants for NaN, +∞, and −∞, as well as other values (MaxValue, MinValue, and Epsilon); for example: Console.WriteLine (double.NegativeInfinity); // -Infinity The constants that represent special values for double and float are as follows: Special value Double constant Float constant NaN double.NaN float.NaN +∞ double.PositiveInfinity float.PositiveInfinity −∞ double.NegativeInfinity float.NegativeInfinity −0 −0.0 −0.0f Dividing a nonzero number by zero results in an infinite value: Console.WriteLine ( 1.0 / 0.0); // Infinity Console.WriteLine (−1.0 / 0.0); // -Infinity Console.WriteLine ( 1.0 / −0.0); // -Infinity Console.WriteLine (−1.0 / −0.0); // Infinity Dividing zero by zero, or subtracting infinity from infinity, results in a NaN: Console.WriteLine ( 0.0 / 0.0); // NaN Console.WriteLine ((1.0 / 0.0) − (1.0 / 0.0)); // NaN When using ==, a NaN value is never equal to another value, even another NaN value: Console.WriteLine (0.0 / 0.0 == double.NaN); // False Numeric Types | 47 C# Language Basics
- 66. To test whether a value is NaN, you must use the float.IsNaN or double.IsNaN method: Console.WriteLine (double.IsNaN (0.0 / 0.0)); // True When using object.Equals, however, two NaN values are equal: Console.WriteLine (object.Equals (0.0 / 0.0, double.NaN)); // True NaNs are sometimes useful in representing special values. In Windows Presentation Foundation (WPF), double.NaN repre‐ sents a measurement whose value is “Automatic.” Another way to represent such a value is with a nullable type (Chapter 4); another is with a custom struct that wraps a numeric type and adds an additional field (Chapter 3). float and double follow the specification of the IEEE 754 format types, supported natively by almost all processors. You can find detailed information on the behavior of these types at http://www.ieee.org. double Versus decimal double is useful for scientific computations (such as computing spatial coordinates). decimal is useful for financial computations and values that are “human-made” rather than the result of real-world measurements. Here’s a summary of the differences: Category double decimal Internal representation Base 2 Base 10 Decimal precision 15–16 significant figures 28–29 significant figures Range ±(~10−324 to ~10308) ±(~10−28 to ~1028) Special values +0, −0, +∞, −∞, and NaN None Speed Native to processor Non-native to processor (about 10 times slower than double) Real Number Rounding Errors float and double internally represent numbers in base 2. For this reason, only numbers expressible in base 2 are represented precisely. Practically, this means most literals with a fractional component (which are in base 10) will not be represented precisely; for example: float x = 0.1f; // Not quite 0.1 Console.WriteLine (x + x + x + x + x + x + x + x + x + x); // 1.0000001 This is why float and double are bad for financial calculations. In contrast, decimal works in base 10 and so can precisely represent numbers expressible in base 10 (as well as its factors, base 2 and base 5). Because real literals are in base 10, decimal can precisely represent numbers such as 0.1. However, neither double nor 48 | Chapter 2: C# Language Basics
- 67. 3 It’s possible to overload these operators (Chapter 4) such that they return a non-bool type, but this is almost never done in practice. decimal can precisely represent a fractional number whose base 10 representation is recurring: decimal m = 1M / 6M; // 0.1666666666666666666666666667M double d = 1.0 / 6.0; // 0.16666666666666666 This leads to accumulated rounding errors: decimal notQuiteWholeM = m+m+m+m+m+m; // 1.0000000000000000000000000002M double notQuiteWholeD = d+d+d+d+d+d; // 0.99999999999999989 which break equality and comparison operations: Console.WriteLine (notQuiteWholeM == 1M); // False Console.WriteLine (notQuiteWholeD < 1.0); // True Boolean Type and Operators C#’s bool type (aliasing the System.Boolean type) is a logical value that can be assigned the literal true or false. Although a Boolean value requires only one bit of storage, the runtime will use one byte of memory because this is the minimum chunk that the runtime and processor can efficiently work with. To avoid space inefficiency in the case of arrays, .NET provides a BitArray class in the System.Collections namespace that is designed to use just one bit per Boolean value. bool Conversions No casting conversions can be made from the bool type to numeric types, or vice versa. Equality and Comparison Operators == and != test for equality and inequality of any type but always return a bool value.3 Value types typically have a very simple notion of equality: int x = 1; int y = 2; int z = 1; Console.WriteLine (x == y); // False Console.WriteLine (x == z); // True For reference types, equality, by default, is based on reference, as opposed to the actual value of the underlying object (more on this in Chapter 6): Dude d1 = new Dude ("John"); Dude d2 = new Dude ("John"); Console.WriteLine (d1 == d2); // False Dude d3 = d1; Boolean Type and Operators | 49 C# Language Basics
- 68. Console.WriteLine (d1 == d3); // True public class Dude { public string Name; public Dude (string n) { Name = n; } } The equality and comparison operators, ==, !=, <, >, >=, and <=, work for all numeric types, but you should use them with caution with real numbers (as we saw in “Real Number Rounding Errors” on page 48). The comparison operators also work on enum type members by comparing their underlying integral-type values. We describe this in “Enums” on page 140. We explain the equality and comparison operators in greater detail in “Operator Overloading” on page 239, “Equality Comparison” on page 324, and “Order Com‐ parison” on page 335. Conditional Operators The && and || operators test for and and or conditions. They are frequently used in conjunction with the ! operator, which expresses not. In the following example, the UseUmbrella method returns true if it’s rainy or sunny (to protect us from the rain or the sun), as long as it’s not also windy (umbrellas are useless in the wind): static bool UseUmbrella (bool rainy, bool sunny, bool windy) { return !windy && (rainy || sunny); } The && and || operators short-circuit evaluation when possible. In the preceding example, if it is windy, the expression (rainy || sunny) is not even evaluated. Short-circuiting is essential in allowing expressions such as the following to run without throwing a NullReferenceException: if (sb != null && sb.Length > 0) ... The & and | operators also test for and and or conditions: return !windy & (rainy | sunny); The difference is that they do not short-circuit. For this reason, they are rarely used in place of conditional operators. Unlike in C and C++, the & and | operators perform (non- short-circuiting) Boolean comparisons when applied to bool expressions. The & and | operators perform bitwise operations only when applied to numbers. 50 | Chapter 2: C# Language Basics
- 69. Conditional operator (ternary operator) The conditional operator (more commonly called the ternary operator because it’s the only operator that takes three operands) has the form q ? a : b; thus, if condition q is true, a is evaluated, otherwise b is evaluated: static int Max (int a, int b) { return (a > b) ? a : b; } The conditional operator is particularly useful in LINQ expressions (Chapter 8). Strings and Characters C#’s char type (aliasing the System.Char type) represents a Unicode character and occupies 2 bytes (UTF-16). A char literal is specified within single quotes: char c = 'A'; // Simple character Escape sequences express characters that cannot be expressed or interpreted literally. An escape sequence is a backslash followed by a character with a special meaning; for example: char newLine = 'n'; char backSlash = ''; Table 2-2 shows the escape sequence characters. Table 2-2. Escape sequence characters Char Meaning Value ' Single quote 0x0027 " Double quote 0x0022 Backslash 0x005C 0 Null 0x0000 a Alert 0x0007 b Backspace 0x0008 f Form feed 0x000C n New line 0x000A r Carriage return 0x000D t Horizontal tab 0x0009 v Vertical tab 0x000B The u (or x) escape sequence lets you specify any Unicode character via its four-digit hexadecimal code: Strings and Characters | 51 C# Language Basics
- 70. char copyrightSymbol = 'u00A9'; char omegaSymbol = 'u03A9'; char newLine = 'u000A'; Char Conversions An implicit conversion from a char to a numeric type works for the numeric types that can accommodate an unsigned short. For other numeric types, an explicit conversion is required. String Type C#’s string type (aliasing the System.String type, covered in depth in Chapter 6) represents an immutable (unmodifiable) sequence of Unicode characters. A string literal is specified within double quotes: string a = "Heat"; string is a reference type rather than a value type. Its equality operators, however, follow value-type semantics: string a = "test"; string b = "test"; Console.Write (a == b); // True The escape sequences that are valid for char literals also work inside strings: string a = "Here's a tab:t"; The cost of this is that whenever you need a literal backslash, you must write it twice: string a1 = "serverfilesharehelloworld.cs"; To avoid this problem, C# allows verbatim string literals. A verbatim string literal is prefixed with @ and does not support escape sequences. The following verbatim string is identical to the preceding one: string a2 = @"serverfilesharehelloworld.cs"; A verbatim string literal can also span multiple lines: string escaped = "First LinernSecond Line"; string verbatim = @"First Line Second Line"; // True if your text editor uses CR-LF line separators: Console.WriteLine (escaped == verbatim); You can include the double-quote character in a verbatim literal by writing it twice: string xml = @"<customer id=""123""></customer>"; 52 | Chapter 2: C# Language Basics
- 71. String concatenation The + operator concatenates two strings: string s = "a" + "b"; One of the operands might be a nonstring value, in which case ToString is called on that value: string s = "a" + 5; // a5 Using the + operator repeatedly to build up a string is inefficient: a better solution is to use the System.Text.StringBuilder type (described in Chapter 6). String interpolation A string preceded with the $ character is called an interpolated string. Interpolated strings can include expressions enclosed in braces: int x = 4; Console.Write ($"A square has {x} sides"); // Prints: A square has 4 sides Any valid C# expression of any type can appear within the braces, and C# will convert the expression to a string by calling its ToString method or equivalent. You can change the formatting by appending the expression with a colon and a format string (format strings are described in “String.Format and composite format strings” on page 276): string s = $"255 in hex is {byte.MaxValue:X2}"; // X2 = 2-digit hexadecimal // Evaluates to "255 in hex is FF" Should you need to use a colon for another purpose (such as a ternary condi‐ tional operator, which we’ll cover later), you must wrap the entire expression in parentheses: bool b = true; Console.WriteLine ($"The answer in binary is {(b ? 1 : 0)}"); Interpolated strings must complete on a single line, unless you also specify the verbatim string operator: int x = 2; // Note that $ must appear before @ prior to C# 8: string s = $@"this interpolation spans { x} lines"; To include a brace literal in an interpolated string, repeat the desired brace character. String comparisons string does not support < and > operators for comparisons. You must use the string’s CompareTo method, described in Chapter 6. Strings and Characters | 53 C# Language Basics
- 72. Constant interpolated strings (C# 10) From C# 10, interpolated strings can be constants, as long as the interpolated values are constants: const string greeting = "Hello"; const string message = $"{greeting}, world"; Arrays An array represents a fixed number of variables (called elements) of a particular type. The elements in an array are always stored in a contiguous block of memory, providing highly efficient access. An array is denoted with square brackets after the element type: char[] vowels = new char[5]; // Declare an array of 5 characters Square brackets also index the array, accessing a particular element by position: vowels[0] = 'a'; vowels[1] = 'e'; vowels[2] = 'i'; vowels[3] = 'o'; vowels[4] = 'u'; Console.WriteLine (vowels[1]); // e This prints “e” because array indexes start at 0. You can use a for loop statement to iterate through each element in the array. The for loop in this example cycles the integer i from 0 to 4: for (int i = 0; i < vowels.Length; i++) Console.Write (vowels[i]); // aeiou The Length property of an array returns the number of elements in the array. After an array has been created, you cannot change its length. The System.Collec tion namespace and subnamespaces provide higher-level data structures, such as dynamically sized arrays and dictionaries. An array initialization expression lets you declare and populate an array in a single step: char[] vowels = new char[] {'a','e','i','o','u'}; Or simply: char[] vowels = {'a','e','i','o','u'}; All arrays inherit from the System.Array class, providing common services for all arrays. These members include methods to get and set elements regardless of the array type. We describe them in “The Array Class” on page 355. 54 | Chapter 2: C# Language Basics
- 73. Default Element Initialization Creating an array always preinitializes the elements with default values. The default value for a type is the result of a bitwise zeroing of memory. For example, consider creating an array of integers. Because int is a value type, this allocates 1,000 integers in one contiguous block of memory. The default value for each element will be 0: int[] a = new int[1000]; Console.Write (a[123]); // 0 Value types versus reference types Whether an array element type is a value type or a reference type has important performance implications. When the element type is a value type, each element value is allocated as part of the array, as shown here: Point[] a = new Point[1000]; int x = a[500].X; // 0 public struct Point { public int X, Y; } Had Point been a class, creating the array would have merely allocated 1,000 null references: Point[] a = new Point[1000]; int x = a[500].X; // Runtime error, NullReferenceException public class Point { public int X, Y; } To avoid this error, we must explicitly instantiate 1,000 Points after instantiating the array: Point[] a = new Point[1000]; for (int i = 0; i < a.Length; i++) // Iterate i from 0 to 999 a[i] = new Point(); // Set array element i with new point An array itself is always a reference type object, regardless of the element type. For instance, the following is legal: int[] a = null; Indices and Ranges Indices and ranges (introduced in C# 8) simplify working with elements or portions of an array. Indices and ranges also work with the CLR types Span<T> and ReadOnlySpan<T> (see Chapter 23). You can also make your own types work with indices and ranges, by defining an indexer of type Index or Range (see “Indexers” on page 109). Arrays | 55 C# Language Basics
- 74. Indices Indices let you refer to elements relative to the end of an array, with the ^ operator. ^1 refers to the last element, ^2 refers to the second-to-last element, and so on: char[] vowels = new char[] {'a','e','i','o','u'}; char lastElement = vowels [^1]; // 'u' char secondToLast = vowels [^2]; // 'o' (^0 equals the length of the array, so vowels[^0] generates an error.) C# implements indices with the help of the Index type, so you can also do the following: Index first = 0; Index last = ^1; char firstElement = vowels [first]; // 'a' char lastElement = vowels [last]; // 'u' Ranges Ranges let you “slice” an array by using the .. operator: char[] firstTwo = vowels [..2]; // 'a', 'e' char[] lastThree = vowels [2..]; // 'i', 'o', 'u' char[] middleOne = vowels [2..3]; // 'i' The second number in the range is exclusive, so ..2 returns the elements before vowels[2]. You can also use the ^ symbol in ranges. The following returns the last two characters: char[] lastTwo = vowels [^2..]; // 'o', 'u' C# implements ranges with the help of the Range type, so you can also do the following: Range firstTwoRange = 0..2; char[] firstTwo = vowels [firstTwoRange]; // 'a', 'e' Multidimensional Arrays Multidimensional arrays come in two varieties: rectangular and jagged. Rectangular arrays represent an n-dimensional block of memory, and jagged arrays are arrays of arrays. Rectangular arrays Rectangular arrays are declared using commas to separate each dimension. The following declares a rectangular two-dimensional array for which the dimensions are 3 by 3: int[,] matrix = new int[3,3]; 56 | Chapter 2: C# Language Basics
- 75. The GetLength method of an array returns the length for a given dimension (start‐ ing at 0): for (int i = 0; i < matrix.GetLength(0); i++) for (int j = 0; j < matrix.GetLength(1); j++) matrix[i,j] = i * 3 + j; You can initialize a rectangular array with explicit values. The following code creates an array identical to the previous example: int[,] matrix = new int[,] { {0,1,2}, {3,4,5}, {6,7,8} }; Jagged arrays Jagged arrays are declared using successive square brackets to represent each dimension. Here is an example of declaring a jagged two-dimensional array for which the outermost dimension is 3: int[][] matrix = new int[3][]; Interestingly, this is new int[3][] and not new int[][3]. Eric Lippert has written an excellent article on why this is so. The inner dimensions aren’t specified in the declaration because, unlike a rectangu‐ lar array, each inner array can be an arbitrary length. Each inner array is implicitly initialized to null rather than an empty array. You must manually create each inner array: for (int i = 0; i < matrix.Length; i++) { matrix[i] = new int[3]; // Create inner array for (int j = 0; j < matrix[i].Length; j++) matrix[i][j] = i * 3 + j; } You can initialize a jagged array with explicit values. The following code creates an array identical to the previous example with an additional element at the end: int[][] matrix = new int[][] { new int[] {0,1,2}, new int[] {3,4,5}, new int[] {6,7,8,9} }; Simplified Array Initialization Expressions There are two ways to shorten array initialization expressions. The first is to omit the new operator and type qualifications: Arrays | 57 C# Language Basics
- 76. char[] vowels = {'a','e','i','o','u'}; int[,] rectangularMatrix = { {0,1,2}, {3,4,5}, {6,7,8} }; int[][] jaggedMatrix = { new int[] {0,1,2}, new int[] {3,4,5}, new int[] {6,7,8,9} }; The second approach is to use the var keyword, which instructs the compiler to implicitly type a local variable: var i = 3; // i is implicitly of type int var s = "sausage"; // s is implicitly of type string // Therefore: var rectMatrix = new int[,] // rectMatrix is implicitly of type int[,] { {0,1,2}, {3,4,5}, {6,7,8} }; var jaggedMat = new int[][] // jaggedMat is implicitly of type int[][] { new int[] {0,1,2}, new int[] {3,4,5}, new int[] {6,7,8,9} }; Implicit typing can be taken one stage further with arrays: you can omit the type qualifier after the new keyword and have the compiler infer the array type: var vowels = new[] {'a','e','i','o','u'}; // Compiler infers char[] For this to work, the elements must all be implicitly convertible to a single type (and at least one of the elements must be of that type, and there must be exactly one best type), as in the following example: var x = new[] {1,10000000000}; // all convertible to long Bounds Checking All array indexing is bounds checked by the runtime. An IndexOutOfRange Exception is thrown if you use an invalid index: int[] arr = new int[3]; arr[3] = 1; // IndexOutOfRangeException thrown 58 | Chapter 2: C# Language Basics
- 77. Array bounds checking is necessary for type safety and simplifies debugging. Generally, the performance hit from bounds checking is minor, and the Just-in-Time (JIT) compiler can perform opti‐ mizations, such as determining in advance whether all indexes will be safe before entering a loop, thus avoiding a check on each iteration. In addition, C# provides “unsafe” code that can explicitly bypass bounds checking (see “Unsafe Code and Pointers” on page 243). Variables and Parameters A variable represents a storage location that has a modifiable value. A variable can be a local variable, parameter (value, ref, out, or in), field (instance or static), or array element. The Stack and the Heap The stack and the heap are the places where variables reside. Each has very different lifetime semantics. Stack The stack is a block of memory for storing local variables and parameters. The stack logically grows and shrinks as a method or function is entered and exited. Consider the following method (to avoid distraction, input argument checking is ignored): static int Factorial (int x) { if (x == 0) return 1; return x * Factorial (x-1); } This method is recursive, meaning that it calls itself. Each time the method is entered, a new int is allocated on the stack, and each time the method exits, the int is deallocated. Heap The heap is the memory in which objects (i.e., reference-type instances) reside. Whenever a new object is created, it is allocated on the heap, and a reference to that object is returned. During a program’s execution, the heap begins filling up as new objects are created. The runtime has a garbage collector that periodically deallocates objects from the heap, so your program does not run out of memory. An object is eligible for deallocation as soon as it’s not referenced by anything that’s itself “alive.” In the following example, we begin by creating a StringBuilder object referenced by the variable ref1 and then write out its content. That StringBuilder object is then immediately eligible for garbage collection because nothing subsequently uses it. Variables and Parameters | 59 C# Language Basics
- 78. Then, we create another StringBuilder referenced by variable ref2 and copy that reference to ref3. Even though ref2 is not used after that point, ref3 keeps the same StringBuilder object alive—ensuring that it doesn’t become eligible for collection until we’ve finished using ref3: using System; using System.Text; StringBuilder ref1 = new StringBuilder ("object1"); Console.WriteLine (ref1); // The StringBuilder referenced by ref1 is now eligible for GC. StringBuilder ref2 = new StringBuilder ("object2"); StringBuilder ref3 = ref2; // The StringBuilder referenced by ref2 is NOT yet eligible for GC. Console.WriteLine (ref3); // object2 Value-type instances (and object references) live wherever the variable was declared. If the instance was declared as a field within a class type, or as an array element, that instance lives on the heap. You can’t explicitly delete objects in C#, as you can in C++. An unreferenced object is eventually collected by the garbage collector. The heap also stores static fields. Unlike objects allocated on the heap (which can be garbage-collected), these live until the process ends. Definite Assignment C# enforces a definite assignment policy. In practice, this means that outside of an unsafe or interop context, you can’t accidentally access uninitialized memory. Definite assignment has three implications: • Local variables must be assigned a value before they can be read. • • Function arguments must be supplied when a method is called (unless marked • as optional; see “Optional parameters” on page 65). • All other variables (such as fields and array elements) are automatically initial‐ • ized by the runtime. For example, the following code results in a compile-time error: int x; Console.WriteLine (x); // Compile-time error Fields and array elements are automatically initialized with the default values for their type. The following code outputs 0 because array elements are implicitly assigned to their default values: 60 | Chapter 2: C# Language Basics
- 79. int[] ints = new int[2]; Console.WriteLine (ints[0]); // 0 The following code outputs 0, because fields are implicitly assigned a default value (whether instance or static): Console.WriteLine (Test.X); // 0 class Test { public static int X; } // field Default Values All type instances have a default value. The default value for the predefined types is the result of a bitwise zeroing of memory: Type Default value Reference types (and nullable value types) null Numeric and enum types 0 char type '0' bool type false You can obtain the default value for any type via the default keyword: Console.WriteLine (default (decimal)); // 0 You can optionally omit the type when it can be inferred: decimal d = default; The default value in a custom value type (i.e., struct) is the same as the default value for each field defined by the custom type. Parameters A method may have a sequence of parameters. Parameters define the set of argu‐ ments that must be provided for that method. In the following example, the method Foo has a single parameter named p, of type int: Foo (8); // 8 is an argument static void Foo (int p) {...} // p is a parameter You can control how parameters are passed with the ref, in, and out modifiers: Parameter modifier Passed by Variable must be definitely assigned (None) Value Going in ref Reference Going in in Reference (read-only) Going in out Reference Going out Variables and Parameters | 61 C# Language Basics
- 80. Passing arguments by value By default, arguments in C# are passed by value, which is by far the most common case. This means that a copy of the value is created when passed to the method: int x = 8; Foo (x); // Make a copy of x Console.WriteLine (x); // x will still be 8 static void Foo (int p) { p = p + 1; // Increment p by 1 Console.WriteLine (p); // Write p to screen } Assigning p a new value does not change the contents of x, because p and x reside in different memory locations. Passing a reference-type argument by value copies the reference but not the object. In the following example, Foo sees the same StringBuilder object we instantiated (sb) but has an independent reference to it. In other words, sb and fooSB are separate variables that reference the same StringBuilder object: StringBuilder sb = new StringBuilder(); Foo (sb); Console.WriteLine (sb.ToString()); // test static void Foo (StringBuilder fooSB) { fooSB.Append ("test"); fooSB = null; } Because fooSB is a copy of a reference, setting it to null doesn’t make sb null. (If, however, fooSB was declared and called with the ref modifier, sb would become null.) The ref modifier To pass by reference, C# provides the ref parameter modifier. In the following example, p and x refer to the same memory locations: int x = 8; Foo (ref x); // Ask Foo to deal directly with x Console.WriteLine (x); // x is now 9 static void Foo (ref int p) { p = p + 1; // Increment p by 1 Console.WriteLine (p); // Write p to screen } 62 | Chapter 2: C# Language Basics
- 81. 4 An exception to this rule is when calling Component Object Model (COM) methods. We discuss this in Chapter 25. Now assigning p a new value changes the contents of x. Notice how the ref modifier is required both when writing and when calling the method.4 This makes it very clear what’s going on. The ref modifier is essential in implementing a swap method (in “Generics” on page 145, we show how to write a swap method that works with any type): string x = "Penn"; string y = "Teller"; Swap (ref x, ref y); Console.WriteLine (x); // Teller Console.WriteLine (y); // Penn static void Swap (ref string a, ref string b) { string temp = a; a = b; b = temp; } A parameter can be passed by reference or by value, regardless of whether the parameter type is a reference type or a value type. The out modifier An out argument is like a ref argument except for the following: • It need not be assigned before going into the function. • • It must be assigned before it comes out of the function. • The out modifier is most commonly used to get multiple return values back from a method; for example: string a, b; Split ("Stevie Ray Vaughn", out a, out b); Console.WriteLine (a); // Stevie Ray Console.WriteLine (b); // Vaughn void Split (string name, out string firstNames, out string lastName) { int i = name.LastIndexOf (' '); firstNames = name.Substring (0, i); lastName = name.Substring (i + 1); } Like a ref parameter, an out parameter is passed by reference. Variables and Parameters | 63 C# Language Basics
- 82. Out variables and discards You can declare variables on the fly when calling methods with out parameters. We can replace the first two lines in our preceding example with this: Split ("Stevie Ray Vaughan", out string a, out string b); When calling methods with multiple out parameters, sometimes you’re not interes‐ ted in receiving values from all the parameters. In such cases, you can “discard” the ones in which you’re uninterested by using an underscore: Split ("Stevie Ray Vaughan", out string a, out _); // Discard 2nd param Console.WriteLine (a); In this case, the compiler treats the underscore as a special symbol, called a discard. You can include multiple discards in a single call. Assuming SomeBigMethod has been defined with seven out parameters, we can ignore all but the fourth, as follows: SomeBigMethod (out _, out _, out _, out int x, out _, out _, out _); For backward compatibility, this language feature will not take effect if a real underscore variable is in scope: string _; Split ("Stevie Ray Vaughan", out string a, out _); Console.WriteLine (_); // Vaughan Implications of passing by reference When you pass an argument by reference, you alias the storage location of an existing variable rather than create a new storage location. In the following example, the variables x and y represent the same instance: class Test { static int x; static void Main() { Foo (out x); } static void Foo (out int y) { Console.WriteLine (x); // x is 0 y = 1; // Mutate y Console.WriteLine (x); // x is 1 } } The in modifier An in parameter is similar to a ref parameter except that the argument’s value cannot be modified by the method (doing so generates a compile-time error). This modifier is most useful when passing a large value type to the method because it allows the compiler to avoid the overhead of copying the argument prior to passing it in while still protecting the original value from modification. 64 | Chapter 2: C# Language Basics
- 83. Overloading solely on the presence of in is permitted: void Foo ( SomeBigStruct a) { ... } void Foo (in SomeBigStruct a) { ... } To call the second overload, the caller must use the in modifier: SomeBigStruct x = ...; Foo (x); // Calls the first overload Foo (in x); // Calls the second overload When there’s no ambiguity, void Bar (in SomeBigStruct a) { ... } use of the in modifier is optional for the caller: Bar (x); // OK (calls the 'in' overload) Bar (in x); // OK (calls the 'in' overload) To make this example meaningful, SomeBigStruct would be defined as a struct (see “Structs” on page 129). The params modifier The params modifier, if applied to the last parameter of a method, allows the method to accept any number of arguments of a particular type. The parameter type must be declared as an (single-dimensional) array, as shown in the following example: int total = Sum (1, 2, 3, 4); Console.WriteLine (total); // 10 // The call to Sum above is equivalent to: int total2 = Sum (new int[] { 1, 2, 3, 4 }); int Sum (params int[] ints) { int sum = 0; for (int i = 0; i < ints.Length; i++) sum += ints [i]; // Increase sum by ints[i] return sum; } If there are zero arguments in the params position, a zero-length array is created. You can also supply a params argument as an ordinary array. The first line in our example is semantically equivalent to this: int total = Sum (new int[] { 1, 2, 3, 4 } ); Optional parameters Methods, constructors, and indexers (Chapter 3) can declare optional parameters. A parameter is optional if it specifies a default value in its declaration: Variables and Parameters | 65 C# Language Basics
- 84. void Foo (int x = 23) { Console.WriteLine (x); } You can omit optional parameters when calling the method: Foo(); // 23 The default argument of 23 is actually passed to the optional parameter x—the com‐ piler bakes the value 23 into the compiled code at the calling side. The preceding call to Foo is semantically identical to Foo (23); because the compiler simply substitutes the default value of an optional parameter wherever it is used. Adding an optional parameter to a public method that’s called from another assembly requires recompilation of both assem‐ blies—just as though the parameter were mandatory. The default value of an optional parameter must be specified by a constant expres‐ sion, a parameterless constructor of a value type, or a default expression. Optional parameters cannot be marked with ref or out. Mandatory parameters must occur before optional parameters in both the method declaration and the method call (the exception is with params arguments, which still always come last). In the following example, the explicit value of 1 is passed to x, and the default value of 0 is passed to y: Foo (1); // 1, 0 void Foo (int x = 0, int y = 0) { Console.WriteLine (x + ", " + y); } You can do the converse (pass a default value to x and an explicit value to y) by combining optional parameters with named arguments. Named arguments Rather than identifying an argument by position, you can identify an argument by name: Foo (x:1, y:2); // 1, 2 void Foo (int x, int y) { Console.WriteLine (x + ", " + y); } Named arguments can occur in any order. The following calls to Foo are semanti‐ cally identical: Foo (x:1, y:2); Foo (y:2, x:1); 66 | Chapter 2: C# Language Basics
- 85. A subtle difference is that argument expressions are evalu‐ ated in the order in which they appear at the calling site. In general, this makes a difference only with interdependent side-effecting expressions such as the following, which writes 0, 1: int a = 0; Foo (y: ++a, x: --a); // ++a is evaluated first Of course, you would almost certainly avoid writing such code in practice! You can mix named and positional arguments: Foo (1, y:2); However, there is a restriction: positional arguments must come before named arguments unless they are used in the correct position. So, you could call Foo like this: Foo (x:1, 2); // OK. Arguments in the declared positions But not like this: Foo (y:2, 1); // Compile-time error. y isn't in the first position Named arguments are particularly useful in conjunction with optional parameters. For instance, consider the following method: void Bar (int a = 0, int b = 0, int c = 0, int d = 0) { ... } You can call this supplying only a value for d, as follows: Bar (d:3); This is particularly useful when calling COM APIs, which we discuss in detail in Chapter 24. Ref Locals A somewhat esoteric feature of C# is that you can define a local variable that references an element in an array or field in an object (from C# 7): int[] numbers = { 0, 1, 2, 3, 4 }; ref int numRef = ref numbers [2]; In this example, numRef is a reference to numbers[2]. When we modify numRef, we modify the array element: numRef *= 10; Console.WriteLine (numRef); // 20 Console.WriteLine (numbers [2]); // 20 The target for a ref local must be an array element, field, or local variable; it cannot be a property (Chapter 3). Ref locals are intended for specialized micro-optimization scenarios and are typically used in conjunction with ref returns. Variables and Parameters | 67 C# Language Basics
- 86. Ref Returns The Span<T> and ReadOnlySpan<T> types that we describe in Chapter 23 use ref returns to implement a highly effi‐ cient indexer. Outside such scenarios, ref returns are not com‐ monly used, and you can consider them a micro-optimization feature. You can return a ref local from a method. This is called a ref return: class Program { static string x = "Old Value"; static ref string GetX() => ref x; // This method returns a ref static void Main() { ref string xRef = ref GetX(); // Assign result to a ref local xRef = "New Value"; Console.WriteLine (x); // New Value } } If you omit the ref modifier on the calling side, it reverts to returning an ordinary value: string localX = GetX(); // Legal: localX is an ordinary non-ref variable. You also can use ref returns when defining a property or indexer: static ref string Prop => ref x; Such a property is implicitly writable, despite there being no set accessor: Prop = "New Value"; You can prevent such modification by using ref readonly: static ref readonly string Prop => ref x; The ref readonly modifier prevents modification while still enabling the perfor‐ mance gain of returning by reference. The gain would be very small in this case, because x is of type string (a reference type): no matter how long the string, the only inefficiency that you can hope to avoid is the copying of a single 32- or 64-bit reference. Real gains can occur with custom value types (see “Structs” on page 129), but only if the struct is marked as readonly (otherwise, the compiler will perform a defensive copy). Attempting to define an explicit set accessor on a ref return property or indexer is illegal. 68 | Chapter 2: C# Language Basics
- 87. var—Implicitly Typed Local Variables It is often the case that you declare and initialize a variable in one step. If the compiler is able to infer the type from the initialization expression, you can use the keyword var in place of the type declaration; for example: var x = "hello"; var y = new System.Text.StringBuilder(); var z = (float)Math.PI; This is precisely equivalent to the following: string x = "hello"; System.Text.StringBuilder y = new System.Text.StringBuilder(); float z = (float)Math.PI; Because of this direct equivalence, implicitly typed variables are statically typed. For example, the following generates a compile-time error: var x = 5; x = "hello"; // Compile-time error; x is of type int var can decrease code readability when you can’t deduce the type purely by looking at the variable declaration; for example: Random r = new Random(); var x = r.Next(); What type is x? In “Anonymous Types” on page 205, we will describe a scenario in which the use of var is mandatory. Target-Typed new Expressions Another way to reduce lexical repetition is with target-typed new expressions (from C# 9): System.Text.StringBuilder sb1 = new(); System.Text.StringBuilder sb2 = new ("Test"); This is precisely equivalent to: System.Text.StringBuilder sb1 = new System.Text.StringBuilder(); System.Text.StringBuilder sb2 = new System.Text.StringBuilder ("Test"); The principle is that you can call new without specifying a type name if the compiler is able to unambiguously infer it. Target-typed new expressions are particularly useful when the variable declaration and initialization are in different parts of your code. A common example is when you want to initialize a field in a constructor: class Foo { System.Text.StringBuilder sb; public Foo (string initialValue) { Variables and Parameters | 69 C# Language Basics
- 88. sb = new (initialValue); } } Target-typed new expressions are also helpful in the following scenario: MyMethod (new ("test")); void MyMethod (System.Text.StringBuilder sb) { ... } Expressions and Operators An expression essentially denotes a value. The simplest kinds of expressions are constants and variables. Expressions can be transformed and combined using oper‐ ators. An operator takes one or more input operands to output a new expression. Here is an example of a constant expression: 12 We can use the * operator to combine two operands (the literal expressions 12 and 30), as follows: 12 * 30 We can build complex expressions because an operand can itself be an expression, such as the operand (12 * 30) in the following example: 1 + (12 * 30) Operators in C# can be classed as unary, binary, or ternary, depending on the number of operands they work on (one, two, or three). The binary operators always use infix notation in which the operator is placed between the two operands. Primary Expressions Primary expressions include expressions composed of operators that are intrinsic to the basic plumbing of the language. Here is an example: Math.Log (1) This expression is composed of two primary expressions. The first expression per‐ forms a member lookup (with the . operator), and the second expression performs a method call (with the () operator). Void Expressions A void expression is an expression that has no value, such as this: Console.WriteLine (1) Because it has no value, you cannot use a void expression as an operand to build more complex expressions: 1 + Console.WriteLine (1) // Compile-time error 70 | Chapter 2: C# Language Basics
- 89. Assignment Expressions An assignment expression uses the = operator to assign the result of another expres‐ sion to a variable; for example: x = x * 5 An assignment expression is not a void expression—it has a value of whatever was assigned, and so can be incorporated into another expression. In the following example, the expression assigns 2 to x and 10 to y: y = 5 * (x = 2) You can use this style of expression to initialize multiple values: a = b = c = d = 0 The compound assignment operators are syntactic shortcuts that combine assign‐ ment with another operator: x *= 2 // equivalent to x = x * 2 x <<= 1 // equivalent to x = x << 1 (A subtle exception to this rule is with events, which we describe in Chapter 4: the += and -= operators here are treated specially and map to the event’s add and remove accessors.) Operator Precedence and Associativity When an expression contains multiple operators, precedence and associativity deter‐ mine the order of their evaluation. Operators with higher precedence execute before operators with lower precedence. If the operators have the same precedence, the operator’s associativity determines the order of evaluation. Precedence The following expression 1 + 2 * 3 is evaluated as follows because * has a higher precedence than +: 1 + (2 * 3) Left-associative operators Binary operators (except for assignment, lambda, and null-coalescing operators) are left-associative; in other words, they are evaluated from left to right. For example, the following expression 8 / 4 / 2 is evaluated as follows: ( 8 / 4 ) / 2 // 1 Expressions and Operators | 71 C# Language Basics
- 90. You can insert parentheses to change the actual order of evaluation: 8 / ( 4 / 2 ) // 4 Right-associative operators The assignment operators as well as the lambda, null-coalescing, and conditional operators are right-associative; in other words, they are evaluated from right to left. Right associativity allows multiple assignments, such as the following, to compile: x = y = 3; This first assigns 3 to y and then assigns the result of that expression (3) to x. Operator Table Table 2-3 lists C#’s operators in order of precedence. Operators in the same category have the same precedence. We explain user-overloadable operators in “Operator Overloading” on page 239. Table 2-3. C# operators (categories in order of precedence) Category Operator symbol Operator name Example User- overloadable Primary . Member access x.y No ?. and ?[] Null-conditional x?.y or x?[0] No ! (postfix) Null-forgiving x!.y or x![0] No -> (unsafe) Pointer to struct x->y No () Function call x() No [] Array/index a[x] Via indexer ++ Post-increment x++ Yes −− Post-decrement x−− Yes new Create instance new Foo() No stackalloc Stack allocation stackalloc(10) No typeof Get type from identifier typeof(int) No nameof Get name of identifier nameof(x) No checked Integral overflow check on checked(x) No unchecked Integral overflow check off unchecked(x) No default Default value default(char) No Unary await Await await myTask No 72 | Chapter 2: C# Language Basics
- 91. Category Operator symbol Operator name Example User- overloadable sizeof Get size of struct sizeof(int) No + Positive value of +x Yes − Negative value of −x Yes ! Not !x Yes ~ Bitwise complement ~x Yes ++ Pre-increment ++x Yes −− Pre-decrement −−x Yes () Cast (int)x No ^ Index from end array[^1] No * (unsafe) Value at address *x No & (unsafe) Address of value &x No Range .. ..^ Range of indices x..y x..^y No Switch & with switch Switch expression num switch { 1 => true, _ => false } No with With expression rec with { X = 123 } No Multiplicative * Multiply x * y Yes / Divide x / y Yes % Remainder x % y Yes Additive + Add x + y Yes − Subtract x − y Yes Shift << Shift left x << 1 Yes >> Shift right x >> 1 Yes Relational < Less than x < y Yes > Greater than x > y Yes <= Less than or equal to x <= y Yes >= Greater than or equal to x >= y Yes is Type is or is subclass of x is y No as Type conversion x as y No Equality == Equals x == y Yes Expressions and Operators | 73 C# Language Basics
- 92. Category Operator symbol Operator name Example User- overloadable != Not equals x != y Yes Logical And & And x & y Yes Logical Xor ^ Exclusive Or x ^ y Yes Logical Or | Or x | y Yes Conditional And && Conditional And x && y Via & Conditional Or || Conditional Or x || y Via | Null coalescing ?? Null coalescing x ?? y No Conditional ?: Conditional isTrue ? thenThis : elseThis No Assignment and lambda = Assign x = y No *= Multiply self by x *= 2 Via * /= Divide self by x /= 2 Via / %= Remainder & assign to self x %= 2 += Add to self x += 2 Via + −= Subtract from self x −= 2 Via − <<= Shift self left by x <<= 2 Via << >>= Shift self right by x >>= 2 Via >> &= And self by x &= 2 Via & ^= Exclusive-Or self by x ^= 2 Via ^ |= Or self by x |= 2 Via | ??= Null-coalescing assignment x ??= 0 No => Lambda x => x + 1 No Null Operators C# provides three operators to make it easier to work with nulls: the null-coalescing operator, the null-coalescing assignment operator, and the null-conditional operator. Null-Coalescing Operator The ?? operator is the null-coalescing operator. It says, “If the operand to the left is non-null, give it to me; otherwise, give me another value.” For example: string s1 = null; string s2 = s1 ?? "nothing"; // s2 evaluates to "nothing" 74 | Chapter 2: C# Language Basics
- 93. If the lefthand expression is non-null, the righthand expression is never evaluated. The null-coalescing operator also works with nullable value types (see “Nullable Value Types” on page 194). Null-Coalescing Assignment Operator The ??= operator (introduced in C# 8) is the null-coalescing assignment operator. It says, “If the operand to the left is null, assign the right operand to the left operand.” Consider the following: myVariable ??= someDefault; This is equivalent to: if (myVariable == null) myVariable = someDefault; The ??= operator is particularly useful in implementing lazily calculated properties. We’ll cover this topic later, in “Calculated Fields and Lazy Evaluation” on page 217. Null-Conditional Operator The ?. operator is the null-conditional or “Elvis” operator (after the Elvis emoticon). It allows you to call methods and access members just like the standard dot operator except that if the operand on the left is null, the expression evaluates to null instead of throwing a NullReferenceException: System.Text.StringBuilder sb = null; string s = sb?.ToString(); // No error; s instead evaluates to null The last line is equivalent to the following: string s = (sb == null ? null : sb.ToString()); Null-conditional expressions also work with indexers: string foo = null; char? c = foo?[1]; // c is null Upon encountering a null, the Elvis operator short-circuits the remainder of the expression. In the following example, s evaluates to null, even with a standard dot operator between ToString() and ToUpper(): System.Text.StringBuilder sb = null; string s = sb?.ToString().ToUpper(); // s evaluates to null without error Repeated use of Elvis is necessary only if the operand immediately to its left might be null. The following expression is robust to both x being null and x.y being null: x?.y?.z It is equivalent to the following (except that x.y is evaluated only once): x == null ? null : (x.y == null ? null : x.y.z) The final expression must be capable of accepting a null. The following is illegal: Null Operators | 75 C# Language Basics
- 94. System.Text.StringBuilder sb = null; int length = sb?.ToString().Length; // Illegal : int cannot be null We can fix this with the use of nullable value types (see “Nullable Value Types” on page 194). If you’re already familiar with nullable value types, here’s a preview: int? length = sb?.ToString().Length; // OK: int? can be null You can also use the null-conditional operator to call a void method: someObject?.SomeVoidMethod(); If someObject is null, this becomes a “no-operation” rather than throwing a Null ReferenceException. You can use the null-conditional operator with the commonly used type members that we describe in Chapter 3, including methods, fields, properties, and indexers. It also combines well with the null-coalescing operator: System.Text.StringBuilder sb = null; string s = sb?.ToString() ?? "nothing"; // s evaluates to "nothing" Statements Functions comprise statements that execute sequentially in the textual order in which they appear. A statement block is a series of statements appearing between braces (the {} tokens). Declaration Statements A variable declaration introduces a new variable, optionally initializing it with an expression. You may declare multiple variables of the same type in a comma- separated list: string someWord = "rosebud"; int someNumber = 42; bool rich = true, famous = false; A constant declaration is like a variable declaration except that it cannot be changed after it has been declared, and the initialization must occur with the declaration (see “Constants” on page 96): const double c = 2.99792458E08; c += 10; // Compile-time Error Local variables The scope of a local variable or local constant extends throughout the current block. You cannot declare another local variable with the same name in the current block or in any nested blocks: int x; { int y; int x; // Error - x already defined 76 | Chapter 2: C# Language Basics
- 95. } { int y; // OK - y not in scope } Console.Write (y); // Error - y is out of scope A variable’s scope extends in both directions throughout its code block. This means that if we moved the initial declara‐ tion of x in this example to the bottom of the method, we’d get the same error. This is in contrast to C++ and is somewhat peculiar, given that it’s not legal to refer to a variable or con‐ stant before it’s declared. Expression Statements Expression statements are expressions that are also valid statements. An expression statement must either change state or call something that might change state. Changing state essentially means changing a variable. Following are the possible expression statements: • Assignment expressions (including increment and decrement expressions) • • Method call expressions (both void and nonvoid) • • Object instantiation expressions • Here are some examples: // Declare variables with declaration statements: string s; int x, y; System.Text.StringBuilder sb; // Expression statements x = 1 + 2; // Assignment expression x++; // Increment expression y = Math.Max (x, 5); // Assignment expression Console.WriteLine (y); // Method call expression sb = new StringBuilder(); // Assignment expression new StringBuilder(); // Object instantiation expression When you call a constructor or a method that returns a value, you’re not obliged to use the result. However, unless the constructor or method changes state, the statement is completely useless: new StringBuilder(); // Legal, but useless new string ('c', 3); // Legal, but useless x.Equals (y); // Legal, but useless Statements | 77 C# Language Basics
- 96. Selection Statements C# has the following mechanisms to conditionally control the flow of program execution: • Selection statements (if, switch) • • Conditional operator (?:) • • Loop statements (while, do-while, for, foreach) • This section covers the simplest two constructs: the if statement and the switch statement. The if statement An if statement executes a statement if a bool expression is true: if (5 < 2 * 3) Console.WriteLine ("true"); // true The statement can be a code block: if (5 < 2 * 3) { Console.WriteLine ("true"); Console.WriteLine ("Let’s move on!"); } The else clause An if statement can optionally feature an else clause: if (2 + 2 == 5) Console.WriteLine ("Does not compute"); else Console.WriteLine ("False"); // False Within an else clause, you can nest another if statement: if (2 + 2 == 5) Console.WriteLine ("Does not compute"); else if (2 + 2 == 4) Console.WriteLine ("Computes"); // Computes Changing the flow of execution with braces An else clause always applies to the immediately preceding if statement in the statement block: if (true) if (false) Console.WriteLine(); 78 | Chapter 2: C# Language Basics
- 97. else Console.WriteLine ("executes"); This is semantically identical to the following: if (true) { if (false) Console.WriteLine(); else Console.WriteLine ("executes"); } We can change the execution flow by moving the braces: if (true) { if (false) Console.WriteLine(); } else Console.WriteLine ("does not execute"); With braces, you explicitly state your intention. This can improve the readability of nested if statements—even when not required by the compiler. A notable exception is with the following pattern: void TellMeWhatICanDo (int age) { if (age >= 35) Console.WriteLine ("You can be president!"); else if (age >= 21) Console.WriteLine ("You can drink!"); else if (age >= 18) Console.WriteLine ("You can vote!"); else Console.WriteLine ("You can wait!"); } Here, we’ve arranged the if and else statements to mimic the “elseif” construct of other languages (and C#’s #elif preprocessor directive). Visual Studio’s auto- formatting recognizes this pattern and preserves the indentation. Semantically, though, each if statement following an else statement is functionally nested within the else clause. The switch statement switch statements let you branch program execution based on a selection of possi‐ ble values that a variable might have. switch statements can result in cleaner code than multiple if statements because switch statements require an expression to be evaluated only once: void ShowCard (int cardNumber) { switch (cardNumber) Statements | 79 C# Language Basics
- 98. { case 13: Console.WriteLine ("King"); break; case 12: Console.WriteLine ("Queen"); break; case 11: Console.WriteLine ("Jack"); break; case -1: // Joker is -1 goto case 12; // In this game joker counts as queen default: // Executes for any other cardNumber Console.WriteLine (cardNumber); break; } } This example demonstrates the most common scenario, which is switching on constants. When you specify a constant, you’re restricted to the built-in integral types; the bool, char, and enum types; and the string type. At the end of each case clause, you must specify explicitly where execution is to go next, with some kind of jump statement (unless your code ends in an infinite loop). Here are the options: • break (jumps to the end of the switch statement) • • goto case x (jumps to another case clause) • • goto default (jumps to the default clause) • • Any other jump statement—namely, return, throw, continue, or goto label • When more than one value should execute the same code, you can list the common cases sequentially: switch (cardNumber) { case 13: case 12: case 11: Console.WriteLine ("Face card"); break; default: Console.WriteLine ("Plain card"); break; } This feature of a switch statement can be pivotal in terms of producing cleaner code than multiple if-else statements. 80 | Chapter 2: C# Language Basics
- 99. Switching on types Switching on a type is a special case of switching on a pattern. A number of other patterns have been introduced in recent versions of C#; see “Patterns” on page 222 for a full discussion. You can also switch on types (from C# 7): TellMeTheType (12); TellMeTheType ("hello"); TellMeTheType (true); void TellMeTheType (object x) // object allows any type. { switch (x) { case int i: Console.WriteLine ("It's an int!"); Console.WriteLine ($"The square of {i} is {i * i}"); break; case string s: Console.WriteLine ("It's a string"); Console.WriteLine ($"The length of {s} is {s.Length}"); break; default: Console.WriteLine ("I don't know what x is"); break; } } (The object type allows for a variable of any type; we discuss this fully in “Inheri‐ tance” on page 115 and “The object Type” on page 125.) Each case clause specifies a type upon which to match, and a variable upon which to assign the typed value if the match succeeds (the “pattern” variable). Unlike with constants, there’s no restriction on what types you can use. You can predicate a case with the when keyword: switch (x) { case bool b when b == true: // Fires only when b is true Console.WriteLine ("True!"); break; case bool b: Console.WriteLine ("False!"); break; } The order of the case clauses can matter when switching on type (unlike when switching on constants). This example would give a different result if we reversed the two cases (in fact, it would not even compile, because the compiler would determine that the second case is unreachable). An exception to this rule is the default clause, which is always executed last, regardless of where it appears. Statements | 81 C# Language Basics
- 100. If you want to switch on a type, but are uninterested in its value, you can use a discard (_): case DateTime _: Console.WriteLine ("It's a DateTime"); You can stack multiple case clauses. The Console.WriteLine in the following code will execute for any floating-point type greater than 1,000: switch (x) { case float f when f > 1000: case double d when d > 1000: case decimal m when m > 1000: Console.WriteLine ("We can refer to x here but not f or d or m"); break; } In this example, the compiler lets us consume the pattern variables f, d, and m, only in the when clauses. When we call Console.WriteLine, its unknown which one of those three variables will be assigned, so the compiler puts all of them out of scope. You can mix and match constants and patterns in the same switch statement. And you can also switch on the null value: case null: Console.WriteLine ("Nothing here"); break; Switch expressions From C# 8, you can use switch in the context of an expression. Assuming that cardNumber is of type int, the following illustrates its use: string cardName = cardNumber switch { 13 => "King", 12 => "Queen", 11 => "Jack", _ => "Pip card" // equivalent to 'default' }; Notice that the switch keyword appears after the variable name and that the case clauses are expressions (terminated by commas) rather than statements. Switch expressions are more compact than their switch statement counterparts, and you can use them in LINQ queries (Chapter 8). If you omit the default expression (_) and the switch fails to match, an exception is thrown. You can also switch on multiple values (the tuple pattern): int cardNumber = 12; string suite = "spades"; 82 | Chapter 2: C# Language Basics
- 101. string cardName = (cardNumber, suite) switch { (13, "spades") => "King of spades", (13, "clubs") => "King of clubs", ... }; Many more options are possible through the use of patterns (see “Patterns” on page 222). Iteration Statements C# enables a sequence of statements to execute repeatedly with the while, do-while, for, and foreach statements. while and do-while loops while loops repeatedly execute a body of code while a bool expression is true. The expression is tested before the body of the loop is executed. For example, the following writes 012: int i = 0; while (i < 3) { Console.Write (i); i++; } do-while loops differ in functionality from while loops only in that they test the expression after the statement block has executed (ensuring that the block is always executed at least once). Here’s the preceding example rewritten with a do-while loop: int i = 0; do { Console.WriteLine (i); i++; } while (i < 3); for loops for loops are like while loops with special clauses for initialization and iteration of a loop variable. A for loop contains three clauses as follows: for (initialization-clause; condition-clause; iteration-clause) statement-or-statement-block Statements | 83 C# Language Basics
- 102. Here’s what each clause does: Initialization clause Executed before the loop begins; used to initialize one or more iteration vari‐ ables Condition clause The bool expression that, while true, will execute the body Iteration clause Executed after each iteration of the statement block; used typically to update the iteration variable For example, the following prints the numbers 0 through 2: for (int i = 0; i < 3; i++) Console.WriteLine (i); The following prints the first 10 Fibonacci numbers (in which each number is the sum of the previous two): for (int i = 0, prevFib = 1, curFib = 1; i < 10; i++) { Console.WriteLine (prevFib); int newFib = prevFib + curFib; prevFib = curFib; curFib = newFib; } Any of the three parts of the for statement can be omitted. You can implement an infinite loop such as the following (though while(true) can be used, instead): for (;;) Console.WriteLine ("interrupt me"); foreach loops The foreach statement iterates over each element in an enumerable object. Most of the .NET types that represent a set or list of elements are enumerable. For example, both an array and a string are enumerable. Here is an example of enumerating over the characters in a string, from the first character through to the last: foreach (char c in "beer") // c is the iteration variable Console.WriteLine (c); Here’s the output: b e e r We define enumerable objects in “Enumeration and Iterators” on page 189. 84 | Chapter 2: C# Language Basics
- 103. Jump Statements The C# jump statements are break, continue, goto, return, and throw. Jump statements obey the reliability rules of try statements (see “try Statements and Exceptions” on page 181). This means that: • A jump out of a try block always executes the try’s • finally block before reaching the target of the jump. • A jump cannot be made from the inside to the outside of • a finally block (except via throw). The break statement The break statement ends the execution of the body of an iteration or switch statement: int x = 0; while (true) { if (x++ > 5) break; // break from the loop } // execution continues here after break ... The continue statement The continue statement forgoes the remaining statements in a loop and makes an early start on the next iteration. The following loop skips even numbers: for (int i = 0; i < 10; i++) { if ((i % 2) == 0) // If i is even, continue; // continue with next iteration Console.Write (i + " "); } OUTPUT: 1 3 5 7 9 The goto statement The goto statement transfers execution to another label within a statement block. The form is as follows: goto statement-label; Or, when used within a switch statement: goto case case-constant; // (Only works with constants, not patterns) Statements | 85 C# Language Basics
- 104. A label is a placeholder in a code block that precedes a statement, denoted with a colon suffix. The following iterates the numbers 1 through 5, mimicking a for loop: int i = 1; startLoop: if (i <= 5) { Console.Write (i + " "); i++; goto startLoop; } OUTPUT: 1 2 3 4 5 The goto case case-constant transfers execution to another case in a switch block (see “The switch statement” on page 79). The return statement The return statement exits the method and must return an expression of the method’s return type if the method is nonvoid: decimal AsPercentage (decimal d) { decimal p = d * 100m; return p; // Return to the calling method with value } A return statement can appear anywhere in a method (except in a finally block) and can be used more than once. The throw statement The throw statement throws an exception to indicate an error has occurred (see “try Statements and Exceptions” on page 181): if (w == null) throw new ArgumentNullException (...); Miscellaneous Statements The using statement provides an elegant syntax for calling Dispose on objects that implement IDisposable, within a finally block (see “try Statements and Exceptions” on page 181 and “IDisposable, Dispose, and Close” on page 557). C# overloads the using keyword to have independent mean‐ ings in different contexts. Specifically, the using directive is different from the using statement. The lock statement is a shortcut for calling the Enter and Exit methods of the Monitor class (see Chapters 14 and 23). 86 | Chapter 2: C# Language Basics
- 105. Namespaces A namespace is a domain for type names. Types are typically organized into hier‐ archical namespaces, making them easier to find and preventing conflicts. For example, the RSA type that handles public key encryption is defined within the following namespace: System.Security.Cryptography A namespace forms an integral part of a type’s name. The following code calls RSA’s Create method: System.Security.Cryptography.RSA rsa = System.Security.Cryptography.RSA.Create(); Namespaces are independent of assemblies, which are .dll files that serve as units of deployment (described in Chapter 17). Namespaces also have no impact on member visibility—pub lic, internal, private, and so on. The namespace keyword defines a namespace for types within that block; for example: namespace Outer.Middle.Inner { class Class1 {} class Class2 {} } The dots in the namespace indicate a hierarchy of nested namespaces. The code that follows is semantically identical to the preceding example: namespace Outer { namespace Middle { namespace Inner { class Class1 {} class Class2 {} } } } You can refer to a type with its fully qualified name, which includes all namespaces from the outermost to the innermost. For example, we could refer to Class1 in the preceding example as Outer.Middle.Inner.Class1. Types not defined in any namespace are said to reside in the global namespace. The global namespace also includes top-level namespaces, such as Outer in our example. Namespaces | 87 C# Language Basics
- 106. File-Scoped Namespaces (C# 10) Often, you will want all the types in a file to be defined in one namespace: namespace MyNamespace { class Class1 {} class Class2 {} } From C# 10, you can accomplish this with a file-scoped namespace: namespace MyNamespace; // Applies to everything that follows in the file. class Class1 {} // inside MyNamespace class Class2 {} // inside MyNamespace File-scoped namespaces reduce clutter and eliminate an unnecessary level of indentation. The using Directive The using directive imports a namespace, allowing you to refer to types without their fully qualified names. The following imports the previous example’s Outer.Mid dle.Inner namespace: using Outer.Middle.Inner; Class1 c; // Don’t need fully qualified name It’s legal (and often desirable) to define the same type name in different namespaces. However, you’d typically do so only if it was unlikely for a consumer to want to import both namespaces at once. A good example is the TextBox class, which is defined both in System.Windows.Controls (WPF) and System.Windows.Forms (Windows Forms). A using directive can be nested within a namespace itself to limit the scope of the directive. The global using Directive (C# 10) From C# 10, if you prefix a using directive with the global keyword, the directive will apply to all files in the project or compilation unit: global using System; global using System.Collection.Generic; This lets you centralize common imports and avoid repeating the same directives in every file. global using directives must precede nonglobal directives and cannot appear inside namespace declarations. The global directive can be used with using static. 88 | Chapter 2: C# Language Basics
- 107. Implicit global usings From .NET 6, project files allow for implicit global using directives. If the ImplicitUsings element is set to true in the project file (the default for new projects), the following namespaces are automatically imported: System System.Collections.Generic System.IO System.Linq System.Net.Http System.Threading System.Threading.Tasks Additional namespaces are imported, based on the project SDK (Web, Windows Forms, WPF, and so on). using static The using static directive imports a type rather than a namespace. All static members of the imported type can then be used without qualification. In the follow‐ ing example, we call the Console class’s static WriteLine method without needing to refer to the type: using static System.Console; WriteLine ("Hello"); The using static directive imports all accessible static members of the type, including fields, properties, and nested types (Chapter 3). You can also apply this directive to enum types (Chapter 3), in which case their members are imported. So, if we import the following enum type using static System.Windows.Visibility; we can specify Hidden instead of Visibility.Hidden: var textBox = new TextBox { Visibility = Hidden }; // XAML-style Should an ambiguity arise between multiple static imports, the C# compiler is not smart enough to infer the correct type from the context and will generate an error. Rules Within a Namespace Name scoping Names declared in outer namespaces can be used unqualified within inner name‐ spaces. In this example, Class1 does not need qualification within Inner: namespace Outer { class Class1 {} namespace Inner Namespaces | 89 C# Language Basics
- 108. { class Class2 : Class1 {} } } If you want to refer to a type in a different branch of your namespace hierarchy, you can use a partially qualified name. In the following example, we base SalesReport on Common.ReportBase: namespace MyTradingCompany { namespace Common { class ReportBase {} } namespace ManagementReporting { class SalesReport : Common.ReportBase {} } } Name hiding If the same type name appears in both an inner and an outer namespace, the inner name wins. To refer to the type in the outer namespace, you must qualify its name: namespace Outer { class Foo { } namespace Inner { class Foo { } class Test { Foo f1; // = Outer.Inner.Foo Outer.Foo f2; // = Outer.Foo } } } All type names are converted to fully qualified names at compile time. Intermediate Language (IL) code contains no unqualified or partially qualified names. Repeated namespaces You can repeat a namespace declaration, as long as the type names within the namespaces don’t conflict: namespace Outer.Middle.Inner { class Class1 {} } 90 | Chapter 2: C# Language Basics
- 109. namespace Outer.Middle.Inner { class Class2 {} } We can even break the example into two source files such that we could compile each class into a different assembly. Source file 1: namespace Outer.Middle.Inner { class Class1 {} } Source file 2: namespace Outer.Middle.Inner { class Class2 {} } Nested using directives You can nest a using directive within a namespace. This allows you to scope the using directive within a namespace declaration. In the following example, Class1 is visible in one scope but not in another: namespace N1 { class Class1 {} } namespace N2 { using N1; class Class2 : Class1 {} } namespace N2 { class Class3 : Class1 {} // Compile-time error } Aliasing Types and Namespaces Importing a namespace can result in type-name collision. Rather than importing the entire namespace, you can import just the specific types that you need, giving each type an alias: using PropertyInfo2 = System.Reflection.PropertyInfo; class Program { PropertyInfo2 p; } Namespaces | 91 C# Language Basics
- 110. An entire namespace can be aliased, as follows: using R = System.Reflection; class Program { R.PropertyInfo p; } Advanced Namespace Features Extern Extern aliases allow your program to reference two types with the same fully qualified name (i.e., the namespace and type name are identical). This is an unusual scenario and can occur only when the two types come from different assemblies. Consider the following example. Library 1, compiled to Widgets1.dll: namespace Widgets { public class Widget {} } Library 2, compiled to Widgets2.dll: namespace Widgets { public class Widget {} } Application, which references Widgets1.dll and Widgets2.dll: using Widgets; Widget w = new Widget(); The application cannot compile, because Widget is ambiguous. Extern aliases can resolve the ambiguity. The first step is to modify the application’s .csproj file, assign‐ ing a unique alias to each reference: <ItemGroup> <Reference Include="Widgets1"> <Aliases>W1</Aliases> </Reference> <Reference Include="Widgets2"> <Aliases>W2</Aliases> </Reference> </ItemGroup> The second step is to use the extern alias directive: extern alias W1; extern alias W2; W1.Widgets.Widget w1 = new W1.Widgets.Widget(); W2.Widgets.Widget w2 = new W2.Widgets.Widget(); 92 | Chapter 2: C# Language Basics
- 111. Namespace alias qualifiers As we mentioned earlier, names in inner namespaces hide names in outer namespa‐ ces. However, sometimes even the use of a fully qualified type name does not resolve the conflict. Consider the following example: namespace N { class A { static void Main() => new A.B(); // Instantiate class B public class B {} // Nested type } } namespace A { class B {} } The Main method could be instantiating either the nested class B or the class B within the namespace A. The compiler always gives higher precedence to identifiers in the current namespace, in this case the nested B class. To resolve such conflicts, a namespace name can be qualified, relative to one of the following: • The global namespace—the root of all namespaces (identified with the contex‐ • tual keyword global) • The set of extern aliases • The :: token performs namespace alias qualification. In this example, we qualify using the global namespace (this is most commonly seen in autogenerated code to avoid name conflicts): namespace N { class A { static void Main() { System.Console.WriteLine (new A.B()); System.Console.WriteLine (new global::A.B()); } public class B {} } } namespace A { class B {} } Namespaces | 93 C# Language Basics
- 112. Here is an example of qualifying with an alias (adapted from the example in “Extern”): extern alias W1; extern alias W2; W1::Widgets.Widget w1 = new W1::Widgets.Widget(); W2::Widgets.Widget w2 = new W2::Widgets.Widget(); 94 | Chapter 2: C# Language Basics
- 113. 3 Creating Types in C# In this chapter, we delve into types and type members. Classes A class is the most common kind of reference type. The simplest possible class declaration is as follows: class YourClassName { } A more complex class optionally has the following: Preceding the keyword class Attributes and class modifiers. The non-nested class modifiers are public, internal, abstract, sealed, static, unsafe, and partial. Following Your ClassName Generic type parameters and constraints, a base class, and interfaces. Within the braces Class members (these are methods, properties, indexers, events, fields, constructors, overloaded operators, nested types, and a finalizer). This chapter covers all of these constructs except attributes, operator functions, and the unsafe keyword, which are covered in Chapter 4. The following sections enumerate each of the class members. Fields A field is a variable that is a member of a class or struct; for example: class Octopus { string name; public int Age = 10; } 95
- 114. Fields allow the following modifiers: Static modifier static Access modifiers public internal private protected Inheritance modifier new Unsafe code modifier unsafe Read-only modifier readonly Threading modifier volatile There are two popular naming conventions for private fields: camel-cased (e.g., firstName) and camel-cased with an underscore (_firstName). The latter conven‐ tion lets you instantly distinguish private fields from parameters and local variables. The readonly modifier The readonly modifier prevents a field from being modified after construction. A read-only field can be assigned only in its declaration or within the enclosing type’s constructor. Field initialization Field initialization is optional. An uninitialized field has a default value (0, '0', null, false). Field initializers run before constructors: public int Age = 10; A field initializer can contain expressions and call methods: static readonly string TempFolder = System.IO.Path.GetTempPath(); Declaring multiple fields together For convenience, you can declare multiple fields of the same type in a comma- separated list. This is a convenient way for all the fields to share the same attributes and field modifiers: static readonly int legs = 8, eyes = 2; Constants A constant is evaluated statically at compile time, and the compiler literally substi‐ tutes its value whenever used (rather like a macro in C++). A constant can be bool, char, string, any of the built-in numeric types, or an enum type. A constant is declared with the const keyword and must be initialized with a value. For example: 96 | Chapter 3: Creating Types in C#
- 115. public class Test { public const string Message = "Hello World"; } A constant can serve a similar role to a static readonly field, but it is much more restrictive—both in the types you can use and in field initialization semantics. A constant also differs from a static readonly field in that the evaluation of the constant occurs at compile time; thus public static double Circumference (double radius) { return 2 * System.Math.PI * radius; } is compiled to: public static double Circumference (double radius) { return 6.2831853071795862 * radius; } It makes sense for PI to be a constant because its value is predetermined at compile time. In contrast, a static readonly field’s value can potentially differ each time the program is run: static readonly DateTime StartupTime = DateTime.Now; A static readonly field is also advantageous when exposing to other assemblies a value that might change in a later ver‐ sion. For instance, suppose that assembly X exposes a constant as follows: public const decimal ProgramVersion = 2.3; If assembly Y references X and uses this constant, the value 2.3 will be baked into assembly Y when compiled. This means that if X is later recompiled with the constant set to 2.4, Y will still use the old value of 2.3 until Y is recompiled. A static readonly field avoids this problem. Another way of looking at this is that any value that might change in the future is not constant by definition; thus, it should not be represented as one. Constants can also be declared local to a method: void Test() { const double twoPI = 2 * System.Math.PI; ... } Classes | 97 Creating Types in C#
- 116. Nonlocal constants allow the following modifiers: Access modifiers public internal private protected Inheritance modifier new Methods A method performs an action in a series of statements. A method can receive input data from the caller by specifying parameters, and output data back to the caller by specifying a return type. A method can specify a void return type, indicating that it doesn’t return any value to its caller. A method can also output data back to the caller via ref/out parameters. A method’s signature must be unique within the type. A method’s signature compri‐ ses its name and parameter types in order (but not the parameter names, nor the return type). Methods allow the following modifiers: Static modifier static Access modifiers public internal private protected Inheritance modifiers new virtual abstract override sealed Partial method modifier partial Unmanaged code modifiers unsafe extern Asynchronous code modifier async Expression-bodied methods A method that comprises a single expression, such as int Foo (int x) { return x * 2; } can be written more tersely as an expression-bodied method. A fat arrow replaces the braces and return keyword: int Foo (int x) => x * 2; Expression-bodied functions can also have a void return type: void Foo (int x) => Console.WriteLine (x); Local methods You can define a method within another method: void WriteCubes() { Console.WriteLine (Cube (3)); Console.WriteLine (Cube (4)); Console.WriteLine (Cube (5)); 98 | Chapter 3: Creating Types in C#
- 117. int Cube (int value) => value * value * value; } The local method (Cube, in this case) is visible only to the enclosing method (Write Cubes). This simplifies the containing type and instantly signals to anyone looking at the code that Cube is used nowhere else. Another benefit of local methods is that they can access the local variables and parameters of the enclosing method. This has a number of consequences, which we describe in detail in “Capturing Outer Variables” on page 176. Local methods can appear within other function kinds, such as property accessors, constructors, and so on. You can even put local methods inside other local methods and inside lambda expressions that use a statement block (Chapter 4). Local meth‐ ods can be iterators (Chapter 4) or asynchronous (Chapter 14). Static local methods Adding the static modifier to a local method (from C# 8) prevents it from seeing the local variables and parameters of the enclosing method. This helps to reduce coupling and prevents the local method from accidentally referring to variables in the containing method. Local methods and top-level statements Any methods that you declare in top-level statements are treated as local methods. This means that (unless marked as static) they can access the variables in the top-level statements: int x = 3; Foo(); void Foo() => Console.WriteLine (x); Overloading methods Local methods cannot be overloaded. This means that meth‐ ods declared in top-level statements (which are treated as local methods) cannot be overloaded. A type can overload methods (define multiple methods with the same name) as long as the signatures are different. For example, the following methods can all coexist in the same type: void Foo (int x) {...} void Foo (double x) {...} void Foo (int x, float y) {...} void Foo (float x, int y) {...} However, the following pairs of methods cannot coexist in the same type, because the return type and the params modifier are not part of a method’s signature: Classes | 99 Creating Types in C#
- 118. void Foo (int x) {...} float Foo (int x) {...} // Compile-time error void Goo (int[] x) {...} void Goo (params int[] x) {...} // Compile-time error Whether a parameter is pass-by-value or pass-by-reference is also part of the signa‐ ture. For example, Foo(int) can coexist with either Foo(ref int) or Foo(out int). However, Foo(ref int) and Foo(out int) cannot coexist: void Foo (int x) {...} void Foo (ref int x) {...} // OK so far void Foo (out int x) {...} // Compile-time error Instance Constructors Constructors run initialization code on a class or struct. A constructor is defined like a method, except that the method name and return type are reduced to the name of the enclosing type: Panda p = new Panda ("Petey"); // Call constructor public class Panda { string name; // Define field public Panda (string n) // Define constructor { name = n; // Initialization code (set up field) } } Instance constructors allow the following modifiers: Access modifiers public internal private protected Unmanaged code modifiers unsafe extern Single-statement constructors can also be written as expression-bodied members: public Panda (string n) => name = n; Overloading constructors A class or struct may overload constructors. To avoid code duplication, one con‐ structor can call another, using the this keyword: using System; public class Wine { public decimal Price; public int Year; public Wine (decimal price) { Price = price; } 100 | Chapter 3: Creating Types in C#
- 119. public Wine (decimal price, int year) : this (price) { Year = year; } } When one constructor calls another, the called constructor executes first. You can pass an expression into another constructor, as follows: public Wine (decimal price, DateTime year) : this (price, year.Year) { } The expression itself cannot make use of the this reference, for example, to call an instance method. (This is enforced because the object has not been initialized by the constructor at this stage, so any methods that you call on it are likely to fail.) It can, however, call static methods. Implicit parameterless constructors For classes, the C# compiler automatically generates a parameterless public con‐ structor if and only if you do not define any constructors. However, as soon as you define at least one constructor, the parameterless constructor is no longer automatically generated. Constructor and field initialization order We previously saw that fields can be initialized with default values in their declaration: class Player { int shields = 50; // Initialized first int health = 100; // Initialized second } Field initializations occur before the constructor is executed, and in the declaration order of the fields. Nonpublic constructors Constructors do not need to be public. A common reason to have a nonpublic con‐ structor is to control instance creation via a static method call. The static method could be used to return an object from a pool rather than creating a new object, or to return various subclasses based on input arguments: public class Class1 { Class1() {} // Private constructor public static Class1 Create (...) { // Perform custom logic here to return an instance of Class1 ... } } Classes | 101 Creating Types in C#
- 120. Deconstructors A deconstructor (also called a deconstructing method) acts as an approximate oppo‐ site to a constructor: whereas a constructor typically takes a set of values (as parameters) and assigns them to fields, a deconstructor does the reverse and assigns fields back to a set of variables. A deconstruction method must be called Deconstruct and have one or more out parameters, such as in the following class: class Rectangle { public readonly float Width, Height; public Rectangle (float width, float height) { Width = width; Height = height; } public void Deconstruct (out float width, out float height) { width = Width; height = Height; } } The following special syntax calls the deconstructor: var rect = new Rectangle (3, 4); (float width, float height) = rect; // Deconstruction Console.WriteLine (width + " " + height); // 3 4 The second line is the deconstructing call. It creates two local variables and then calls the Deconstruct method. Our deconstructing call is equivalent to the following: float width, height; rect.Deconstruct (out width, out height); Or: rect.Deconstruct (out var width, out var height); Deconstructing calls allow implicit typing, so we could shorten our call to this: (var width, var height) = rect; Or simply this: var (width, height) = rect; 102 | Chapter 3: Creating Types in C#
- 121. You can use C#’s discard symbol (_) if you’re uninterested in one or more variables: var (_, height) = rect; This better indicates your intention than declaring a variable that you never use. If the variables into which you’re deconstructing are already defined, omit the types altogether: float width, height; (width, height) = rect; This is called a deconstructing assignment. You can use a deconstructing assignment to simplify your class’s constructor: public Rectangle (float width, float height) => (Width, Height) = (width, height); You can offer the caller a range of deconstruction options by overloading the Deconstruct method. The Deconstruct method can be an extension method (see “Extension Methods” on page 202). This is a useful trick if you want to deconstruct types that you did not author. From C# 10, you can mix and match existing and new variables when deconstructing: double x1 = 0; (x1, double y2) = rect; Object Initializers To simplify object initialization, any accessible fields or properties of an object can be set via an object initializer directly after construction. For example, consider the following class: public class Bunny { public string Name; public bool LikesCarrots; public bool LikesHumans; public Bunny () {} public Bunny (string n) { Name = n; } } Using object initializers, you can instantiate Bunny objects as follows: // Note parameterless constructors can omit empty parentheses Bunny b1 = new Bunny { Name="Bo", LikesCarrots=true, LikesHumans=false }; Bunny b2 = new Bunny ("Bo") { LikesCarrots=true, LikesHumans=false }; Classes | 103 Creating Types in C#
- 122. The code to construct b1 and b2 is precisely equivalent to the following: Bunny temp1 = new Bunny(); // temp1 is a compiler-generated name temp1.Name = "Bo"; temp1.LikesCarrots = true; temp1.LikesHumans = false; Bunny b1 = temp1; Bunny temp2 = new Bunny ("Bo"); temp2.LikesCarrots = true; temp2.LikesHumans = false; Bunny b2 = temp2; The temporary variables are to ensure that if an exception is thrown during initiali‐ zation, you can’t end up with a half-initialized object. Object Initializers Versus Optional Parameters Instead of using object initializers, we could make Bunny’s constructor accept optional parameters: public Bunny (string name, bool likesCarrots = false, bool likesHumans = false) { Name = name; LikesCarrots = likesCarrots; LikesHumans = likesHumans; } This would allow us to construct a Bunny as follows: Bunny b1 = new Bunny (name: "Bo", likesCarrots: true); Historically, this approach could be advantageous in that it allowed us to make Bunny’s fields (or properties, which we’ll explain shortly) read-only. Making fields or properties read-only is good practice when there’s no valid reason for them to change throughout the life of the object. However, as we’ll see soon in our discussion on properties, the init modifier that was introduced in C# 9 lets us achieve this goal with object initializers. Optional parameters have two drawbacks. The first is that while their use in constructors allows for read-only types, they don’t (easily) allow for nondestruc‐ tive mutation. (We’ll cover nondestructive mutation—and the solution to this problem—in “Records” on page 211.) The second drawback of optional parameters is that when used in public libraries, they hinder backward compatibility. This is because the act of adding an optional parameter at a later date breaks the assembly’s binary compatibility with existing consumers. (This is particularly important when a library is published on NuGet: the problem becomes intractable when a consumer references packages A and B, if A and B each depend on incompatible versions of L.) 104 | Chapter 3: Creating Types in C#
- 123. The difficulty is that each optional parameter value is baked into the calling site. In other words, C# translates our constructor call into this: Bunny b1 = new Bunny ("Bo", true, false); This is problematic if we instantiate the Bunny class from another assembly and later modify Bunny by adding another optional parameter—such as likesCats. Unless the referencing assembly is also recompiled, it will continue to call the (now nonexistent) constructor with three parameters and fail at runtime. (A subtler problem is that if we changed the value of one of the optional parameters, callers in other assemblies would continue to use the old optional value until they were recompiled.) The this Reference The this reference refers to the instance itself. In the following example, the Marry method uses this to set the partner’s mate field: public class Panda { public Panda Mate; public void Marry (Panda partner) { Mate = partner; partner.Mate = this; } } The this reference also disambiguates a local variable or parameter from a field; for example: public class Test { string name; public Test (string name) { this.name = name; } } The this reference is valid only within nonstatic members of a class or struct. Properties Properties look like fields from the outside, but internally they contain logic, like methods do. For example, you can’t tell by looking at the following code whether CurrentPrice is a field or a property: Stock msft = new Stock(); msft.CurrentPrice = 30; msft.CurrentPrice -= 3; Console.WriteLine (msft.CurrentPrice); A property is declared like a field but with a get/set block added. Here’s how to implement CurrentPrice as a property: Classes | 105 Creating Types in C#
- 124. public class Stock { decimal currentPrice; // The private "backing" field public decimal CurrentPrice // The public property { get { return currentPrice; } set { currentPrice = value; } } } get and set denote property accessors. The get accessor runs when the property is read. It must return a value of the property’s type. The set accessor runs when the property is assigned. It has an implicit parameter named value of the property’s type that you typically assign to a private field (in this case, currentPrice). Although properties are accessed in the same way as fields, they differ in that they give the implementer complete control over getting and setting its value. This con‐ trol enables the implementer to choose whatever internal representation is needed without exposing the internal details to the user of the property. In this example, the set method could throw an exception if value was outside a valid range of values. Throughout this book, we use public fields extensively to keep the examples free of distraction. In a real application, you would typically favor public properties over public fields in order to promote encapsulation. Properties allow the following modifiers: Static modifier static Access modifiers public internal private protected Inheritance modifiers new virtual abstract override sealed Unmanaged code modifiers unsafe extern Read-only and calculated properties A property is read-only if it specifies only a get accessor, and it is write-only if it specifies only a set accessor. Write-only properties are rarely used. A property typically has a dedicated backing field to store the underlying data. However, a property can also be computed from other data: decimal currentPrice, sharesOwned; public decimal Worth { get { return currentPrice * sharesOwned; } } 106 | Chapter 3: Creating Types in C#
- 125. Expression-bodied properties You can declare a read-only property, such as the one in the preceding example, more tersely as an expression-bodied property. A fat arrow replaces all the braces and the get and return keywords: public decimal Worth => currentPrice * sharesOwned; With a little extra syntax, set accessors can also be expression-bodied: public decimal Worth { get => currentPrice * sharesOwned; set => sharesOwned = value / currentPrice; } Automatic properties The most common implementation for a property is a getter and/or setter that sim‐ ply reads and writes to a private field of the same type as the property. An automatic property declaration instructs the compiler to provide this implementation. We can improve the first example in this section by declaring CurrentPrice as an automatic property: public class Stock { ... public decimal CurrentPrice { get; set; } } The compiler automatically generates a private backing field of a compiler- generated name that cannot be referred to. The set accessor can be marked private or protected if you want to expose the property as read-only to other types. Automatic properties were introduced in C# 3.0. Property initializers You can add a property initializer to automatic properties, just as with fields: public decimal CurrentPrice { get; set; } = 123; This gives CurrentPrice an initial value of 123. Properties with an initializer can be read-only: public int Maximum { get; } = 999; Just as with read-only fields, read-only automatic properties can also be assigned in the type’s constructor. This is useful in creating immutable (read-only) types. get and set accessibility The get and set accessors can have different access levels. The typical use case for this is to have a public property with an internal or private access modifier on the setter: Classes | 107 Creating Types in C#
- 126. public class Foo { private decimal x; public decimal X { get { return x; } private set { x = Math.Round (value, 2); } } } Notice that you declare the property itself with the more permissive access level (public, in this case) and add the modifier to the accessor you want to be less accessible. Init-only setters From C# 9, you can declare a property accessor with init instead of set: public class Note { public int Pitch { get; init; } = 20; // “Init-only” property public int Duration { get; init; } = 100; // “Init-only” property } These init-only properties act like read-only properties, except that they can also be set via an object initializer: var note = new Note { Pitch = 50 }; After that, the property cannot be altered: note.Pitch = 200; // Error – init-only setter! Init-only properties cannot even be set from inside their class, except via their property initializer, the constructor, or another init-only accessor. The alternative to init-only properties is to have read-only properties that you populate via a constructor: public class Note { public int Pitch { get; } public int Duration { get; } public Note (int pitch = 20, int duration = 100) { Pitch = pitch; Duration = duration; } } Should the class be part of a public library, this approach makes versioning difficult, in that adding an optional parameter to the constructor at a later date breaks binary compatibility with consumers (whereas adding a new init-only property breaks nothing). 108 | Chapter 3: Creating Types in C#
- 127. Init-only properties have another significant advantage, which is that they allow for nondestructive mutation when used in conjunction with records (see “Records” on page 211). As with ordinary set accessors, init-only accessors can provide an implementation: public class Note { readonly int _pitch; public int Pitch { get => _pitch; init => _pitch = value; } ... Notice that the _pitch field is read-only: init-only setters are permitted to modify readonly fields in their own class. (Without this feature, _pitch would need to be writable, and the class would fail at being internally immutable.) Changing a property’s accessor from init to set (or vice versa) is a binary breaking change: anyone that references your assembly will need to recompile their assembly. This should not be an issue when creating wholly immutable types, in that your type will never require properties with a (writable) set accessor. CLR property implementation C# property accessors internally compile to methods called get_XXX and set_XXX: public decimal get_CurrentPrice {...} public void set_CurrentPrice (decimal value) {...} An init accessor is processed like a set accessor, but with an extra flag encoded into the set accessor’s “modreq” metadata (see “Init-only properties” on page 794). Simple nonvirtual property accessors are inlined by the Just-in-Time (JIT) compiler, eliminating any performance difference between accessing a property and a field. Inlining is an optimization in which a method call is replaced with the body of that method. Indexers Indexers provide a natural syntax for accessing elements in a class or struct that encapsulate a list or dictionary of values. Indexers are similar to properties but are accessed via an index argument rather than a property name. The string class has an indexer that lets you access each of its char values via an int index: string s = "hello"; Console.WriteLine (s[0]); // 'h' Console.WriteLine (s[3]); // 'l' The syntax for using indexers is like that for using arrays, except that the index argument(s) can be of any type(s). Classes | 109 Creating Types in C#
- 128. Indexers have the same modifiers as properties (see “Properties” on page 105) and can be called null-conditionally by inserting a question mark before the square bracket (see “Null Operators” on page 74): string s = null; Console.WriteLine (s?[0]); // Writes nothing; no error. Implementing an indexer To write an indexer, define a property called this, specifying the arguments in square brackets: class Sentence { string[] words = "The quick brown fox".Split(); public string this [int wordNum] // indexer { get { return words [wordNum]; } set { words [wordNum] = value; } } } Here’s how we could use this indexer: Sentence s = new Sentence(); Console.WriteLine (s[3]); // fox s[3] = "kangaroo"; Console.WriteLine (s[3]); // kangaroo A type can declare multiple indexers, each with parameters of different types. An indexer can also take more than one parameter: public string this [int arg1, string arg2] { get { ... } set { ... } } If you omit the set accessor, an indexer becomes read-only, and you can use expression-bodied syntax to shorten its definition: public string this [int wordNum] => words [wordNum]; CLR indexer implementation Indexers internally compile to methods called get_Item and set_Item, as follows: public string get_Item (int wordNum) {...} public void set_Item (int wordNum, string value) {...} Using indices and ranges with indexers You can support indices and ranges (see “Indices and ranges” on page 13) in your own classes by defining an indexer with a parameter type of Index or Range. 110 | Chapter 3: Creating Types in C#
- 129. We could extend our previous example by adding the following indexers to the Sentence class: public string this [Index index] => words [index]; public string[] this [Range range] => words [range]; This then enables the following: Sentence s = new Sentence(); Console.WriteLine (s [^1]); // fox string[] firstTwoWords = s [..2]; // (The, quick) Static Constructors A static constructor executes once per type rather than once per instance. A type can define only one static constructor, and it must be parameterless and have the same name as the type: class Test { static Test() { Console.WriteLine ("Type Initialized"); } } The runtime automatically invokes a static constructor just prior to the type being used. Two things trigger this: • Instantiating the type • • Accessing a static member in the type • The only modifiers allowed by static constructors are unsafe and extern. If a static constructor throws an unhandled exception (Chap‐ ter 4), that type becomes unusable for the life of the application. From C# 9, you can also define module initializers, which execute once per assembly (when the assembly is first loaded). To define a module initializer, write a static void method and then apply the [ModuleInitializer] attribute to that method: [System.Runtime.CompilerServices.ModuleInitializer] internal static void InitAssembly() { ... } Static constructors and field initialization order Static field initializers run just before the static constructor is called. If a type has no static constructor, static field initializers will execute just prior to the type being used—or anytime earlier at the whim of the runtime. Classes | 111 Creating Types in C#
- 130. Static field initializers run in the order in which the fields are declared. The follow‐ ing example illustrates this. X is initialized to 0, and Y is initialized to 3: class Foo { public static int X = Y; // 0 public static int Y = 3; // 3 } If we swap the two field initializers around, both fields are initialized to 3. The next example prints 0 followed by 3 because the field initializer that instantiates a Foo executes before X is initialized to 3: Console.WriteLine (Foo.X); // 3 class Foo { public static Foo Instance = new Foo(); public static int X = 3; Foo() => Console.WriteLine (X); // 0 } If we swap the two lines in boldface, the example prints 3 followed by 3. Static Classes A class marked static cannot be instantiated or subclassed, and must be composed solely of static members. The System.Console and System.Math classes are good examples of static classes. Finalizers Finalizers are class-only methods that execute before the garbage collector reclaims the memory for an unreferenced object. The syntax for a finalizer is the name of the class prefixed with the ~ symbol: class Class1 { ~Class1() { ... } } This is actually C# syntax for overriding Object’s Finalize method, and the com‐ piler expands it into the following method declaration: protected override void Finalize() { ... base.Finalize(); } We discuss garbage collection and finalizers fully in Chapter 12. 112 | Chapter 3: Creating Types in C#
- 131. Finalizers allow the following modifier: Unmanaged code modifier unsafe You can write single-statement finalizers using expression-bodied syntax: ~Class1() => Console.WriteLine ("Finalizing"); Partial Types and Methods Partial types allow a type definition to be split—typically across multiple files. A common scenario is for a partial class to be autogenerated from some other source (such as a Visual Studio template or designer), and for that class to be augmented with additional hand-authored methods: // PaymentFormGen.cs - auto-generated partial class PaymentForm { ... } // PaymentForm.cs - hand-authored partial class PaymentForm { ... } Each participant must have the partial declaration; the following is illegal: partial class PaymentForm {} class PaymentForm {} Participants cannot have conflicting members. A constructor with the same param‐ eters, for instance, cannot be repeated. Partial types are resolved entirely by the compiler, which means that each participant must be available at compile time and must reside in the same assembly. You can specify a base class on one or more partial class declarations, as long as the base class, if specified, is the same. In addition, each participant can inde‐ pendently specify interfaces to implement. We cover base classes and interfaces in “Inheritance” on page 115 and “Interfaces” on page 134. The compiler makes no guarantees with regard to field initialization order between partial type declarations. Partial methods A partial type can contain partial methods. These let an autogenerated partial type provide customizable hooks for manual authoring; for example: partial class PaymentForm // In auto-generated file { ... partial void ValidatePayment (decimal amount); } partial class PaymentForm // In hand-authored file { ... Classes | 113 Creating Types in C#
- 132. partial void ValidatePayment (decimal amount) { if (amount > 100) ... } } A partial method consists of two parts: a definition and an implementation. The definition is typically written by a code generator, and the implementation is typi‐ cally manually authored. If an implementation is not provided, the definition of the partial method is compiled away (as is the code that calls it). This allows autogener‐ ated code to be liberal in providing hooks without having to worry about bloat. Partial methods must be void and are implicitly private. They cannot include out parameters. Extended partial methods Extended partial methods (from C# 9) are designed for the reverse code generation scenario, where a programmer defines hooks that a code generator implements. An example of where this might occur is with source generators, a Roslyn feature that lets you feed the compiler an assembly that automatically generates portions of your code. A partial method declaration is extended if it begins with an accessibility modifier: public partial class Test { public partial void M1(); // Extended partial method private partial void M2(); // Extended partial method } The presence of the accessibility modifier doesn’t just affect accessibility: it tells the compiler to treat the declaration differently. Extended partial methods must have implementations; they do not melt away if unimplemented. In this example, both M1 and M2 must have implementations because they each specify accessibility modifiers (public and private). Because they cannot melt away, extended partial methods can return any type and can include out parameters: public partial class Test { public partial bool IsValid (string identifier); internal partial bool TryParse (string number, out int result); } The nameof operator The nameof operator returns the name of any symbol (type, member, variable, and so on) as a string: 114 | Chapter 3: Creating Types in C#
- 133. int count = 123; string name = nameof (count); // name is "count" Its advantage over simply specifying a string is that of static type checking. Tools such as Visual Studio can understand the symbol reference, so if you rename the symbol in question, all of its references will be renamed, too. To specify the name of a type member such as a field or property, include the type as well. This works with both static and instance members: string name = nameof (StringBuilder.Length); This evaluates to Length. To return StringBuilder.Length, you would do this: nameof (StringBuilder) + "." + nameof (StringBuilder.Length); Inheritance A class can inherit from another class to extend or customize the original class. Inheriting from a class lets you reuse the functionality in that class instead of building it from scratch. A class can inherit from only a single class but can itself be inherited by many classes, thus forming a class hierarchy. In this example, we begin by defining a class called Asset: public class Asset { public string Name; } Next, we define classes called Stock and House, which will inherit from Asset. Stock and House get everything an Asset has, plus any additional members that they define: public class Stock : Asset // inherits from Asset { public long SharesOwned; } public class House : Asset // inherits from Asset { public decimal Mortgage; } Here’s how we can use these classes: Stock msft = new Stock { Name="MSFT", SharesOwned=1000 }; Console.WriteLine (msft.Name); // MSFT Console.WriteLine (msft.SharesOwned); // 1000 House mansion = new House { Name="Mansion", Mortgage=250000 }; Inheritance | 115 Creating Types in C#
- 134. Console.WriteLine (mansion.Name); // Mansion Console.WriteLine (mansion.Mortgage); // 250000 The derived classes, Stock and House, inherit the Name field from the base class, Asset. A derived class is also called a subclass. A base class is also called a superclass. Polymorphism References are polymorphic. This means a variable of type x can refer to an object that subclasses x. For instance, consider the following method: public static void Display (Asset asset) { System.Console.WriteLine (asset.Name); } This method can display both a Stock and a House because they are both Assets: Stock msft = new Stock ... ; House mansion = new House ... ; Display (msft); Display (mansion); Polymorphism works on the basis that subclasses (Stock and House) have all the features of their base class (Asset). The converse, however, is not true. If Display was modified to accept a House, you could not pass in an Asset: Display (new Asset()); // Compile-time error public static void Display (House house) // Will not accept Asset { System.Console.WriteLine (house.Mortgage); } Casting and Reference Conversions An object reference can be: • Implicitly upcast to a base class reference • • Explicitly downcast to a subclass reference • Upcasting and downcasting between compatible reference types performs reference conversions: a new reference is (logically) created that points to the same object. An upcast always succeeds; a downcast succeeds only if the object is suitably typed. 116 | Chapter 3: Creating Types in C#
- 135. Upcasting An upcast operation creates a base class reference from a subclass reference: Stock msft = new Stock(); Asset a = msft; // Upcast After the upcast, variable a still references the same Stock object as variable msft. The object being referenced is not itself altered or converted: Console.WriteLine (a == msft); // True Although a and msft refer to the identical object, a has a more restrictive view on that object: Console.WriteLine (a.Name); // OK Console.WriteLine (a.SharesOwned); // Compile-time error The last line generates a compile-time error because the variable a is of type Asset, even though it refers to an object of type Stock. To get to its SharesOwned field, you must downcast the Asset to a Stock. Downcasting A downcast operation creates a subclass reference from a base class reference: Stock msft = new Stock(); Asset a = msft; // Upcast Stock s = (Stock)a; // Downcast Console.WriteLine (s.SharesOwned); // <No error> Console.WriteLine (s == a); // True Console.WriteLine (s == msft); // True As with an upcast, only references are affected—not the underlying object. A down‐ cast requires an explicit cast because it can potentially fail at runtime: House h = new House(); Asset a = h; // Upcast always succeeds Stock s = (Stock)a; // Downcast fails: a is not a Stock If a downcast fails, an InvalidCastException is thrown. This is an example of runtime type checking (we elaborate on this concept in “Static and Runtime Type Checking” on page 127). The as operator The as operator performs a downcast that evaluates to null (rather than throwing an exception) if the downcast fails: Asset a = new Asset(); Stock s = a as Stock; // s is null; no exception thrown This is useful when you’re going to subsequently test whether the result is null: if (s != null) Console.WriteLine (s.SharesOwned); Inheritance | 117 Creating Types in C#
- 136. Without such a test, a cast is advantageous, because if it fails, a more helpful exception is thrown. We can illustrate by com‐ paring the following two lines of code: long shares = ((Stock)a).SharesOwned; // Approach #1 long shares = (a as Stock).SharesOwned; // Approach #2 If a is not a Stock, the first line throws an InvalidCastExcep tion, which is an accurate description of what went wrong. The second line throws a NullReferenceException, which is ambiguous. Was a not a Stock, or was a null? Another way of looking at it is that with the cast operator, you’re saying to the compiler, “I’m certain of a value’s type; if I’m wrong, there’s a bug in my code, so throw an exception!” Whereas with the as operator, you’re uncertain of its type and want to branch according to the outcome at runtime. The as operator cannot perform custom conversions (see “Operator Overloading” on page 239), and it cannot do numeric conversions: long x = 3 as long; // Compile-time error The as and cast operators will also perform upcasts, although this is not terribly useful because an implicit conversion will do the job. The is operator The is operator tests whether a variable matches a pattern. C# supports several kinds of patterns, the most important being a type pattern, where a type name follows the is keyword. In this context, the is operator tests whether a reference conversion would suc‐ ceed—in other words, whether an object derives from a specified class (or imple‐ ments an interface). It is often used to test before downcasting: if (a is Stock) Console.WriteLine (((Stock)a).SharesOwned); The is operator also evaluates to true if an unboxing conversion would succeed (see “The object Type” on page 125). However, it does not consider custom or numeric conversions. The is operator works with many other patterns introduced in recent versions of C#. For a full discussion, see “Patterns” on page 222. Introducing a pattern variable You can introduce a variable while using the is operator: if (a is Stock s) Console.WriteLine (s.SharesOwned); 118 | Chapter 3: Creating Types in C#
- 137. This is equivalent to the following: Stock s; if (a is Stock) { s = (Stock) a; Console.WriteLine (s.SharesOwned); } The variable that you introduce is available for “immediate” consumption, so the following is legal: if (a is Stock s && s.SharesOwned > 100000) Console.WriteLine ("Wealthy"); And it remains in scope outside the is expression, allowing this: if (a is Stock s && s.SharesOwned > 100000) Console.WriteLine ("Wealthy"); Else s = new Stock(); // s is in scope Console.WriteLine (s.SharesOwned); // Still in scope Virtual Function Members A function marked as virtual can be overridden by subclasses wanting to provide a specialized implementation. Methods, properties, indexers, and events can all be declared virtual: public class Asset { public string Name; public virtual decimal Liability => 0; // Expression-bodied property } (Liability => 0 is a shortcut for { get { return 0; } }. For more details on this syntax, see “Expression-bodied properties” on page 107.) A subclass overrides a virtual method by applying the override modifier: public class Stock : Asset { public long SharesOwned; } public class House : Asset { public decimal Mortgage; public override decimal Liability => Mortgage; } By default, the Liability of an Asset is 0. A Stock does not need to specialize this behavior. However, the House specializes the Liability property to return the value of the Mortgage: Inheritance | 119 Creating Types in C#
- 138. House mansion = new House { Name="McMansion", Mortgage=250000 }; Asset a = mansion; Console.WriteLine (mansion.Liability); // 250000 Console.WriteLine (a.Liability); // 250000 The signatures, return types, and accessibility of the virtual and overridden methods must be identical. An overridden method can call its base class implementation via the base keyword (we cover this in “The base Keyword” on page 122). Calling virtual methods from a constructor is potentially dan‐ gerous because authors of subclasses are unlikely to know, when overriding the method, that they are working with a partially initialized object. In other words, the overriding method might end up accessing methods or properties that rely on fields not yet initialized by the constructor. Covariant return types From C# 9, you can override a method (or property get accessor) such that it returns a more derived (subclassed) type. For example: public class Asset { public string Name; public virtual Asset Clone() => new Asset { Name = Name }; } public class House : Asset { public decimal Mortgage; public override House Clone() => new House { Name = Name, Mortgage = Mortgage }; } This is permitted because it does not break the contract that Clone must return an Asset: it returns a House, which is an Asset (and more). Prior to C# 9, you had to override methods with the identical return type: public override Asset Clone() => new House { ... } This still does the job, because the overridden Clone method instantiates a House rather than an Asset. However, to treat the returned object as a House, you must then perform a downcast: House mansion1 = new House { Name="McMansion", Mortgage=250000 }; House mansion2 = (House) mansion1.Clone(); Abstract Classes and Abstract Members A class declared as abstract can never be instantiated. Instead, only its concrete subclasses can be instantiated. 120 | Chapter 3: Creating Types in C#
- 139. Abstract classes are able to define abstract members. Abstract members are like vir‐ tual members except that they don’t provide a default implementation. That imple‐ mentation must be provided by the subclass unless that subclass is also declared abstract: public abstract class Asset { // Note empty implementation public abstract decimal NetValue { get; } } public class Stock : Asset { public long SharesOwned; public decimal CurrentPrice; // Override like a virtual method. public override decimal NetValue => CurrentPrice * SharesOwned; } Hiding Inherited Members A base class and a subclass can define identical members. For example: public class A { public int Counter = 1; } public class B : A { public int Counter = 2; } The Counter field in class B is said to hide the Counter field in class A. Usually, this happens by accident, when a member is added to the base type after an identical member was added to the subtype. For this reason, the compiler generates a warn‐ ing and then resolves the ambiguity as follows: • References to A (at compile time) bind to A.Counter. • • References to B (at compile time) bind to B.Counter. • Occasionally, you want to hide a member deliberately, in which case you can apply the new modifier to the member in the subclass. The new modifier does nothing more than suppress the compiler warning that would otherwise result: public class A { public int Counter = 1; } public class B : A { public new int Counter = 2; } The new modifier communicates your intent to the compiler—and other program‐ mers—that the duplicate member is not an accident. C# overloads the new keyword to have independent meanings in different contexts. Specifically, the new operator is different from the new member modifier. Inheritance | 121 Creating Types in C#
- 140. new versus override Consider the following class hierarchy: public class BaseClass { public virtual void Foo() { Console.WriteLine ("BaseClass.Foo"); } } public class Overrider : BaseClass { public override void Foo() { Console.WriteLine ("Overrider.Foo"); } } public class Hider : BaseClass { public new void Foo() { Console.WriteLine ("Hider.Foo"); } } The differences in behavior between Overrider and Hider are demonstrated in the following code: Overrider over = new Overrider(); BaseClass b1 = over; over.Foo(); // Overrider.Foo b1.Foo(); // Overrider.Foo Hider h = new Hider(); BaseClass b2 = h; h.Foo(); // Hider.Foo b2.Foo(); // BaseClass.Foo Sealing Functions and Classes An overridden function member can seal its implementation with the sealed key‐ word to prevent it from being overridden by further subclasses. In our earlier virtual function member example, we could have sealed House’s implementation of Liability, preventing a class that derives from House from overriding Liability, as follows: public sealed override decimal Liability { get { return Mortgage; } } You can also apply the sealed modifier to the class itself, to prevent subclassing. Sealing a class is more common than sealing a function member. Although you can seal a function member against overriding, you can’t seal a member against being hidden. The base Keyword The base keyword is similar to the this keyword. It serves two essential purposes: • Accessing an overridden function member from the subclass • 122 | Chapter 3: Creating Types in C#
- 141. • Calling a base-class constructor (see the next section) • In this example, House uses the base keyword to access Asset’s implementation of Liability: public class House : Asset { ... public override decimal Liability => base.Liability + Mortgage; } With the base keyword, we access Asset’s Liability property nonvirtually. This means that we will always access Asset’s version of this property—regardless of the instance’s actual runtime type. The same approach works if Liability is hidden rather than overridden. (You can also access hidden members by casting to the base class before invoking the function.) Constructors and Inheritance A subclass must declare its own constructors. The base class’s constructors are accessible to the derived class but are never automatically inherited. For example, if we define Baseclass and Subclass as follows: public class Baseclass { public int X; public Baseclass () { } public Baseclass (int x) { this.X = x; } } public class Subclass : Baseclass { } the following is illegal: Subclass s = new Subclass (123); Subclass must hence “redefine” any constructors it wants to expose. In doing so, however, it can call any of the base class’s constructors via the base keyword: public class Subclass : Baseclass { public Subclass (int x) : base (x) { } } The base keyword works rather like the this keyword except that it calls a con‐ structor in the base class. Base-class constructors always execute first; this ensures that base initialization occurs before specialized initialization. Inheritance | 123 Creating Types in C#
- 142. Implicit calling of the parameterless base-class constructor If a constructor in a subclass omits the base keyword, the base type’s parameterless constructor is implicitly called: public class BaseClass { public int X; public BaseClass() { X = 1; } } public class Subclass : BaseClass { public Subclass() { Console.WriteLine (X); } // 1 } If the base class has no accessible parameterless constructor, subclasses are forced to use the base keyword in their constructors. Constructor and field initialization order When an object is instantiated, initialization takes place in the following order: 1. From subclass to base class: 1. a. Fields are initialized. a. b. Arguments to base-class constructor calls are evaluated. b. 2. From base class to subclass: 2. a. Constructor bodies execute. a. For example: public class B { int x = 1; // Executes 3rd public B (int x) { ... // Executes 4th } } public class D : B { int y = 1; // Executes 1st public D (int x) : base (x + 1) // Executes 2nd { ... // Executes 5th } } 124 | Chapter 3: Creating Types in C#
- 143. Overloading and Resolution Inheritance has an interesting impact on method overloading. Consider the follow‐ ing two overloads: static void Foo (Asset a) { } static void Foo (House h) { } When an overload is called, the most specific type has precedence: House h = new House (...); Foo(h); // Calls Foo(House) The particular overload to call is determined statically (at compile time) rather than at runtime. The following code calls Foo(Asset), even though the runtime type of a is House: Asset a = new House (...); Foo(a); // Calls Foo(Asset) If you cast Asset to dynamic (Chapter 4), the decision as to which overload to call is deferred until runtime and is then based on the object’s actual type: Asset a = new House (...); Foo ((dynamic)a); // Calls Foo(House) The object Type object (System.Object) is the ultimate base class for all types. Any type can be upcast to object. To illustrate how this is useful, consider a general-purpose stack. A stack is a data structure based on the principle of LIFO—“last in, first out.” A stack has two operations: push an object on the stack, and pop an object off the stack. Here is a simple implementation that can hold up to 10 objects: public class Stack { int position; object[] data = new object[10]; public void Push (object obj) { data[position++] = obj; } public object Pop() { return data[--position]; } } Because Stack works with the object type, we can Push and Pop instances of any type to and from the Stack: Stack stack = new Stack(); stack.Push ("sausage"); string s = (string) stack.Pop(); // Downcast, so explicit cast is needed Console.WriteLine (s); // sausage The object Type | 125 Creating Types in C#
- 144. 1 The reference type can also be System.ValueType or System.Enum (Chapter 6). object is a reference type, by virtue of being a class. Despite this, value types, such as int, can also be cast to and from object, and so be added to our stack. This feature of C# is called type unification and is demonstrated here: stack.Push (3); int three = (int) stack.Pop(); When you cast between a value type and object, the CLR must perform some special work to bridge the difference in semantics between value and reference types. This process is called boxing and unboxing. In “Generics” on page 145, we describe how to improve our Stack class to better handle stacks with same-typed elements. Boxing and Unboxing Boxing is the act of converting a value-type instance to a reference-type instance. The reference type can be either the object class or an interface (which we visit later in the chapter).1 In this example, we box an int into an object: int x = 9; object obj = x; // Box the int Unboxing reverses the operation by casting the object back to the original value type: int y = (int)obj; // Unbox the int Unboxing requires an explicit cast. The runtime checks that the stated value type matches the actual object type, and throws an InvalidCastException if the check fails. For instance, the following throws an exception because long does not exactly match int: object obj = 9; // 9 is inferred to be of type int long x = (long) obj; // InvalidCastException The following succeeds, however: object obj = 9; long x = (int) obj; As does this: object obj = 3.5; // 3.5 is inferred to be of type double int x = (int) (double) obj; // x is now 3 In the last example, (double) performs an unboxing, and then (int) performs a numeric conversion. 126 | Chapter 3: Creating Types in C#
- 145. Boxing conversions are crucial in providing a unified type system. The system is not perfect, however: we’ll see in “Generics” on page 145 that variance with arrays and generics supports only reference conversions and not boxing conversions: object[] a1 = new string[3]; // Legal object[] a2 = new int[3]; // Error Copying semantics of boxing and unboxing Boxing copies the value-type instance into the new object, and unboxing copies the contents of the object back into a value-type instance. In the following example, changing the value of i doesn’t change its previously boxed copy: int i = 3; object boxed = i; i = 5; Console.WriteLine (boxed); // 3 Static and Runtime Type Checking C# programs are type-checked both statically (at compile time) and at runtime (by the CLR). Static type checking enables the compiler to verify the correctness of your program without running it. The following code will fail because the compiler enforces static typing: int x = "5"; Runtime type checking is performed by the CLR when you downcast via a reference conversion or unboxing: object y = "5"; int z = (int) y; // Runtime error, downcast failed Runtime type checking is possible because each object on the heap internally stores a little type token. You can retrieve this token by calling the GetType method of object. The GetType Method and typeof Operator All types in C# are represented at runtime with an instance of System.Type. There are two basic ways to get a System.Type object: • Call GetType on the instance. • • Use the typeof operator on a type name. • GetType is evaluated at runtime; typeof is evaluated statically at compile time (when generic type parameters are involved, it’s resolved by the JIT compiler). The object Type | 127 Creating Types in C#
- 146. System.Type has properties for such things as the type’s name, assembly, base type, and so on: Point p = new Point(); Console.WriteLine (p.GetType().Name); // Point Console.WriteLine (typeof (Point).Name); // Point Console.WriteLine (p.GetType() == typeof(Point)); // True Console.WriteLine (p.X.GetType().Name); // Int32 Console.WriteLine (p.Y.GetType().FullName); // System.Int32 public class Point { public int X, Y; } System.Type also has methods that act as a gateway to the runtime’s reflection model, described in Chapter 18. The ToString Method The ToString method returns the default textual representation of a type instance. This method is overridden by all built-in types. Here is an example of using the int type’s ToString method: int x = 1; string s = x.ToString(); // s is "1" You can override the ToString method on custom types as follows: Panda p = new Panda { Name = "Petey" }; Console.WriteLine (p); // Petey public class Panda { public string Name; public override string ToString() => Name; } If you don’t override ToString, the method returns the type name. When you call an overridden object member such as ToString directly on a value type, boxing doesn’t occur. Box‐ ing then occurs only if you cast: int x = 1; string s1 = x.ToString(); // Calling on nonboxed value object box = x; string s2 = box.ToString(); // Calling on boxed value Object Member Listing Here are all the members of object: public class Object { public Object(); public extern Type GetType(); 128 | Chapter 3: Creating Types in C#
- 147. public virtual bool Equals (object obj); public static bool Equals (object objA, object objB); public static bool ReferenceEquals (object objA, object objB); public virtual int GetHashCode(); public virtual string ToString(); protected virtual void Finalize(); protected extern object MemberwiseClone(); } We describe the Equals, ReferenceEquals, and GetHashCode methods in “Equality Comparison” on page 324. Structs A struct is similar to a class, with the following key differences: • A struct is a value type, whereas a class is a reference type. • • A struct does not support inheritance (other than implicitly deriving from • object, or more precisely, System.ValueType). A struct can have all of the members that a class can, except for a finalizer. And because it cannot be subclassed, members cannot be marked as virtual, abstract, or protected. Prior to C# 10, structs were further prohibited from defining fields initializers and parameterless constructors. Although this prohibition has now been relaxed—primarily for the ben‐ efit of record structs (see “Records” on page 211)—it’s worth thinking carefully before defining these constructs, as they can result in confusing behavior that we’ll describe in “Struct Construction Semantics” on page 129. A struct is appropriate when value-type semantics are desirable. Good examples of structs are numeric types, where it is more natural for assignment to copy a value rather than a reference. Because a struct is a value type, each instance does not require instantiation of an object on the heap; this results in useful savings when creating many instances of a type. For instance, creating an array of value type elements requires only a single heap allocation. Because structs are value types, an instance cannot be null. The default value for a struct is an empty instance, with all fields empty (set to their default values). Struct Construction Semantics Unlike with classes, every field in a struct must be explicitly assigned in the con‐ structor (or field initializer). For example: Structs | 129 Creating Types in C#
- 148. struct Point { int x, y; public Point (int x, int y) { this.x = x; this.y = y; } // OK } If we added the following constructor, the struct would not compile, because y would remain unassigned: public Point (int x) { this.x = x; } // Not OK The default constructor In addition to any constructors that you define, a struct always has an implicit parameterless constructor that performs a bitwise-zeroing of its fields (setting them to their default values): Point p = new Point(); // p.x and p.y will be 0 struct Point { int x, y; } Even when you define a parameterless constructor of your own, the implicit param‐ eterless constructor still exists and can be accessed via the default keyword: Point p1 = new Point(); // p1.x and p1.y will be 1 Point p2 = default; // p2.x and p2.y will be 0 struct Point { int x = 1; int y; public Point() => y = 1; } In this example, we initialized x to 1 via a field initializer, and we initialized y to 1 via the parameterless constructor. And yet with the default keyword, we were still able to create a Point that bypassed both initializations. The default constructor can be accessed other ways, too, as the following example illustrates: var points = new Point[10]; // Each point in the array will be (0,0) var test = new Test(); // test.p will be (0,0) class Test { Point p; } Having what amounts to two parameterless constructors can be a source of confusion and is arguably a good reason to avoid defining field initializers and explicit parameterless con‐ structors in structs. A good strategy with structs is to design them such that their default value is a valid state, thereby making initialization redundant. For example, rather than this: struct WebOptions { public string Protocol { get; set; } = "https"; } consider the following: 130 | Chapter 3: Creating Types in C#
- 149. struct WebOptions { string protocol; public string Protocol { get => protocol ?? "https"; set => protocol = value; } } Read-Only Structs and Functions You can apply the readonly modifier to a struct to enforce that all fields are readonly; this aids in declaring intent as well as affording the compiler more optimization freedom: readonly struct Point { public readonly int X, Y; // X and Y must be readonly } If you need to apply readonly at a more granular level, you can apply the readonly modifier (from C# 8) to a struct’s functions. This ensures that if the function attempts to modify any field, a compile-time error is generated: struct Point { public int X, Y; public readonly void ResetX() => X = 0; // Error! } If a readonly function calls a non-readonly function, the compiler generates a warning (and defensively copies the struct to avoid the possibility of a mutation). Ref Structs Ref structs were introduced in C# 7.2 as a niche feature pri‐ marily for the benefit of the Span<T> and ReadOnlySpan<T> structs that we describe in Chapter 23 (and the highly opti‐ mized Utf8JsonReader that we describe in Chapter 11). These structs help with a micro-optimization technique that aims to reduce memory allocations. Unlike reference types, whose instances always live on the heap, value types live in-place (wherever the variable was declared). If a value type appears as a parameter or local variable, it will reside on the stack: void SomeMethod() { Point p; // p will reside on the stack } struct Point { public int X, Y; } Structs | 131 Creating Types in C#
- 150. But if a value type appears as a field in a class, it will reside on the heap: class MyClass { Point p; // Lives on heap, because MyClass instances live on the heap } Similarly, arrays of structs live on the heap, and boxing a struct sends it to the heap. Adding the ref modifier to a struct’s declaration ensures that it can only ever reside on the stack. Attempting to use a ref struct in such a way that it could reside on the heap generates a compile-time error: var points = new Point [100]; // Error: will not compile! ref struct Point { public int X, Y; } class MyClass { Point P; } // Error: will not compile! Ref structs were introduced mainly for the benefit of the Span<T> and ReadOnly Span<T> structs. Because Span<T> and ReadOnlySpan<T> instances can exist only on the stack, it’s possible for them to safely wrap stack-allocated memory. Ref structs cannot partake in any C# feature that directly or indirectly introduces the possibility of existing on the heap. This includes a number of advanced C# features that we describe in Chapter 4, namely lambda expressions, iterators, and asynchronous functions (because, behind the scenes, these features all create hidden classes with fields). Also, ref structs cannot appear inside non-ref structs, and they cannot implement interfaces (because this could result in boxing). Access Modifiers To promote encapsulation, a type or type member can limit its accessibility to other types and other assemblies by adding one of five access modifiers to the declaration: public Fully accessible. This is the implicit accessibility for members of an enum or interface. internal Accessible only within the containing assembly or friend assemblies. This is the default accessibility for non-nested types. private Accessible only within the containing type. This is the default accessibility for members of a class or struct. protected Accessible only within the containing type or subclasses. protected internal The union of protected and internal accessibility. A member that is protected internal is accessible in two ways. 132 | Chapter 3: Creating Types in C#
- 151. private protected (from C# 7.2) The intersection of protected and internal accessibility. A member that is private protected is accessible only within the containing type, or subclasses that reside in the same assembly (making it less accessible than protected or internal alone). Examples Class2 is accessible from outside its assembly; Class1 is not: class Class1 {} // Class1 is internal (default) public class Class2 {} ClassB exposes field x to other types in the same assembly; ClassA does not: class ClassA { int x; } // x is private (default) class ClassB { internal int x; } Functions within Subclass can call Bar but not Foo: class BaseClass { void Foo() {} // Foo is private (default) protected void Bar() {} } class Subclass : BaseClass { void Test1() { Foo(); } // Error - cannot access Foo void Test2() { Bar(); } // OK } Friend Assemblies You can expose internal members to other friend assemblies by adding the System.Runtime.CompilerServices.InternalsVisibleTo assembly attribute, specifying the name of the friend assembly as follows: [assembly: InternalsVisibleTo ("Friend")] If the friend assembly has a strong name (see Chapter 17), you must specify its full 160-byte public key: [assembly: InternalsVisibleTo ("StrongFriend, PublicKey=0024f000048c...")] You can extract the full public key from a strongly named assembly with a LINQ query (we explain LINQ in detail in Chapter 8): string key = string.Join ("", Assembly.GetExecutingAssembly().GetName().GetPublicKey() .Select (b => b.ToString ("x2"))); The companion sample in LINQPad invites you to browse to an assembly and then copies the assembly’s full public key to the clipboard. Access Modifiers | 133 Creating Types in C#
- 152. Accessibility Capping A type caps the accessibility of its declared members. The most common example of capping is when you have an internal type with public members. For example, consider this: class C { public void Foo() {} } C’s (default) internal accessibility caps Foo’s accessibility, effectively making Foo internal. A common reason Foo would be marked public is to make for easier refactoring should C later be changed to public. Restrictions on Access Modifiers When overriding a base class function, accessibility must be identical on the over‐ ridden function; for example: class BaseClass { protected virtual void Foo() {} } class Subclass1 : BaseClass { protected override void Foo() {} } // OK class Subclass2 : BaseClass { public override void Foo() {} } // Error (An exception is when overriding a protected internal method in another assem‐ bly, in which case the override must simply be protected.) The compiler prevents any inconsistent use of access modifiers. For example, a subclass itself can be less accessible than a base class, but not more: internal class A {} public class B : A {} // Error Interfaces An interface is similar to a class, but only specifies behavior and does not hold state (data). Consequently: • An interface can define only functions and not fields. • • Interface members are implicitly abstract. (Although nonabstract functions are • permitted from C# 8, this is considered a special case, which we describe in “Default Interface Members” on page 139.) • A class (or struct) can implement multiple interfaces. In contrast, a class can • inherit from only a single class, and a struct cannot inherit at all (aside from deriving from System.ValueType). An interface declaration is like a class declaration, but it (typically) provides no implementation for its members because its members are implicitly abstract. These members will be implemented by the classes and structs that implement the inter‐ face. An interface can contain only functions, that is, methods, properties, events, and indexers (which noncoincidentally are precisely the members of a class that can be abstract). 134 | Chapter 3: Creating Types in C#
- 153. As defined in System.Collections, here is the definition of the IEnumerator interface: public interface IEnumerator { bool MoveNext(); object Current { get; } void Reset(); } Interface members are always implicitly public and cannot declare an access modi‐ fier. Implementing an interface means providing a public implementation for all of its members: internal class Countdown : IEnumerator { int count = 11; public bool MoveNext() => count-- > 0; public object Current => count; public void Reset() { throw new NotSupportedException(); } } You can implicitly cast an object to any interface that it implements: IEnumerator e = new Countdown(); while (e.MoveNext()) Console.Write (e.Current); // 109876543210 Even though Countdown is an internal class, its members that implement IEnumerator can be called publicly by casting an instance of Countdown to IEnumerator. For instance, if a pub‐ lic type in the same assembly defined a method as follows: public static class Util { public static object GetCountDown() => new CountDown(); } a caller from another assembly could do this: IEnumerator e = (IEnumerator) Util.GetCountDown(); e.MoveNext(); If IEnumerator were itself defined as internal, this wouldn’t be possible. Extending an Interface Interfaces can derive from other interfaces; for instance: public interface IUndoable { void Undo(); } public interface IRedoable : IUndoable { void Redo(); } IRedoable “inherits” all the members of IUndoable. In other words, types that implement IRedoable must also implement the members of IUndoable. Interfaces | 135 Creating Types in C#
- 154. Explicit Interface Implementation Implementing multiple interfaces can sometimes result in a collision between member signatures. You can resolve such collisions by explicitly implementing an interface member. Consider the following example: interface I1 { void Foo(); } interface I2 { int Foo(); } public class Widget : I1, I2 { public void Foo() { Console.WriteLine ("Widget's implementation of I1.Foo"); } int I2.Foo() { Console.WriteLine ("Widget's implementation of I2.Foo"); return 42; } } Because I1 and I2 have conflicting Foo signatures, Widget explicitly implements I2’s Foo method. This lets the two methods coexist in one class. The only way to call an explicitly implemented member is to cast to its interface: Widget w = new Widget(); w.Foo(); // Widget's implementation of I1.Foo ((I1)w).Foo(); // Widget's implementation of I1.Foo ((I2)w).Foo(); // Widget's implementation of I2.Foo Another reason to explicitly implement interface members is to hide members that are highly specialized and distracting to a type’s normal use case. For example, a type that implements ISerializable would typically want to avoid flaunting its ISerializable members unless explicitly cast to that interface. Implementing Interface Members Virtually An implicitly implemented interface member is, by default, sealed. It must be marked virtual or abstract in the base class in order to be overridden: public interface IUndoable { void Undo(); } public class TextBox : IUndoable { public virtual void Undo() => Console.WriteLine ("TextBox.Undo"); } public class RichTextBox : TextBox { public override void Undo() => Console.WriteLine ("RichTextBox.Undo"); } 136 | Chapter 3: Creating Types in C#
- 155. Calling the interface member through either the base class or the interface calls the subclass’s implementation: RichTextBox r = new RichTextBox(); r.Undo(); // RichTextBox.Undo ((IUndoable)r).Undo(); // RichTextBox.Undo ((TextBox)r).Undo(); // RichTextBox.Undo An explicitly implemented interface member cannot be marked virtual, nor can it be overridden in the usual manner. It can, however, be reimplemented. Reimplementing an Interface in a Subclass A subclass can reimplement any interface member already implemented by a base class. Reimplementation hijacks a member implementation (when called through the interface) and works whether or not the member is virtual in the base class. It also works whether a member is implemented implicitly or explicitly—although it works best in the latter case, as we will demonstrate. In the following example, TextBox implements IUndoable.Undo explicitly, and so it cannot be marked as virtual. To “override” it, RichTextBox must reimplement IUndoable’s Undo method: public interface IUndoable { void Undo(); } public class TextBox : IUndoable { void IUndoable.Undo() => Console.WriteLine ("TextBox.Undo"); } public class RichTextBox : TextBox, IUndoable { public void Undo() => Console.WriteLine ("RichTextBox.Undo"); } Calling the reimplemented member through the interface calls the subclass’s implementation: RichTextBox r = new RichTextBox(); r.Undo(); // RichTextBox.Undo Case 1 ((IUndoable)r).Undo(); // RichTextBox.Undo Case 2 Assuming the same RichTextBox definition, suppose that TextBox implemented Undo implicitly: public class TextBox : IUndoable { public void Undo() => Console.WriteLine ("TextBox.Undo"); } This would give us another way to call Undo, which would “break” the system, as shown in Case 3: RichTextBox r = new RichTextBox(); r.Undo(); // RichTextBox.Undo Case 1 Interfaces | 137 Creating Types in C#
- 156. ((IUndoable)r).Undo(); // RichTextBox.Undo Case 2 ((TextBox)r).Undo(); // TextBox.Undo Case 3 Case 3 demonstrates that reimplementation hijacking is effective only when a mem‐ ber is called through the interface and not through the base class. This is usually undesirable in that it can create inconsistent semantics. This makes reimplementa‐ tion most appropriate as a strategy for overriding explicitly implemented interface members. Alternatives to interface reimplementation Even with explicit member implementation, interface reimplementation is problem‐ atic for a couple of reasons: • The subclass has no way to call the base class method. • • The base class author might not anticipate that a method would be reimple‐ • mented and might not allow for the potential consequences. Reimplementation can be a good last resort when subclassing hasn’t been anticipa‐ ted. A better option, however, is to design a base class such that reimplementation will never be required. There are two ways to achieve this: • When implicitly implementing a member, mark it virtual if appropriate. • • When explicitly implementing a member, use the following pattern if you • anticipate that subclasses might need to override any logic: public class TextBox : IUndoable { void IUndoable.Undo() => Undo(); // Calls method below protected virtual void Undo() => Console.WriteLine ("TextBox.Undo"); } public class RichTextBox : TextBox { protected override void Undo() => Console.WriteLine("RichTextBox.Undo"); } If you don’t anticipate any subclassing, you can mark the class as sealed to preempt interface reimplementation. Interfaces and Boxing Converting a struct to an interface causes boxing. Calling an implicitly implemented member on a struct does not cause boxing: interface I { void Foo(); } struct S : I { public void Foo() {} } ... S s = new S(); s.Foo(); // No boxing. 138 | Chapter 3: Creating Types in C#
- 157. I i = s; // Box occurs when casting to interface. i.Foo(); Default Interface Members From C# 8, you can add a default implementation to an interface member, making it optional to implement: interface ILogger { void Log (string text) => Console.WriteLine (text); } This is advantageous if you want to add a member to an interface defined in a popular library without breaking (potentially thousands of) implementations. Default implementations are always explicit, so if a class implementing ILogger fails to define a Log method, the only way to call it is through the interface: class Logger : ILogger { } ... ((ILogger)new Logger()).Log ("message"); This prevents a problem of multiple implementation inheritance: if the same default member is added to two interfaces that a class implements, there is never an ambiguity as to which member is called. Interfaces can also now define static members (including fields), which can be accessed from code inside default implementations: interface ILogger { void Log (string text) => Console.WriteLine (Prefix + text); static string Prefix = ""; } Because interface members are implicitly public, you can also access static members from the outside: ILogger.Prefix = "File log: "; You can restrict this by adding an accessibility modifier to the static interface member (such as private, protected, or internal). Instance fields are (still) prohibited. This is in line with the principle of interfaces, which is to define behavior, not state. Interfaces | 139 Creating Types in C#
- 158. Writing a Class Versus an Interface As a guideline: • Use classes and subclasses for types that naturally share an implementation. • • Use interfaces for types that have independent implementations. • Consider the following classes: abstract class Animal {} abstract class Bird : Animal {} abstract class Insect : Animal {} abstract class FlyingCreature : Animal {} abstract class Carnivore : Animal {} // Concrete classes: class Ostrich : Bird {} class Eagle : Bird, FlyingCreature, Carnivore {} // Illegal class Bee : Insect, FlyingCreature {} // Illegal class Flea : Insect, Carnivore {} // Illegal The Eagle, Bee, and Flea classes do not compile because inheriting from multiple classes is prohibited. To resolve this, we must convert some of the types to inter‐ faces. The question then arises, which types? Following our general rule, we could say that insects share an implementation, and birds share an implementation, so they remain classes. In contrast, flying creatures have independent mechanisms for flying, and carnivores have independent strategies for eating animals, so we would convert FlyingCreature and Carnivore to interfaces: interface IFlyingCreature {} interface ICarnivore {} In a typical scenario, Bird and Insect might correspond to a Windows control and a web control; FlyingCreature and Carnivore might correspond to IPrintable and IUndoable. Enums An enum is a special value type that lets you specify a group of named numeric constants. For example: public enum BorderSide { Left, Right, Top, Bottom } We can use this enum type as follows: BorderSide topSide = BorderSide.Top; bool isTop = (topSide == BorderSide.Top); // true Each enum member has an underlying integral value. These are by default: • Underlying values are of type int. • 140 | Chapter 3: Creating Types in C#
- 159. • The constants 0, 1, 2... are automatically assigned in the declaration order of the • enum members. You can specify an alternative integral type, as follows: public enum BorderSide : byte { Left, Right, Top, Bottom } You can also specify an explicit underlying value for each enum member: public enum BorderSide : byte { Left=1, Right=2, Top=10, Bottom=11 } The compiler also lets you explicitly assign some of the enum members. The unassigned enum members keep incrementing from the last explicit value. The preceding example is equiva‐ lent to the following: public enum BorderSide : byte { Left=1, Right, Top=10, Bottom } Enum Conversions You can convert an enum instance to and from its underlying integral value with an explicit cast: int i = (int) BorderSide.Left; BorderSide side = (BorderSide) i; bool leftOrRight = (int) side <= 2; You can also explicitly cast one enum type to another. Suppose that Horizontal Alignment is defined as follows: public enum HorizontalAlignment { Left = BorderSide.Left, Right = BorderSide.Right, Center } A translation between the enum types uses the underlying integral values: HorizontalAlignment h = (HorizontalAlignment) BorderSide.Right; // same as: HorizontalAlignment h = (HorizontalAlignment) (int) BorderSide.Right; The numeric literal 0 is treated specially by the compiler in an enum expression and does not require an explicit cast: BorderSide b = 0; // No cast required if (b == 0) ... There are two reasons for the special treatment of 0: • The first member of an enum is often used as the “default” value. • • For combined enum types, 0 means “no flags.” • Enums | 141 Creating Types in C#
- 160. Flags Enums You can combine enum members. To prevent ambiguities, members of a combina‐ ble enum require explicitly assigned values, typically in powers of two: [Flags] enum BorderSides { None=0, Left=1, Right=2, Top=4, Bottom=8 } or: enum BorderSides { None=0, Left=1, Right=1<<1, Top=1<<2, Bottom=1<<3 } To work with combined enum values, you use bitwise operators such as | and &. These operate on the underlying integral values: BorderSides leftRight = BorderSides.Left | BorderSides.Right; if ((leftRight & BorderSides.Left) != 0) Console.WriteLine ("Includes Left"); // Includes Left string formatted = leftRight.ToString(); // "Left, Right" BorderSides s = BorderSides.Left; s |= BorderSides.Right; Console.WriteLine (s == leftRight); // True s ^= BorderSides.Right; // Toggles BorderSides.Right Console.WriteLine (s); // Left By convention, the Flags attribute should always be applied to an enum type when its members are combinable. If you declare such an enum without the Flags attribute, you can still combine members, but calling ToString on an enum instance will emit a number rather than a series of names. By convention, a combinable enum type is given a plural rather than singular name. For convenience, you can include combination members within an enum declara‐ tion itself: [Flags] enum BorderSides { None=0, Left=1, Right=1<<1, Top=1<<2, Bottom=1<<3, LeftRight = Left | Right, TopBottom = Top | Bottom, All = LeftRight | TopBottom } Enum Operators The operators that work with enums are: = == != < > <= >= + - ^ & | ˜ += -= ++ -- sizeof 142 | Chapter 3: Creating Types in C#
- 161. The bitwise, arithmetic, and comparison operators return the result of processing the underlying integral values. Addition is permitted between an enum and an integral type, but not between two enums. Type-Safety Issues Consider the following enum: public enum BorderSide { Left, Right, Top, Bottom } Because an enum can be cast to and from its underlying integral type, the actual value it can have might fall outside the bounds of a legal enum member: BorderSide b = (BorderSide) 12345; Console.WriteLine (b); // 12345 The bitwise and arithmetic operators can produce similarly invalid values: BorderSide b = BorderSide.Bottom; b++; // No errors An invalid BorderSide would break the following code: void Draw (BorderSide side) { if (side == BorderSide.Left) {...} else if (side == BorderSide.Right) {...} else if (side == BorderSide.Top) {...} else {...} // Assume BorderSide.Bottom } One solution is to add another else clause: ... else if (side == BorderSide.Bottom) ... else throw new ArgumentException ("Invalid BorderSide: " + side, "side"); Another workaround is to explicitly check an enum value for validity. The static Enum.IsDefined method does this job: BorderSide side = (BorderSide) 12345; Console.WriteLine (Enum.IsDefined (typeof (BorderSide), side)); // False Unfortunately, Enum.IsDefined does not work for flagged enums. However, the following helper method (a trick dependent on the behavior of Enum.ToString()) returns true if a given flagged enum is valid: for (int i = 0; i <= 16; i++) { BorderSides side = (BorderSides)i; Console.WriteLine (IsFlagDefined (side) + " " + side); } bool IsFlagDefined (Enum e) { decimal d; return !decimal.TryParse(e.ToString(), out d); Enums | 143 Creating Types in C#
- 162. } [Flags] public enum BorderSides { Left=1, Right=2, Top=4, Bottom=8 } Nested Types A nested type is declared within the scope of another type: public class TopLevel { public class Nested { } // Nested class public enum Color { Red, Blue, Tan } // Nested enum } A nested type has the following features: • It can access the enclosing type’s private members and everything else the • enclosing type can access. • You can declare it with the full range of access modifiers rather than just • public and internal. • The default accessibility for a nested type is private rather than internal. • • Accessing a nested type from outside the enclosing type requires qualification • with the enclosing type’s name (like when accessing static members). For example, to access Color.Red from outside our TopLevel class, we’d need to do this: TopLevel.Color color = TopLevel.Color.Red; All types (classes, structs, interfaces, delegates, and enums) can be nested within either a class or a struct. Here is an example of accessing a private member of a type from a nested type: public class TopLevel { static int x; class Nested { static void Foo() { Console.WriteLine (TopLevel.x); } } } Here is an example of applying the protected access modifier to a nested type: public class TopLevel { protected class Nested { } } public class SubTopLevel : TopLevel { 144 | Chapter 3: Creating Types in C#
- 163. static void Foo() { new TopLevel.Nested(); } } Here is an example of referring to a nested type from outside the enclosing type: public class TopLevel { public class Nested { } } class Test { TopLevel.Nested n; } Nested types are used heavily by the compiler itself when it generates private classes that capture state for constructs such as iterators and anonymous methods. If the sole reason for using a nested type is to avoid cluttering a namespace with too many types, consider using a nested namespace instead. A nested type should be used because of its stronger access control restrictions, or when the nested class must access private members of the containing class. Generics C# has two separate mechanisms for writing code that is reusable across different types: inheritance and generics. Whereas inheritance expresses reusability with a base type, generics express reusability with a “template” that contains “placeholder” types. Generics, when compared to inheritance, can increase type safety and reduce casting and boxing. C# generics and C++ templates are similar concepts, but they work differently. We explain this difference in “C# Generics Versus C++ Templates” on page 157. Generic Types A generic type declares type parameters—placeholder types to be filled in by the consumer of the generic type, which supplies the type arguments. Here is a generic type Stack<T>, designed to stack instances of type T. Stack<T> declares a single type parameter T: public class Stack<T> { int position; T[] data = new T[100]; public void Push (T obj) => data[position++] = obj; public T Pop() => data[--position]; } Generics | 145 Creating Types in C#
- 164. We can use Stack<T> as follows: var stack = new Stack<int>(); stack.Push (5); stack.Push (10); int x = stack.Pop(); // x is 10 int y = stack.Pop(); // y is 5 Stack<int> fills in the type parameter T with the type argument int, implicitly cre‐ ating a type on the fly (the synthesis occurs at runtime). Attempting to push a string onto our Stack<int> would, however, produce a compile-time error. Stack<int> effectively has the following definition (substitutions appear in bold, with the class name hashed out to avoid confusion): public class ### { int position; int[] data = new int[100]; public void Push (int obj) => data[position++] = obj; public int Pop() => data[--position]; } Technically, we say that Stack<T> is an open type, whereas Stack<int> is a closed type. At runtime, all generic type instances are closed—with the placeholder types filled in. This means that the following statement is illegal: var stack = new Stack<T>(); // Illegal: What is T? However, it’s legal if it’s within a class or method that itself defines T as a type parameter: public class Stack<T> { ... public Stack<T> Clone() { Stack<T> clone = new Stack<T>(); // Legal ... } } Why Generics Exist Generics exist to write code that is reusable across different types. Suppose that we need a stack of integers, but we don’t have generic types. One solution would be to hardcode a separate version of the class for every required element type (e.g., IntStack, StringStack, etc.). Clearly, this would cause considerable code duplication. Another solution would be to write a stack that is generalized by using object as the element type: public class ObjectStack { int position; object[] data = new object[10]; 146 | Chapter 3: Creating Types in C#
- 165. public void Push (object obj) => data[position++] = obj; public object Pop() => data[--position]; } An ObjectStack, however, wouldn’t work as well as a hardcoded IntStack for specifically stacking integers. An ObjectStack would require boxing and downcast‐ ing that could not be checked at compile time: // Suppose we just want to store integers here: ObjectStack stack = new ObjectStack(); stack.Push ("s"); // Wrong type, but no error! int i = (int)stack.Pop(); // Downcast - runtime error What we need is both a general implementation of a stack that works for all element types as well as a way to easily specialize that stack to a specific element type for increased type safety and reduced casting and boxing. Generics give us precisely this by allowing us to parameterize the element type. Stack<T> has the benefits of both ObjectStack and IntStack. Like ObjectStack, Stack<T> is written once to work generally across all types. Like IntStack, Stack<T> is specialized for a particular type—the beauty is that this type is T, which we substitute on the fly. ObjectStack is functionally equivalent to Stack<object>. Generic Methods A generic method declares type parameters within the signature of a method. With generic methods, many fundamental algorithms can be implemented in a general-purpose way. Here is a generic method that swaps the contents of two variables of any type T: static void Swap<T> (ref T a, ref T b) { T temp = a; a = b; b = temp; } Swap<T> is called as follows: int x = 5; int y = 10; Swap (ref x, ref y); Generally, there is no need to supply type arguments to a generic method, because the compiler can implicitly infer the type. If there is ambiguity, generic methods can be called with type arguments as follows: Swap<int> (ref x, ref y); Generics | 147 Creating Types in C#
- 166. Within a generic type, a method is not classed as generic unless it introduces type parameters (with the angle bracket syntax). The Pop method in our generic stack merely uses the type’s existing type parameter, T, and is not classed as a generic method. Methods and types are the only constructs that can introduce type parameters. Properties, indexers, events, fields, constructors, operators, and so on cannot declare type parameters, although they can partake in any type parameters already declared by their enclosing type. In our generic stack example, for instance, we could write an indexer that returns a generic item: public T this [int index] => data [index]; Similarly, constructors can partake in existing type parameters but not introduce them: public Stack<T>() { } // Illegal Declaring Type Parameters Type parameters can be introduced in the declaration of classes, structs, interfaces, delegates (covered in Chapter 4), and methods. Other constructs, such as proper‐ ties, cannot introduce a type parameter, but they can use one. For example, the property Value uses T: public struct Nullable<T> { public T Value { get; } } A generic type or method can have multiple parameters: class Dictionary<TKey, TValue> {...} To instantiate: Dictionary<int,string> myDict = new Dictionary<int,string>(); Or: var myDict = new Dictionary<int,string>(); Generic type names and method names can be overloaded as long as the number of type parameters is different. For example, the following three type names do not conflict: class A {} class A<T> {} class A<T1,T2> {} By convention, generic types and methods with a single type parameter typically name their parameter T, as long as the intent of the parameter is clear. When using multiple type parameters, each parameter is prefixed with T but has a more descriptive name. 148 | Chapter 3: Creating Types in C#
- 167. typeof and Unbound Generic Types Open generic types do not exist at runtime: they are closed as part of compilation. However, it is possible for an unbound generic type to exist at runtime—purely as a Type object. The only way to specify an unbound generic type in C# is via the typeof operator: class A<T> {} class A<T1,T2> {} ... Type a1 = typeof (A<>); // Unbound type (notice no type arguments). Type a2 = typeof (A<,>); // Use commas to indicate multiple type args. Open generic types are used in conjunction with the Reflection API (Chapter 18). You can also use the typeof operator to specify a closed type: Type a3 = typeof (A<int,int>); Or, you can specify an open type (which is closed at runtime): class B<T> { void X() { Type t = typeof (T); } } The default Generic Value You can use the default keyword to get the default value for a generic type parame‐ ter. The default value for a reference type is null, and the default value for a value type is the result of bitwise-zeroing the value type’s fields: static void Zap<T> (T[] array) { for (int i = 0; i < array.Length; i++) array[i] = default(T); } From C# 7.1, you can omit the type argument for cases in which the compiler is able to infer it. We could replace the last line of code with this: array[i] = default; Generic Constraints By default, you can substitute a type parameter with any type whatsoever. Con‐ straints can be applied to a type parameter to require more specific type arguments. These are the possible constraints: where T : base-class // Base-class constraint where T : interface // Interface constraint where T : class // Reference-type constraint where T : class? // (see "Nullable Reference Types" in Chapter 1) where T : struct // Value-type constraint (excludes Nullable types) where T : unmanaged // Unmanaged constraint where T : new() // Parameterless constructor constraint where U : T // Naked type constraint Generics | 149 Creating Types in C#
- 168. where T : notnull // Non-nullable value type, or (from C# 8) // a non-nullable reference type In the following example, GenericClass<T,U> requires T to derive from (or be identical to) SomeClass and implement Interface1, and requires U to provide a parameterless constructor: class SomeClass {} interface Interface1 {} class GenericClass<T,U> where T : SomeClass, Interface1 where U : new() {...} You can apply constraints wherever type parameters are defined, in both methods and type definitions. A base-class constraint specifies that the type parameter must subclass (or match) a particular class; an interface constraint specifies that the type parameter must implement that interface. These constraints allow instances of the type parameter to be implicitly converted to that class or interface. For example, suppose that we want to write a generic Max method, which returns the maximum of two values. We can take advantage of the generic interface defined in the System namespace called IComparable<T>: public interface IComparable<T> // Simplified version of interface { int CompareTo (T other); } CompareTo returns a positive number if this is greater than other. Using this interface as a constraint, we can write a Max method as follows (to avoid distraction, null checking is omitted): static T Max <T> (T a, T b) where T : IComparable<T> { return a.CompareTo (b) > 0 ? a : b; } The Max method can accept arguments of any type implementing IComparable<T> (which includes most built-in types such as int and string): int z = Max (5, 10); // 10 string last = Max ("ant", "zoo"); // zoo The class constraint and struct constraint specify that T must be a reference type or (non-nullable) value type. A great example of the struct constraint is the Sys tem.Nullable<T> struct (we discuss this class in depth in “Nullable Value Types” on page 194): struct Nullable<T> where T : struct {...} 150 | Chapter 3: Creating Types in C#
- 169. The unmanaged constraint (introduced in C# 7.3) is a stronger version of a struct constraint: T must be a simple value type or a struct that is (recursively) free of any reference types. The parameterless constructor constraint requires T to have a public parameterless constructor. If this constraint is defined, you can call new() on T: static void Initialize<T> (T[] array) where T : new() { for (int i = 0; i < array.Length; i++) array[i] = new T(); } The naked type constraint requires one type parameter to derive from (or match) another type parameter. In this example, the method FilteredStack returns another Stack, containing only the subset of elements where the type parameter U is of the type parameter T: class Stack<T> { Stack<U> FilteredStack<U>() where U : T {...} } Subclassing Generic Types A generic class can be subclassed just like a nongeneric class. The subclass can leave the base class’s type parameters open, as in the following example: class Stack<T> {...} class SpecialStack<T> : Stack<T> {...} Or, the subclass can close the generic type parameters with a concrete type: class IntStack : Stack<int> {...} A subtype can also introduce fresh type arguments: class List<T> {...} class KeyedList<T,TKey> : List<T> {...} Technically, all type arguments on a subtype are fresh: you could say that a subtype closes and then reopens the base type arguments. This means that a subclass can give new (and potentially more meaningful) names to the type arguments that it reopens: class List<T> {...} class KeyedList<TElement,TKey> : List<TElement> {...} Self-Referencing Generic Declarations A type can name itself as the concrete type when closing a type argument: public interface IEquatable<T> { bool Equals (T obj); } public class Balloon : IEquatable<Balloon> Generics | 151 Creating Types in C#
- 170. { public string Color { get; set; } public int CC { get; set; } public bool Equals (Balloon b) { if (b == null) return false; return b.Color == Color && b.CC == CC; } } The following are also legal: class Foo<T> where T : IComparable<T> { ... } class Bar<T> where T : Bar<T> { ... } Static Data Static data is unique for each closed type: Console.WriteLine (++Bob<int>.Count); // 1 Console.WriteLine (++Bob<int>.Count); // 2 Console.WriteLine (++Bob<string>.Count); // 1 Console.WriteLine (++Bob<object>.Count); // 1 class Bob<T> { public static int Count; } Type Parameters and Conversions C#’s cast operator can perform several kinds of conversion, including: • Numeric conversion • • Reference conversion • • Boxing/unboxing conversion • • Custom conversion (via operator overloading; see Chapter 4) • The decision as to which kind of conversion will take place happens at compile time, based on the known types of the operands. This creates an interesting scenario with generic type parameters, because the precise operand types are unknown at compile time. If this leads to ambiguity, the compiler generates an error. The most common scenario is when you want to perform a reference conversion: StringBuilder Foo<T> (T arg) { if (arg is StringBuilder) return (StringBuilder) arg; // Will not compile ... } Without knowledge of T’s actual type, the compiler is concerned that you might have intended this to be a custom conversion. The simplest solution is to instead 152 | Chapter 3: Creating Types in C#
- 171. use the as operator, which is unambiguous because it cannot perform custom conversions: StringBuilder Foo<T> (T arg) { StringBuilder sb = arg as StringBuilder; if (sb != null) return sb; ... } A more general solution is to first cast to object. This works because conversions to/from object are assumed not to be custom conversions, but reference or box‐ ing/unboxing conversions. In this case, StringBuilder is a reference type, so it must be a reference conversion: return (StringBuilder) (object) arg; Unboxing conversions can also introduce ambiguities. The following could be an unboxing, numeric, or custom conversion: int Foo<T> (T x) => (int) x; // Compile-time error The solution, again, is to first cast to object and then to int (which then unambig‐ uously signals an unboxing conversion in this case): int Foo<T> (T x) => (int) (object) x; Covariance Assuming A is convertible to B, X has a covariant type parameter if X<A> is converti‐ ble to X<B>. With C#’s notion of covariance (and contravariance), “conver‐ tible” means convertible via an implicit reference conversion— such as A subclassing B, or A implementing B. Numeric conver‐ sions, boxing conversions, and custom conversions are not included. For instance, type IFoo<T> has a covariant T if the following is legal: IFoo<string> s = ...; IFoo<object> b = s; Interfaces permit covariant type parameters (as do delegates; see Chapter 4), but classes do not. Arrays also allow covariance (A[] can be converted to B[] if A has an implicit reference conversion to B) and are discussed here for comparison. Covariance and contravariance (or simply “variance”) are advanced concepts. The motivation behind introducing and enhancing variance in C# was to allow generic interface and generic types (in particular, those defined in .NET, such as IEnumerable<T>) to work more as you’d expect. You can benefit from this without understanding the details behind covariance and contravariance. Generics | 153 Creating Types in C#
- 172. Variance is not automatic To ensure static type safety, type parameters are not automatically variant. Consider the following: class Animal {} class Bear : Animal {} class Camel : Animal {} public class Stack<T> // A simple Stack implementation { int position; T[] data = new T[100]; public void Push (T obj) => data[position++] = obj; public T Pop() => data[--position]; } The following fails to compile: Stack<Bear> bears = new Stack<Bear>(); Stack<Animal> animals = bears; // Compile-time error That restriction prevents the possibility of runtime failure with the following code: animals.Push (new Camel()); // Trying to add Camel to bears Lack of covariance, however, can hinder reusability. Suppose, for instance, that we wanted to write a method to Wash a stack of animals: public class ZooCleaner { public static void Wash (Stack<Animal> animals) {...} } Calling Wash with a stack of bears would generate a compile-time error. One work‐ around is to redefine the Wash method with a constraint: class ZooCleaner { public static void Wash<T> (Stack<T> animals) where T : Animal { ... } } We can now call Wash as follows: Stack<Bear> bears = new Stack<Bear>(); ZooCleaner.Wash (bears); Another solution is to have Stack<T> implement an interface with a covariant type parameter, as you’ll see shortly. Arrays For historical reasons, array types support covariance. This means that B[] can be cast to A[] if B subclasses A (and both are reference types): Bear[] bears = new Bear[3]; Animal[] animals = bears; // OK 154 | Chapter 3: Creating Types in C#
- 173. The downside of this reusability is that element assignments can fail at runtime: animals[0] = new Camel(); // Runtime error Declaring a covariant type parameter Type parameters on interfaces and delegates can be declared covariant by marking them with the out modifier. This modifier ensures that, unlike with arrays, cova‐ riant type parameters are fully type-safe. We can illustrate this with our Stack<T> class by having it implement the following interface: public interface IPoppable<out T> { T Pop(); } The out modifier on T indicates that T is used only in output positions (e.g., return types for methods). The out modifier flags the type parameter as covariant and allows us to do this: var bears = new Stack<Bear>(); bears.Push (new Bear()); // Bears implements IPoppable<Bear>. We can convert to IPoppable<Animal>: IPoppable<Animal> animals = bears; // Legal Animal a = animals.Pop(); The conversion from bears to animals is permitted by the compiler—by virtue of the type parameter being covariant. This is type-safe because the case the compiler is trying to avoid—pushing a Camel onto the stack—can’t occur, because there’s no way to feed a Camel into an interface where T can appear only in output positions. Covariance (and contravariance) in interfaces is something that you typically consume: it’s less common that you need to write variant interfaces. Curiously, method parameters marked as out are not eligible for covariance, due to a limitation in the CLR. We can leverage the ability to cast covariantly to solve the reusability problem described earlier: public class ZooCleaner { public static void Wash (IPoppable<Animal> animals) { ... } } The IEnumerator<T> and IEnumerable<T> interfaces described in Chapter 7 have a covariant T. This allows you to cast IEnumerable<string> to IEnumerable<object>, for instance. The compiler will generate an error if you use a covariant type parameter in an input position (e.g., a parameter to a method or a writable property). Generics | 155 Creating Types in C#
- 174. Covariance (and contravariance) works only for elements with reference conversions—not boxing conversions. (This applies both to type parameter variance and array variance.) So, if you wrote a method that accepted a parameter of type IPoppable<object>, you could call it with IPoppa ble<string> but not IPoppable<int>. Contravariance We previously saw that, assuming that A allows an implicit reference conversion to B, a type X has a covariant type parameter if X<A> allows a reference conversion to X<B>. Contravariance is when you can convert in the reverse direction—from X<B> to X<A>. This is supported if the type parameter appears only in input positions and is designated with the in modifier. Extending our previous example, suppose the Stack<T> class implements the following interface: public interface IPushable<in T> { void Push (T obj); } We can now legally do this: IPushable<Animal> animals = new Stack<Animal>(); IPushable<Bear> bears = animals; // Legal bears.Push (new Bear()); No member in IPushable outputs a T, so we can’t get into trouble by casting animals to bears (there’s no way to Pop, for instance, through that interface). Our Stack<T> class can implement both IPushable<T> and IPoppable<T>—despite T having opposing variance annota‐ tions in the two interfaces! This works because you must exercise variance through the interface and not the class; therefore, you must commit to the lens of either IPoppable or IPushable before performing a variant conversion. This lens then restricts you to the operations that are legal under the appropriate variance rules. This also illustrates why classes do not allow variant type parameters: concrete implementations typically require data to flow in both directions. To give another example, consider the following interface, defined in the System namespace: public interface IComparer<in T> { // Returns a value indicating the relative ordering of a and b int Compare (T a, T b); } 156 | Chapter 3: Creating Types in C#
- 175. Because the interface has a contravariant T, we can use an IComparer<object> to compare two strings: var objectComparer = Comparer<object>.Default; // objectComparer implements IComparer<object> IComparer<string> stringComparer = objectComparer; int result = stringComparer.Compare ("Brett", "Jemaine"); Mirroring covariance, the compiler will report an error if you try to use a contravar‐ iant type parameter in an output position (e.g., as a return value or in a readable property). C# Generics Versus C++ Templates C# generics are similar in application to C++ templates, but they work very differ‐ ently. In both cases, a synthesis between the producer and consumer must take place in which the placeholder types of the producer are filled in by the consumer. However, with C# generics, producer types (i.e., open types such as List<T>) can be compiled into a library (such as mscorlib.dll). This works because the synthesis between the producer and the consumer that produces closed types doesn’t actually happen until runtime. With C++ templates, this synthesis is performed at compile time. This means that in C++ you don’t deploy template libraries as .dlls—they exist only as source code. It also makes it difficult to dynamically inspect, let alone create, parameterized types on the fly. To dig deeper into why this is the case, consider again the Max method in C#: static T Max <T> (T a, T b) where T : IComparable<T> => a.CompareTo (b) > 0 ? a : b; Why couldn’t we have implemented it like this? static T Max <T> (T a, T b) => (a > b ? a : b); // Compile error The reason is that Max needs to be compiled once and work for all possible values of T. Compilation cannot succeed because there is no single meaning for > across all values of T—in fact, not every T even has a > operator. In contrast, the following code shows the same Max method written with C++ templates. This code will be compiled separately for each value of T, taking on whatever semantics > has for a particular T and failing to compile if a particular T does not support the > operator: template <class T> T Max (T a, T b) { return a > b ? a : b; } Generics | 157 Creating Types in C#
- 177. 4 Advanced C# In this chapter, we cover advanced C# topics that build on concepts explored in Chapters 2 and 3. You should read the first four sections sequentially; you can read the remaining sections in any order. Delegates A delegate is an object that knows how to call a method. A delegate type defines the kind of method that delegate instances can call. Specif‐ ically, it defines the method’s return type and its parameter types. The following defines a delegate type called Transformer: delegate int Transformer (int x); Transformer is compatible with any method with an int return type and a single int parameter, such as this: int Square (int x) { return x * x; } Or, more tersely: int Square (int x) => x * x; Assigning a method to a delegate variable creates a delegate instance: Transformer t = Square; You can invoke a delegate instance in the same way as a method: int answer = t(3); // answer is 9 Here’s a complete example: Transformer t = Square; // Create delegate instance int result = t(3); // Invoke delegate Console.WriteLine (result); // 9 int Square (int x) => x * x; 159
- 178. delegate int Transformer (int x); // Delegate type declaration A delegate instance literally acts as a delegate for the caller: the caller invokes the delegate, and then the delegate calls the target method. This indirection decouples the caller from the target method. The statement Transformer t = Square; is shorthand for: Transformer t = new Transformer (Square); Technically, we are specifying a method group when we refer to Square without brackets or arguments. If the method is overloaded, C# will pick the correct overload based on the signature of the delegate to which it’s being assigned. The expression t(3) is shorthand for: t.Invoke(3) A delegate is similar to a callback, a general term that captures constructs such as C function pointers. Writing Plug-In Methods with Delegates A delegate variable is assigned a method at runtime. This is useful for writing plug-in methods. In this example, we have a utility method named Transform that applies a transform to each element in an integer array. The Transform method has a delegate parameter, which you can use for specifying a plug-in transform: int[] values = { 1, 2, 3 }; Transform (values, Square); // Hook in the Square method foreach (int i in values) Console.Write (i + " "); // 1 4 9 void Transform (int[] values, Transformer t) { for (int i = 0; i < values.Length; i++) values[i] = t (values[i]); } int Square (int x) => x * x; int Cube (int x) => x * x * x; delegate int Transformer (int x); 160 | Chapter 4: Advanced C#
- 179. We can change the transformation just by changing Square to Cube in the second line of code. Our Transform method is a higher-order function because it’s a function that takes a function as an argument. (A method that returns a delegate would also be a higher-order function.) Instance and Static Method Targets A delegate’s target method can be a local, static, or instance method. The following illustrates a static target method: Transformer t = Test.Square; Console.WriteLine (t(10)); // 100 class Test { public static int Square (int x) => x * x; } delegate int Transformer (int x); The following illustrates an instance target method: Test test = new Test(); Transformer t = test.Square; Console.WriteLine (t(10)); // 100 class Test { public int Square (int x) => x * x; } delegate int Transformer (int x); When an instance method is assigned to a delegate object, the latter maintains a reference not only to the method but also to the instance to which the method belongs. The System.Delegate class’s Target property represents this instance (and will be null for a delegate referencing a static method). Here’s an example: MyReporter r = new MyReporter(); r.Prefix = "%Complete: "; ProgressReporter p = r.ReportProgress; p(99); // %Complete: 99 Console.WriteLine (p.Target == r); // True Console.WriteLine (p.Method); // Void ReportProgress(Int32) r.Prefix = ""; p(99); // 99 public delegate void ProgressReporter (int percentComplete); class MyReporter { public string Prefix = ""; public void ReportProgress (int percentComplete) => Console.WriteLine (Prefix + percentComplete); } Because the instance is stored in the delegate’s Target property, its lifetime is extended to (at least as long as) the delegate’s lifetime. Delegates | 161 Advanced C#
- 180. Multicast Delegates All delegate instances have multicast capability. This means that a delegate instance can reference not just a single target method but also a list of target methods. The + and += operators combine delegate instances: SomeDelegate d = SomeMethod1; d += SomeMethod2; The last line is functionally the same as the following: d = d + SomeMethod2; Invoking d will now call both SomeMethod1 and SomeMethod2. Delegates are invoked in the order in which they are added. The - and -= operators remove the right delegate operand from the left delegate operand: d -= SomeMethod1; Invoking d will now cause only SomeMethod2 to be invoked. Calling + or += on a delegate variable with a null value works, and it is equivalent to assigning the variable to a new value: SomeDelegate d = null; d += SomeMethod1; // Equivalent (when d is null) to d = SomeMethod1; Similarly, calling -= on a delegate variable with a single matching target is equivalent to assigning null to that variable. Delegates are immutable, so when you call += or -=, you’re in fact creating a new delegate instance and assigning it to the existing variable. If a multicast delegate has a nonvoid return type, the caller receives the return value from the last method to be invoked. The preceding methods are still called, but their return values are discarded. For most scenarios in which multicast delegates are used, they have void return types, so this subtlety does not arise. All delegate types implicitly derive from System.Multicast Delegate, which inherits from System.Delegate. C# compiles +, -, +=, and -= operations made on a delegate to the static Combine and Remove methods of the System.Delegate class. Multicast delegate example Suppose that you wrote a method that took a long time to execute. That method could regularly report progress to its caller by invoking a delegate. In this example, the HardWork method has a ProgressReporter delegate parameter, which it invokes to indicate progress: 162 | Chapter 4: Advanced C#
- 181. public delegate void ProgressReporter (int percentComplete); public class Util { public static void HardWork (ProgressReporter p) { for (int i = 0; i < 10; i++) { p (i * 10); // Invoke delegate System.Threading.Thread.Sleep (100); // Simulate hard work } } } To monitor progress, we can create a multicast delegate instance p, such that pro‐ gress is monitored by two independent methods: ProgressReporter p = WriteProgressToConsole; p += WriteProgressToFile; Util.HardWork (p); void WriteProgressToConsole (int percentComplete) => Console.WriteLine (percentComplete); void WriteProgressToFile (int percentComplete) => System.IO.File.WriteAllText ("progress.txt", percentComplete.ToString()); Generic Delegate Types A delegate type can contain generic type parameters: public delegate T Transformer<T> (T arg); With this definition, we can write a generalized Transform utility method that works on any type: int[] values = { 1, 2, 3 }; Util.Transform (values, Square); // Hook in Square foreach (int i in values) Console.Write (i + " "); // 1 4 9 int Square (int x) => x * x; public class Util { public static void Transform<T> (T[] values, Transformer<T> t) { for (int i = 0; i < values.Length; i++) values[i] = t (values[i]); } } Delegates | 163 Advanced C#
- 182. The Func and Action Delegates With generic delegates, it becomes possible to write a small set of delegate types that are so general they can work for methods of any return type and any (reasonable) number of arguments. These delegates are the Func and Action delegates, defined in the System namespace (the in and out annotations indicate variance, which we cover in the context of delegates shortly): delegate TResult Func <out TResult> (); delegate TResult Func <in T, out TResult> (T arg); delegate TResult Func <in T1, in T2, out TResult> (T1 arg1, T2 arg2); ... and so on, up to T16 delegate void Action (); delegate void Action <in T> (T arg); delegate void Action <in T1, in T2> (T1 arg1, T2 arg2); ... and so on, up to T16 These delegates are extremely general. The Transformer delegate in our previous example can be replaced with a Func delegate that takes a single argument of type T and returns a same-typed value: public static void Transform<T> (T[] values, Func<T,T> transformer) { for (int i = 0; i < values.Length; i++) values[i] = transformer (values[i]); } The only practical scenarios not covered by these delegates are ref/out and pointer parameters. When C# was first introduced, the Func and Action delegates did not exist (because generics did not exist). It’s for this his‐ torical reason that much of .NET uses custom delegate types rather than Func and Action. Delegates Versus Interfaces A problem that you can solve with a delegate can also be solved with an interface. For instance, we can rewrite our original example with an interface called ITrans former instead of a delegate: int[] values = { 1, 2, 3 }; Util.TransformAll (values, new Squarer()); foreach (int i in values) Console.WriteLine (i); public interface ITransformer { int Transform (int x); } public class Util { 164 | Chapter 4: Advanced C#
- 183. public static void TransformAll (int[] values, ITransformer t) { for (int i = 0; i < values.Length; i++) values[i] = t.Transform (values[i]); } } class Squarer : ITransformer { public int Transform (int x) => x * x; } A delegate design might be a better choice than an interface design if one or more of these conditions are true: • The interface defines only a single method. • • Multicast capability is needed. • • The subscriber needs to implement the interface multiple times. • In the ITransformer example, we don’t need to multicast. However, the interface defines only a single method. Furthermore, our subscriber might need to implement ITransformer multiple times, to support different transforms, such as square or cube. With interfaces, we’re forced into writing a separate type per transform because a class can implement ITransformer only once. This is quite cumbersome: int[] values = { 1, 2, 3 }; Util.TransformAll (values, new Cuber()); foreach (int i in values) Console.WriteLine (i); class Squarer : ITransformer { public int Transform (int x) => x * x; } class Cuber : ITransformer { public int Transform (int x) => x * x * x; } Delegate Compatibility Type compatibility Delegate types are all incompatible with one another, even if their signatures are the same: D1 d1 = Method1; D2 d2 = d1; // Compile-time error void Method1() { } Delegates | 165 Advanced C#
- 184. delegate void D1(); delegate void D2(); The following, however, is permitted: D2 d2 = new D2 (d1); Delegate instances are considered equal if they have the same method targets: D d1 = Method1; D d2 = Method1; Console.WriteLine (d1 == d2); // True void Method1() { } delegate void D(); Multicast delegates are considered equal if they reference the same methods in the same order. Parameter compatibility When you call a method, you can supply arguments that have more specific types than the parameters of that method. This is ordinary polymorphic behavior. For the same reason, a delegate can have more specific parameter types than its method target. This is called contravariance. Here’s an example: StringAction sa = new StringAction (ActOnObject); sa ("hello"); void ActOnObject (object o) => Console.WriteLine (o); // hello delegate void StringAction (string s); (As with type parameter variance, delegates are variant only for reference conversions.) A delegate merely calls a method on someone else’s behalf. In this case, the String Action is invoked with an argument of type string. When the argument is then relayed to the target method, the argument is implicitly upcast to an object. The standard event pattern is designed to help you utilize contravariance through its use of the common EventArgs base class. For example, you can have a single method invoked by two different delegates, one passing a MouseEventArgs and the other passing a KeyEventArgs. Return type compatibility If you call a method, you might get back a type that is more specific than what you asked for. This is ordinary polymorphic behavior. For the same reason, a delegate’s target method might return a more specific type than described by the delegate. This is called covariance: 166 | Chapter 4: Advanced C#
- 185. ObjectRetriever o = new ObjectRetriever (RetriveString); object result = o(); Console.WriteLine (result); // hello string RetriveString() => "hello"; delegate object ObjectRetriever(); ObjectRetriever expects to get back an object, but an object subclass will also do: delegate return types are covariant. Generic delegate type parameter variance In Chapter 3, we saw how generic interfaces support covariant and contravariant type parameters. The same capability exists for delegates, too. If you’re defining a generic delegate type, it’s good practice to do the following: • Mark a type parameter used only on the return value as covariant (out). • • Mark any type parameters used only on parameters as contravariant (in). • Doing so allows conversions to work naturally by respecting inheritance relation‐ ships between types. The following delegate (defined in the System namespace) has a covariant TResult: delegate TResult Func<out TResult>(); This allows: Func<string> x = ...; Func<object> y = x; The following delegate (defined in the System namespace) has a contravariant T: delegate void Action<in T> (T arg); This allows: Action<object> x = ...; Action<string> y = x; Events When using delegates, two emergent roles commonly appear: broadcaster and subscriber. The broadcaster is a type that contains a delegate field. The broadcaster decides when to broadcast, by invoking the delegate. The subscribers are the method target recipients. A subscriber decides when to start and stop listening by calling += and -= on the broadcaster’s delegate. A subscriber does not know about, or interfere with, other subscribers. Events | 167 Advanced C#
- 186. Events are a language feature that formalizes this pattern. An event is a construct that exposes just the subset of delegate features required for the broadcaster/sub‐ scriber model. The main purpose of events is to prevent subscribers from interfering with one another. The easiest way to declare an event is to put the event keyword in front of a delegate member: // Delegate definition public delegate void PriceChangedHandler (decimal oldPrice, decimal newPrice); public class Broadcaster { // Event declaration public event PriceChangedHandler PriceChanged; } Code within the Broadcaster type has full access to PriceChanged and can treat it as a delegate. Code outside of Broadcaster can perform only += and -= operations on the PriceChanged event. How Do Events Work on the Inside? Three things happen under the hood when you declare an event as follows: public class Broadcaster { public event PriceChangedHandler PriceChanged; } First, the compiler translates the event declaration into something close to the following: PriceChangedHandler priceChanged; // private delegate public event PriceChangedHandler PriceChanged { add { priceChanged += value; } remove { priceChanged -= value; } } The add and remove keywords denote explicit event accessors—which act rather like property accessors. We describe how to write these later. Second, the compiler looks within the Broadcaster class for references to Price Changed that perform operations other than += or -= and redirects them to the underlying priceChanged delegate field. Third, the compiler translates += and -= operations on the event to calls to the event’s add and remove accessors. Interestingly, this makes the behavior of += and -= unique when applied to events: unlike in other scenarios, it’s not simply a shortcut for + and - followed by an assignment. 168 | Chapter 4: Advanced C#
- 187. Consider the following example. The Stock class fires its PriceChanged event every time the Price of the Stock changes: public delegate void PriceChangedHandler (decimal oldPrice, decimal newPrice); public class Stock { string symbol; decimal price; public Stock (string symbol) => this.symbol = symbol; public event PriceChangedHandler PriceChanged; public decimal Price { get => price; set { if (price == value) return; // Exit if nothing has changed decimal oldPrice = price; price = value; if (PriceChanged != null) // If invocation list not PriceChanged (oldPrice, price); // empty, fire event. } } } If we remove the event keyword from our example so that PriceChanged becomes an ordinary delegate field, our example would give the same results. However, Stock would be less robust insomuch as subscribers could do the following things to interfere with one another: • Replace other subscribers by reassigning PriceChanged (instead of using the += • operator) • Clear all subscribers (by setting PriceChanged to null) • • Broadcast to other subscribers by invoking the delegate • Standard Event Pattern In almost all cases for which events are defined in the .NET libraries, their defini‐ tion adheres to a standard pattern designed to provide consistency across library and user code. At the core of the standard event pattern is System.EventArgs, a pre‐ defined .NET class with no members (other than the static Empty field). EventArgs is a base class for conveying information for an event. In our Stock example, we would subclass EventArgs to convey the old and new prices when a PriceChanged event is fired: Events | 169 Advanced C#
- 188. public class PriceChangedEventArgs : System.EventArgs { public readonly decimal LastPrice; public readonly decimal NewPrice; public PriceChangedEventArgs (decimal lastPrice, decimal newPrice) { LastPrice = lastPrice; NewPrice = newPrice; } } For reusability, the EventArgs subclass is named according to the information it contains (rather than the event for which it will be used). It typically exposes data as properties or as read-only fields. With an EventArgs subclass in place, the next step is to choose or define a delegate for the event. There are three rules: • It must have a void return type. • • It must accept two arguments: the first of type object and the second a • subclass of EventArgs. The first argument indicates the event broadcaster, and the second argument contains the extra information to convey. • Its name must end with EventHandler. • .NET defines a generic delegate called System.EventHandler<> to help with this: public delegate void EventHandler<TEventArgs> (object source, TEventArgs e) Before generics existed in the language (prior to C# 2.0), we would have had to instead write a custom delegate as follows: public delegate void PriceChangedHandler (object sender, PriceChangedEventArgs e); For historical reasons, most events within the .NET libraries use delegates defined in this way. The next step is to define an event of the chosen delegate type. Here, we use the generic EventHandler delegate: public class Stock { ... public event EventHandler<PriceChangedEventArgs> PriceChanged; } Finally, the pattern requires that you write a protected virtual method that fires the event. The name must match the name of the event, prefixed with the word On, and then accept a single EventArgs argument: 170 | Chapter 4: Advanced C#
- 189. public class Stock { ... public event EventHandler<PriceChangedEventArgs> PriceChanged; protected virtual void OnPriceChanged (PriceChangedEventArgs e) { if (PriceChanged != null) PriceChanged (this, e); } } To work robustly in multithreaded scenarios (Chapter 14), you need to assign the delegate to a temporary variable before testing and invoking it: var temp = PriceChanged; if (temp != null) temp (this, e); We can achieve the same functionality without the temp vari‐ able with the null-conditional operator: PriceChanged?.Invoke (this, e); Being both thread-safe and succinct, this is the best general way to invoke events. This provides a central point from which subclasses can invoke or override the event (assuming the class is not sealed). Here’s the complete example: using System; Stock stock = new Stock ("THPW"); stock.Price = 27.10M; // Register with the PriceChanged event stock.PriceChanged += stock_PriceChanged; stock.Price = 31.59M; void stock_PriceChanged (object sender, PriceChangedEventArgs e) { if ((e.NewPrice - e.LastPrice) / e.LastPrice > 0.1M) Console.WriteLine ("Alert, 10% stock price increase!"); } public class PriceChangedEventArgs : EventArgs { public readonly decimal LastPrice; public readonly decimal NewPrice; public PriceChangedEventArgs (decimal lastPrice, decimal newPrice) { LastPrice = lastPrice; NewPrice = newPrice; } } public class Stock Events | 171 Advanced C#
- 190. { string symbol; decimal price; public Stock (string symbol) => this.symbol = symbol; public event EventHandler<PriceChangedEventArgs> PriceChanged; protected virtual void OnPriceChanged (PriceChangedEventArgs e) { PriceChanged?.Invoke (this, e); } public decimal Price { get => price; set { if (price == value) return; decimal oldPrice = price; price = value; OnPriceChanged (new PriceChangedEventArgs (oldPrice, price)); } } } The predefined nongeneric EventHandler delegate can be used when an event doesn’t carry extra information. In this example, we rewrite Stock such that the PriceChanged event is fired after the price changes, and no information about the event is necessary, other than it happened. We also make use of the EventArgs.Empty property to avoid unnecessarily instantiating an instance of EventArgs: public class Stock { string symbol; decimal price; public Stock (string symbol) { this.symbol = symbol; } public event EventHandler PriceChanged; protected virtual void OnPriceChanged (EventArgs e) { PriceChanged?.Invoke (this, e); } public decimal Price { get { return price; } set { if (price == value) return; price = value; 172 | Chapter 4: Advanced C#
- 191. OnPriceChanged (EventArgs.Empty); } } } Event Accessors An event’s accessors are the implementations of its += and -= functions. By default, accessors are implemented implicitly by the compiler. Consider this event declaration: public event EventHandler PriceChanged; The compiler converts this to the following: • A private delegate field • • A public pair of event accessor functions (add_PriceChanged and • remove_PriceChanged) whose implementations forward the += and -= opera‐ tions to the private delegate field You can take over this process by defining explicit event accessors. Here’s a manual implementation of the PriceChanged event from our previous example: private EventHandler priceChanged; // Declare a private delegate public event EventHandler PriceChanged { add { priceChanged += value; } remove { priceChanged -= value; } } This example is functionally identical to C#’s default accessor implementation (except that C# also ensures thread safety around updating the delegate via a lock- free compare-and-swap algorithm; see http://albahari.com/threading). By defining event accessors ourselves, we instruct C# not to generate default field and accessor logic. With explicit event accessors, you can apply more complex strategies to the storage and access of the underlying delegate. There are three scenarios for which this is useful: • When the event accessors are merely relays for another class that is broadcast‐ • ing the event. • When the class exposes many events, for which most of the time very few • subscribers exist, such as a Windows control. In such cases, it is better to store the subscriber’s delegate instances in a dictionary because a dictionary will contain less storage overhead than dozens of null delegate field references. • When explicitly implementing an interface that declares an event. • Events | 173 Advanced C#
- 192. Here is an example that illustrates the last point: public interface IFoo { event EventHandler Ev; } class Foo : IFoo { private EventHandler ev; event EventHandler IFoo.Ev { add { ev += value; } remove { ev -= value; } } } The add and remove parts of an event are compiled to add_XXX and remove_XXX methods. Event Modifiers Like methods, events can be virtual, overridden, abstract, or sealed. Events can also be static: public class Foo { public static event EventHandler<EventArgs> StaticEvent; public virtual event EventHandler<EventArgs> VirtualEvent; } Lambda Expressions A lambda expression is an unnamed method written in place of a delegate instance. The compiler immediately converts the lambda expression to either of the following: • A delegate instance. • • An expression tree, of type Expression<TDelegate>, representing the code • inside the lambda expression in a traversable object model. This allows the lambda expression to be interpreted later at runtime (see “Building Query Expressions” on page 442). In the following example, x => x * x is a lambda expression: Transformer sqr = x => x * x; Console.WriteLine (sqr(3)); // 9 delegate int Transformer (int i); Internally, the compiler resolves lambda expressions of this type by writing a private method and then moving the expres‐ sion’s code into that method. 174 | Chapter 4: Advanced C#
- 193. A lambda expression has the following form: (parameters) => expression-or-statement-block For convenience, you can omit the parentheses if and only if there is exactly one parameter of an inferable type. In our example, there is a single parameter, x, and the expression is x * x: x => x * x; Each parameter of the lambda expression corresponds to a delegate parameter, and the type of the expression (which may be void) corresponds to the return type of the delegate. In our example, x corresponds to parameter i, and the expression x * x corre‐ sponds to the return type int, therefore being compatible with the Transformer delegate: delegate int Transformer (int i); A lambda expression’s code can be a statement block instead of an expression. We can rewrite our example as follows: x => { return x * x; }; Lambda expressions are used most commonly with the Func and Action delegates, so you will most often see our earlier expression written as follows: Func<int,int> sqr = x => x * x; Here’s an example of an expression that accepts two parameters: Func<string,string,int> totalLength = (s1, s2) => s1.Length + s2.Length; int total = totalLength ("hello", "world"); // total is 10; If you do not need to use the parameters, you can discard them with an underscore (from C# 9): Func<string,string,int> totalLength = (_,_) => ... Here’s an example of an expression that takes zero arguments: Func<string> greetor = () => "Hello, world"; From C# 10, the compiler permits implicit typing with lambda expressions that can be resolved via the Func and Action delegates, so we can shorten this statement to: var greeter = () => "Hello, world"; Explicitly Specifying Lambda Parameter and Return Types The compiler can usually infer the type of lambda parameters contextually. When this is not the case, you must specify the type of each parameter explicitly. Consider the following two methods: void Foo<T> (T x) {} void Bar<T> (Action<T> a) {} Lambda Expressions | 175 Advanced C#
- 194. The following code will fail to compile, because the compiler cannot infer the type of x: Bar (x => Foo (x)); // What type is x? We can fix this by explicitly specifying x’s type as follows: Bar ((int x) => Foo (x)); This particular example is simple enough that it can be fixed in two other ways: Bar<int> (x => Foo (x)); // Specify type parameter for Bar Bar<int> (Foo); // As above, but with method group The following example illustrates another use for explicit parameter types (from C# 10): var sqr = (int x) => x * x; The compiler infers sqr to be of type Func<int,int>. (Without specifying int, implicit typing would fail: the compiler would know that sqr should be Func<T,T>, but it wouldn’t know what T should be.) From C# 10, you can also specify the lambda return type: var sqr = int (int x) => x; Specifying a return type can improve compiler performance with complex nested lambdas. Capturing Outer Variables A lambda expression can reference any variables that are accessible where the lambda expression is defined. These are called outer variables, and can include local variables, parameters, and fields: int factor = 2; Func<int, int> multiplier = n => n * factor; Console.WriteLine (multiplier (3)); // 6 Outer variables referenced by a lambda expression are called captured variables. A lambda expression that captures variables is called a closure. Variables can also be captured by anonymous methods and local methods. The rules for captured variables, in these cases, are the same. Captured variables are evaluated when the delegate is actually invoked, not when the variables were captured: int factor = 2; Func<int, int> multiplier = n => n * factor; factor = 10; Console.WriteLine (multiplier (3)); // 30 Lambda expressions can themselves update captured variables: 176 | Chapter 4: Advanced C#
- 195. int seed = 0; Func<int> natural = () => seed++; Console.WriteLine (natural()); // 0 Console.WriteLine (natural()); // 1 Console.WriteLine (seed); // 2 Captured variables have their lifetimes extended to that of the delegate. In the following example, the local variable seed would ordinarily disappear from scope when Natural finished executing. But because seed has been captured, its lifetime is extended to that of the capturing delegate, natural: static Func<int> Natural() { int seed = 0; return () => seed++; // Returns a closure } static void Main() { Func<int> natural = Natural(); Console.WriteLine (natural()); // 0 Console.WriteLine (natural()); // 1 } A local variable instantiated within a lambda expression is unique per invocation of the delegate instance. If we refactor our previous example to instantiate seed within the lambda expression, we get a different (in this case, undesirable) result: static Func<int> Natural() { return() => { int seed = 0; return seed++; }; } static void Main() { Func<int> natural = Natural(); Console.WriteLine (natural()); // 0 Console.WriteLine (natural()); // 0 } Capturing is internally implemented by “hoisting” the cap‐ tured variables into fields of a private class. When the method is called, the class is instantiated and lifetime-bound to the delegate instance. Static lambdas When you capture local variables, parameters, instance fields, or the this reference, the compiler may need to create and instantiate a private class to store a reference to the captured data. This incurs a small performance cost, because memory must be allocated (and subsequently collected). In situations where performance is critical, one micro-optimization strategy is to minimize the load on the garbage collector by ensuring that code hot paths incur few or no allocations. Lambda Expressions | 177 Advanced C#
- 196. From C# 9, you can ensure that a lambda expression, local function, or anonymous method doesn’t capture state by applying the static keyword. This can be useful in micro-optimization scenarios to prevent unintentional memory allocations. For example, we can apply the static modifier to a lambda expression as follows: Func<int, int> multiplier = static n => n * 2; If we later try to modify the lambda expression such that it captures a local variable, the compiler will generate an error: int factor = 2; Func<int, int> multiplier = static n => n * factor; // will not compile The lambda itself evaluates to a delegate instance, which requires a memory allocation. However, if the lambda doesn’t capture variables, the compiler will reuse a single cached instance across the life of the application, so there will be no cost in practice. This feature can also be used with local methods. In the following example, the Multiply method cannot access the factor variable: void Foo() { int factor = 123; static int Multiply (int x) => x * 2; // Local static method } Of course, the Multiply method could still explicitly allocate memory by calling new. What this protects us from is a potential allocation by stealth. Applying static here is also arguably useful as a documentation tool, indicating a reduced level of coupling. Static lambdas can still access static variables and constants (because these do not require a closure). The static keyword acts merely as a check; it has no effect on the IL that the compiler produces. Without the static keyword, the compiler does not generate a closure unless it needs to (and even then, it has tricks to mitigate the cost). Capturing iteration variables When you capture the iteration variable of a for loop, C# treats that variable as though it were declared outside the loop. This means that the same variable is captured in each iteration. The following program writes 333 instead of 012: Action[] actions = new Action[3]; for (int i = 0; i < 3; i++) actions [i] = () => Console.Write (i); foreach (Action a in actions) a(); // 333 178 | Chapter 4: Advanced C#
- 197. Each closure (shown in boldface) captures the same variable, i. (This actually makes sense when you consider that i is a variable whose value persists between loop iterations; you can even explicitly change i within the loop body if you want.) The consequence is that when the delegates are later invoked, each delegate sees i’s value at the time of invocation—which is 3. We can illustrate this better by expanding the for loop, as follows: Action[] actions = new Action[3]; int i = 0; actions[0] = () => Console.Write (i); i = 1; actions[1] = () => Console.Write (i); i = 2; actions[2] = () => Console.Write (i); i = 3; foreach (Action a in actions) a(); // 333 The solution, if we want to write 012, is to assign the iteration variable to a local variable that’s scoped within the loop: Action[] actions = new Action[3]; for (int i = 0; i < 3; i++) { int loopScopedi = i; actions [i] = () => Console.Write (loopScopedi); } foreach (Action a in actions) a(); // 012 Because loopScopedi is freshly created on every iteration, each closure captures a different variable. Prior to C# 5.0, foreach loops worked in the same way. This caused considerable confusion: unlike with a for loop, the iteration variable in a foreach loop is immutable, and so you would expect it to be treated as local to the loop body. The good news is that it’s now fixed, and you can safely capture a foreach loop’s iteration variable without surprises. Lambda Expressions Versus Local Methods The functionality of local methods (see “Local methods” on page 19) overlaps with that of lambda expressions. Local methods have the following three advantages: • They can be recursive (they can call themselves) without ugly hacks. • • They avoid the clutter of specifying a delegate type. • • They incur slightly less overhead. • Local methods are more efficient because they avoid the indirection of a delegate (which costs some CPU cycles and a memory allocation). They can also access local variables of the containing method without the compiler having to “hoist” the captured variables into a hidden class. Lambda Expressions | 179 Advanced C#
- 198. However, in many cases you need a delegate—most commonly when calling a higher-order function, that is, a method with a delegate-typed parameter: public void Foo (Func<int,bool> predicate) { ... } (You can see plenty more of these in Chapter 8.) In such cases, you need a delegate anyway, and it’s in precisely these cases that lambda expressions are usually terser and cleaner. Anonymous Methods Anonymous methods are a C# 2.0 feature that was mostly subsumed by C# 3.0’s lambda expressions. An anonymous method is like a lambda expression, but it lacks the following features: • Implicitly typed parameters • • Expression syntax (an anonymous method must always be a statement block) • • The ability to compile to an expression tree, by assigning to Expression<T> • An anonymous method uses the delegate keyword followed (optionally) by a parameter declaration and then a method body. For example: Transformer sqr = delegate (int x) {return x * x;}; Console.WriteLine (sqr(3)); // 9 delegate int Transformer (int i); The first line is semantically equivalent to the following lambda expression: Transformer sqr = (int x) => {return x * x;}; Or simply: Transformer sqr = x => x * x; Anonymous methods capture outer variables in the same way lambda expressions do, and can be preceded by the static keyword to make them behave like static lambdas. A unique feature of anonymous methods is that you can omit the parameter declaration entirely—even if the delegate expects it. This can be useful in declaring events with a default empty handler: public event EventHandler Clicked = delegate { }; This avoids the need for a null check before firing the event. The following is also legal: // Notice that we omit the parameters: Clicked += delegate { Console.WriteLine ("clicked"); }; 180 | Chapter 4: Advanced C#
- 199. try Statements and Exceptions A try statement specifies a code block subject to error-handling or cleanup code. The try block must be followed by one or more catch blocks and/or a finally block, or both. The catch block executes when an error is thrown in the try block. The finally block executes after execution leaves the try block (or, if present, the catch block) to perform cleanup code, regardless of whether an exception was thrown. A catch block has access to an Exception object that contains information about the error. You use a catch block to either compensate for the error or rethrow the exception. You rethrow an exception if you merely want to log the problem or if you want to rethrow a new, higher-level exception type. A finally block adds determinism to your program: the CLR endeavors to always execute it. It’s useful for cleanup tasks such as closing network connections. A try statement looks like this: try { ... // exception may get thrown within execution of this block } catch (ExceptionA ex) { ... // handle exception of type ExceptionA } catch (ExceptionB ex) { ... // handle exception of type ExceptionB } finally { ... // cleanup code } Consider the following program: int y = Calc (0); Console.WriteLine (y); int Calc (int x) => 10 / x; Because x is zero, the runtime throws a DivideByZeroException, and our program terminates. We can prevent this by catching the exception as follows: try { int y = Calc (0); Console.WriteLine (y); } catch (DivideByZeroException ex) { Console.WriteLine ("x cannot be zero"); try Statements and Exceptions | 181 Advanced C#
- 200. } Console.WriteLine ("program completed"); int Calc (int x) => 10 / x; Here’s the output: x cannot be zero program completed This is a simple example to illustrate exception handling. We could deal with this particular scenario better in practice by checking explicitly for the divisor being zero before calling Calc. Checking for preventable errors is preferable to relying on try/catch blocks because exceptions are relatively expensive to handle, taking hundreds of clock cycles or more. When an exception is thrown within a try statement, the CLR performs a test: Does the try statement have any compatible catch blocks? • If so, execution jumps to the compatible catch block, followed by the finally • block (if present), and then execution continues normally. • If not, execution jumps directly to the finally block (if present), then the CLR • looks up the call stack for other try blocks; if found, it repeats the test. If no function in the call stack takes responsibility for the exception, the program terminates. The catch Clause A catch clause specifies what type of exception to catch. This must either be System.Exception or a subclass of System.Exception. Catching System.Exception catches all possible errors. This is useful in the follow‐ ing circumstances: • Your program can potentially recover regardless of the specific exception type. • • You plan to rethrow the exception (perhaps after logging it). • • Your error handler is the last resort, prior to termination of the program. • More typically, though, you catch specific exception types to avoid having to deal with circumstances for which your handler wasn’t designed (e.g., an OutOfMemory Exception). You can handle multiple exception types with multiple catch clauses (again, this example could be written with explicit argument checking rather than exception handling): 182 | Chapter 4: Advanced C#
- 201. class Test { static void Main (string[] args) { try { byte b = byte.Parse (args[0]); Console.WriteLine (b); } catch (IndexOutOfRangeException) { Console.WriteLine ("Please provide at least one argument"); } catch (FormatException) { Console.WriteLine ("That's not a number!"); } catch (OverflowException) { Console.WriteLine ("You've given me more than a byte!"); } } } Only one catch clause executes for a given exception. If you want to include a safety net to catch more general exceptions (such as System.Exception), you must put the more-specific handlers first. An exception can be caught without specifying a variable, if you don’t need to access its properties: catch (OverflowException) // no variable { ... } Furthermore, you can omit both the variable and the type (meaning that all excep‐ tions will be caught): catch { ... } Exception filters You can specify an exception filter in a catch clause by adding a when clause: catch (WebException ex) when (ex.Status == WebExceptionStatus.Timeout) { ... } If a WebException is thrown in this example, the Boolean expression following the when keyword is then evaluated. If the result is false, the catch block in question is ignored and any subsequent catch clauses are considered. With exception filters, it can be meaningful to catch the same exception type again: try Statements and Exceptions | 183 Advanced C#
- 202. catch (WebException ex) when (ex.Status == WebExceptionStatus.Timeout) { ... } catch (WebException ex) when (ex.Status == WebExceptionStatus.SendFailure) { ... } The Boolean expression in the when clause can be side-effecting, such as a method that logs the exception for diagnostic purposes. The finally Block A finally block always executes—regardless of whether an exception is thrown and whether the try block runs to completion. You typically use finally blocks for cleanup code. A finally block executes after any of the following: • A catch block finishes (or throws a new exception). • • The try block finishes (or throws an exception for which there’s no catch • block). • Control leaves the try block because of a jump statement (e.g., return or goto). • The only things that can defeat a finally block are an infinite loop or the process ending abruptly. A finally block helps add determinism to a program. In the following example, the file that we open always gets closed, regardless of whether: • The try block finishes normally. • • Execution returns early because the file is empty (EndOfStream). • • An IOException is thrown while reading the file. • void ReadFile() { StreamReader reader = null; // In System.IO namespace try { reader = File.OpenText ("file.txt"); if (reader.EndOfStream) return; Console.WriteLine (reader.ReadToEnd()); } finally { if (reader != null) reader.Dispose(); } } In this example, we closed the file by calling Dispose on the StreamReader. Calling Dispose on an object, within a finally block, is a standard convention and is supported explicitly in C# through the using statement. 184 | Chapter 4: Advanced C#
- 203. The using statement Many classes encapsulate unmanaged resources, such as file handles, graphics handles, or database connections. These classes implement System.IDisposable, which defines a single parameterless method named Dispose to clean up these resources. The using statement provides an elegant syntax for calling Dispose on an IDisposable object within a finally block. Thus, using (StreamReader reader = File.OpenText ("file.txt")) { ... } is precisely equivalent to the following: { StreamReader reader = File.OpenText ("file.txt"); try { ... } finally { if (reader != null) ((IDisposable)reader).Dispose(); } } using declarations If you omit the brackets and statement block following a using statement (C# 8+), it becomes a using declaration. The resource is then disposed when execution falls outside the enclosing statement block: if (File.Exists ("file.txt")) { using var reader = File.OpenText ("file.txt"); Console.WriteLine (reader.ReadLine()); ... } In this case, reader will be disposed when execution falls outside the if statement block. Throwing Exceptions Exceptions can be thrown either by the runtime or in user code. In this example, Display throws a System.ArgumentNullException: try { Display (null); } catch (ArgumentNullException ex) { Console.WriteLine ("Caught the exception"); try Statements and Exceptions | 185 Advanced C#
- 204. } void Display (string name) { if (name == null) throw new ArgumentNullException (nameof (name)); Console.WriteLine (name); } Because null-checking an argument and throwing an ArgumentNullException is such a common code path, there’s actually a shortcut for it, from .NET 6: void Display (string name) { ArgumentNullException.ThrowIfNull (name); Console.WriteLine (name); } Notice that we didn’t need to specify the name of the parame‐ ter. We’ll explain why later, in “CallerArgumentExpression (C# 10)” on page 231. throw expressions throw can also appear as an expression in expression-bodied functions: public string Foo() => throw new NotImplementedException(); A throw expression can also appear in a ternary conditional expression: string ProperCase (string value) => value == null ? throw new ArgumentException ("value") : value == "" ? "" : char.ToUpper (value[0]) + value.Substring (1); Rethrowing an exception You can capture and rethrow an exception as follows: try { ... } catch (Exception ex) { // Log error ... throw; // Rethrow same exception } If we replaced throw with throw ex, the example would still work, but the StackTrace property of the newly propagated exception would no longer reflect the original error. Rethrowing in this manner lets you log an error without swallowing it. It also lets you back out of handling an exception should circumstances turn out to be beyond 186 | Chapter 4: Advanced C#
- 205. what you expected. The other common scenario is to rethrow a more specific exception type: try { ... // Parse a DateTime from XML element data } catch (FormatException ex) { throw new XmlException ("Invalid DateTime", ex); } Notice that when we constructed XmlException, we passed in the original excep‐ tion, ex, as the second argument. This argument populates the InnerException property of the new exception and aids debugging. Nearly all types of exception offer a similar constructor. Rethrowing a less-specific exception is something you might do when crossing a trust boundary, so as not to leak technical information to potential hackers. Key Properties of System.Exception The most important properties of System.Exception are the following: StackTrace A string representing all the methods that are called from the origin of the exception to the catch block. Message A string with a description of the error. InnerException The inner exception (if any) that caused the outer exception. This, itself, can have another InnerException. All exceptions in C# are runtime exceptions—there is no equivalent to Java’s compile-time checked exceptions. Common Exception Types The following exception types are used widely throughout the CLR and .NET libra‐ ries. You can throw these yourself or use them as base classes for deriving custom exception types: System.ArgumentException Thrown when a function is called with a bogus argument. This generally indicates a program bug. System.ArgumentNullException Subclass of ArgumentException that’s thrown when a function argument is (unexpectedly) null. try Statements and Exceptions | 187 Advanced C#
- 206. System.ArgumentOutOfRangeException Subclass of ArgumentException that’s thrown when a (usually numeric) argu‐ ment is too big or too small. For example, this is thrown when passing a negative number into a function that accepts only positive values. System.InvalidOperationException Thrown when the state of an object is unsuitable for a method to successfully execute, regardless of any particular argument values. Examples include read‐ ing an unopened file or getting the next element from an enumerator for which the underlying list has been modified partway through the iteration. System.NotSupportedException Thrown to indicate that a particular functionality is not supported. A good example is calling the Add method on a collection for which IsReadOnly returns true. System.NotImplementedException Thrown to indicate that a function has not yet been implemented. System.ObjectDisposedException Thrown when the object upon which the function is called has been disposed. Another commonly encountered exception type is NullReferenceException. The CLR throws this exception when you attempt to access a member of an object whose value is null (indicating a bug in your code). You can throw a NullReferen ceException directly (for testing purposes) as follows: throw null; The TryXXX Method Pattern When writing a method, you have a choice, when something goes wrong, to return some kind of failure code or throw an exception. In general, you throw an exception when the error is outside the normal workflow—or if you expect that the immediate caller won’t be able to cope with it. Occasionally, though, it can be best to offer both choices to the consumer. An example of this is the int type, which defines two versions of its Parse method: public int Parse (string input); public bool TryParse (string input, out int returnValue); If parsing fails, Parse throws an exception; TryParse returns false. You can implement this pattern by having the XXX method call the TryXXX method as follows: public return-type XXX (input-type input) { return-type returnValue; if (!TryXXX (input, out returnValue)) throw new YYYException (...) 188 | Chapter 4: Advanced C#
- 207. return returnValue; } Alternatives to Exceptions As with int.TryParse, a function can communicate failure by sending an error code back to the calling function via a return type or parameter. Although this can work with simple and predictable failures, it becomes clumsy when extended to all errors, polluting method signatures and creating unnecessary complexity and clutter. It also cannot generalize to functions that are not methods, such as operators (e.g., the division operator) or properties. An alternative is to place the error in a common place where all functions in the call stack can see it (e.g., a static method that stores the current error per thread). This, though, requires each function to participate in an error-propagation pattern, which is cumbersome and, ironically, itself error prone. Enumeration and Iterators Enumeration An enumerator is a read-only, forward-only cursor over a sequence of values. C# treats a type as an enumerator if it does any of the following: • Has a public parameterless method named MoveNext and property called • Current • Implements System.Collections.Generic.IEnumerator<T> • • Implements System.Collections.IEnumerator • The foreach statement iterates over an enumerable object. An enumerable object is the logical representation of a sequence. It is not itself a cursor, but an object that produces cursors over itself. C# treats a type as enumerable if it does any of the following (the check is performed in this order): • Has a public parameterless method named GetEnumerator that returns an • enumerator • Implements System.Collections.Generic.IEnumerable<T> • • Implements System.Collections.IEnumerable • • (From C# 9) Can bind to an extension method named GetEnumerator that • returns an enumerator (see “Extension Methods” on page 202) The enumeration pattern is as follows: class Enumerator // Typically implements IEnumerator or IEnumerator<T> { public IteratorVariableType Current { get {...} } public bool MoveNext() {...} Enumeration and Iterators | 189 Advanced C#
- 208. } class Enumerable // Typically implements IEnumerable or IEnumerable<T> { public Enumerator GetEnumerator() {...} } Here is the high-level way of iterating through the characters in the word beer using a foreach statement: foreach (char c in "beer") Console.WriteLine (c); Here is the low-level way of iterating through the characters in beer without using a foreach statement: using (var enumerator = "beer".GetEnumerator()) while (enumerator.MoveNext()) { var element = enumerator.Current; Console.WriteLine (element); } If the enumerator implements IDisposable, the foreach statement also acts as a using statement, implicitly disposing the enumerator object. Chapter 7 explains the enumeration interfaces in further detail. Collection Initializers You can instantiate and populate an enumerable object in a single step: using System.Collections.Generic; ... List<int> list = new List<int> {1, 2, 3}; The compiler translates this to the following: using System.Collections.Generic; ... List<int> list = new List<int>(); list.Add (1); list.Add (2); list.Add (3); This requires that the enumerable object implements the System.Collec tions.IEnumerable interface, and that it has an Add method that has the appropri‐ ate number of parameters for the call. You can similarly initialize dictionaries (see “Dictionaries” on page 372) as follows: var dict = new Dictionary<int, string>() { { 5, "five" }, 190 | Chapter 4: Advanced C#
- 209. { 10, "ten" } }; Or, more succinctly: var dict = new Dictionary<int, string>() { [3] = "three", [10] = "ten" }; The latter is valid not only with dictionaries but with any type for which an indexer exists. Iterators Whereas a foreach statement is a consumer of an enumerator, an iterator is a producer of an enumerator. In this example, we use an iterator to return a sequence of Fibonacci numbers (where each number is the sum of the previous two): using System; using System.Collections.Generic; foreach (int fib in Fibs(6)) Console.Write (fib + " "); } IEnumerable<int> Fibs (int fibCount) { for (int i = 0, prevFib = 1, curFib = 1; i < fibCount; i++) { yield return prevFib; int newFib = prevFib+curFib; prevFib = curFib; curFib = newFib; } } OUTPUT: 1 1 2 3 5 8 Whereas a return statement expresses, “Here’s the value you asked me to return from this method,” a yield return statement expresses, “Here’s the next element you asked me to yield from this enumerator.” On each yield statement, control is returned to the caller, but the callee’s state is maintained so that the method can continue executing as soon as the caller enumerates the next element. The lifetime of this state is bound to the enumerator such that the state can be released when the caller has finished enumerating. Enumeration and Iterators | 191 Advanced C#
- 210. The compiler converts iterator methods into private classes that implement IEnumerable<T> and/or IEnumerator<T>. The logic within the iterator block is “inverted” and spliced into the MoveNext method and Current property on the compiler- written enumerator class. This means that when you call an iterator method, all you’re doing is instantiating the compiler- written class; none of your code actually runs! Your code runs only when you start enumerating over the resultant sequence, typically with a foreach statement. Iterators can be local methods (see “Local methods” on page 19). Iterator Semantics An iterator is a method, property, or indexer that contains one or more yield statements. An iterator must return one of the following four interfaces (otherwise, the compiler will generate an error): // Enumerable interfaces System.Collections.IEnumerable System.Collections.Generic.IEnumerable<T> // Enumerator interfaces System.Collections.IEnumerator System.Collections.Generic.IEnumerator<T> An iterator has different semantics, depending on whether it returns an enumerable interface or an enumerator interface. We describe this in Chapter 7. Multiple yield statements are permitted: foreach (string s in Foo()) Console.WriteLine(s); // Prints "One","Two","Three" IEnumerable<string> Foo() { yield return "One"; yield return "Two"; yield return "Three"; } yield break A return statement is illegal in an iterator block; instead you must use the yield break statement to indicate that the iterator block should exit early, without return‐ ing more elements. We can modify Foo as follows to demonstrate: IEnumerable<string> Foo (bool breakEarly) { yield return "One"; yield return "Two"; if (breakEarly) yield break; 192 | Chapter 4: Advanced C#
- 211. yield return "Three"; } Iterators and try/catch/finally blocks A yield return statement cannot appear in a try block that has a catch clause: IEnumerable<string> Foo() { try { yield return "One"; } // Illegal catch { ... } } Nor can yield return appear in a catch or finally block. These restrictions are due to the fact that the compiler must translate iterators into ordinary classes with MoveNext, Current, and Dispose members, and translating exception-handling blocks would create excessive complexity. You can, however, yield within a try block that has (only) a finally block: IEnumerable<string> Foo() { try { yield return "One"; } // OK finally { ... } } The code in the finally block executes when the consuming enumerator reaches the end of the sequence or is disposed. A foreach statement implicitly disposes the enumerator if you break early, making this a safe way to consume enumerators. When working with enumerators explicitly, a trap is to abandon enumeration early without disposing it, circumventing the finally block. You can avoid this risk by wrapping explicit use of enumerators in a using statement: string firstElement = null; var sequence = Foo(); using (var enumerator = sequence.GetEnumerator()) if (enumerator.MoveNext()) firstElement = enumerator.Current; Composing Sequences Iterators are highly composable. We can extend our example, this time to output even Fibonacci numbers only: using System; using System.Collections.Generic; foreach (int fib in EvenNumbersOnly (Fibs(6))) Console.WriteLine (fib); IEnumerable<int> Fibs (int fibCount) { for (int i = 0, prevFib = 1, curFib = 1; i < fibCount; i++) Enumeration and Iterators | 193 Advanced C#
- 212. { yield return prevFib; int newFib = prevFib+curFib; prevFib = curFib; curFib = newFib; } } IEnumerable<int> EvenNumbersOnly (IEnumerable<int> sequence) { foreach (int x in sequence) if ((x % 2) == 0) yield return x; } Each element is not calculated until the last moment—when requested by a Move Next() operation. Figure 4-1 shows the data requests and data output over time. Figure 4-1. Composing sequences The composability of the iterator pattern is extremely useful in LINQ; we discuss the subject again in Chapter 8. Nullable Value Types Reference types can represent a nonexistent value with a null reference. Value types, however, cannot ordinarily represent null values: string s = null; // OK, Reference Type int i = null; // Compile Error, Value Type cannot be null 194 | Chapter 4: Advanced C#
- 213. To represent null in a value type, you must use a special construct called a nullable type. A nullable type is denoted with a value type followed by the ? symbol: int? i = null; // OK, Nullable Type Console.WriteLine (i == null); // True Nullable<T> Struct T? translates into System.Nullable<T>, which is a lightweight immutable structure, having only two fields, to represent Value and HasValue. The essence of System .Nullable<T> is very simple: public struct Nullable<T> where T : struct { public T Value {get;} public bool HasValue {get;} public T GetValueOrDefault(); public T GetValueOrDefault (T defaultValue); ... } The code int? i = null; Console.WriteLine (i == null); // True translates to the following: Nullable<int> i = new Nullable<int>(); Console.WriteLine (! i.HasValue); // True Attempting to retrieve Value when HasValue is false throws an InvalidOperation Exception. GetValueOrDefault() returns Value if HasValue is true; otherwise, it returns new T() or a specified custom default value. The default value of T? is null. Implicit and Explicit Nullable Conversions The conversion from T to T? is implicit, while the conversion from T? to T is explicit: int? x = 5; // implicit int y = (int)x; // explicit The explicit cast is directly equivalent to calling the nullable object’s Value property. Hence, an InvalidOperationException is thrown if HasValue is false. Boxing and Unboxing Nullable Values When T? is boxed, the boxed value on the heap contains T, not T?. This optimiza‐ tion is possible because a boxed value is a reference type that can already express null. Nullable Value Types | 195 Advanced C#
- 214. C# also permits the unboxing of nullable value types with the as operator. The result will be null if the cast fails: object o = "string"; int? x = o as int?; Console.WriteLine (x.HasValue); // False Operator Lifting The Nullable<T> struct does not define operators such as <, >, or even ==. Despite this, the following code compiles and executes correctly: int? x = 5; int? y = 10; bool b = x < y; // true This works because the compiler borrows or “lifts” the less-than operator from the underlying value type. Semantically, it translates the preceding comparison expression into this: bool b = (x.HasValue && y.HasValue) ? (x.Value < y.Value) : false; In other words, if both x and y have values, it compares via int’s less-than operator; otherwise, it returns false. Operator lifting means that you can implicitly use T’s operators on T?. You can define operators for T? in order to provide special-purpose null behavior, but in the vast majority of cases, it’s best to rely on the compiler automatically applying systematic nullable logic for you. Here are some examples: int? x = 5; int? y = null; // Equality operator examples Console.WriteLine (x == y); // False Console.WriteLine (x == null); // False Console.WriteLine (x == 5); // True Console.WriteLine (y == null); // True Console.WriteLine (y == 5); // False Console.WriteLine (y != 5); // True // Relational operator examples Console.WriteLine (x < 6); // True Console.WriteLine (y < 6); // False Console.WriteLine (y > 6); // False // All other operator examples Console.WriteLine (x + 5); // 10 Console.WriteLine (x + y); // null (prints empty line) The compiler performs null logic differently depending on the category of operator. The following sections explain these different rules. 196 | Chapter 4: Advanced C#
- 215. Equality operators (== and !=) Lifted equality operators handle nulls just like reference types do. This means that two null values are equal: Console.WriteLine ( null == null); // True Console.WriteLine ((bool?)null == (bool?)null); // True Further: • If exactly one operand is null, the operands are unequal. • • If both operands are non-null, their Values are compared. • Relational operators (<, <=, >=, >) The relational operators work on the principle that it is meaningless to compare null operands. This means comparing a null value to either a null or a non-null value returns false: bool b = x < y; // Translation: bool b = (x.HasValue && y.HasValue) ? (x.Value < y.Value) : false; // b is false (assuming x is 5 and y is null) All other operators (+, −, *, /, %, &, |, ^, <<, >>, +, ++, --, !, ~) These operators return null when any of the operands are null. This pattern should be familiar to SQL users: int? c = x + y; // Translation: int? c = (x.HasValue && y.HasValue) ? (int?) (x.Value + y.Value) : null; // c is null (assuming x is 5 and y is null) An exception is when the & and | operators are applied to bool?, which we discuss shortly. Mixing nullable and non-nullable operators You can mix and match nullable and non-nullable value types (this works because there is an implicit conversion from T to T?): int? a = null; int b = 2; int? c = a + b; // c is null - equivalent to a + (int?)b Nullable Value Types | 197 Advanced C#
- 216. bool? with & and | Operators When supplied operands of type bool?, the & and | operators treat null as an unknown value. So, null | true is true because: • If the unknown value is false, the result would be true. • • If the unknown value is true, the result would be true. • Similarly, null & false is false. This behavior should be familiar to SQL users. The following example enumerates other combinations: bool? n = null; bool? f = false; bool? t = true; Console.WriteLine (n | n); // (null) Console.WriteLine (n | f); // (null) Console.WriteLine (n | t); // True Console.WriteLine (n & n); // (null) Console.WriteLine (n & f); // False Console.WriteLine (n & t); // (null) Nullable Value Types and Null Operators Nullable value types work particularly well with the ?? operator (see “Null- Coalescing Operator” on page 74), as illustrated in this example: int? x = null; int y = x ?? 5; // y is 5 int? a = null, b = 1, c = 2; Console.WriteLine (a ?? b ?? c); // 1 (first non-null value) Using ?? on a nullable value type is equivalent to calling GetValueOrDefault with an explicit default value except that the expression for the default value is never evaluated if the variable is not null. Nullable value types also work well with the null-conditional operator (see “Null- Conditional Operator” on page 75). In the following example, length evaluates to null: System.Text.StringBuilder sb = null; int? length = sb?.ToString().Length; We can combine this with the null-coalescing operator to evaluate to zero instead of null: int length = sb?.ToString().Length ?? 0; // Evaluates to 0 if sb is null Scenarios for Nullable Value Types One of the most common scenarios for nullable value types is to represent unknown values. This frequently occurs in database programming, where a class is mapped to a table with nullable columns. If these columns are strings (e.g., an 198 | Chapter 4: Advanced C#
- 217. EmailAddress column on a Customer table), there is no problem because string is a reference type in the CLR, which can be null. However, most other SQL column types map to CLR struct types, making nullable value types very useful when mapping SQL to the CLR: // Maps to a Customer table in a database public class Customer { ... public decimal? AccountBalance; } A nullable type can also be used to represent the backing field of what’s sometimes called an ambient property. An ambient property, if null, returns the value of its parent: public class Row { ... Grid parent; Color? color; public Color Color { get { return color ?? parent.Color; } set { color = value == parent.Color ? (Color?)null : value; } } } Alternatives to Nullable Value Types Before nullable value types were part of the C# language (i.e., before C# 2.0), there were many strategies to deal with them, examples of which still appear in the .NET libraries for historical reasons. One of these strategies is to designate a particular non-null value as the “null value”; an example is in the string and array classes. String.IndexOf returns the magic value of −1 when the character is not found: int i = "Pink".IndexOf ('b'); Console.WriteLine (i); // −1 However, Array.IndexOf returns −1 only if the index is 0-bounded. The more general formula is that IndexOf returns one less than the lower bound of the array. In the next example, IndexOf returns 0 when an element is not found: // Create an array whose lower bound is 1 instead of 0: Array a = Array.CreateInstance (typeof (string), new int[] {2}, new int[] {1}); a.SetValue ("a", 1); a.SetValue ("b", 2); Console.WriteLine (Array.IndexOf (a, "c")); // 0 Nullable Value Types | 199 Advanced C#
- 218. Nominating a “magic value” is problematic for several reasons: • It means that each value type has a different representation of null. In contrast, • nullable value types provide one common pattern that works for all value types. • There might be no reasonable designated value. In the previous example, −1 • could not always be used. The same is true for our earlier example representing an unknown account balance. • Forgetting to test for the magic value results in an incorrect value that might • go unnoticed until later in execution—when it pulls an unintended magic trick. Forgetting to test HasValue on a null value, however, throws an InvalidOpera tionException on the spot. • The ability for a value to be null is not captured in the type. Types communicate • the intention of a program, allow the compiler to check for correctness, and enable a consistent set of rules enforced by the compiler. Nullable Reference Types Whereas nullable value types bring nullability to value types, nullable reference types (C# 8+) do the opposite. When enabled, they bring (a degree of) non-nullability to reference types, with the purpose of helping to avoid NullReferenceExceptions. Nullable reference types introduce a level of safety that’s enforced purely by the compiler, in the form of warnings when it detects code that’s at risk of generating a NullReferenceException. To enable nullable reference types, you must either add the Nullable element to your .csproj project file (if you want to enable it for the entire project): <PropertyGroup> <Nullable>enable</Nullable> </PropertyGroup> or/and use the following directives in your code, in the places where it should take effect: #nullable enable // enables nullable reference types from this point on #nullable disable // disables nullable reference types from this point on #nullable restore // resets nullable reference types to project setting After being enabled, the compiler makes non-nullability the default: if you want a reference type to accept nulls without the compiler generating a warning, you must apply the ? suffix to indicate a nullable reference type. In the following example, s1 is non-nullable, whereas s2 is nullable: #nullable enable // Enable nullable reference types string s1 = null; // Generates a compiler warning! string? s2 = null; // OK: s2 is nullable reference type 200 | Chapter 4: Advanced C#
- 219. Because nullable reference types are compile-time constructs, there’s no runtime difference between string and string?. In contrast, nullable value types introduce something concrete into the type system, namely the Nullable<T> struct. The following also generates a warning because x is not initialized: class Foo { string x; } The warning disappears if you initialize x, either via a field initializer or via code in the constructor. The Null-Forgiving Operator The compiler also warns you upon dereferencing a nullable reference type, if it thinks a NullReferenceException might occur. In the following example, accessing the string’s Length property generates a warning: void Foo (string? s) => Console.Write (s.Length); You can remove the warning with the null-forgiving operator (!): void Foo (string? s) => Console.Write (s!.Length); Our use of the null-forgiving operator in this example is dangerous in that we could end up throwing the very NullReferenceException we were trying to avoid in the first place. We could fix it as follows: void Foo (string? s) { if (s != null) Console.Write (s.Length); } Notice now that we don’t need the null-forgiving operator. This is because the compiler performs static flow analysis and is smart enough to infer—at least in simple cases—when a dereference is safe and there’s no chance of a NullReference Exception. The compiler’s ability to detect and warn is not bulletproof, and there are also limits to what’s possible in terms of coverage. For instance, it’s unable to know whether an array’s elements have been populated, and so the following does not generate a warning: var strings = new string[10]; Console.WriteLine (strings[0].Length); Separating the Annotation and Warning Contexts Enabling nullable reference types via the #nullable enable directive (or the <Nullable>enable</Nullable> project setting) does two things: Nullable Reference Types | 201 Advanced C#
- 220. • It enables the nullable annotation context, which tells the compiler to treat all • reference-type variable declarations as non-nullable unless suffixed by the ? symbol. • It enables the nullable warning context, which tells the compiler to gener‐ • ate warnings upon encountering code at risk of throwing a NullReference Exception. It can sometimes be useful to separate these two concepts and enable just the annotation context, or (less usefully) just the warning context: #nullable enable annotations // Enable the annotation context // OR: #nullable enable warnings // Enable the warning context (The same trick works with #nullable disable and #nullable restore.) You can also do it via the project file: <Nullable>annotations</Nullable> <!-- OR --> <Nullable>warnings</Nullable> Enabling just the annotation context for a particular class or assembly can be a good first step in introducing nullable reference types into a legacy codebase. By correctly annotating public members, your class or assembly can act as a “good citizen” to other classes or assemblies—so that they can benefit fully from nullable reference types—without having to deal with warnings in your own class or assembly. Treating Nullable Warnings as Errors In greenfield projects, it makes sense to fully enable the nullable context from the outset. You might want to take the additional step of treating nullable warnings as errors so that your project cannot compile until all null warnings have been resolved: <PropertyGroup> <Nullable>enable</Nullable> <WarningsAsErrors>CS8600;CS8602;CS8603</WarningsAsErrors> </PropertyGroup> Extension Methods Extension methods allow an existing type to be extended with new methods without altering the definition of the original type. An extension method is a static method of a static class, where the this modifier is applied to the first parameter. The type of the first parameter will be the type that is extended: public static class StringHelper { public static bool IsCapitalized (this string s) { 202 | Chapter 4: Advanced C#
- 221. if (string.IsNullOrEmpty(s)) return false; return char.IsUpper (s[0]); } } The IsCapitalized extension method can be called as though it were an instance method on a string, as follows: Console.WriteLine ("Perth".IsCapitalized()); An extension method call, when compiled, is translated back into an ordinary static method call: Console.WriteLine (StringHelper.IsCapitalized ("Perth")); The translation works as follows: arg0.Method (arg1, arg2, ...); // Extension method call StaticClass.Method (arg0, arg1, arg2, ...); // Static method call Interfaces can be extended, too: public static T First<T> (this IEnumerable<T> sequence) { foreach (T element in sequence) return element; throw new InvalidOperationException ("No elements!"); } ... Console.WriteLine ("Seattle".First()); // S Extension Method Chaining Extension methods, like instance methods, provide a tidy way to chain functions. Consider the following two functions: public static class StringHelper { public static string Pluralize (this string s) {...} public static string Capitalize (this string s) {...} } x and y are equivalent, and both evaluate to "Sausages", but x uses extension methods, whereas y uses static methods: string x = "sausage".Pluralize().Capitalize(); string y = StringHelper.Capitalize (StringHelper.Pluralize ("sausage")); Ambiguity and Resolution Namespaces An extension method cannot be accessed unless its class is in scope, typically by its namespace being imported. Consider the extension method IsCapitalized in the following example: Extension Methods | 203 Advanced C#
- 222. using System; namespace Utils { public static class StringHelper { public static bool IsCapitalized (this string s) { if (string.IsNullOrEmpty(s)) return false; return char.IsUpper (s[0]); } } } To use IsCapitalized, the following application must import Utils to avoid a compile-time error: namespace MyApp { using Utils; class Test { static void Main() => Console.WriteLine ("Perth".IsCapitalized()); } } Extension methods versus instance methods Any compatible instance method will always take precedence over an extension method. In the following example, Test’s Foo method will always take precedence, even when called with an argument x of type int: class Test { public void Foo (object x) { } // This method always wins } static class Extensions { public static void Foo (this Test t, int x) { } } The only way to call the extension method in this case is via normal static syntax, in other words, Extensions.Foo(...). Extension methods versus extension methods If two extension methods have the same signature, the extension method must be called as an ordinary static method to disambiguate the method to call. If one extension method has more specific arguments, however, the more specific method takes precedence. To illustrate, consider the following two classes: 204 | Chapter 4: Advanced C#
- 223. static class StringHelper { public static bool IsCapitalized (this string s) {...} } static class ObjectHelper { public static bool IsCapitalized (this object s) {...} } The following code calls StringHelper’s IsCapitalized method: bool test1 = "Perth".IsCapitalized(); Classes and structs are considered more specific than interfaces. Demoting an extension method An interesting scenario can arise when Microsoft adds an extension method to a .NET runtime library that conflicts with an extension method in some existing third-party library. As the author of the third-party library, you might want to “withdraw” your extension method, but without removing it, and without breaking binary compatibility with existing consumers. Fortunately, this is easy to accomplish, simply by removing the this keyword from your extension method’s definition. This demotes your extension method to an ordinary static method. The beauty of this solution is that any assembly that was compiled against your old library will continue to work (and bind to your method, as before). This is because extension method calls are converted to static method calls during compilation. Consumers will be affected by your demotion only when they recompile, at which time calls to your former extension method will now bind to Microsoft’s version (if the namespace has been imported). Should the consumer still want to call your method, they can do so by invoking it as a static method. Anonymous Types An anonymous type is a simple class created by the compiler on the fly to store a set of values. To create an anonymous type, use the new keyword followed by an object initializer, specifying the properties and values the type will contain; for example: var dude = new { Name = "Bob", Age = 23 }; The compiler translates this to (approximately) the following: internal class AnonymousGeneratedTypeName { private string name; // Actual field name is irrelevant private int age; // Actual field name is irrelevant public AnonymousGeneratedTypeName (string name, int age) { this.name = name; this.age = age; Anonymous Types | 205 Advanced C#
- 224. } public string Name { get { return name; } } public int Age { get { return age; } } // The Equals and GetHashCode methods are overridden (see Chapter 6). // The ToString method is also overridden. } ... var dude = new AnonymousGeneratedTypeName ("Bob", 23); You must use the var keyword to reference an anonymous type because it doesn’t have a name. The property name of an anonymous type can be inferred from an expression that is itself an identifier (or ends with one); thus int Age = 23; var dude = new { Name = "Bob", Age, Age.ToString().Length }; is equivalent to the following: var dude = new { Name = "Bob", Age = Age, Length = Age.ToString().Length }; Two anonymous type instances declared within the same assembly will have the same underlying type if their elements are named and typed identically: var a1 = new { X = 2, Y = 4 }; var a2 = new { X = 2, Y = 4 }; Console.WriteLine (a1.GetType() == a2.GetType()); // True Additionally, the Equals method is overridden to perform structural equality com‐ parison (comparison of the data): Console.WriteLine (a1.Equals (a2)); // True whereas the equality operator (==) performs referential comparison: Console.WriteLine (a1 == a2); // False You can create arrays of anonymous types as follows: var dudes = new[] { new { Name = "Bob", Age = 30 }, new { Name = "Tom", Age = 40 } }; A method cannot (usefully) return an anonymously typed object, because it is illegal to write a method whose return type is var: var Foo() => new { Name = "Bob", Age = 30 }; // Not legal! Instead, you must use object or dynamic, and then whoever calls Foo must rely on dynamic binding, with loss of static type safety (and IntelliSense in Visual Studio): dynamic Foo() => new { Name = "Bob", Age = 30 }; // No static type safety. 206 | Chapter 4: Advanced C#
- 225. Anonymous types are immutable, so instances cannot be modified after creation. However, from C# 10, you can use the with keyword to create a copy with variations (nondestructive mutation): var a1 = new { A = 1, B = 2, C = 3, D = 4, E = 5 }; var a2 = a1 with { E = 10 }; Console.WriteLine (a2); // { A = 1, B = 2, C = 3, D = 4, E = 10 } Anonymous types are particularly useful when writing LINQ queries (see Chapter 8). Tuples Like anonymous types, tuples provide a simple way to store a set of values. The main purpose of tuples is to safely return multiple values from a method without resorting to out parameters (something you cannot do with anonymous types). Tuples do almost everything that anonymous types do and more. Their one disadvantage—as you’ll see soon—is runtime type erasure with named elements. The simplest way to create a tuple literal is to list the desired values in parentheses. This creates a tuple with unnamed elements, which you refer to as Item1, Item2, and so on: var bob = ("Bob", 23); // Allow compiler to infer the element types Console.WriteLine (bob.Item1); // Bob Console.WriteLine (bob.Item2); // 23 Tuples are value types, with mutable (read/write) elements: var joe = bob; // joe is a *copy* of bob joe.Item1 = "Joe"; // Change joe’s Item1 from Bob to Joe Console.WriteLine (bob); // (Bob, 23) Console.WriteLine (joe); // (Joe, 23) Unlike with anonymous types, you can specify a tuple type explicitly. Just list each of the element types in parentheses: (string,int) bob = ("Bob", 23); This means that you can usefully return a tuple from a method: (string,int) person = GetPerson(); // Could use 'var' instead if we want Console.WriteLine (person.Item1); // Bob Console.WriteLine (person.Item2); // 23 (string,int) GetPerson() => ("Bob", 23); Tuples play well with generics, so the following types are all legal: Task<(string,int)> Dictionary<(string,int),Uri> IEnumerable<(int id, string name)> // See below for naming elements Tuples | 207 Advanced C#
- 226. Naming Tuple Elements You can optionally give meaningful names to elements when creating tuple literals: var tuple = (name:"Bob", age:23); Console.WriteLine (tuple.name); // Bob Console.WriteLine (tuple.age); // 23 You can do the same when specifying tuple types: var person = GetPerson(); Console.WriteLine (person.name); // Bob Console.WriteLine (person.age); // 23 (string name, int age) GetPerson() => ("Bob", 23); Note that you can still treat the elements as unnamed and refer to them as Item1, Item2, etc. (although Visual Studio hides these fields from IntelliSense). Element names are automatically inferred from property or field names: var now = DateTime.Now; var tuple = (now.Day, now.Month, now.Year); Console.WriteLine (tuple.Day); // OK Tuples are type compatible with one another if their element types match up (in order). Their element names need not: (string name, int age, char sex) bob1 = ("Bob", 23, 'M'); (string age, int sex, char name) bob2 = bob1; // No error! Our particular example leads to confusing results: Console.WriteLine (bob2.name); // M Console.WriteLine (bob2.age); // Bob Console.WriteLine (bob2.sex); // 23 Type erasure We stated previously that the C# compiler handles anonymous types by building custom classes with named properties for each of the elements. With tuples, C# works differently and uses a preexisting family of generic structs: public struct ValueTuple<T1> public struct ValueTuple<T1,T2> public struct ValueTuple<T1,T2,T3> ... Each of the ValueTuple<> structs has fields named Item1, Item2, and so on. Hence, (string,int) is an alias for ValueTuple<string,int>, and this means that named tuple elements have no corresponding property names in the underlying types. Instead, the names exist only in the source code, and in the imagination of the compiler. At runtime, the names mostly disappear, so if you decompile a program that refers to named tuple elements, you’ll see references to just Item1, 208 | Chapter 4: Advanced C#
- 227. Item2, and so on. Further, when you examine a tuple variable in a debugger after having assigned it to an object (or Dump it in LINQPad), the element names are not there. And for the most part, you cannot use reflection (Chapter 18) to determine a tuple’s element names at runtime. We said that the names mostly disappear because there’s an exception. With methods/properties that return named tuple types, the compiler emits the element names by applying a custom attribute called TupleElementNamesAttribute (see “Attributes” on page 227) to the member’s return type. This allows named elements to work when calling methods in a different assembly (for which the compiler does not have the source code). ValueTuple.Create You can also create tuples via a factory method on the (nongeneric) ValueTuple type: ValueTuple<string,int> bob1 = ValueTuple.Create ("Bob", 23); (string,int) bob2 = ValueTuple.Create ("Bob", 23); (string name, int age) bob3 = ValueTuple.Create ("Bob", 23); Deconstructing Tuples Tuples implicitly support the deconstruction pattern (see “Deconstructors” on page 20), so you can easily deconstruct a tuple into individual variables. Consider the following: var bob = ("Bob", 23); string name = bob.Item1; int age = bob.Item2; With the tuple’s deconstructor, you can simplify the code to this: var bob = ("Bob", 23); (string name, int age) = bob; // Deconstruct the bob tuple into // separate variables (name and age). Console.WriteLine (name); Console.WriteLine (age); The syntax for deconstruction is confusingly similar to the syntax for declaring a tuple with named elements. The following highlights the difference: (string name, int age) = bob; // Deconstructing a tuple (string name, int age) bob2 = bob; // Declaring a new tuple Here’s another example, this time when calling a method and with type inference (var): Tuples | 209 Advanced C#
- 228. var (name, age, sex) = GetBob(); Console.WriteLine (name); // Bob Console.WriteLine (age); // 23 Console.WriteLine (sex); // M string, int, char) GetBob() => ( "Bob", 23, 'M'); You can also deconstruct directly into fields and properties, which provides a nice shortcut for populating multiple fields or properties in a constructor: class Point { public readonly int X, Y; public Point (int x, int y) => (X, Y) = (x, y); } Equality Comparison As with anonymous types, the Equals method performs structural equality compar‐ ison. This means that it compares the underlying data rather than the reference: var t1 = ("one", 1); var t2 = ("one", 1); Console.WriteLine (t1.Equals (t2)); // True In addition, ValueTuple<> overloads the == and != operators: Console.WriteLine (t1 == t2); // True (from C# 7.3) Tuples also override the GetHashCode method, making it practical to use tuples as keys in dictionaries. We cover equality comparison in detail in “Equality Compari‐ son” on page 324, and dictionaries in Chapter 7. The ValueTuple<> types also implement IComparable (see “Order Comparison” on page 335), making it possible to use tuples as a sorting key. The System.Tuple classes You’ll find another family of generic types in the System namespace called Tuple (rather than ValueTuple). These were introduced back in 2010 and were defined as classes (whereas the ValueTuple types are structs). Defining tuples as classes was in retrospect considered a mistake: in the scenarios in which tuples are commonly used, structs have a slight performance advantage (in that they avoid unnecessary memory allocations), with almost no downside. Hence, when Microsoft added language support for tuples in C# 7, it ignored the existing Tuple types in favor of the new ValueTuple. You might still come across the Tuple classes in code written prior to C# 7. They have no special language support and are used as follows: Tuple<string,int> t = Tuple.Create ("Bob", 23); // Factory method Console.WriteLine (t.Item1); // Bob Console.WriteLine (t.Item2); // 23 210 | Chapter 4: Advanced C#
- 229. Records A record is a special kind of class or struct that’s designed to work well with immut‐ able (read-only) data. Its most useful feature is nondestructive mutation; however, records are also useful in creating types that just combine or hold data. In simple cases, they eliminate boilerplate code while honoring the equality semantics most suitable for immutable types. Records are purely a C# compile-time construct. At runtime, the CLR sees them just as classes or structs (with a bunch of extra “synthesized” members added by the compiler). Background Writing immutable types (whose fields cannot be modified after initialization) is a popular strategy for simplifying software and reducing bugs. It’s also a core aspect of functional programming, where mutable state is avoided and functions are treated as data. LINQ is inspired by this principle. In order to “modify” an immutable object, you must create a new one and copy over the data while incorporating your modifications (this is called nondestructive muta‐ tion). In terms of performance, this is not as inefficient as you might expect, because a shallow copy will always suffice (a deep copy, where you also copy subobjects and collections, is unnecessary when data is immutable). But in terms of coding effort, implementing nondestructive mutation can be very inefficient, especially when there are many properties. Records solve this problem via a language-supported pattern. A second issue is that programmers—particularly functional programmers—some‐ times use immutable types just to combine data (without adding behavior). Defin‐ ing such types is more work than it should be, requiring a constructor to assign each parameter to each property (a deconstructor may also be useful). With records, the compiler can do this work for you. Finally, one of the consequences of an object being immutable is that its identity cannot change, which means that it’s more useful for such types to implement struc‐ tural equality than referential equality. Structural equality means that two instances are the same if their data is the same (as with tuples). Records give you structural equality by default—regardless of whether the underlying type is a class or struct— without any boilerplate code. Defining a Record A record definition is like a class or struct definition, and can contain the same kinds of members, including fields, properties, methods, and so on. Records can implement interfaces, and (class-based) records can subclass other (class-based) records. Records | 211 Advanced C#
- 230. By default, the underlying type of a record is a class: record Point { } // Point is a class From C# 10, the underlying type of a record can also be a struct: record struct Point { } // Point is a struct (record class is also legal and has the same meaning as record.) A simple record might contain just a bunch of init-only properties and perhaps a constructor: record Point { public Point (double x, double y) => (X, Y) = (x, y); public double X { get; init; } public double Y { get; init; } } Our constructor employs a shortcut that we described in the preceding section. (X, Y) = (x, y); is equivalent (in this case) to the following: { this.X = x; this.Y = y; } Upon compilation, C# transforms the record definition into a class (or struct) and performs the following additional steps: • It writes a protected copy constructor (and a hidden Clone method) to facilitate • nondestructive mutation. • It overrides/overloads the equality-related functions to implement structural • equality. • It overrides the ToString() method (to expand the record’s public properties, • as with anonymous types). The preceding record declaration expands into something like this: class Point { public Point (double x, double y) => (X, Y) = (x, y); public double X { get; init; } public double Y { get; init; } protected Point (Point original) // “Copy constructor” { this.X = original.X; this.Y = original.Y } // This method has a strange compiler-generated name: public virtual Point <Clone>$() => new Point (this); // Clone method 212 | Chapter 4: Advanced C#
- 231. // Additional code to override Equals, ==, !=, GetHashCode, ToString() // ... } While there’s nothing to stop you from putting optional parameters into the constructor, a good pattern (at least in public libraries) is to leave them out of the constructor and expose them purely as init-only properties: new Foo (123, 234) { Optional2 = 345 }; record Foo { public Foo (int required1, int required2) { ... } public int Required1 { get; init; } public int Required2 { get; init; } public int Optional1 { get; init; } public int Optional2 { get; init; } } The advantage of this pattern is that you can safely add init- only properties later without breaking binary compatibility with consumers who have compiled against older versions of your assembly. Parameter lists A record definition can also include a parameter list: record Point (double X, double Y) { // You can optionally define additional class members here... } Parameters can include the in and params modifiers, but not out or ref. If a parameter list is specified, the compiler performs the following extra steps: • It writes an init-only property per parameter. • • It writes a primary constructor to populate the properties. • • It writes a deconstructor. • Mutability with Record Structs When you define a parameter list in a record struct, the compiler emits writable properties instead of init-only properties, unless you prefix the record declaration with readonly: readonly record struct Point (double X, double Y); Records | 213 Advanced C#
- 232. The rationale is that in typical use cases, the safety benefits of immutability arise not from a struct being immutable but from its home being immutable. In the following example, we are unable to mutate field X, even though X is writable: var test = new Immutable(); test.Field.X++; // Prohibited, because Field is readonly test.Prop.X++; // Prohibited, because Prop is {get;} only class Immutable { public readonly Mutable Field; public Mutable Prop { get; } } struct Mutable { public int X, Y; } And while we could do the following: var test = new Immutable(); Mutable m = test.Prop; m.X++; all that we would achieve is to mutate a local variable (a copy of test.Prop). Mutat‐ ing a local variable can be a useful optimization and doesn’t invalidate the benefits of an immutable type system. Conversely, if we made Field a writable field, and Prop a writable property, we could simply replace their contents—regardless of how the Mutable struct was declared. This means that if we declare our Point record simply as record Point (double X, double Y); the compiler will end up generating (almost) exactly what we listed in the preceding expansion. A minor difference is that the parameter names in the primary construc‐ tor will end up as X and Y instead of x and y: public Point (double X, double Y) // “Primary constructor” { this.X = X; this.Y = Y; } Also, due to being a primary constructor, parameters X and Y become magically available to any field or property initializers in your record. We discuss the subtleties of this later, in “Pri‐ mary Constructors” on page 219. Another difference, when you define a parameter list, is that the compiler also generates a deconstructor: public void Deconstruct (out double X, out double Y) // Deconstructor { X = this.X; Y = this.Y; } Records with parameter lists can be subclassed using the following syntax: 214 | Chapter 4: Advanced C#
- 233. record Point3D (double X, double Y, double Z) : Point (X, Y); The compiler then emits a primary constructor as follows: class Point3D : Point { public double Z { get; init; } public Point3D (double X, double Y, double Z) : base (X, Y) => this.Z = Z; } Parameter lists offer a nice shortcut when you need a class that simply groups together a bunch of values (a product type in functional programming), and can also be useful for prototyp‐ ing. As we’ll see later, they’re not so helpful when you need to add logic to the init accessors (such as argument validation). Nondestructive Mutation The most important step that the compiler performs with all records is to write a copy constructor (and a hidden Clone method). This enables nondestructive muta‐ tion via the with keyword: Point p1 = new Point (3, 3); Point p2 = p1 with { Y = 4 }; Console.WriteLine (p2); // Point { X = 3, Y = 4 } record Point (double X, double Y); In this example, p2 is a copy of p1, but with its Y property set to 4. The benefit is more apparent when there are more properties: Test t1 = new Test (1, 2, 3, 4, 5, 6, 7, 8); Test t2 = t1 with { A = 10, C = 30 }; Console.WriteLine (t2); record Test (int A, int B, int C, int D, int E, int F, int G, int H); Here’s the output: Test { A = 10, B = 2, C = 30, D = 4, E = 5, F = 6, G = 7, H = 8 } Nondestructive mutation occurs in two phases: 1. First, the copy constructor clones the record. By default, it copies each of 1. the record’s underlying fields, creating a faithful replica while bypassing (the overhead of) any logic in the init accessors. All fields are included (public and private, as well as the hidden fields that back automatic properties). 2. Then, each property in the member initializer list is updated (this time using 2. the init accessors). The compiler translates: Records | 215 Advanced C#
- 234. Test t2 = t1 with { A = 10, C = 30 }; into something functionally equivalent to the following: Test t2 = new Test(t1); // Use copy constructor to clone t1 field by field t2.A = 10; // Update property A t2.C = 30; // Update property C (The same code would not compile if you wrote it explicitly because A and C are init-only properties. Furthermore, the copy constructor is protected; C# works around this by invoking it via a public hidden method that it writes into the record called <Clone>$.) If necessary, you can define your own copy constructor. C# will then use your definition instead of writing one itself: protected Point (Point original) { this.X = original.X; this.Y = original.Y; } Writing a custom copy constructor might be useful if your record contains mutable subobjects or collections that you wish to clone, or if there are computed fields that you wish to clear. Unfortunately, you can only replace, not enhance, the default implementation. When subclassing another record, the copy constructor is responsible for copying only its own fields. To copy the base record’s fields, delegate to the base: protected Point (Point original) : base (original) { ... } Property Validation With explicit properties, you can write validation logic into the init accessors. In the following example, we ensure that X can never be NaN (Not a Number): record Point { // Notice that we assign x to the X property (and not the _x field): public Point (double x, double y) => (X, Y) = (x, y); double _x; public double X { get => _x; init { if (double.IsNaN (value)) throw new ArgumentException ("X Cannot be NaN"); _x = value; } } 216 | Chapter 4: Advanced C#
- 235. public double Y { get; init; } } Our design ensures that validation occurs both during construction and when the object is nondestructively mutated: Point p1 = new Point (2, 3); Point p2 = p1 with { X = double.NaN }; // throws an exception Recall that the automatically generated copy constructor copies over all fields and automatic properties. This means that the generated copy constructor will now look like this: protected Point (Point original) { _x = original._x; Y = original.Y; } Notice that the copying of the _x field circumvents the X property accessor. How‐ ever, this cannot break anything, because it’s faithfully copying an object that will have already been safely populated via X’s init accessor. Calculated Fields and Lazy Evaluation A popular functional programming pattern that works well with immutable types is lazy evaluation, where a value is not computed until required, and is then cached for reuse. Suppose, for instance, that we want to define a property in our Point record that returns the distance from the origin (0, 0): record Point (double X, double Y) { public double DistanceFromOrigin => Math.Sqrt (X*X + Y*Y); } Let’s now try to refactor this to avoid the cost of recomputing DistanceFromOrigin every time the property is accessed. We’ll start by removing the property list and defining X, Y, and DistanceFromOrigin as read-only properties. Then we can calcu‐ late the latter in the constructor: record Point { public double X { get; } public double Y { get; } public double DistanceFromOrigin { get; } public Point (double x, double y) => (X, Y, DistanceFromOrigin) = (x, y, Math.Sqrt (x*x + y*y)); } This works, but it doesn’t allow for nondestructive mutation (changing X and Y to init-only properties would break the code because DistanceFromOrigin would become stale after the init accessors execute). It’s also suboptimal in that the calcu‐ lation is always performed, regardless of whether the DistanceFromOrigin property Records | 217 Advanced C#
- 236. is ever read. The optimal solution is to cache its value in a field and populate it lazily (on first use): record Point { ... double? _distance; public double DistanceFromOrigin { get { if (_distance == null) _distance = Math.Sqrt (X*X + Y*Y); return _distance.Value; } } } Technically, we mutate _distance in this code. It’s still fair, though, to call Point an immutable type. Mutating a field purely to populate a lazy value does not invalidate the prin‐ ciples or benefits of immutability, and can even be masked through the use of the Lazy<T> type that we describe in Chapter 21. With C#’s null-coalescing assignment operator (??=) we can reduce the entire prop‐ erty declaration to one line of code: public double DistanceFromOrigin => _distance ??= Math.Sqrt (X*X + Y*Y); (This says, return _distance if it’s non-null; otherwise return Math.Sqrt (X*X + Y*Y) while assigning it to _distance.) To make this work with init-only properties, we need one further step, which is to clear the cached _distance field when X or Y is updated via the init accessor. Here’s the complete code: record Point { public Point (double x, double y) => (X, Y) = (x, y); double _x, _y; public double X { get => _x; init { _x = value; _distance = null; } } public double Y { get => _y; init { _y = value; _distance = null; } } double? _distance; public double DistanceFromOrigin => _distance ??= Math.Sqrt (X*X + Y*Y); } Point can now be mutated nondestructively: Point p1 = new Point (2, 3); Console.WriteLine (p1.DistanceFromOrigin); // 3.605551275463989 218 | Chapter 4: Advanced C#
- 237. Point p2 = p1 with { Y = 4 }; Console.WriteLine (p2.DistanceFromOrigin); // 4.47213595499958 A nice bonus is that the autogenerated copy constructor copies over the cached _distance field. This means that should a record have other properties that aren’t involved in the calculation, a nondestructive mutation of those properties wouldn’t trigger an unnecessary loss of the cached value. If you don’t care for this bonus, an alternative to clearing the cached value in the init accessors is to write a custom copy constructor that ignores the cached field. This is more concise because it works with parameter lists, and the custom copy constructor can leverage the deconstructor: record Point (double X, double Y) { double? _distance; public double DistanceFromOrigin => _distance ??= Math.Sqrt (X*X + Y*Y); protected Point (Point other) => (X, Y) = other; } Note that with either solution, the addition of lazy calculated fields breaks the default structural equality comparison (because such fields may or may not be populated), although we’ll see shortly that it’s relatively easy to fix. Primary Constructors When you define a record with a parameter list, the compiler generates property declarations automatically, as well as a primary constructor (and a deconstructor). As we’ve seen, this works well in simple cases, and in more complex cases you can omit the parameter list and write the property declarations and constructor manually. C# also offers a mildly useful intermediate option—if you’re willing to deal with the curious semantics of primary constructors—which is to define a parameter list while writing some or all of the property declarations yourself: record Student (string ID, string LastName, string GivenName) { public string ID { get; } = ID; } In this case, we “took over” the ID property definition, defining it as read-only (instead of init-only), preventing it from partaking in nondestructive mutation. If you never need to nondestructively mutate a particular property, making it read-only lets you store computed data in the record without having to code up a refresh mechanism. Notice that we needed to include a property initializer (in boldface): public string ID { get; } = ID; Records | 219 Advanced C#
- 238. When you “take over” a property declaration, you become responsible for initializ‐ ing its value; the primary constructor no longer does this automatically. Note that the ID in boldface refers to the primary constructor parameter, not the ID property. With record structs, it’s legal to redefine a property as a field: record struct Student (string ID) { public string ID = ID; } A unique feature of primary constructors is that their parameters (ID, LastName, and GivenName in this case) are magically visible to all field and property initializers. We can illustrate this by extending our example as follows: record Student (string ID, string LastName, string FirstName) { public string ID { get; } = ID; readonly int _enrollmentYear = int.Parse (ID.Substring (0, 4)); } Again, the ID in boldface refers to the primary constructor parameter, not the property. (The reason for there not being an ambiguity is that it’s illegal to access properties from initializers.) In this example, we calculated _enrollmentYear from the first four digits of the ID. While it’s safe to store this in a read-only field (because the ID property is read-only and so cannot be nondestructively mutated), this code would not work so well in the real world. This is because without an explicit constructor, there’s no central place in which to validate ID and throw a meaningful exception should it be invalid (a common requirement). Validation is also a good reason for needing to write explicit init-only accessors (as we discussed in “Property Validation” on page 216). Unfortunately, primary constructors do not play well in this scenario. To illustrate, consider the following record, where an init accessor performs a null validation check: record Person (string Name) { string _name = Name; public string Name { get => _name; init => _name = value ?? throw new ArgumentNullException ("Name"); } } Because Name is not an automatic property, it cannot define an initializer. The best we can do is put the initializer on the backing field (in boldface). Unfortunately, doing so bypasses the null check: var p = new Person (null); // Succeeds! (bypasses the null check) The difficulty is that there’s no way to assign a primary constructor parameter to a property without writing the constructor ourselves. While there are workarounds 220 | Chapter 4: Advanced C#
- 239. (such as factoring the init validation logic into a separate static method that we call twice), the simplest workaround is to avoid the parameter list altogether and write an ordinary constructor manually (and deconstructor, should you need it): record Person { public Person (string name) => Name = name; // Assign to *PROPERTY* string _name; public string Name { get => _name; init => ... } } Records and Equality Comparison Just as with structs, anonymous types, and tuples, records provide structural equal‐ ity out of the box, meaning that two records are equal if their fields (and automatic properties) are equal: var p1 = new Point (1, 2); var p2 = new Point (1, 2); Console.WriteLine (p1.Equals (p2)); // True record Point (double X, double Y); The equality operator also works with records (as it does with tuples): Console.WriteLine (p1 == p2); // True The default equality implementation for records is unavoidably fragile. In particular, it breaks if the record contains lazy values, transient values, arrays, or collection types (which require special handling for equality comparison). Fortunately, it’s relatively easy to fix (should you need equality to work), and doing so is less work than adding full equality behavior to classes or structs. Unlike with classes and structs, you do not (and cannot) override the object.Equals method; instead, you define a public Equals method with the following signature: record Point (double X, double Y) { double _someOtherField; public virtual bool Equals (Point other) => other != null && X == other.X && Y == other.Y; } The Equals method must be virtual (not override), and it must be strongly typed such that it accepts the actual record type (Point in this case, not object). Once you get the signature right, the compiler will automatically patch in your method. In our example, we changed the equality logic such that we compare only X and Y (and ignore _someOtherField). Should you subclass another record, you can call the base.Equals method: public virtual bool Equals (Point other) => base.Equals (other) && ... Records | 221 Advanced C#
- 240. As with any type, if you take over equality comparison, you should also override GetHashCode(). A nice feature of records is that you don’t overload != or ==; nor do you implement IEquatable<T>: this is all done for you. We cover this topic of equality comparison fully in “Equality Comparison” on page 324. Patterns In Chapter 3, we demonstrated how to use the is operator to test whether a reference conversion will succeed: if (obj is string) Console.WriteLine (((string)obj).Length); Or, more concisely: if (obj is string s) Console.WriteLine (s.Length); This employs one kind of pattern called a type pattern. The is operator also sup‐ ports other patterns that were introduced in recent versions of C#, such as the property pattern: if (obj is string { Length:4 }) Console.WriteLine ("A string with 4 characters"); Patterns are supported in the following contexts: • After the is operator (variable is pattern) • • In switch statements • • In switch expressions • We’ve already covered the type pattern (and briefly, the tuple pattern) in “Switching on types” on page 81 and “The is operator” on page 118. In this section, we cover more advanced patterns that were introduced in recent versions of C#. Some of the more specialized patterns are intended for use in switch state‐ ments/expressions. Here, they reduce the need for when clauses and let you use switches where you couldn’t previously. The patterns in this section are mild-to-moderately useful in some scenarios. Remember that you can always replace highly patterned switch expressions with simple if statements— or, in some cases, the ternary conditional operator—often without much extra code. var Pattern The var pattern is a variation of the type pattern whereby you replace the type name with the var keyword. The conversion always succeeds, so its purpose is merely to let you reuse the variable that follows: 222 | Chapter 4: Advanced C#
- 241. bool IsJanetOrJohn (string name) => name.ToUpper() is var upper && (upper == "JANET" || upper == "JOHN"); This is equivalent to: bool IsJanetOrJohn (string name) { string upper = name.ToUpper(); return upper == "JANET" || upper == "JOHN"; } The ability to introduce and reuse an intermediate variable (upper, in this case) in an expression-bodied method is convenient. Unfortunately, it tends to be useful only when the method in question has a bool return type. Constant Pattern The constant pattern lets you match directly to a constant, and is useful when working with the object type: void Foo (object obj) { if (obj is 3) ... } This expression in boldface is equivalent to the following: obj is int && (int)obj == 3 (Being a static operator, C# won’t let you use == to compare an object directly to a constant, because the compiler needs to know the types in advance.) On its own, this pattern is only marginally useful in that there’s a reasonable alternative: if (3.Equals (obj)) ... As we’ll see soon, the constant pattern can become more useful with pattern combinators. Relational Patterns From C# 9, you can use the <, >, <=, and >= operators in patterns: if (x is > 100) Console.WriteLine ("x is greater than 100"); This becomes meaningfully useful in a switch: string GetWeightCategory (decimal bmi) => bmi switch { < 18.5m => "underweight", < 25m => "normal", < 30m => "overweight", _ => "obese" }; Patterns | 223 Advanced C#
- 242. Relational patterns become even more useful in conjunction with pattern combinators. The relational pattern also works when the variable has a compile-time type of object, but you have to be extremely careful with your use of numeric constants. In the following example, the last line prints False because we are attempting to match a decimal value to an integer literal: object obj = 2m; // obj is decimal Console.WriteLine (obj is < 3m); // True Console.WriteLine (obj is < 3); // False Pattern Combinators From C# 9, you can use the and, or, and not keywords to combine patterns: bool IsJanetOrJohn (string name) => name.ToUpper() is "JANET" or "JOHN"; bool IsVowel (char c) => c is 'a' or 'e' or 'i' or 'o' or 'u'; bool Between1And9 (int n) => n is >= 1 and <= 9; bool IsLetter (char c) => c is >= 'a' and <= 'z' or >= 'A' and <= 'Z'; As with the && and || operators, and has higher precedence than or. You can override this with parentheses. A nice trick is to combine the not combinator with the type pattern to test whether an object is (not) a type: if (obj is not string) ... This looks nicer than: if (!(obj is string)) ... Tuple and Positional Patterns The tuple pattern (introduced in C# 8) matches tuples: var p = (2, 3); Console.WriteLine (p is (2, 3)); // True You can use this to switch on multiple values: int AverageCelsiusTemperature (Season season, bool daytime) => (season, daytime) switch { (Season.Spring, true) => 20, (Season.Spring, false) => 16, (Season.Summer, true) => 27, (Season.Summer, false) => 22, (Season.Fall, true) => 18, (Season.Fall, false) => 12, (Season.Winter, true) => 10, 224 | Chapter 4: Advanced C#
- 243. (Season.Winter, false) => -2, _ => throw new Exception ("Unexpected combination") }; enum Season { Spring, Summer, Fall, Winter }; The tuple pattern can be considered a special case of the positional pattern (C# 8+), which matches any type that exposes a Deconstruct method (see “Deconstructors” on page 20). In the following example, we leverage the Point record’s compiler- generated deconstructor: var p = new Point (2, 2); Console.WriteLine (p is (2, 2)); // True record Point (int X, int Y); // Has compiler-generated deconstructor You can deconstruct as you match, using the following syntax: Console.WriteLine (p is (var x, var y) && x == y); // True Here’s a switch expression that combines a type pattern with a positional pattern: string Print (object obj) => obj switch { Point (0, 0) => "Empty point", Point (var x, var y) when x == y => "Diagonal" ... }; Property Patterns A property pattern (C# 8+) matches on one or more of an object’s property values. We gave a simple example previously in the context of the is operator: if (obj is string { Length:4 }) ... However, this doesn’t save much over the following: if (obj is string s && s.Length == 4) ... With switch statements and expressions, property patterns are more useful. Con‐ sider the System.Uri class, which represents a URI. It has properties that include Scheme, Host, Port, and IsLoopback. In writing a firewall, we could decide whether to allow or block a URI by employing a switch expression that uses property patterns: bool ShouldAllow (Uri uri) => uri switch { { Scheme: "http", Port: 80 } => true, { Scheme: "https", Port: 443 } => true, { Scheme: "ftp", Port: 21 } => true, { IsLoopback: true } => true, _ => false }; Patterns | 225 Advanced C#
- 244. You can nest properties, making the following clause legal: { Scheme: { Length: 4 }, Port: 80 } => true, which, from C# 10, can be simplified to: { Scheme.Length: 4, Port: 80 } => true, You can use other patterns inside property patterns, including the relational pattern: { Host: { Length: < 1000 }, Port: > 0 } => true, More elaborate conditions can be expressed with a when clause: { Scheme: "http" } when string.IsNullOrWhiteSpace (uri.Query) => true, You can also combine the property pattern with the type pattern: bool ShouldAllow (object uri) => uri switch { Uri { Scheme: "http", Port: 80 } => true, Uri { Scheme: "https", Port: 443 } => true, ... As you might expect with type patterns, you can introduce a variable at the end of a clause and then consume that variable: Uri { Scheme: "http", Port: 80 } httpUri => httpUri.Host.Length < 1000, You can also use that variable in a when clause: Uri { Scheme: "http", Port: 80 } httpUri when httpUri.Host.Length < 1000 => true, A somewhat bizarre twist with property patterns is that you can also introduce variables at the property level: { Scheme: "http", Port: 80, Host: string host } => host.Length < 1000, Implicit typing is permitted, so you can substitute string with var. Here’s a com‐ plete example: bool ShouldAllow (Uri uri) => uri switch { { Scheme: "http", Port: 80, Host: var host } => host.Length < 1000, { Scheme: "https", Port: 443 } => true, { Scheme: "ftp", Port: 21 } => true, { IsLoopback: true } => true, _ => false }; It’s difficult to invent examples for which this saves more than a few characters. In our case, the alternative is actually shorter: { Scheme: "http", Port: 80 } => uri.Host.Length < 1000 => ... Or: { Scheme: "http", Port: 80, Host: { Length: < 1000 } } => ... 226 | Chapter 4: Advanced C#
- 245. Attributes You’re already familiar with the notion of attributing code elements of a program with modifiers, such as virtual or ref. These constructs are built into the language. Attributes are an extensible mechanism for adding custom information to code elements (assemblies, types, members, return values, parameters, and generic type parameters). This extensibility is useful for services that integrate deeply into the type system, without requiring special keywords or constructs in the C# language. A good scenario for attributes is serialization—the process of converting arbitrary objects to and from a particular format for storage or transmission. In this scenario, an attribute on a field can specify the translation between C#’s representation of the field and the format’s representation of the field. Attribute Classes An attribute is defined by a class that inherits (directly or indirectly) from the abstract class System.Attribute. To attach an attribute to a code element, specify the attribute’s type name in square brackets, before the code element. For example, the following attaches the ObsoleteAttribute to the Foo class: [ObsoleteAttribute] public class Foo {...} This particular attribute is recognized by the compiler and will cause compiler warnings if a type or member marked as obsolete is referenced. By convention, all attribute types end in the word Attribute. C# recognizes this and allows you to omit the suffix when attaching an attribute: [Obsolete] public class Foo {...} ObsoleteAttribute is a type declared in the System namespace as follows (simpli‐ fied for brevity): public sealed class ObsoleteAttribute : Attribute {...} The .NET libraries include many predefined attributes. We describe how to write your own attributes in Chapter 18. Named and Positional Attribute Parameters Attributes can have parameters. In the following example, we apply XmlTypeAttri bute to a class. This attribute instructs the XML serializer (in System.Xml.Seriali zation) as to how an object is represented in XML and accepts several attribute parameters. The following attribute maps the CustomerEntity class to an XML element named Customer, which belongs to the http://oreilly.com namespace: [XmlType ("Customer", Namespace="http://oreilly.com")] public class CustomerEntity { ... } Attributes | 227 Advanced C#
- 246. Attribute parameters fall into one of two categories: positional or named. In the preceding example, the first argument is a positional parameter; the second is a named parameter. Positional parameters correspond to parameters of the attribute type’s public constructors. Named parameters correspond to public fields or public properties on the attribute type. When specifying an attribute, you must include positional parameters that corre‐ spond to one of the attribute’s constructors. Named parameters are optional. In Chapter 18, we describe the valid parameter types and rules for their evaluation. Applying Attributes to Assemblies and Backing Fields Implicitly, the target of an attribute is the code element it immediately precedes, which is typically a type or type member. You can also attach attributes, however, to an assembly. This requires that you explicitly specify the attribute’s target. Here is how you can use the AssemblyFileVersion attribute to attach a version to the assembly: [assembly: AssemblyFileVersion ("1.2.3.4")] From C# 7.3, you can use the field: prefix to apply an attribute to the backing fields of an automatic property. This can be useful in controlling serialization: [field:NonSerialized] public int MyProperty { get; set; } Applying Attributes to Lambda Expressions (C# 10) From C# 10, you can apply attributes to the method, parameters, and return value of a lambda expression: Action<int> a = [Description ("Method")] [return: Description ("Return value")] ([Description ("Parameter")]int x) => Console.Write (x); This is useful when working with frameworks—such as ASP.NET—that rely on you placing attributes on methods that you write. With this feature, you can avoid having to create named methods for simple operations. These attributes are applied to the compiler-generated method to which the delegate points. In Chapter 18, we’ll describe how to reflect over attributes in code. For now, here’s the extra code you need to resolve that indirection: var methodAtt = a.GetMethodInfo().GetCustomAttributes(); var paramAtt = a.GetMethodInfo().GetParameters()[0].GetCustomAttributes(); var returnAtt = a.GetMethodInfo().ReturnParameter.GetCustomAttributes(); To avoid syntactical ambiguity when applying attributes to a parameter on a lambda expression, parentheses are always required. Attributes are not permitted on expression-tree lambdas. 228 | Chapter 4: Advanced C#
- 247. Specifying Multiple Attributes You can specify multiple attributes for a single code element. You can list each attribute either within the same pair of square brackets (separated by a comma) or in separate pairs of square brackets (or a combination of the two). The following three examples are semantically identical: [Serializable, Obsolete, CLSCompliant(false)] public class Bar {...} [Serializable] [Obsolete] [CLSCompliant(false)] public class Bar {...} [Serializable, Obsolete] [CLSCompliant(false)] public class Bar {...} Caller Info Attributes You can tag optional parameters with one of three caller info attributes, which instruct the compiler to feed information obtained from the caller’s source code into the parameter’s default value: • [CallerMemberName] applies the caller’s member name. • • [CallerFilePath] applies the path to the caller’s source code file. • • [CallerLineNumber] applies the line number in the caller’s source code file. • The Foo method in the following program demonstrates all three: using System; using System.Runtime.CompilerServices; class Program { static void Main() => Foo(); static void Foo ( [CallerMemberName] string memberName = null, [CallerFilePath] string filePath = null, [CallerLineNumber] int lineNumber = 0) { Console.WriteLine (memberName); Console.WriteLine (filePath); Console.WriteLine (lineNumber); } } Assuming that our program resides in c:sourcetestProgram.cs, the output would be: Caller Info Attributes | 229 Advanced C#
- 248. Main c:sourcetestProgram.cs 6 As with standard optional parameters, the substitution is done at the calling site. Hence, our Main method is syntactic sugar for this: static void Main() => Foo ("Main", @"c:sourcetestProgram.cs", 6); Caller info attributes are useful for logging—and for implementing patterns such as firing a single change notification event whenever any property on an object changes. In fact, there’s a standard interface for this in the System.ComponentModel namespace, called INotifyPropertyChanged: public interface INotifyPropertyChanged { event PropertyChangedEventHandler PropertyChanged; } public delegate void PropertyChangedEventHandler (object sender, PropertyChangedEventArgs e); public class PropertyChangedEventArgs : EventArgs { public PropertyChangedEventArgs (string propertyName); public virtual string PropertyName { get; } } Notice that PropertyChangedEventArgs requires the name of the property that changed. By applying the [CallerMemberName] attribute, however, we can imple‐ ment this interface and invoke the event without ever specifying property names: public class Foo : INotifyPropertyChanged { public event PropertyChangedEventHandler PropertyChanged = delegate { }; void RaisePropertyChanged ([CallerMemberName] string propertyName = null) => PropertyChanged (this, new PropertyChangedEventArgs (propertyName)); string customerName; public string CustomerName { get => customerName; set { if (value == customerName) return; customerName = value; RaisePropertyChanged(); // The compiler converts the above line to: // RaisePropertyChanged ("CustomerName"); } } } 230 | Chapter 4: Advanced C#
- 249. CallerArgumentExpression (C# 10) A method parameter to which you apply the [CallerArgumentExpression] attribute captures an argument expression from the call site: Print (Math.PI * 2); void Print (double number, [CallerArgumentExpression("number")] string expr = null) => Console.WriteLine (expr); // Output: Math.PI * 2 The compiler feeds in the calling expression’s source code literally, including comments: Print (Math.PI /*(π)*/ * 2); // Output: Math.PI /*(π)*/ * 2 The main application for this feature is when writing validation and assertion libraries. In the following example, an exception is thrown, whose message includes the text “2 + 2 == 5”. This aids in debugging: Assert (2 + 2 == 5); void Assert (bool condition, [CallerArgumentExpression ("condition")] string message = null) { if (!condition) throw new Exception ("Assertion failed: " + message); } Another example is the static ThrowIfNull method on the ArgumentNullException class. This method was introduced in .NET 6 and is defined as follows: public static void ThrowIfNull (object argument, [CallerArgumentExpression("argument")] string paramName = null) { if (argument == null) throw new ArgumentNullException (paramName); } It is used as follows: void Print (string message) { ArgumentNullException.ThrowIfNull (message); ... } You can use [CallerArgumentExpression] multiple times, to capture multiple argument expressions. Caller Info Attributes | 231 Advanced C#
- 250. Dynamic Binding Dynamic binding defers binding—the process of resolving types, members, and operators—from compile time to runtime. Dynamic binding is useful when at compile time you know that a certain function, member, or operation exists, but the compiler does not. This commonly occurs when you are interoperating with dynamic languages (such as IronPython) and COM as well as for scenarios in which you might otherwise use reflection. A dynamic type is declared with the contextual keyword dynamic: dynamic d = GetSomeObject(); d.Quack(); A dynamic type tells the compiler to relax. We expect the runtime type of d to have a Quack method. We just can’t prove it statically. Because d is dynamic, the compiler defers binding Quack to d until runtime. To understand what this means requires distinguishing between static binding and dynamic binding. Static Binding Versus Dynamic Binding The canonical binding example is mapping a name to a specific function when compiling an expression. To compile the following expression, the compiler needs to find the implementation of the method named Quack: d.Quack(); Let’s suppose that the static type of d is Duck: Duck d = ... d.Quack(); In the simplest case, the compiler does the binding by looking for a parameterless method named Quack on Duck. Failing that, the compiler extends its search to meth‐ ods taking optional parameters, methods on base classes of Duck, and extension methods that take Duck as its first parameter. If no match is found, you’ll get a compilation error. Regardless of what method is bound, the bottom line is that the binding is done by the compiler, and the binding utterly depends on statically knowing the types of the operands (in this case, d). This makes it static binding. Now let’s change the static type of d to object: object d = ... d.Quack(); Calling Quack gives us a compilation error, because although the value stored in d can contain a method called Quack, the compiler cannot know it, because the only information it has is the type of the variable, which in this case is object. But let’s now change the static type of d to dynamic: dynamic d = ... d.Quack(); 232 | Chapter 4: Advanced C#
- 251. A dynamic type is like object—it’s equally nondescriptive about a type. The dif‐ ference is that it lets you use it in ways that aren’t known at compile time. A dynamic object binds at runtime based on its runtime type, not its compile-time type. When the compiler sees a dynamically bound expression (which in general is an expression that contains any value of type dynamic), it merely packages up the expression such that the binding can be done later at runtime. At runtime, if a dynamic object implements IDynamicMetaObjectProvider, that interface is used to perform the binding. If not, binding occurs in almost the same way as it would have had the compiler known the dynamic object’s runtime type. These two alternatives are called custom binding and language binding. Custom Binding Custom binding occurs when a dynamic object implements IDynamicMetaObject Provider (IDMOP). Although you can implement IDMOP on types that you write in C#, and that is useful to do, the more common case is that you have acquired an IDMOP object from a dynamic language that is implemented in .NET on the Dynamic Language Runtime (DLR), such as IronPython or IronRuby. Objects from those languages implicitly implement IDMOP as a means by which to directly control the meanings of operations performed on them. We discuss custom binders in greater detail in Chapter 19, but for now, let’s write a simple one to demonstrate the feature: using System; using System.Dynamic; dynamic d = new Duck(); d.Quack(); // Quack method was called d.Waddle(); // Waddle method was called public class Duck : DynamicObject { public override bool TryInvokeMember ( InvokeMemberBinder binder, object[] args, out object result) { Console.WriteLine (binder.Name + " method was called"); result = null; return true; } } The Duck class doesn’t actually have a Quack method. Instead, it uses custom bind‐ ing to intercept and interpret all method calls. Language Binding Language binding occurs when a dynamic object does not implement IDynamic MetaObjectProvider. It is useful when working around imperfectly designed types or inherent limitations in the .NET type system (we explore more scenarios in Dynamic Binding | 233 Advanced C#
- 252. Chapter 19). A typical problem when using numeric types is that they have no common interface. We have seen that we can bind methods dynamically; the same is true for operators: int x = 3, y = 4; Console.WriteLine (Mean (x, y)); dynamic Mean (dynamic x, dynamic y) => (x + y) / 2; The benefit is obvious—you don’t need to duplicate code for each numeric type. However, you lose static type safety, risking runtime exceptions rather than compile-time errors. Dynamic binding circumvents static type safety but not run‐ time type safety. Unlike with reflection (Chapter 18), you can’t circumvent member accessibility rules with dynamic binding. By design, language runtime binding behaves as similarly as possible to static binding, had the runtime types of the dynamic objects been known at compile time. In our previous example, the behavior of our program would be identical if we hardcoded Mean to work with the int type. The most notable exception in parity between static and dynamic binding is for extension methods, which we discuss in “Uncallable Functions” on page 238. Dynamic binding also incurs a performance hit. Because of the DLR’s caching mechanisms, however, repeated calls to the same dynamic expression are optimized—allowing you to effi‐ ciently call dynamic expressions in a loop. This optimization brings the typical overhead for a simple dynamic expression on today’s hardware down to less than 100 ns. RuntimeBinderException If a member fails to bind, a RuntimeBinderException is thrown. You can think of this like a compile-time error at runtime: dynamic d = 5; d.Hello(); // throws RuntimeBinderException The exception is thrown because the int type has no Hello method. Runtime Representation of Dynamic There is a deep equivalence between the dynamic and object types. The runtime treats the following expression as true: typeof (dynamic) == typeof (object) This principle extends to constructed types and array types: typeof (List<dynamic>) == typeof (List<object>) typeof (dynamic[]) == typeof (object[]) 234 | Chapter 4: Advanced C#
- 253. Like an object reference, a dynamic reference can point to an object of any type (except pointer types): dynamic x = "hello"; Console.WriteLine (x.GetType().Name); // String x = 123; // No error (despite same variable) Console.WriteLine (x.GetType().Name); // Int32 Structurally, there is no difference between an object reference and a dynamic reference. A dynamic reference simply enables dynamic operations on the object it points to. You can convert from object to dynamic to perform any dynamic operation you want on an object: object o = new System.Text.StringBuilder(); dynamic d = o; d.Append ("hello"); Console.WriteLine (o); // hello Reflecting on a type exposing (public) dynamic members reveals that those members are represented as annotated objects; for example public class Test { public dynamic Foo; } is equivalent to: public class Test { [System.Runtime.CompilerServices.DynamicAttribute] public object Foo; } This allows consumers of that type to know that Foo should be treated as dynamic while allowing languages that don’t support dynamic binding to fall back to object. Dynamic Conversions The dynamic type has implicit conversions to and from all other types: int i = 7; dynamic d = i; long j = d; // No cast required (implicit conversion) For the conversion to succeed, the runtime type of the dynamic object must be implicitly convertible to the target static type. The preceding example worked because an int is implicitly convertible to a long. The following example throws a RuntimeBinderException because an int is not implicitly convertible to a short: int i = 7; dynamic d = i; short j = d; // throws RuntimeBinderException Dynamic Binding | 235 Advanced C#
- 254. var Versus dynamic The var and dynamic types bear a superficial resemblance, but the difference is deep: • var says, “Let the compiler figure out the type.” • • dynamic says, “Let the runtime figure out the type.” • To illustrate: dynamic x = "hello"; // Static type is dynamic; runtime type is string var y = "hello"; // Static type is string; runtime type is string int i = x; // Runtime error (cannot convert string to int) int j = y; // Compile-time error (cannot convert string to int) The static type of a variable declared with var can be dynamic: dynamic x = "hello"; var y = x; // Static type of y is dynamic int z = y; // Runtime error (cannot convert string to int) Dynamic Expressions Fields, properties, methods, events, constructors, indexers, operators, and conver‐ sions can all be called dynamically. Trying to consume the result of a dynamic expression with a void return type is prohibited—just as with a statically typed expression. The difference is that the error occurs at runtime: dynamic list = new List<int>(); var result = list.Add (5); // RuntimeBinderException thrown Expressions involving dynamic operands are typically themselves dynamic because the effect of absent type information is cascading: dynamic x = 2; var y = x * 3; // Static type of y is dynamic There are a couple of obvious exceptions to this rule. First, casting a dynamic expression to a static type yields a static expression: dynamic x = 2; var y = (int)x; // Static type of y is int Second, constructor invocations always yield static expressions—even when called with dynamic arguments. In this example, x is statically typed to a StringBuilder: dynamic capacity = 10; var x = new System.Text.StringBuilder (capacity); In addition, there are a few edge cases for which an expression containing a dynamic argument is static, including passing an index to an array and delegate creation expressions. 236 | Chapter 4: Advanced C#
- 255. Dynamic Calls Without Dynamic Receivers The canonical use case for dynamic involves a dynamic receiver. This means that a dynamic object is the receiver of a dynamic function call: dynamic x = ...; x.Foo(); // x is the receiver However, you can also call statically known functions with dynamic arguments. Such calls are subject to dynamic overload resolution and can include the following: • Static methods • • Instance constructors • • Instance methods on receivers with a statically known type • In the following example, the particular Foo that gets dynamically bound is depen‐ dent on the runtime type of the dynamic argument: class Program { static void Foo (int x) => Console.WriteLine ("int"); static void Foo (string x) => Console.WriteLine ("string"); static void Main() { dynamic x = 5; dynamic y = "watermelon"; Foo (x); // int Foo (y); // string } } Because a dynamic receiver is not involved, the compiler can statically perform a basic check to see whether the dynamic call will succeed. It checks whether a function with the correct name and number of parameters exists. If no candidate is found, you get a compile-time error: class Program { static void Foo (int x) => Console.WriteLine ("int"); static void Foo (string x) => Console.WriteLine ("string"); static void Main() { dynamic x = 5; Foo (x, x); // Compiler error - wrong number of parameters Fook (x); // Compiler error - no such method name } } Dynamic Binding | 237 Advanced C#
- 256. Static Types in Dynamic Expressions It’s obvious that dynamic types are used in dynamic binding. It’s not so obvious that static types are also used—wherever possible—in dynamic binding. Consider the following: class Program { static void Foo (object x, object y) { Console.WriteLine ("oo"); } static void Foo (object x, string y) { Console.WriteLine ("os"); } static void Foo (string x, object y) { Console.WriteLine ("so"); } static void Foo (string x, string y) { Console.WriteLine ("ss"); } static void Main() { object o = "hello"; dynamic d = "goodbye"; Foo (o, d); // os } } The call to Foo (o,d) is dynamically bound because one of its arguments, d, is dynamic. But because o is statically known, the binding—even though it occurs dynamically—will make use of that. In this case, overload resolution will pick the second implementation of Foo due to the static type of o and the runtime type of d. In other words, the compiler is “as static as it can possibly be.” Uncallable Functions Some functions cannot be called dynamically. You cannot call the following: • Extension methods (via extension method syntax) • • Members of an interface, if you need to cast to that interface to do so • • Base members hidden by a subclass • Understanding why this is so is useful in understanding dynamic binding. Dynamic binding requires two pieces of information: the name of the function to call and the object upon which to call the function. However, in each of the three uncallable scenarios, an additional type is involved, which is known only at compile time. As of this writing, there’s no way to specify these additional types dynamically. When calling extension methods, that additional type is implicit. It’s the static class on which the extension method is defined. The compiler searches for it given the using directives in your source code. This makes extension methods compile-time-only concepts because using directives melt away upon compilation (after they’ve done their job in the binding process in mapping simple names to namespace-qualified names). 238 | Chapter 4: Advanced C#
- 257. When calling members via an interface, you specify that additional type via an implicit or explicit cast. There are two scenarios for which you might want to do this: when calling explicitly implemented interface members and when calling interface members implemented in a type internal to another assembly. We can illustrate the former with the following two types: interface IFoo { void Test(); } class Foo : IFoo { void IFoo.Test() {} } To call the Test method, we must cast to the IFoo interface. This is easy with static typing: IFoo f = new Foo(); // Implicit cast to interface f.Test(); Now consider the situation with dynamic typing: IFoo f = new Foo(); dynamic d = f; d.Test(); // Exception thrown The implicit cast shown in bold tells the compiler to bind subsequent member calls on f to IFoo rather than Foo—in other words, to view that object through the lens of the IFoo interface. However, that lens is lost at runtime, so the DLR cannot complete the binding. The loss is illustrated as follows: Console.WriteLine (f.GetType().Name); // Foo A similar situation arises when calling a hidden base member: you must specify an additional type via either a cast or the base keyword—and that additional type is lost at runtime. Should you need to invoke interface members dynamically, a workaround is to use the Uncapsulator open source library, available on NuGet and GitHub. Uncapsulator was written by the author to address this problem, and leverages custom binding to provide a better dynamic than dynamic: IFoo f = new Foo(); dynamic uf = f.Uncapsulate(); uf.Test(); Uncapsulator also lets you cast to base types and interfaces by name, dynamically call static members, and access nonpublic members of a type. Operator Overloading You can overload operators to provide more natural syntax for custom types. Oper‐ ator overloading is most appropriately used for implementing custom structs that represent fairly primitive data types. For example, a custom numeric type is an excellent candidate for operator overloading. Operator Overloading | 239 Advanced C#
- 258. The following symbolic operators can be overloaded: + (unary) - (unary) ! ˜ ++ -- + - * / % & | ^ << >> == != > < >= <= The following operators are also overloadable: • Implicit and explicit conversions (with the implicit and explicit keywords) • • The true and false operators (not literals) • The following operators are indirectly overloaded: • The compound assignment operators (e.g., +=, /=) are implicitly overridden by • overriding the noncompound operators (e.g., +, /). • The conditional operators && and || are implicitly overridden by overriding • the bitwise operators & and |. Operator Functions You overload an operator by declaring an operator function. An operator function has the following rules: • The name of the function is specified with the operator keyword followed by • an operator symbol. • The operator function must be marked static and public. • • The parameters of the operator function represent the operands. • • The return type of an operator function represents the result of an expression. • • At least one of the operands must be the type in which the operator function is • declared. In the following example, we define a struct called Note representing a musical note and then overload the + operator: public struct Note { int value; public Note (int semitonesFromA) { value = semitonesFromA; } public static Note operator + (Note x, int semitones) { return new Note (x.value + semitones); 240 | Chapter 4: Advanced C#
- 259. } } This overload allows us to add an int to a Note: Note B = new Note (2); Note CSharp = B + 2; Overloading an operator automatically overloads the corresponding compound assignment operator. In our example, because we overrode +, we can use += too: CSharp += 2; Just as with methods and properties, C# allows operator functions comprising a single expression to be written more tersely with expression-bodied syntax: public static Note operator + (Note x, int semitones) => new Note (x.value + semitones); Overloading Equality and Comparison Operators Equality and comparison operators are sometimes overridden when writing structs, and in rare cases when writing classes. Special rules and obligations come with overloading the equality and comparison operators, which we explain in Chapter 6. A summary of these rules is as follows: Pairing The C# compiler enforces operators that are logical pairs to both be defined. These operators are (== !=), (< >), and (<= >=). Equals and GetHashCode In most cases, if you overload (==) and (!=), you must override the Equals and GetHashCode methods defined on object to get meaningful behavior. The C# compiler will give a warning if you do not do this. (See “Equality Comparison” on page 324 for more details.) IComparable and IComparable<T> If you overload (< >) and (<= >=), you should implement IComparable and IComparable<T>. Custom Implicit and Explicit Conversions Implicit and explicit conversions are overloadable operators. These conversions are typically overloaded to make converting between strongly related types (such as numeric types) concise and natural. To convert between weakly related types, the following strategies are more suitable: • Write a constructor that has a parameter of the type to convert from. • • Write ToXXX and (static) FromXXX methods to convert between types. • Operator Overloading | 241 Advanced C#
- 260. As explained in the discussion on types, the rationale behind implicit conversions is that they are guaranteed to succeed and not lose information during the conversion. Conversely, an explicit conversion should be required either when runtime circum‐ stances will determine whether the conversion will succeed or if information might be lost during the conversion. In this example, we define conversions between our musical Note type and a double (which represents the frequency in hertz of that note): ... // Convert to hertz public static implicit operator double (Note x) => 440 * Math.Pow (2, (double) x.value / 12 ); // Convert from hertz (accurate to the nearest semitone) public static explicit operator Note (double x) => new Note ((int) (0.5 + 12 * (Math.Log (x/440) / Math.Log(2) ) )); ... Note n = (Note)554.37; // explicit conversion double x = n; // implicit conversion Following our own guidelines, this example might be better implemented with a ToFrequency method (and a static From Frequency method) instead of implicit and explicit operators. Custom conversions are ignored by the as and is operators: Console.WriteLine (554.37 is Note); // False Note n = 554.37 as Note; // Error Overloading true and false The true and false operators are overloaded in the extremely rare case of types that are Boolean “in spirit” but do not have a conversion to bool. An example is a type that implements three-state logic: by overloading true and false, such a type can work seamlessly with conditional statements and operators—namely, if, do, while, for, &&, ||, and ?:. The System.Data.SqlTypes.SqlBoolean struct provides this functionality: SqlBoolean a = SqlBoolean.Null; if (a) Console.WriteLine ("True"); else if (!a) Console.WriteLine ("False"); else Console.WriteLine ("Null"); OUTPUT: Null The following code is a reimplementation of the parts of SqlBoolean necessary to demonstrate the true and false operators: 242 | Chapter 4: Advanced C#
- 261. public struct SqlBoolean { public static bool operator true (SqlBoolean x) => x.m_value == True.m_value; public static bool operator false (SqlBoolean x) => x.m_value == False.m_value; public static SqlBoolean operator ! (SqlBoolean x) { if (x.m_value == Null.m_value) return Null; if (x.m_value == False.m_value) return True; return False; } public static readonly SqlBoolean Null = new SqlBoolean(0); public static readonly SqlBoolean False = new SqlBoolean(1); public static readonly SqlBoolean True = new SqlBoolean(2); private SqlBoolean (byte value) { m_value = value; } private byte m_value; } Unsafe Code and Pointers C# supports direct memory manipulation via pointers within blocks of code marked unsafe and compiled with the /unsafe compiler option. Pointer types are primarily useful for interoperability with C APIs, but you also can use them for accessing memory outside the managed heap or for performance-critical hotspots. Pointer Basics For every value type or reference type V, there is a corresponding pointer type V*. A pointer instance holds the address of a variable. Pointer types can be (unsafely) cast to any other pointer type. Following are the main pointer operators: Operator Meaning & The address-of operator returns a pointer to the address of a variable. * The dereference operator returns the variable at the address of a pointer. -> The pointer-to-member operator is a syntactic shortcut, in which x->y is equivalent to (*x).y. In keeping with C, adding (or subtracting) an integer offset to a pointer generates another pointer. Subtracting one pointer from another generates a 64-bit integer (on both 64-bit and 32-bit platforms). Unsafe Code By marking a type, type member, or statement block with the unsafe keyword, you’re permitted to use pointer types and perform C style pointer operations on Unsafe Code and Pointers | 243 Advanced C#
- 262. memory within that scope. Here is an example of using pointers to quickly process a bitmap: unsafe void BlueFilter (int[,] bitmap) { int length = bitmap.Length; fixed (int* b = bitmap) { int* p = b; for (int i = 0; i < length; i++) *p++ &= 0xFF; } } Unsafe code can run faster than a corresponding safe implementation. In this case, the code would have required a nested loop with array indexing and bounds check‐ ing. An unsafe C# method can also be faster than calling an external C function given that there is no overhead associated with leaving the managed execution environment. The fixed Statement The fixed statement is required to pin a managed object, such as the bitmap in the previous example. During the execution of a program, many objects are allocated and deallocated from the heap. To avoid unnecessary waste or fragmentation of memory, the garbage collector moves objects around. Pointing to an object is futile if its address could change while referencing it, so the fixed statement tells the garbage collector to “pin” the object and not move it around. This can have an impact on the efficiency of the runtime, so you should use fixed blocks only briefly, and you should avoid heap allocation within the fixed block. Within a fixed statement, you can get a pointer to any value type, an array of value types, or a string. In the case of arrays and strings, the pointer will actually point to the first element, which is a value type. Value types declared inline within reference types require the reference type to be pinned, as follows: Test test = new Test(); unsafe { fixed (int* p = &test.X) // Pins test { *p = 9; } Console.WriteLine (test.X); } class Test { public int X; } We describe the fixed statement further in “Mapping a Struct to Unmanaged Memory” on page 973. 244 | Chapter 4: Advanced C#
- 263. The Pointer-to-Member Operator In addition to the & and * operators, C# also provides the C++ style -> operator, which you can use on structs: Test test = new Test(); unsafe { Test* p = &test; p->X = 9; System.Console.WriteLine (test.X); } struct Test { public int X; } The stackalloc Keyword You can allocate memory in a block on the stack explicitly by using the stackalloc keyword. Because it is allocated on the stack, its lifetime is limited to the execution of the method, just as with any other local variable (whose life hasn’t been extended by virtue of being captured by a lambda expression, iterator block, or asynchronous function). The block can use the [] operator to index into memory: int* a = stackalloc int [10]; for (int i = 0; i < 10; ++i) Console.WriteLine (a[i]); In Chapter 23, we describe how you can use Span<T> to manage stack-allocated memory without using the unsafe keyword: Span<int> a = stackalloc int [10]; for (int i = 0; i < 10; ++i) Console.WriteLine (a[i]); Fixed-Size Buffers The fixed keyword has another use, which is to create fixed-size buffers within structs (this can be useful when calling an unmanaged function; see Chapter 24): new UnsafeClass ("Christian Troy"); unsafe struct UnsafeUnicodeString { public short Length; public fixed byte Buffer[30]; // Allocate block of 30 bytes } unsafe class UnsafeClass { UnsafeUnicodeString uus; public UnsafeClass (string s) { uus.Length = (short)s.Length; Unsafe Code and Pointers | 245 Advanced C#
- 264. fixed (byte* p = uus.Buffer) for (int i = 0; i < s.Length; i++) p[i] = (byte) s[i]; } } Fixed-size buffers are not arrays: if Buffer were an array, it would consist of a reference to an object stored on the (managed) heap, rather than 30 bytes within the struct itself. The fixed keyword is also used in this example to pin the object on the heap that contains the buffer (which will be the instance of UnsafeClass). Hence, fixed means two different things: fixed in size and fixed in place. The two are often used together, in that a fixed-size buffer must be fixed in place to be used. void* A void pointer (void*) makes no assumptions about the type of the underlying data and is useful for functions that deal with raw memory. An implicit conversion exists from any pointer type to void*. A void* cannot be dereferenced, and arithmetic operations cannot be performed on void pointers. Here’s an example: short[] a = { 1, 1, 2, 3, 5, 8, 13, 21, 34, 55 }; unsafe { fixed (short* p = a) { //sizeof returns size of value-type in bytes Zap (p, a.Length * sizeof (short)); } } foreach (short x in a) System.Console.WriteLine (x); // Prints all zeros unsafe void Zap (void* memory, int byteCount) { byte* b = (byte*)memory; for (int i = 0; i < byteCount; i++) *b++ = 0; } Native-Sized Integers The nint and nuint native-sized integer types (introduced in C# 9) are sized to match the address space of the process at runtime (in practice, 32 or 64 bits). Native-sized integers can improve efficiency, overflow safety, and convenience when performing pointer arithmetic. The gain in efficiency arises because when you subtract two pointers in C#, the result is always a 64-bit integer (long), which is inefficient on 32-bit platforms. By first casting the pointers to nint, the result of a subtraction is also nint (which will be 32 bits on a 32-bit platform): 246 | Chapter 4: Advanced C#
- 265. unsafe nint AddressDif (char* x, char* y) => (nint)x - (nint)y; The gain in overflow safety and convenience arises when you need a type to repre‐ sent an offset in memory or a buffer length. This is because the historical alternative to using nint/nuint has been to repurpose System.IntPtr and System.UIntPtr, types whose intended purpose is to wrap operating system handles or address pointers, allowing interop outside an unsafe context. Although they are natively sized, these types have limited support for arithmetic—and the support they have is always unchecked (so overflows fail silently). In contrast, native-sized integers behave much like standard integers, with full support for arithmetic operations and overflow checking: nint x = 123, y = 234; checked { nint sum = x + y, product = x * y; Console.WriteLine (product); } Native-sized integers can be assigned 32-bit integer constants (but not 64-bit integer constants, because these might overflow at runtime). You can use an explicit cast to convert to or from other integral types. At runtime, nint and nuint map to the IntPtr and UIntPtr structs, so you can convert between them without casting (an identity conversion): nint x = 123; IntPtr p = x; nint y = p; For reasons described previously, nint/nuint are not merely shortcuts for IntPtr/ UIntPtr, despite their runtime equivalence. Specifically, the compiler treats a vari‐ able of type nint/nuint as a numeric type, allowing you to perform arithmetic operations not implemented by IntPtr and UIntPtr (and with checked blocks honored). A nint/nuint variable is like an IntPtr/UIntPtr wearing a special hat. This hat is recognized by the compiler to mean “please treat me as a safe numeric type.” This multi-hat behavior is unique to native-sized integers. For example, int acts as a pure synonym for System.Int32, and the two can be freely interchanged. This nonequivalence means that both constructs are useful: • nint/nuint are useful for representing a memory offset or buffer length. • • IntPtr/UIntPtr are useful for wrapping handles and pointers for interop. • Using the types in this manner also correctly signals your intention. Unsafe Code and Pointers | 247 Advanced C#
- 266. A good example of the real-world use of nint and nuint is in the implementation of Buffer.MemoryCopy. You can see this in the .NET source code for Buffer.cs on GitHub or by decompiling the method in ILSpy. A simplified version has also been included in the LINQPad samples for C# 10 in a Nutshell. Function Pointers A function pointer (from C# 9) is like a delegate, but without the indirection of a delegate instance; instead, it points directly to a method. A function pointer can point only to static methods, lacks multicast capability, and requires an unsafe con‐ text (because it bypasses runtime type safety). Its main purpose is to simplify and optimize interop with unmanaged APIs (see “Callbacks from Unmanaged Code” on page 967). A function pointer type is declared as follows (with the return type appearing last): delegate*<int, char, string, void> // (void refers to the return type) This matches a function with this signature: void SomeFunction (int x, char y, string z) The & operator creates a function pointer from a method group. Here’s a complete example: unsafe { delegate*<string, int> functionPointer = &GetLength; int length = functionPointer ("Hello, world"); static int GetLength (string s) => s.Length; } In this example, functionPointer is not an object upon which you can call a method such as Invoke (or with a reference to a Target object). Instead, it’s a variable that points directly to the target method’s address in memory: Console.WriteLine ((IntPtr)functionPointer); Like any other pointer, it’s not subject to runtime type checking. The following treats our function’s return value as a decimal (which, being longer than an int, means that we incorporate some random memory into the output): var pointer2 = (delegate*<string, decimal>) (IntPtr) functionPointer; Console.WriteLine (pointer2 ("Hello, unsafe world")); [SkipLocalsInit] When C# compiles a method, it emits a flag that instructs the runtime to initial‐ ize the method’s local variables to their default values (by zeroing the memory). From C# 9, you can ask the compiler not to emit this flag by applying the 248 | Chapter 4: Advanced C#
- 267. [SkipLocalInit] attribute to a method (in the System.Runtime.CompilerServices namespace): [SkipLocalsInit] void Foo() ... You can also apply this attribute to a type—which is equivalent to applying it to all of the type’s methods—or even an entire module (the container for an assembly): [module: System.Runtime.CompilerServices.SkipLocalsInit] In normal safe scenarios, [SkipLocalsInit] has little effect on functionality or performance, because C#’s definite assignment policy requires that you explicitly assign local variables before they can be read. This means that the JIT optimizer is likely to emit the same machine code, whether or not the attribute is applied. In an unsafe context, however, use of [SkipLocalsInit] can usefully save the CLR from the overhead of initializing value-typed local variables, creating a small performance gain with methods that make extensive use of the stack (through a large stackalloc). The following example prints uninitialized memory when [SkipLocalsInit] is applied (instead of all zeros): [SkipLocalsInit] unsafe void Foo() { int local; int* ptr = &local; Console.WriteLine (*ptr); int* a = stackalloc int [100]; for (int i = 0; i < 100; ++i) Console.WriteLine (a [i]); } Interestingly, you can achieve the same result in a “safe” context through the use of Span<T>: [SkipLocalsInit] void Foo() { Span<int> a = stackalloc int [100]; for (int i = 0; i < 100; ++i) Console.WriteLine (a [i]); } Consequently, use of [SkipLocalsInit] requires that you compile your assembly with the unsafe option—even if none of your methods are marked as unsafe. Preprocessor Directives Preprocessor directives supply the compiler with additional information about regions of code. The most common preprocessor directives are the conditional directives, which provide a way to include or exclude regions of code from compilation: Preprocessor Directives | 249 Advanced C#
- 268. #define DEBUG class MyClass { int x; void Foo() { #if DEBUG Console.WriteLine ("Testing: x = {0}", x); #endif } ... } In this class, the statement in Foo is compiled as conditionally dependent upon the presence of the DEBUG symbol. If we remove the DEBUG symbol, the statement is not compiled. You can define preprocessor symbols within a source file (as we have done) or at a project level in the .csproj file: <PropertyGroup> <DefineConstants>DEBUG;ANOTHERSYMBOL</DefineConstants> </PropertyGroup> With the #if and #elif directives, you can use the ||, &&, and ! operators to perform or, and, and not operations on multiple symbols. The following directive instructs the compiler to include the code that follows if the TESTMODE symbol is defined and the DEBUG symbol is not defined: #if TESTMODE && !DEBUG ... Keep in mind, however, that you’re not building an ordinary C# expression, and the symbols upon which you operate have absolutely no connection to variables—static or otherwise. The #error and #warning symbols prevent accidental misuse of conditional direc‐ tives by making the compiler generate a warning or error given an undesirable set of compilation symbols. Table 4-1 lists the preprocessor directives. Table 4-1. Preprocessor directives Preprocessor directive Action #define symbol Defines symbol #undef symbol Undefines symbol #if symbol [operator symbol2]... symbol to test operators are ==, !=, &&, and ||, followed by #else, #elif, and #endif #else Executes code to subsequent #endif #elif symbol [operator symbol2] Combines #else branch and #if test 250 | Chapter 4: Advanced C#
- 269. Preprocessor directive Action #endif Ends conditional directives #warning text text of the warning to appear in compiler output #error text text of the error to appear in compiler output #error version Reports the compiler version and exits #pragma warning [disable | restore] Disables/restores compiler warning(s) #line [ number ["file"] | hidden] number specifies the line in source code (a column can also be specified from C# 10); file is the filename to appear in computer output; hidden instructs debuggers to skip over code from this point until the next #line directive #region name Marks the beginning of an outline #endregion Ends an outline region #nullable option See “Nullable reference types” on page 16 Conditional Attributes An attribute decorated with the Conditional attribute will be compiled only if a given preprocessor symbol is present: // file1.cs #define DEBUG using System; using System.Diagnostics; [Conditional("DEBUG")] public class TestAttribute : Attribute {} // file2.cs #define DEBUG [Test] class Foo { [Test] string s; } The compiler will incorporate the [Test] attributes only if the DEBUG symbol is in scope for file2.cs. Pragma Warning The compiler generates a warning when it spots something in your code that seems unintentional. Unlike errors, warnings don’t ordinarily prevent your application from compiling. Compiler warnings can be extremely valuable in spotting bugs. Their usefulness, however, is undermined when you get false warnings. In a large application, Preprocessor Directives | 251 Advanced C#
- 270. maintaining a good signal-to-noise ratio is essential if the “real” warnings are to be noticed. To this effect, the compiler allows you to selectively suppress warnings by using the #pragma warning directive. In this example, we instruct the compiler not to warn us about the field Message not being used: public class Foo { static void Main() { } #pragma warning disable 414 static string Message = "Hello"; #pragma warning restore 414 } Omitting the number in the #pragma warning directive disables or restores all warning codes. If you are thorough in applying this directive, you can compile with the /warnaser ror switch—this instructs the compiler to treat any residual warnings as errors. XML Documentation A documentation comment is a piece of embedded XML that documents a type or member. A documentation comment comes immediately before a type or member declaration and starts with three slashes: /// <summary>Cancels a running query.</summary> public void Cancel() { ... } Multiline comments can be done like this: /// <summary> /// Cancels a running query /// </summary> public void Cancel() { ... } Or like this (notice the extra star at the start): /** <summary> Cancels a running query. </summary> */ public void Cancel() { ... } If you add the following option to your .csproj file <PropertyGroup> <DocumentationFile>SomeFile.xml</DocumentationFile> </PropertyGroup> the compiler extracts and collates documentation comments into the specified XML file. This has two main uses: 252 | Chapter 4: Advanced C#
- 271. • If placed in the same folder as the compiled assembly, tools such as Visual • Studio and LINQPad automatically read the XML file and use the information to provide IntelliSense member listings to consumers of the assembly of the same name. • Third-party tools (such as Sandcastle and NDoc) can transform the XML file • into an HTML help file. Standard XML Documentation Tags Here are the standard XML tags that Visual Studio and documentation generators recognize: <summary> <summary>...</summary> Indicates the tool tip that IntelliSense should display for the type or member, typically a single phrase or sentence. <remarks> <remarks>...</remarks> Additional text that describes the type or member. Documentation generators pick this up and merge it into the bulk of a type’s or member’s description. <param> <param name="name">...</param> Explains a parameter on a method. <returns> <returns>...</returns> Explains the return value for a method. <exception> <exception [cref="type"]>...</exception> Lists an exception that a method can throw (cref refers to the exception type). <example> <example>...</example> Denotes an example (used by documentation generators). This usually contains both description text and source code (source code is typically within a <c> or <code> tag). <c> <c>...</c> Indicates an inline code snippet. This tag is usually used within an <example> block. XML Documentation | 253 Advanced C#
- 272. <code> <code>...</code> Indicates a multiline code sample. This tag is usually used within an <example> block. <see> <see cref="member">...</see> Inserts an inline cross-reference to another type or member. HTML documen‐ tation generators typically convert this to a hyperlink. The compiler emits a warning if the type or member name is invalid. To refer to generic types, use curly braces; for example, cref="Foo{T,U}". <seealso> <seealso cref="member">...</seealso> Cross-references another type or member. Documentation generators typically write this into a separate “See Also” section at the bottom of the page. <paramref> <paramref name="name"/> References a parameter from within a <summary> or <remarks> tag. <list> <list type=[ bullet | number | table ]> <listheader> <term>...</term> <description>...</description> </listheader> <item> <term>...</term> <description>...</description> </item> </list> Instructs documentation generators to emit a bulleted, numbered, or table- style list. <para> <para>...</para> Instructs documentation generators to format the contents into a separate paragraph. <include> <include file='filename' path='tagpath[@name="id"]'>...</include> Merges an external XML file that contains documentation. The path attribute denotes an XPath query to a specific element in that file. 254 | Chapter 4: Advanced C#
- 273. User-Defined Tags Little is special about the predefined XML tags recognized by the C# compiler, and you are free to define your own. The only special processing done by the compiler is on the <param> tag (in which it verifies the parameter name and that all the parameters on the method are documented) and the cref attribute (in which it verifies that the attribute refers to a real type or member and expands it to a fully qualified type or member ID). You can also use the cref attribute in your own tags; it is verified and expanded just as it is in the predefined <exception>, <permission>, <see>, and <seealso> tags. Type or Member Cross-References Type names and type or member cross-references are translated into IDs that uniquely define the type or member. These names are composed of a prefix that defines what the ID represents and a signature of the type or member. Following are the member prefixes: XML type prefix ID prefixes applied to… N Namespace T Type (class, struct, enum, interface, delegate) F Field P Property (includes indexers) M Method (includes special methods) E Event ! Error The rules describing how the signatures are generated are well documented, although fairly complex. Here is an example of a type and the IDs that are generated: // Namespaces do not have independent signatures namespace NS { /// T:NS.MyClass class MyClass { /// F:NS.MyClass.aField string aField; /// P:NS.MyClass.aProperty short aProperty {get {...} set {...}} /// T:NS.MyClass.NestedType class NestedType {...}; /// M:NS.MyClass.X() XML Documentation | 255 Advanced C#
- 274. void X() {...} /// M:NS.MyClass.Y(System.Int32,System.Double@,System.Decimal@) void Y(int p1, ref double p2, out decimal p3) {...} /// M:NS.MyClass.Z(System.Char[ ],System.Single[0:,0:]) void Z(char[ ] p1, float[,] p2) {...} /// M:NS.MyClass.op_Addition(NS.MyClass,NS.MyClass) public static MyClass operator+(MyClass c1, MyClass c2) {...} /// M:NS.MyClass.op_Implicit(NS.MyClass)˜System.Int32 public static implicit operator int(MyClass c) {...} /// M:NS.MyClass.#ctor MyClass() {...} /// M:NS.MyClass.Finalize ˜MyClass() {...} /// M:NS.MyClass.#cctor static MyClass() {...} } } 256 | Chapter 4: Advanced C#
- 275. 5 .NET Overview Almost all of the capabilities of the .NET 6 runtime are exposed via a vast set of managed types. These types are organized into hierarchical namespaces and packaged into a set of assemblies. Some of the .NET types are used directly by the CLR and are essential for the managed hosting environment. These types reside in an assembly called System.Pri‐ vate.CoreLib.dll and include C#’s built-in types as well as the basic collection classes, and types for stream processing, serialization, reflection, threading, and native interoperability. System.Private.CoreLib.dll replaces .NET Framework’s mscor‐ lib.dll. Many places in the official documentation still refer to mscorlib. At a level above this are additional types that “flesh out” the CLR-level functional‐ ity, providing features such as XML, JSON, networking, and Language-Integrated Query. These constitute the Base Class Library (BCL). Sitting above this are applica‐ tion layers, which provide APIs for developing particular kinds of applications such as web or rich client. In this chapter, we provide the following: • An overview of the BCL (which we cover in the rest of the book) • • A high-level summary of the application layers • 257
- 276. What’s New in .NET 6 In .NET 6, the Base Class Library has numerous new features. In particular: • New DateOnly/TimeOnly structs cleanly capture a date or time for scenarios • such as recording a birthday or alarm time (see “DateOnly and TimeOnly” on page 292). • A new BitOperations static class provides access to low-level base-2 numeric • operations (see “BitOperations” on page 320). • The following new LINQ methods have been added: Chunk, DistinctBy, • UnionBy, IntersectBy, ExceptBy, MinBy, and MaxBy (see Chapter 9). Take also now accepts a Range variable. • The new JsonNode API provides a fluent writable DOM whose types don’t • require disposal (see “JsonNode” on page 551. Utf8JsonWriter also now has a WriteRawValue method. • The new RandomAccess class provides performant thread-safe file I/O • operations. • The new NullabilityInfoContext class in System.Reflection queries nullabil‐ • ity annotations (see “NullabilityContextInfo” on page 794). • The RandomNumberGenerator class in System.Security.Cryptography now has • a static GetBytes(int) method that returns an array of random bytes in one operation. There are also new methods to simplify encryption and decryption (see “Encrypting in Memory” on page 856). • A new ForEachAsync method on the Parallel class limits asynchronous con‐ • currency (see “Parallel.ForEachAsync” on page 882). Task also now has a WaitAsync method, which applies a timeout to any asynchronous operation, and there’s a new timer designed to work with await (see “PeriodicTimer” on page 902). • The new NativeMemory class provides a lightweight wrapper around low-level • memory allocation operations such as malloc. .NET 6 also features numerous performance improvements to the runtime and enhanced support for Windows ARM64 and Apple M1/M2 processors. In the application layers, the biggest change is in the introduction of MAUI (Multi- platform App UI, early 2022), which replaces Xamarin for cross-platform mobile development. MAUI also supports cross-platform desktop application development for macOS and Windows, and targets a unified .NET 6 CLR/BCL. UWP also has a successor, Windows App SDK (with WinUI 3 as its presentation layer), and a new technology has emerged called Blazor Desktop for writing HTML-based desktop and mobile applications. 258 | Chapter 5: .NET Overview
- 277. .NET Standard The wealth of public libraries that are available on NuGet wouldn’t be as valuable if they supported only .NET 6. When writing a library, you’ll often want to support a variety of platforms and runtime versions. To achieve that goal without creating a separate build for each runtime, you must target the lowest common denominator. This is relatively easy if you wish to support only .NET 6’s direct predecessors: for example, if you target .NET Core 3.0, your library will run on .NET Core 3.0, .NET Core 3.1, and .NET 5+. The situation becomes messier if you also want to support .NET Framework or Xamarin. This is because each of these runtimes has a CLR and BCL with overlap‐ ping features—no one runtime is a pure subset of the others. .NET Standard solves this problem by defining artificial subsets that work across an entire range of legacy runtimes. By targeting .NET Standard, you can easily write libraries with extensive reach. .NET Standard is not a runtime; it’s merely a specification describing a minimum baseline of functionality (types and members) that guarantees compatibility with a certain set of runtimes. The concept is similar to C# interfaces: .NET Stan‐ dard is like an interface that concrete types (runtimes) can implement. .NET Standard 2.0 The most useful version is .NET Standard 2.0. A library that targets .NET Standard 2.0 instead of a specific runtime will run without modification on most modern and legacy runtimes that are still in use today, including: • .NET Core 2.0+ (including .NET 5 and .NET 6) • • UWP 10.0.16299+ • • Mono 5.4+ (the CLR/BCL used by older versions of Xamarin) • • .NET Framework 4.6.1+ • To target .NET Standard 2.0, add the following to your .csproj file: <PropertyGroup> <TargetFramework>netstandard2.0</TargetFramework> <PropertyGroup> Most of the APIs described in this book are supported by .NET Standard 2.0. .NET Standard | 259 .NET Overview
- 278. .NET Standard 2.1 NET Standard 2.1 is a superset of .NET Standard 2.0 that supports (only) the following platforms: • .NET Core 3+ • • Mono 6.4+ • .NET Standard 2.1 is not supported by any version of .NET Framework (and not even by UWP), making it much less useful than .NET Standard 2.0. The following APIs, in particular, are available in .NET Standard 2.1 (but not .NET Standard 2.0): • Span<T> (Chapter 23) • • Reflection.Emit (Chapter 18) • • ValueTask<T> (Chapter 14) • Older .NET Standards There are also older .NET Standards, most notably 1.1, 1.2, 1.3, and 1.6. A higher- numbered standard is always a strict superset of a lower-numbered standard. For instance, if you write a library that targets .NET Standard 1.6, you will support not only recent versions of the major runtimes but also .NET Core 1.0. And if you target .NET Standard 1.3, you support everything we’ve already mentioned plus .NET Framework 4.6.0. The table that follows elaborates: If you target... You also support... Standard 1.6 .NET Core 1.0 Standard 1.3 Above plus .NET 4.6.0 Standard 1.2 Above plus .NET 4.5.1, Windows Phone 8.1, WinRT for Windows 8.1 Standard 1.1 Above plus .NET 4.5.0, Windows Phone 8.0, WinRT for Windows 8.0 The 1.x standards lack thousands of APIs that are present in 2.0, including much of what we describe in this book. This can make targeting a 1.x standard significantly more challenging, especially if you need to integrate existing code or libraries. You can also think of .NET Standard as a lowest common denominator. In the case of .NET Standard 2.0, the runtimes that implement it have a similar BCL, so the lowest common denominator is big and useful. However, if you also want compati‐ bility with .NET Core 1.0 (with its significantly cut-down BCL), the lowest common denominator—.NET Standard 1.x—becomes much smaller and less useful. 260 | Chapter 5: .NET Overview
- 279. .NET Framework and .NET 6 Compatibility Because .NET Framework has existed for so long, it’s not uncommon to encounter libraries that are available only for .NET Framework (with no .NET Standard, .NET Core, or .NET 6 equivalent). To help mitigate this situation, .NET 5+ and .NET Core projects are permitted to reference .NET Framework assemblies, with the following provisos: • An exception is thrown should the .NET Framework assembly call an API • that’s unsupported. • Nontrivial dependencies might (and often do) fail to resolve. • In practice, it’s most likely to work in simple cases, such as an assembly that wraps an unmanaged DLL. Runtime and C# Language Versions By default, your project’s runtime version determines which C# language version is used: • For .NET 6, it’s C# 10. • • For .NET 5, it’s C# 9. • • For .NET Core 3.x, Xamarin, and .NET Standard 2.1, it’s C# 8. • • For .NET Core 2.x, .NET Framework, and .NET Standard 2.0 and below, it’s • C# 7.3. This is because later versions of C# rely on types that are available only in later runtimes. You can override the language version in your project file with the <LangVersion> element. Using an older runtime (such as .NET 5) with a later language version (such as C# 10) means that the language features that rely on newer .NET types will not work (although in some cases, you can define those types yourself). Reference Assemblies When you target .NET Standard, your project implicitly references an assembly called netstandard.dll, which contains all of the allowable types and members for your chosen version of .NET Standard. This is called a reference assembly because it exists only for the benefit of the compiler and contains no compiled code. At runtime, the “real” assemblies are identified through assembly redirection attributes (the choice of assemblies will depend on which runtime and platform the assembly eventually runs on). Interestingly, a similar thing happens when you target .NET 6. Your project implic‐ itly references a set of reference assemblies whose types mirror what’s in the Runtime and C# Language Versions | 261 .NET Overview
- 280. runtime assemblies for the chosen .NET version. This helps with versioning and cross-platform compatibility, and also allows you to target a different .NET version than what is installed on your machine. The CLR and BCL System Types The most fundamental types live directly in the System namespace. These include C#’s built-in types; the Exception base class; the Enum, Array, and Delegate base classes; and Nullable, Type, DateTime, TimeSpan, and Guid. The System name‐ space also includes types for performing mathematical functions (Math), generating random numbers (Random), and converting between various types (Convert and BitConverter). Chapter 6 describes these types as well as the interfaces that define standard pro‐ tocols used across .NET for such tasks as formatting (IFormattable) and order comparison (IComparable). The System namespace also defines the IDisposable interface and the GC class for interacting with the garbage collector, which we cover in Chapter 12. Text Processing The System.Text namespace contains the StringBuilder class (the editable or mutable cousin of string) and the types for working with text encodings, such as UTF-8 (Encoding and its subtypes). We cover this in Chapter 6. The System.Text.RegularExpressions namespace contains types that perform advanced pattern-based search-and-replace operations; we describe these in Chapter 25. Collections .NET offers a variety of classes for managing collections of items. These include both list- and dictionary-based structures; they work in conjunction with a set of standard interfaces that unify their common characteristics. All collection types are defined in the following namespaces, covered in Chapter 7: System.Collections // Nongeneric collections System.Collections.Generic // Generic collections System.Collections.Specialized // Strongly typed collections System.Collections.ObjectModel // Bases for your own collections System.Collections.Concurrent // Thread-safe collections (Chapter 22) Querying Language-Integrated Query (LINQ) allows you to perform type-safe queries over local and remote collections (e.g., SQL Server tables) and is described in Chapters 8 262 | Chapter 5: .NET Overview
- 281. through 10. A big advantage of LINQ is that it presents a consistent querying API across a variety of domains. The essential types reside in the following namespaces: System.Linq // LINQ to Objects and PLINQ System.Linq.Expressions // For building expressions manually System.Xml.Linq // LINQ to XML XML and JSON XML and JSON are widely supported in .NET. Chapter 10 focuses entirely on LINQ to XML—a lightweight XML Document Object Model (DOM) that can be constructed and queried through LINQ. Chapter 11 covers the performant low-level XML reader/writer classes, XML schemas and stylesheets, and types for working with JSON: System.Xml // XmlReader, XmlWriter System.Xml.Linq // The LINQ to XML DOM System.Xml.Schema // Support for XSD System.Xml.Serialization // Declarative XML serialization for .NET types System.Xml.XPath // XPath query language System.Xml.Xsl // Stylesheet support System.Text.Json // JSON reader/writer and DOM In the online supplement at http://www.albahari.com/nutshell, we cover the JSON serializer. Diagnostics In Chapter 13, we cover logging and assertion and describe how to interact with other processes, write to the Windows event log, and handle performance monitor‐ ing. The types for this are defined in and under System.Diagnostics. Concurrency and Asynchrony Many modern applications need to deal with more than one thing happening at a time. Since C# 5.0, this has become easier through asynchronous functions and high-level constructs such as tasks and task combinators. Chapter 14 explains all of this in detail, after starting with the basics of multithreading. Types for work‐ ing with threads and asynchronous operations are in the System.Threading and System.Threading.Tasks namespaces. Streams and Input/Output .NET provides a stream-based model for low-level input/output (I/O). Streams are typically used to read and write directly to files and network connections, and can be chained or wrapped in decorator streams to add compression or encryption functionality. Chapter 15 describes the stream architecture as well as the specific support for working with files and directories, compression, pipes, and memory- mapped files. The Stream and I/O types are defined in and under the System.IO The CLR and BCL | 263 .NET Overview
- 282. namespace, and the Windows Runtime (WinRT) types for file I/O are in and under Windows.Storage. Networking You can directly access most standard network protocols such as HTTP, TCP/IP, and SMTP via the types in System.Net. In Chapter 16, we demonstrate how to communicate using each of these protocols, starting with simple tasks such as downloading from a web page and finishing with using TCP/IP directly to retrieve POP3 email. Here are the namespaces we cover: System.Net System.Net.Http // HttpClient System.Net.Mail // For sending mail via SMTP System.Net.Sockets // TCP, UDP, and IP Assemblies, Reflection, and Attributes The assemblies into which C# programs compile comprise executable instructions (stored as IL) and metadata, which describes the program’s types, members, and attributes. Through reflection, you can inspect this metadata at runtime and do such things as dynamically invoke methods. With Reflection.Emit, you can construct new code on the fly. In Chapter 17, we describe the makeup of assemblies and how to dynamically load and isolate them. In Chapter 18, we cover reflection and attributes—describing how to inspect metadata, dynamically invoke functions, write custom attributes, emit new types, and parse raw IL. The types for using reflection and working with assemblies reside in the following namespaces: System System.Reflection System.Reflection.Emit Dynamic Programming In Chapter 19, we look at some of the patterns for dynamic programming and uti‐ lizing the Dynamic Language Runtime (DLR). We describe how to implement the Visitor pattern, write custom dynamic objects, and interoperate with IronPython. The types for dynamic programming are in System.Dynamic. Cryptography .NET provides extensive support for popular hashing and encryption protocols. In Chapter 20, we cover hashing, symmetric and public-key encryption, and the Windows Data Protection API. The types for this are defined in: System.Security System.Security.Cryptography 264 | Chapter 5: .NET Overview
- 283. Advanced Threading C#’s asynchronous functions make concurrent programming significantly easier because they lessen the need for lower-level techniques. However, there are still times when you need signaling constructs, thread-local storage, reader/writer locks, and so on. Chapter 21 explains this in depth. Threading types are in the System.Threading namespace. Parallel Programming In Chapter 22, we cover in detail the libraries and types for leveraging multicore processors, including APIs for task parallelism, imperative data parallelism, and functional parallelism (PLINQ). Span<T> and Memory<T> To help with micro-optimizing performance hotspots, the CLR provides a number of types to help you program in such a way as to reduce the load on the memory manager. Two of the key types are Span<T> and Memory<T>, which we describe in Chapter 23. Native and COM Interoperability You can interoperate with both native and Component Object Model (COM) code. Native interoperability allows you to call functions in unmanaged DLLs, register callbacks, map data structures, and interoperate with native data types. COM inter‐ operability allows you to call COM types (on Windows machines) and expose .NET types to COM. The types that support these functions are in System.Runtime.Inter opServices, and we cover them in Chapter 24. Regular Expressions In Chapter 25, we cover how you can use regular expressions to match character patterns in strings. Serialization .NET provides several systems for saving and restoring objects to a binary or text representation. Such systems can be used for communication as well as for saving and restoring objects to a file. In the online supplement at http://www.alba‐ hari.com/nutshell, we cover all four serialization engines: the binary serializer, the (newly updated) JSON serializer, the XML serializer, and the data contract serializer. The CLR and BCL | 265 .NET Overview
- 284. The Roslyn Compiler The C# compiler itself is written in C#—the project is called “Roslyn,” and the libraries are available as NuGet packages. With these libraries, you can utilize the compiler’s functionality in many ways besides compiling source code to an assem‐ bly, such as writing code analysis and refactoring tools. We cover Roslyn in the online supplement, at http://www.albahari.com/nutshell. Application Layers User interface (UI)–based applications can be divided into two categories: thin client, which amounts to a website, and rich client, which is a program the end user must download and install on a computer or mobile device. For writing thin-client applications in C#, there’s ASP.NET Core, which runs on Windows, Linux, and macOS. ASP.NET Core is also designed for writing web APIs. For rich-client applications, there is a choice of APIs: • The Windows Desktop layer includes the WPF and Windows Forms APIs, and • runs on Windows 7/8/10/11 desktop. • UWP lets you write Windows Store apps that run on Windows 10+ desktop • and devices. • WinUI 3 (Windows App SDK) is a recent successor to UWP that runs on • Windows 10+ desktop. • MAUI (formerly Xamarin) runs on iOS and Android mobile devices. MAUI • also allows for cross-platform desktop applications that target macOS and Windows. There are also third-party libraries, such as Avalonia, which offer cross-platform UI support. ASP.NET Core ASP.NET Core is a lightweight modular successor to ASP.NET and is suitable for creating web sites, REST-based web APIs, and microservices. It can also run in con‐ junction with two popular single-page-application frameworks: React and Angular. ASP.NET supports the popular Model-View-Controller (MVC) pattern, as well as a newer technology called Blazor, where client-side code is written in C# instead of JavaScript. ASP.NET Core runs on Windows, Linux, and macOS and can self-host in a custom process. Unlike its .NET Framework predecessor (ASP.NET), ASP.NET Core is not dependent on System.Web and the historical baggage of web forms. 266 | Chapter 5: .NET Overview
- 285. As with any thin-client architecture, ASP.NET Core offers the following general advantages over rich clients: • There is zero deployment at the client end. • • The client can run on any platform that supports a web browser. • • Updates are easily deployed. • Windows Desktop The Windows Desktop application layer offers a choice of two UI APIs for writing rich-client applications: WPF and Windows Forms. Both APIs run on Windows Desktop/Server 7 through 11. WPF WPF was introduced in 2006 and has been enhanced ever since. Unlike its prede‐ cessor, Windows Forms, WPF explicitly renders controls using DirectX, with the following benefits: • It supports sophisticated graphics, such as arbitrary transformations, 3D ren‐ • dering, multimedia, and true transparency. Skinning is supported through styles and templates. • Its primary measurement unit is not pixel based, so applications display cor‐ • rectly at any DPI setting. • It has extensive and flexible layout support, which means that you can localize • an application without danger of elements overlapping. • Its use of DirectX makes rendering fast and able to take advantage of graphics • hardware acceleration. • It offers reliable data binding. • • UIs can be described declaratively in XAML files that can be maintained • independent of the “code-behind” files—this helps to separate appearance from functionality. WPF takes some time to learn due to its size and complexity. The types for writing WPF applications are in the System.Windows namespace and all subnamespaces except for System.Windows.Forms. Windows Forms Windows Forms is a rich-client API that shipped with the first version of .NET Framework in 2000. Compared to WPF, Windows Forms is a relatively simple tech‐ nology that provides most of the features you need for writing a typical Windows application. It also has significant relevancy in maintaining legacy applications. But Application Layers | 267 .NET Overview
- 286. compared to WPF, it has numerous drawbacks, most of which stem from it being a wrapper over GDI+ and the Win32 control library: • Although it provides mechanisms for DPI-awareness, it’s still too easy to • write applications that break on clients whose DPI settings differ from the developer’s. • The API for drawing nonstandard controls is GDI+, which, although reasona‐ • bly flexible, is slow in rendering large areas (and without double buffering, might flicker). • Controls lack true transparency. • • Most controls are noncompositional. For instance, you can’t put an image • control inside a tab control header. Customizing list views, combo boxes, and tab controls in a way that would be trivial with WPF is time consuming and painful in Windows Forms. • Dynamic layout is difficult to correctly implement reliably. • The last point is an excellent reason to favor WPF over Windows Forms—even if you’re writing a business application that needs just a UI and not a “user experience.” The layout elements in WPF, such as Grid, make it easy to assemble labels and text boxes such that they always align—even after language-changing localization—without messy logic and without any flickering. Further, you don’t need to bow to the lowest common denominator in screen resolution—WPF layout elements have been designed from the outset to adapt properly to resizing. On the positive side, Windows Forms is relatively simple to learn and still has a good number of third-party controls. The Windows Forms types are in the System.Windows.Forms (in System.Win‐ dows.Forms.dll) and System.Drawing (in System.Drawing.dll) namespaces. The lat‐ ter also contains the GDI+ types for drawing custom controls. UWP and WinUI 3 UWP is a rich-client API for writing touch-first UIs that target Windows 10+ desktop and devices. The word “Universal” refers to its ability to run on a range of Windows 10 devices, including Xbox, Surface Hub, and HoloLens. However, it’s not compatible with earlier versions of Windows, including Windows 7 and Windows 8/8.1. The UWP API uses XAML and is somewhat similar to WPF. Here are its key differences: • The primary mode of distribution for UWP apps is the Windows Store. • • UWP apps run in a sandbox to lessen the threat of malware, which means that • they cannot perform tasks such as reading or writing arbitrary files, and they cannot run with administrative elevation. 268 | Chapter 5: .NET Overview
- 287. • UWP relies on WinRT types that are part of the operating system (Windows), • not the managed runtime. This means that when writing apps, you must nomi‐ nate a Windows version range (such as Windows 10 build 17763 to Windows 10 build 18362). This means that you either need to target an old API or require that your customers install the latest Windows update. Because of the limitations created by these differences, UWP never succeeded in matching the popularity of WPF and Windows Forms. To address this, Microsoft has introduced a new technology to supersede UWP called Windows App SDK (with a UI layer called WinUI 3). The Windows App SDK transfers the WinRT APIs from the operating system to the runtime, thereby exposing a fully managed interface and removing the necessity to target a specific operating system version range. It also does the following: • Integrates better with the Windows Desktop APIs (Windows Forms and WPF) • • Allows you to write applications that run outside the Windows Store sandbox • • Runs atop the latest .NET (instead of being tied to .NET Core 2.2, as is the case • with UWP) The Windows App SDK, however, does not support Xbox or HoloLens at the time of writing. MAUI MAUI (early 2022) is the new incarnation of Xamarin, and lets you develop mobile apps in C# that target iOS and Android (as well as cross-platform desktop apps that target macOS and Windows). The CLR/BCL that runs on iOS and Android is called Mono (a derivation of the open-source Mono runtime). Historically, Mono hasn’t been fully compatible with .NET, and libraries that ran on both Mono and .NET would target .NET Stan‐ dard. With MAUI, Mono’s public interface has merged with .NET 6, making Mono, in effect, an implementation of .NET 6. Features new to MAUI include a unified project interface, hot reloading, support for Blazor Desktop and hybrid apps, and improved performance and startup times. See https://github.com/dotnet/maui for more information. Application Layers | 269 .NET Overview
- 289. 6 .NET Fundamentals Many of the core facilities that you need when programming are provided not by the C# language but by types in the .NET BCL. In this chapter, we cover types that help with fundamental programming tasks, such as virtual equality comparison, order comparison, and type conversion. We also cover the basic .NET types, such as String, DateTime, and Enum. The types in this section reside in the System namespace, with the following exceptions: • StringBuilder is defined in System.Text, as are the types for text encodings. • • CultureInfo and associated types are defined in System.Globalization. • • XmlConvert is defined in System.Xml. • String and Text Handling Char A C# char represents a single Unicode character and aliases the System.Char struct. In Chapter 2, we described how to express char literals: char c = 'A'; char newLine = 'n'; System.Char defines a range of static methods for working with characters, such as ToUpper, ToLower, and IsWhiteSpace. You can call these through either the System.Char type or its char alias: Console.WriteLine (System.Char.ToUpper ('c')); // C Console.WriteLine (char.IsWhiteSpace ('t')); // True ToUpper and ToLower honor the end user’s locale, which can lead to subtle bugs. The following expression evaluates to false in Turkey: 271
- 290. char.ToUpper ('i') == 'I' The reason is because in Turkey, char.ToUpper ('i') is 'İ' (notice the dot on top!). To avoid this problem, System.Char (and System.String) also provides culture-invariant versions of ToUpper and ToLower ending with the word Invariant. These always apply English culture rules: Console.WriteLine (char.ToUpperInvariant ('i')); // I This is a shortcut for: Console.WriteLine (char.ToUpper ('i', CultureInfo.InvariantCulture)) For more on locales and culture, see “Formatting and Parsing” on page 297. Most of char’s remaining static methods are related to categorizing characters. Table 6-1 lists these. Table 6-1. Static methods for categorizing characters Static method Characters included Unicode categories included IsLetter A–Z, a–z, and letters of other alphabets UpperCaseLetter LowerCaseLetter TitleCaseLetter ModifierLetter OtherLetter IsUpper Uppercase letters UpperCaseLetter IsLower Lowercase letters LowerCaseLetter IsDigit 0–9 plus digits of other alphabets DecimalDigitNumber IsLetterOrDigit Letters plus digits (IsLetter, IsDigit) IsNumber All digits plus Unicode fractions and Roman numeral symbols DecimalDigitNumber LetterNumber OtherNumber IsSeparator Space plus all Unicode separator characters LineSeparator ParagraphSeparator IsWhiteSpace All separators plus n, r, t, f, and v LineSeparator ParagraphSeparator IsPunctuation Symbols used for punctuation in Western and other alphabets DashPunctuation ConnectorPunctuation InitialQuotePunctuation FinalQuotePunctuation IsSymbol Most other printable symbols MathSymbol ModifierSymbol OtherSymbol IsControl Nonprintable “control” characters below 0x20, such as r, n, t, 0, and characters between 0x7F and 0x9A (None) 272 | Chapter 6: .NET Fundamentals
- 291. For more granular categorization, char provides a static method called GetUnicode Category; this returns a UnicodeCategory enumeration whose members are shown in the rightmost column of Table 6-1. By explicitly casting from an integer, it’s possible to produce a char outside the allocated Unicode set. To test a charac‐ ter’s validity, call char.GetUnicodeCategory: if the result is UnicodeCategory.OtherNotAssigned, the character is invalid. A char is 16 bits wide—enough to represent any Unicode character in the Basic Multilingual Plane. To go beyond this, you must use surrogate pairs: we describe the methods for doing this in “Text Encodings and Unicode” on page 281. String A C# string (== System.String) is an immutable (unchangeable) sequence of characters. In Chapter 2, we described how to express string literals, perform equal‐ ity comparisons, and concatenate two strings. This section covers the remaining functions for working with strings, exposed through the static and instance mem‐ bers of the System.String class. Constructing strings The simplest way to construct a string is to assign a literal, as we saw in Chapter 2: string s1 = "Hello"; string s2 = "First LinernSecond Line"; string s3 = @"serverfilesharehelloworld.cs"; To create a repeating sequence of characters, you can use string’s constructor: Console.Write (new string ('*', 10)); // ********** You can also construct a string from a char array. The ToCharArray method does the reverse: char[] ca = "Hello".ToCharArray(); string s = new string (ca); // s = "Hello" string’s constructor is also overloaded to accept various (unsafe) pointer types, in order to create strings from types such as char*. Null and empty strings An empty string has a length of zero. To create an empty string, you can use either a literal or the static string.Empty field; to test for an empty string, you can either perform an equality comparison or test its Length property: string empty = ""; Console.WriteLine (empty == ""); // True Console.WriteLine (empty == string.Empty); // True Console.WriteLine (empty.Length == 0); // True String and Text Handling | 273 .NET Fundamentals
- 292. Because strings are reference types, they can also be null: string nullString = null; Console.WriteLine (nullString == null); // True Console.WriteLine (nullString == ""); // False Console.WriteLine (nullString.Length == 0); // NullReferenceException The static string.IsNullOrEmpty method is a useful shortcut for testing whether a given string is either null or empty. Accessing characters within a string A string’s indexer returns a single character at the given index. As with all functions that operate on strings, this is zero-indexed: string str = "abcde"; char letter = str[1]; // letter == 'b' string also implements IEnumerable<char>, so you can foreach over its characters: foreach (char c in "123") Console.Write (c + ","); // 1,2,3, Searching within strings The simplest methods for searching within strings are StartsWith, EndsWith, and Contains. These all return true or false: Console.WriteLine ("quick brown fox".EndsWith ("fox")); // True Console.WriteLine ("quick brown fox".Contains ("brown")); // True StartsWith and EndsWith are overloaded to let you specify a StringComparison enum or a CultureInfo object to control case and culture sensitivity (see “Ordinal versus culture comparison” on page 278). The default is to perform a case-sensitive match using rules applicable to the current (localized) culture. The following instead performs a case-insensitive search using the invariant culture’s rules: "abcdef".StartsWith ("aBc", StringComparison.InvariantCultureIgnoreCase) The Contains method doesn’t offer the convenience of this overload, although you can achieve the same result with the IndexOf method. IndexOf is more powerful: it returns the first position of a given character or substring (or −1 if the substring isn’t found): Console.WriteLine ("abcde".IndexOf ("cd")); // 2 IndexOf is also overloaded to accept a startPosition (an index from which to begin searching) as well as a StringComparison enum: Console.WriteLine ("abcde abcde".IndexOf ("CD", 6, StringComparison.CurrentCultureIgnoreCase)); // 8 274 | Chapter 6: .NET Fundamentals
- 293. LastIndexOf is like IndexOf, but it works backward through the string. IndexOfAny returns the first matching position of any one of a set of characters: Console.Write ("ab,cd ef".IndexOfAny (new char[] {' ', ','} )); // 2 Console.Write ("pas5w0rd".IndexOfAny ("0123456789".ToCharArray() )); // 3 LastIndexOfAny does the same in the reverse direction. Manipulating strings Because String is immutable, all the methods that “manipulate” a string return a new one, leaving the original untouched (the same goes for when you reassign a string variable). Substring extracts a portion of a string: string left3 = "12345".Substring (0, 3); // left3 = "123"; string mid3 = "12345".Substring (1, 3); // mid3 = "234"; If you omit the length, you get the remainder of the string: string end3 = "12345".Substring (2); // end3 = "345"; Insert and Remove insert or remove characters at a specified position: string s1 = "helloworld".Insert (5, ", "); // s1 = "hello, world" string s2 = s1.Remove (5, 2); // s2 = "helloworld"; PadLeft and PadRight pad a string to a given length with a specified character (or a space if unspecified): Console.WriteLine ("12345".PadLeft (9, '*')); // ****12345 Console.WriteLine ("12345".PadLeft (9)); // 12345 If the input string is longer than the padding length, the original string is returned unchanged. TrimStart and TrimEnd remove specified characters from the beginning or end of a string; Trim does both. By default, these functions remove whitespace characters (including spaces, tabs, new lines, and Unicode variations of these): Console.WriteLine (" abc trn ".Trim().Length); // 3 Replace replaces all (nonoverlapping) occurrences of a particular character or substring: Console.WriteLine ("to be done".Replace (" ", " | ") ); // to | be | done Console.WriteLine ("to be done".Replace (" ", "") ); // tobedone ToUpper and ToLower return uppercase and lowercase versions of the input string. By default, they honor the user’s current language settings; ToUpperInvariant and ToLowerInvariant always apply English alphabet rules. String and Text Handling | 275 .NET Fundamentals
- 294. Splitting and joining strings Split divides a string into pieces: string[] words = "The quick brown fox".Split(); foreach (string word in words) Console.Write (word + "|"); // The|quick|brown|fox| By default, Split uses whitespace characters as delimiters; it’s also overloaded to accept a params array of char or string delimiters. Split also optionally accepts a StringSplitOptions enum, which has an option to remove empty entries: this is useful when words are separated by several delimiters in a row. The static Join method does the reverse of Split. It requires a delimiter and string array: string[] words = "The quick brown fox".Split(); string together = string.Join (" ", words); // The quick brown fox The static Concat method is similar to Join but accepts only a params string array and applies no separator. Concat is exactly equivalent to the + operator (the compiler, in fact, translates + to Concat): string sentence = string.Concat ("The", " quick", " brown", " fox"); string sameSentence = "The" + " quick" + " brown" + " fox"; String.Format and composite format strings The static Format method provides a convenient way to build strings that embed variables. The embedded variables (or values) can be of any type; the Format simply calls ToString on them. The master string that includes the embedded variables is called a composite for‐ mat string. When calling string.Format, you provide a composite format string followed by each of the embedded variables: string composite = "It's {0} degrees in {1} on this {2} morning"; string s = string.Format (composite, 35, "Perth", DateTime.Now.DayOfWeek); // s == "It's 35 degrees in Perth on this Friday morning" (And that’s Celsius!) We can use interpolated string literals to the same effect (see “String Type” on page 52). Just precede the string with the $ symbol and put the expressions in braces: string s = $"It's hot this {DateTime.Now.DayOfWeek} morning"; Each number in curly braces is called a format item. The number corresponds to the argument position and is optionally followed by: 276 | Chapter 6: .NET Fundamentals
- 295. • A comma and a minimum width to apply • • A colon and a format string • The minimum width is useful for aligning columns. If the value is negative, the data is left-aligned; otherwise, it’s right-aligned: string composite = "Name={0,-20} Credit Limit={1,15:C}"; Console.WriteLine (string.Format (composite, "Mary", 500)); Console.WriteLine (string.Format (composite, "Elizabeth", 20000)); Here’s the result: Name=Mary Credit Limit= $500.00 Name=Elizabeth Credit Limit= $20,000.00 Here’s the equivalent without using string.Format: string s = "Name=" + "Mary".PadRight (20) + " Credit Limit=" + 500.ToString ("C").PadLeft (15); The credit limit is formatted as currency by virtue of the "C" format string. We describe format strings in detail in “Formatting and Parsing” on page 297. Comparing Strings In comparing two values, .NET differentiates the concepts of equality comparison and order comparison. Equality comparison tests whether two instances are seman‐ tically the same; order comparison tests which of two (if any) instances comes first when arranging them in ascending or descending sequence. Equality comparison is not a subset of order comparison; the two systems have different purposes. It’s legal, for instance, to have two unequal values in the same ordering position. We resume this topic in “Equality Comparison” on page 324. For string equality comparison, you can use the == operator or one of string’s Equals methods. The latter are more versatile because they allow you to specify options such as case insensitivity. Another difference is that == does not work reliably on strings if the variables are cast to the object type. We explain why this is so in “Equality Comparison” on page 324. For string order comparison, you can use either the CompareTo instance method or the static Compare and CompareOrdinal methods. These return a positive or negative number, or zero, depending on whether the first value comes after, before, or alongside the second. Before going into the details of each, we need to examine .NET’s underlying string comparison algorithms. String and Text Handling | 277 .NET Fundamentals
- 296. Ordinal versus culture comparison There are two basic algorithms for string comparison: ordinal and culture sensitive. Ordinal comparisons interpret characters simply as numbers (according to their numeric Unicode value); culture-sensitive comparisons interpret characters with reference to a particular alphabet. There are two special cultures: the “current cul‐ ture,” which is based on settings picked up from the computer’s control panel, and the “invariant culture,” which is the same on every computer (and closely matches American culture). For equality comparison, both ordinal and culture-specific algorithms are useful. For ordering, however, culture-specific comparison is nearly always preferable: to order strings alphabetically, you need an alphabet. Ordinal relies on the numeric Unicode point values, which happen to put English characters in alphabetical order —but even then, not exactly as you might expect. For example, assuming case sensitivity, consider the strings "Atom", "atom", and "Zamia". The invariant culture puts them in the following order: "atom", "Atom", "Zamia" Ordinal arranges them instead as follows: "Atom", "Zamia", "atom" This is because the invariant culture encapsulates an alphabet, which considers uppercase characters adjacent to their lowercase counterparts (aAbBcCdD...). The ordinal algorithm, however, puts all the uppercase characters first and then all lowercase characters (A...Z, a...z). This is essentially a throwback to the ASCII character set invented in the 1960s. String equality comparison Despite ordinal’s limitations, string’s == operator always performs ordinal case- sensitive comparison. The same goes for the instance version of string.Equals when called without arguments; this defines the “default” equality comparison behavior for the string type. The ordinal algorithm was chosen for string’s == and Equals functions because it’s both highly efficient and deterministic. String equality comparison is considered fundamental and is performed far more frequently than order comparison. A “strict” notion of equality is also consistent with the general use of the == operator. The following methods allow culture-aware or case-insensitive comparisons: public bool Equals(string value, StringComparison comparisonType); public static bool Equals (string a, string b, StringComparison comparisonType); 278 | Chapter 6: .NET Fundamentals
- 297. The static version is advantageous in that it still works if one or both of the strings are null. StringComparison is an enum defined as follows: public enum StringComparison { CurrentCulture, // Case-sensitive CurrentCultureIgnoreCase, InvariantCulture, // Case-sensitive InvariantCultureIgnoreCase, Ordinal, // Case-sensitive OrdinalIgnoreCase } For example: Console.WriteLine (string.Equals ("foo", "FOO", StringComparison.OrdinalIgnoreCase)); // True Console.WriteLine ("ṻ" == "ǖ"); // False Console.WriteLine (string.Equals ("ṻ", "ǖ", StringComparison.CurrentCulture)); // ? (The result of the third example is determined by the computer’s current language settings.) String order comparison String’s CompareTo instance method performs culture-sensitive, case-sensitive order comparison. Unlike the == operator, CompareTo does not use ordinal comparison: for ordering, a culture-sensitive algorithm is much more useful. Here’s the method’s definition: public int CompareTo (string strB); The CompareTo instance method implements the generic IComparable interface, a standard comparison protocol used across the .NET libraries. This means string’s CompareTo defines the default ordering behavior of strings in such appli‐ cations as sorted collections, for instance. For more informa‐ tion on IComparable, see “Order Comparison” on page 335. For other kinds of comparison, you can call the static Compare and CompareOrdinal methods: public static int Compare (string strA, string strB, StringComparison comparisonType); public static int Compare (string strA, string strB, bool ignoreCase, CultureInfo culture); public static int Compare (string strA, string strB, bool ignoreCase); public static int CompareOrdinal (string strA, string strB); String and Text Handling | 279 .NET Fundamentals
- 298. The last two methods are simply shortcuts for calling the first two methods. All of the order comparison methods return a positive number, a negative number, or zero depending on whether the first value comes after, before, or alongside the second value: Console.WriteLine ("Boston".CompareTo ("Austin")); // 1 Console.WriteLine ("Boston".CompareTo ("Boston")); // 0 Console.WriteLine ("Boston".CompareTo ("Chicago")); // -1 Console.WriteLine ("ṻ".CompareTo ("ǖ")); // 0 Console.WriteLine ("foo".CompareTo ("FOO")); // -1 The following performs a case-insensitive comparison using the current culture: Console.WriteLine (string.Compare ("foo", "FOO", true)); // 0 By supplying a CultureInfo object, you can plug in any alphabet: // CultureInfo is defined in the System.Globalization namespace CultureInfo german = CultureInfo.GetCultureInfo ("de-DE"); int i = string.Compare ("Müller", "Muller", false, german); StringBuilder The StringBuilder class (System.Text namespace) represents a mutable (editable) string. With a StringBuilder, you can Append, Insert, Remove, and Replace sub‐ strings without replacing the whole StringBuilder. StringBuilder’s constructor optionally accepts an initial string value as well as a starting size for its internal capacity (default is 16 characters). If you go beyond this, StringBuilder automatically resizes its internal structures to accommodate (at a slight performance cost) up to its maximum capacity (default is int.MaxValue). A popular use of StringBuilder is to build up a long string by repeatedly calling Append. This approach is much more efficient than repeatedly concatenating ordi‐ nary string types: StringBuilder sb = new StringBuilder(); for (int i = 0; i < 50; i++) sb.Append(i).Append(","); To get the final result, call ToString(): Console.WriteLine (sb.ToString()); 0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26, 27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49, AppendLine performs an Append that adds a new line sequence ("rn" in Win‐ dows). AppendFormat accepts a composite format string, just like String.Format. In addition to the Insert, Remove, and Replace methods (Replace works like string’s Replace), StringBuilder defines a Length property and a writable indexer for getting/setting individual characters. 280 | Chapter 6: .NET Fundamentals
- 299. To clear the contents of a StringBuilder, either instantiate a new one or set its Length to zero. Setting a StringBuilder’s Length to zero doesn’t shrink its internal capacity. So, if the StringBuilder previously con‐ tained one million characters, it will continue to occupy around two megabytes of memory after zeroing its Length. If you want to release the memory, you must create a new StringBuilder and allow the old one to drop out of scope (and be garbage collected). Text Encodings and Unicode A character set is an allocation of characters, each with a numeric code, or code point. There are two character sets in common use: Unicode and ASCII. Unicode has an address space of approximately one million characters, of which about 100,000 are currently allocated. Unicode covers most spoken world languages as well as some historical languages and special symbols. The ASCII set is simply the first 128 characters of the Unicode set, which covers most of what you see on a US-style keyboard. ASCII predates Unicode by 30 years and is still sometimes used for its simplicity and efficiency: each character is represented by one byte. The .NET type system is designed to work with the Unicode character set. ASCII is implicitly supported, though, by virtue of being a subset of Unicode. A text encoding maps characters from their numeric code point to a binary repre‐ sentation. In .NET, text encodings come into play primarily when dealing with text files or streams. When you read a text file into a string, a text encoder translates the file data from binary into the internal Unicode representation that the char and string types expect. A text encoding can restrict what characters can be represented as well as affect storage efficiency. There are two categories of text encoding in .NET: • Those that map Unicode characters to another character set • • Those that use standard Unicode encoding schemes • The first category contains legacy encodings such as IBM’s EBCDIC and 8-bit char‐ acter sets with extended characters in the upper-128 region that were popular prior to Unicode (identified by a code page). The ASCII encoding is also in this category: it encodes the first 128 characters and drops everything else. This category contains the nonlegacy GB18030 as well, which is the mandatory standard for applications written in China—or sold to China—since 2000. In the second category are UTF-8, UTF-16, and UTF-32 (and the obsolete UTF-7). Each differs in space efficiency. UTF-8 is the most space-efficient for most kinds of text: it uses between one and four bytes to represent each character. The first 128 characters require only a single byte, making it compatible with ASCII. UTF-8 is the most popular encoding for text files and streams (particularly on the internet), and String and Text Handling | 281 .NET Fundamentals
- 300. it is the default for stream input/output (I/O) in .NET (in fact, it’s the default for almost everything that implicitly uses an encoding). UTF-16 uses one or two 16-bit words to represent each character. This is what .NET uses internally to represent characters and strings. Some programs also write files in UTF-16. UTF-32 is the least space-efficient: it maps each code point directly to 32 bits, so every character consumes four bytes. UTF-32 is rarely used for this reason. It does, however, make random access very easy because every character takes an equal number of bytes. Obtaining an Encoding object The Encoding class in System.Text is the common base type for classes that encap‐ sulate text encodings. There are several subclasses—their purpose is to encapsulate families of encodings with similar features. The most common encodings can be obtained through dedicated static properties on Encoding: Encoding name Static property on Encoding UTF-8 Encoding.UTF8 UTF-16 Encoding.Unicode (not UTF16) UTF-32 Encoding.UTF32 ASCII Encoding.ASCII You can obtain other encodings by calling Encoding.GetEncoding with a standard Internet Assigned Numbers Authority (IANA) Character Set name: // In .NET 5+ and .NET Core, you must first call RegisterProvider: Encoding.RegisterProvider (CodePagesEncodingProvider.Instance); Encoding chinese = Encoding.GetEncoding ("GB18030"); The static GetEncodings method returns a list of all supported encodings along with their standard IANA names: foreach (EncodingInfo info in Encoding.GetEncodings()) Console.WriteLine (info.Name); The other way to obtain an encoding is to directly instantiate an encoding class. Doing so allows you to set various options via constructor arguments, including: • Whether to throw an exception if an invalid byte sequence is encountered • when decoding. The default is false. • Whether to encode/decode UTF-16/UTF-32 with the most significant bytes • first (big endian) or the least significant bytes first (little endian). The default is little endian, the standard on the Windows operating system. 282 | Chapter 6: .NET Fundamentals
- 301. • Whether to emit a byte-order mark (a prefix that indicates endianness). • Encoding for file and stream I/O The most common application for an Encoding object is to control how text is read and written to a file or stream. For example, the following writes “Testing…” to a file called data.txt in UTF-16 encoding: System.IO.File.WriteAllText ("data.txt", "Testing...", Encoding.Unicode); If you omit the final argument, WriteAllText applies the ubiquitous UTF-8 encoding. UTF-8 is the default text encoding for all file and stream I/O. We resume this subject in Chapter 15, in “Stream Adapters” on page 687. Encoding to byte arrays You can also use an Encoding object to go to and from a byte array. The GetBytes method converts from string to byte[] with the given encoding; GetString con‐ verts from byte[] to string: byte[] utf8Bytes = System.Text.Encoding.UTF8.GetBytes ("0123456789"); byte[] utf16Bytes = System.Text.Encoding.Unicode.GetBytes ("0123456789"); byte[] utf32Bytes = System.Text.Encoding.UTF32.GetBytes ("0123456789"); Console.WriteLine (utf8Bytes.Length); // 10 Console.WriteLine (utf16Bytes.Length); // 20 Console.WriteLine (utf32Bytes.Length); // 40 string original1 = System.Text.Encoding.UTF8.GetString (utf8Bytes); string original2 = System.Text.Encoding.Unicode.GetString (utf16Bytes); string original3 = System.Text.Encoding.UTF32.GetString (utf32Bytes); Console.WriteLine (original1); // 0123456789 Console.WriteLine (original2); // 0123456789 Console.WriteLine (original3); // 0123456789 UTF-16 and surrogate pairs Recall that .NET stores characters and strings in UTF-16. Because UTF-16 requires one or two 16-bit words per character, and a char is only 16 bits in length, some Unicode characters require two chars to represent. This has a couple of consequences: • A string’s Length property can be greater than its real character count. • • A single char is not always enough to fully represent a Unicode character. • String and Text Handling | 283 .NET Fundamentals
- 302. Most applications ignore this because nearly all commonly used characters fit into a section of Unicode called the Basic Multilingual Plane (BMP), which requires only one 16-bit word in UTF-16. The BMP covers several dozen world languages and includes more than 30,000 Chinese characters. Excluded are characters of some ancient languages, symbols for musical notation, and some less-common Chinese characters. If you need to support two-word characters, the following static methods in char convert a 32-bit code point to a string of two chars and back again: string ConvertFromUtf32 (int utf32) int ConvertToUtf32 (char highSurrogate, char lowSurrogate) Two-word characters are called surrogates. They are easy to spot because each word is in the range 0xD800 to 0xDFFF. You can use the following static methods in char to assist: bool IsSurrogate (char c) bool IsHighSurrogate (char c) bool IsLowSurrogate (char c) bool IsSurrogatePair (char highSurrogate, char lowSurrogate) The StringInfo class in the System.Globalization namespace also provides a range of methods and properties for working with two-word characters. Characters outside the BMP typically require special fonts and have limited operat‐ ing system support. Dates and Times The following immutable structs in the System namespace do the job of repre‐ senting dates and times: DateTime, DateTimeOffset, TimeSpan, DateOnly, and Time Only. C# doesn’t define any special keywords that map to these types. TimeSpan A TimeSpan represents an interval of time—or a time of the day. In the latter role, it’s simply the “clock” time (without the date), which is equivalent to the time since midnight, assuming no daylight saving transition. A TimeSpan has a resolution of 100 ns, has a maximum value of about 10 million days, and can be positive or negative. There are three ways to construct a TimeSpan: • Through one of the constructors • • By calling one of the static From… methods • • By subtracting one DateTime from another • Here are the constructors: 284 | Chapter 6: .NET Fundamentals
- 303. public TimeSpan (int hours, int minutes, int seconds); public TimeSpan (int days, int hours, int minutes, int seconds); public TimeSpan (int days, int hours, int minutes, int seconds, int milliseconds); public TimeSpan (long ticks); // Each tick = 100ns The static From… methods are more convenient when you want to specify an interval in just a single unit, such as minutes, hours, and so on: public static TimeSpan FromDays (double value); public static TimeSpan FromHours (double value); public static TimeSpan FromMinutes (double value); public static TimeSpan FromSeconds (double value); public static TimeSpan FromMilliseconds (double value); For example: Console.WriteLine (new TimeSpan (2, 30, 0)); // 02:30:00 Console.WriteLine (TimeSpan.FromHours (2.5)); // 02:30:00 Console.WriteLine (TimeSpan.FromHours (-2.5)); // -02:30:00 TimeSpan overloads the < and > operators as well as the + and - operators. The following expression evaluates to a TimeSpan of 2.5 hours: TimeSpan.FromHours(2) + TimeSpan.FromMinutes(30); The next expression evaluates to one second short of 10 days: TimeSpan.FromDays(10) - TimeSpan.FromSeconds(1); // 9.23:59:59 Using this expression, we can illustrate the integer properties Days, Hours, Minutes, Seconds, and Milliseconds: TimeSpan nearlyTenDays = TimeSpan.FromDays(10) - TimeSpan.FromSeconds(1); Console.WriteLine (nearlyTenDays.Days); // 9 Console.WriteLine (nearlyTenDays.Hours); // 23 Console.WriteLine (nearlyTenDays.Minutes); // 59 Console.WriteLine (nearlyTenDays.Seconds); // 59 Console.WriteLine (nearlyTenDays.Milliseconds); // 0 In contrast, the Total... properties return values of type double describing the entire time span: Console.WriteLine (nearlyTenDays.TotalDays); // 9.99998842592593 Console.WriteLine (nearlyTenDays.TotalHours); // 239.999722222222 Console.WriteLine (nearlyTenDays.TotalMinutes); // 14399.9833333333 Console.WriteLine (nearlyTenDays.TotalSeconds); // 863999 Console.WriteLine (nearlyTenDays.TotalMilliseconds); // 863999000 The static Parse method does the opposite of ToString, converting a string to a TimeSpan. TryParse does the same but returns false rather than throwing an exception if the conversion fails. The XmlConvert class also provides TimeSpan/ string conversion methods that follow standard XML formatting protocols. The default value for a TimeSpan is TimeSpan.Zero. Dates and Times | 285 .NET Fundamentals
- 304. TimeSpan can also be used to represent the time of the day (the elapsed time since midnight). To obtain the current time of day, call DateTime.Now.TimeOfDay. DateTime and DateTimeOffset DateTime and DateTimeOffset are immutable structs for representing a date and, optionally, a time. They have a resolution of 100 ns and a range covering the years 0001 through 9999. DateTimeOffset is functionally similar to DateTime. Its distinguishing feature is that it also stores a Coordinated Universal Time (UTC) offset; this allows more meaningful results when comparing values across different time zones. An excellent article on the rationale behind the introduction of DateTimeOffset is available online, titled “A Brief History of DateTime,” by Anthony Moore. Choosing between DateTime and DateTimeOffset DateTime and DateTimeOffset differ in how they handle time zones. A DateTime incorporates a three-state flag indicating whether the DateTime is relative to the following: • The local time on the current computer • • UTC (the modern equivalent of Greenwich Mean Time) • • Unspecified • A DateTimeOffset is more specific—it stores the offset from UTC as a TimeSpan: July 01 2019 03:00:00 -06:00 This influences equality comparisons, which is the main factor in choosing between DateTime and DateTimeOffset. Specifically: • DateTime ignores the three-state flag in comparisons and considers two values • equal if they have the same year, month, day, hour, minute, and so on. • DateTimeOffset considers two values equal if they refer to the same point in • time. Daylight Saving Time can make this distinction important even if your application doesn’t need to handle multiple geo‐ graphic time zones. So, DateTime considers the following two values different, whereas DateTimeOffset considers them equal: July 01 2019 09:00:00 +00:00 (GMT) July 01 2019 03:00:00 -06:00 (local time, Central America) 286 | Chapter 6: .NET Fundamentals
- 305. In most cases, DateTimeOffset’s equality logic is preferable. For example, in calcu‐ lating which of two international events is more recent, a DateTimeOffset implicitly gives the correct answer. Similarly, a hacker plotting a Distributed Denial of Service attack would reach for a DateTimeOffset! To do the same with DateTime requires standardizing on a single time zone (typically UTC) throughout your application. This is problematic for two reasons: • To be friendly to the end user, UTC DateTimes require explicit conversion to • local time prior to formatting. • It’s easy to forget and incorporate a local DateTime. • DateTime is better, though, at specifying a value relative to the local computer at runtime—for example, if you want to schedule an archive at each of your interna‐ tional offices for next Sunday, at 3 A.M. local time (when there’s the least amount of activity). Here, DateTime would be more suitable because it would respect each site’s local time. Internally, DateTimeOffset uses a short integer to store the UTC offset in minutes. It doesn’t store any regional informa‐ tion, so there’s nothing present to indicate whether an offset of +08:00, for instance, refers to Singapore time or Perth time. We revisit time zones and equality comparison in more depth in “DateOnly and TimeOnly” on page 292. SQL Server 2008 introduced direct support for DateTimeOff set through a new data type of the same name. Constructing a DateTime DateTime defines constructors that accept integers for the year, month, and day— and, optionally, the hour, minute, second, and millisecond: public DateTime (int year, int month, int day); public DateTime (int year, int month, int day, int hour, int minute, int second, int millisecond); If you specify only a date, the time is implicitly set to midnight (0:00). The DateTime constructors also allow you to specify a DateTimeKind—an enum with the following values: Unspecified, Local, Utc This corresponds to the three-state flag described in the preceding section. Unspeci fied is the default, and it means that the DateTime is time-zone-agnostic. Local means relative to the local time zone on the current computer. A local DateTime Dates and Times | 287 .NET Fundamentals
- 306. does not include information about which particular time zone it refers to, or, unlike DateTimeOffset, the numeric offset from UTC. A DateTime’s Kind property returns its DateTimeKind. DateTime’s constructors are also overloaded to accept a Calendar object, as well. This allows you to specify a date using any of the Calendar subclasses defined in System.Globalization: DateTime d = new DateTime (5767, 1, 1, new System.Globalization.HebrewCalendar()); Console.WriteLine (d); // 12/12/2006 12:00:00 AM (The formatting of the date in this example depends on your computer’s control panel settings.) A DateTime always uses the default Gregorian calendar—this exam‐ ple, a one-time conversion, takes place during construction. To perform computa‐ tions using another calendar, you must use the methods on the Calendar subclass itself. You can also construct a DateTime with a single ticks value of type long, where ticks is the number of 100 ns intervals from midnight 01/01/0001. For interoperability, DateTime provides the static FromFileTime and FromFile TimeUtc methods for converting from a Windows file time (specified as a long) and FromOADate for converting from an OLE automation date/time (specified as a double). To construct a DateTime from a string, call the static Parse or ParseExact method. Both methods accept optional flags and format providers; ParseExact also accepts a format string. We discuss parsing in greater detail in “Formatting and Parsing” on page 297. Constructing a DateTimeOffset DateTimeOffset has a similar set of constructors. The difference is that you also specify a UTC offset as a TimeSpan: public DateTimeOffset (int year, int month, int day, int hour, int minute, int second, TimeSpan offset); public DateTimeOffset (int year, int month, int day, int hour, int minute, int second, int millisecond, TimeSpan offset); The TimeSpan must amount to a whole number of minutes; otherwise an exception is thrown. DateTimeOffset also has constructors that accept a Calendar object, a long ticks value, and static Parse and ParseExact methods that accept a string. 288 | Chapter 6: .NET Fundamentals
- 307. You can construct a DateTimeOffset from an existing DateTime either by using these constructors: public DateTimeOffset (DateTime dateTime); public DateTimeOffset (DateTime dateTime, TimeSpan offset); or with an implicit cast: DateTimeOffset dt = new DateTime (2000, 2, 3); The implicit cast from DateTime to DateTimeOffset is handy because most of the .NET BCL supports DateTime—not Date TimeOffset. If you don’t specify an offset, it’s inferred from the DateTime value using these rules: • If the DateTime has a DateTimeKind of Utc, the offset is zero. • • If the DateTime has a DateTimeKind of Local or Unspecified (the default), the • offset is taken from the current local time zone. To convert in the other direction, DateTimeOffset provides three properties that return values of type DateTime: • The UtcDateTime property returns a DateTime in UTC time. • • The LocalDateTime property returns a DateTime in the current local time zone • (converting it if necessary). • The DateTime property returns a DateTime in whatever zone it was specified, • with a Kind of Unspecified (i.e., it returns the UTC time plus the offset). The current DateTime/DateTimeOffset Both DateTime and DateTimeOffset have a static Now property that returns the current date and time: Console.WriteLine (DateTime.Now); // 11/11/2019 1:23:45 PM Console.WriteLine (DateTimeOffset.Now); // 11/11/2019 1:23:45 PM -06:00 DateTime also provides a Today property that returns just the date portion: Console.WriteLine (DateTime.Today); // 11/11/2019 12:00:00 AM The static UtcNow property returns the current date and time in UTC: Console.WriteLine (DateTime.UtcNow); // 11/11/2019 7:23:45 AM Console.WriteLine (DateTimeOffset.UtcNow); // 11/11/2019 7:23:45 AM +00:00 The precision of all these methods depends on the operating system and is typically in the 10 to 20 ms region. Dates and Times | 289 .NET Fundamentals
- 308. Working with dates and times DateTime and DateTimeOffset provide a similar set of instance properties that return various date/time elements: DateTime dt = new DateTime (2000, 2, 3, 10, 20, 30); Console.WriteLine (dt.Year); // 2000 Console.WriteLine (dt.Month); // 2 Console.WriteLine (dt.Day); // 3 Console.WriteLine (dt.DayOfWeek); // Thursday Console.WriteLine (dt.DayOfYear); // 34 Console.WriteLine (dt.Hour); // 10 Console.WriteLine (dt.Minute); // 20 Console.WriteLine (dt.Second); // 30 Console.WriteLine (dt.Millisecond); // 0 Console.WriteLine (dt.Ticks); // 630851700300000000 Console.WriteLine (dt.TimeOfDay); // 10:20:30 (returns a TimeSpan) DateTimeOffset also has an Offset property of type TimeSpan. Both types provide the following instance methods to perform computations (most accept an argument of type double or int): AddYears AddMonths AddDays AddHours AddMinutes AddSeconds AddMilliseconds AddTicks These all return a new DateTime or DateTimeOffset, and they take into account such things as leap years. You can pass in a negative value to subtract. The Add method adds a TimeSpan to a DateTime or DateTimeOffset. The + operator is overloaded to do the same job: TimeSpan ts = TimeSpan.FromMinutes (90); Console.WriteLine (dt.Add (ts)); Console.WriteLine (dt + ts); // same as above You can also subtract a TimeSpan from a DateTime/DateTimeOffset and subtract one DateTime/DateTimeOffset from another. The latter gives you a TimeSpan: DateTime thisYear = new DateTime (2015, 1, 1); DateTime nextYear = thisYear.AddYears (1); TimeSpan oneYear = nextYear - thisYear; Formatting and parsing DateTimes Calling ToString on a DateTime formats the result as a short date (all numbers) followed by a long time (including seconds). For example: 11/11/2019 11:50:30 AM The operating system’s control panel, by default, determines such things as whether the day, month, or year comes first; the use of leading zeros; and whether 12- or 24-hour time is used. 290 | Chapter 6: .NET Fundamentals
- 309. Calling ToString on a DateTimeOffset is the same, except that the offset is also returned: 11/11/2019 11:50:30 AM -06:00 The ToShortDateString and ToLongDateString methods return just the date por‐ tion. The long date format is also determined by the control panel; an example is “Wednesday, 11 November 2015.” ToShortTimeString and ToLongTimeString return just the time portion, such as 17:10:10 (the former excludes seconds). These four just-described methods are actually shortcuts to four different format strings. ToString is overloaded to accept a format string and provider, allowing you to specify a wide range of options and control how regional settings are applied. We describe this in “Formatting and Parsing” on page 297. DateTimes and DateTimeOffsets can be misparsed if the cul‐ ture settings differ from those in force when formatting takes place. You can avoid this problem by using ToString in con‐ junction with a format string that ignores culture settings (such as “o”): DateTime dt1 = DateTime.Now; string cannotBeMisparsed = dt1.ToString ("o"); DateTime dt2 = DateTime.Parse (cannotBeMisparsed); The static Parse/TryParse and ParseExact/TryParseExact methods do the reverse of ToString, converting a string to a DateTime or DateTimeOffset. These methods are also overloaded to accept a format provider. The Try* methods return false instead of throwing a FormatException. Null DateTime and DateTimeOffset values Because DateTime and DateTimeOffset are structs, they are not intrinsically nulla‐ ble. When you need nullability, there are two ways around this: • Use a Nullable type (i.e., DateTime? or DateTimeOffset?). • • Use the static field DateTime.MinValue or DateTimeOffset.MinValue (the • default values for these types). A nullable type is usually the best approach because the compiler helps to prevent mistakes. DateTime.MinValue is useful for backward compatibility with code writ‐ ten prior to C# 2.0 (when nullable value types were introduced). Calling ToUniversalTime or ToLocalTime on a DateTime.Min Value can result in it no longer being DateTime.MinValue (depending on which side of GMT you are on). If you’re right on GMT (England, outside daylight saving), the problem won’t arise at all because local and UTC times are the same. This is your compensation for the English winter! Dates and Times | 291 .NET Fundamentals
- 310. DateOnly and TimeOnly The DateOnly and TimeOnly structs (from .NET 6) exist for when you only want to represent a date or time. DateOnly is similar to DateTime, but without a time component. DateOnly also lacks DateTimeKind; in effect, it’s always Unspecified and has no concept of Local or Utc. The historical alternative to DateOnly was to use DateTime with a zero time (midnight). The difficulty with this approach is that equality comparisons fail when a non-zero time finds its way into your code. TimeOnly is similar to DateTime, but without a date component. TimeOnly is intended for capturing the time of day and is suitable for applications such as recording alarm times or opening hours. DateTime and Time Zones DateTime is simplistic in its handling of time zones. Internally, it stores a DateTime using two pieces of information: • A 62-bit number, indicating the number of ticks since 1/1/0001 • • A 2-bit enum, indicating the DateTimeKind (Unspecified, Local, or Utc) • When you compare two DateTime instances, only their ticks values are compared; their DateTimeKinds are ignored: DateTime dt1 = new DateTime (2000, 1, 1, 10, 20, 30, DateTimeKind.Local); DateTime dt2 = new DateTime (2000, 1, 1, 10, 20, 30, DateTimeKind.Utc); Console.WriteLine (dt1 == dt2); // True DateTime local = DateTime.Now; DateTime utc = local.ToUniversalTime(); Console.WriteLine (local == utc); // False The instance methods ToUniversalTime/ToLocalTime convert to universal/local time. These apply the computer’s current time zone settings and return a new DateTime with a DateTimeKind of Utc or Local. No conversion happens if you call ToUniversalTime on a DateTime that’s already Utc, or ToLocalTime on a DateTime that’s already Local. You will get a conversion, however, if you call ToUniversal Time or ToLocalTime on a DateTime that’s Unspecified. You can construct a DateTime that differs from another only in Kind with the static DateTime.SpecifyKind method: DateTime d = new DateTime (2015, 12, 12); // Unspecified DateTime utc = DateTime.SpecifyKind (d, DateTimeKind.Utc); Console.WriteLine (utc); // 12/12/2015 12:00:00 AM 292 | Chapter 6: .NET Fundamentals
- 311. DateTimeOffset and Time Zones Internally, DateTimeOffset comprises a DateTime field whose value is always in UTC, and a 16-bit integer field for the UTC offset in minutes. Comparisons look only at the (UTC) DateTime; the Offset is used primarily for formatting. The ToUniversalTime/ToLocalTime methods return a DateTimeOffset represent‐ ing the same point in time but with a UTC or local offset. Unlike with DateTime, these methods don’t affect the underlying date/time value, only the offset: DateTimeOffset local = DateTimeOffset.Now; DateTimeOffset utc = local.ToUniversalTime(); Console.WriteLine (local.Offset); // -06:00:00 (in Central America) Console.WriteLine (utc.Offset); // 00:00:00 Console.WriteLine (local == utc); // True To include the Offset in the comparison, you must use the EqualsExact method: Console.WriteLine (local.EqualsExact (utc)); // False TimeZoneInfo The TimeZoneInfo class provides information on time zone names, UTC offsets, and Daylight Saving Time rules. TimeZone The static TimeZone.CurrentTimeZone method returns a TimeZone: TimeZone zone = TimeZone.CurrentTimeZone; Console.WriteLine (zone.StandardName); // Pacific Standard Time Console.WriteLine (zone.DaylightName); // Pacific Daylight Time The GetDaylightChanges method returns specific Daylight Saving Time informa‐ tion for a given year: DaylightTime day = zone.GetDaylightChanges (2019); Console.WriteLine (day.Start.ToString ("M")); // 10 March Console.WriteLine (day.End.ToString ("M")); // 03 November Console.WriteLine (day.Delta); // 01:00:00 TimeZoneInfo The static TimeZoneInfo.Local method returns a TimeZoneInfo object based on the current local settings. The following demonstrates the result if run in California: TimeZoneInfo zone = TimeZoneInfo.Local; Console.WriteLine (zone.StandardName); // Pacific Standard Time Console.WriteLine (zone.DaylightName); // Pacific Daylight Time The IsDaylightSavingTime and GetUtcOffset methods work as follows: DateTime and Time Zones | 293 .NET Fundamentals
- 312. DateTime dt1 = new DateTime (2019, 1, 1); // DateTimeOffset works, too DateTime dt2 = new DateTime (2019, 6, 1); Console.WriteLine (zone.IsDaylightSavingTime (dt1)); // True Console.WriteLine (zone.IsDaylightSavingTime (dt2)); // False Console.WriteLine (zone.GetUtcOffset (dt1)); // -08:00:00 Console.WriteLine (zone.GetUtcOffset (dt2)); // -07:00:00 You can obtain a TimeZoneInfo for any of the world’s time zones by calling Find SystemTimeZoneById with the zone ID. We’ll switch to Western Australia for rea‐ sons that will soon become clear: TimeZoneInfo wa = TimeZoneInfo.FindSystemTimeZoneById ("W. Australia Standard Time"); Console.WriteLine (wa.Id); // W. Australia Standard Time Console.WriteLine (wa.DisplayName); // (GMT+08:00) Perth Console.WriteLine (wa.BaseUtcOffset); // 08:00:00 Console.WriteLine (wa.SupportsDaylightSavingTime); // True The Id property corresponds to the value passed to FindSystemTimeZoneById. The static GetSystemTimeZones method returns all world time zones; hence, you can list all valid zone ID strings as follows: foreach (TimeZoneInfo z in TimeZoneInfo.GetSystemTimeZones()) Console.WriteLine (z.Id); You can also create a custom time zone by calling Time ZoneInfo.CreateCustomTimeZone. Because TimeZoneInfo is immutable, you must pass in all the relevant data as method arguments. You can serialize a predefined or custom time zone to a (semi) human-readable string by calling ToSerialized String—and deserialize it by calling TimeZoneInfo.FromSer ializedString. The static ConvertTime method converts a DateTime or DateTimeOffset from one time zone to another. You can include either just a destination TimeZoneInfo, or both source and destination TimeZoneInfo objects. You can also convert directly from or to UTC with the methods ConvertTimeFromUtc and ConvertTimeToUtc. For working with Daylight Saving Time, TimeZoneInfo provides the following addi‐ tional methods: • IsInvalidTime returns true if a DateTime is within the hour (or delta) that’s • skipped when the clocks move forward. • IsAmbiguousTime returns true if a DateTime or DateTimeOffset is within the • hour (or delta) that’s repeated when the clocks move back. • GetAmbiguousTimeOffsets returns an array of TimeSpans representing the • valid offset choices for an ambiguous DateTime or DateTimeOffset. 294 | Chapter 6: .NET Fundamentals
- 313. You can’t obtain simple dates from a TimeZoneInfo indicating the start and end of Daylight Saving Time. Instead, you must call GetAdjustmentRules, which returns a declarative summary of all daylight saving rules that apply to all years. Each rule has a DateStart and DateEnd indicating the date range within which the rule is valid: foreach (TimeZoneInfo.AdjustmentRule rule in wa.GetAdjustmentRules()) Console.WriteLine ("Rule: applies from " + rule.DateStart + " to " + rule.DateEnd); Western Australia first introduced Daylight Saving Time in 2006, midseason (and then rescinded it in 2009). This required a special rule for the first year; hence, there are two rules: Rule: applies from 1/01/2006 12:00:00 AM to 31/12/2006 12:00:00 AM Rule: applies from 1/01/2007 12:00:00 AM to 31/12/2009 12:00:00 AM Each AdjustmentRule has a DaylightDelta property of type TimeSpan (this is one hour in almost every case) and properties called DaylightTransitionStart and DaylightTransitionEnd. The latter two are of type TimeZoneInfo.Transition Time, which has the following properties: public bool IsFixedDateRule { get; } public DayOfWeek DayOfWeek { get; } public int Week { get; } public int Day { get; } public int Month { get; } public DateTime TimeOfDay { get; } A transition time is somewhat complicated in that it needs to represent both fixed and floating dates. An example of a floating date is “the last Sunday in March.” Here are the rules for interpreting a transition time: 1. If, for an end transition, IsFixedDateRule is true, Day is 1, Month is 1, and 1. TimeOfDay is DateTime.MinValue, there is no end to Daylight Saving Time in that year (this can happen only in the southern hemisphere, upon the initial introduction of daylight saving time to a region). 2. Otherwise, if IsFixedDateRule is true, the Month, Day, and TimeOfDay proper‐ 2. ties determine the start or end of the adjustment rule. 3. Otherwise, if IsFixedDateRule is false, the Month, DayOfWeek, Week, and 3. TimeOfDay properties determine the start or end of the adjustment rule. In the last case, Week refers to the week of the month, with “5” meaning the last week. We can demonstrate this by enumerating the adjustment rules for our wa time zone: foreach (TimeZoneInfo.AdjustmentRule rule in wa.GetAdjustmentRules()) { Console.WriteLine ("Rule: applies from " + rule.DateStart + " to " + rule.DateEnd); Console.WriteLine (" Delta: " + rule.DaylightDelta); DateTime and Time Zones | 295 .NET Fundamentals
- 314. Console.WriteLine (" Start: " + FormatTransitionTime (rule.DaylightTransitionStart, false)); Console.WriteLine (" End: " + FormatTransitionTime (rule.DaylightTransitionEnd, true)); Console.WriteLine(); } In FormatTransitionTime, we honor the rules just described: static string FormatTransitionTime (TimeZoneInfo.TransitionTime tt, bool endTime) { if (endTime && tt.IsFixedDateRule && tt.Day == 1 && tt.Month == 1 && tt.TimeOfDay == DateTime.MinValue) return "-"; string s; if (tt.IsFixedDateRule) s = tt.Day.ToString(); else s = "The " + "first second third fourth last".Split() [tt.Week - 1] + " " + tt.DayOfWeek + " in"; return s + " " + DateTimeFormatInfo.CurrentInfo.MonthNames [tt.Month-1] + " at " + tt.TimeOfDay.TimeOfDay; } Daylight Saving Time and DateTime If you use a DateTimeOffset or a UTC DateTime, equality comparisons are unimpe‐ ded by the effects of Daylight Saving Time. But with local DateTimes, daylight saving can be problematic. We can summarize the rules as follows: • Daylight saving affects local time but not UTC time. • • When the clocks turn back, comparisons that rely on time moving forward will • break if (and only if) they use local DateTimes. • You can always reliably round-trip between UTC and local times (on the same • computer)—even as the clocks turn back. The IsDaylightSavingTime tells you whether a given local DateTime is subject to Daylight Saving Time. UTC times always return false: Console.Write (DateTime.Now.IsDaylightSavingTime()); // True or False Console.Write (DateTime.UtcNow.IsDaylightSavingTime()); // Always False 296 | Chapter 6: .NET Fundamentals
- 315. Assuming dto is a DateTimeOffset, the following expression does the same: dto.LocalDateTime.IsDaylightSavingTime The end of Daylight Saving Time presents a particular complication for algorithms that use local time, because when the clocks go back, the same hour (or more precisely, Delta) repeats itself. You can reliably compare any two DateTimes by first call‐ ing ToUniversalTime on each. This strategy fails if (and only if) exactly one of them has a DateTimeKind of Unspeci fied. This potential for failure is another reason for favoring DateTimeOffset. Formatting and Parsing Formatting means converting to a string; parsing means converting from a string. The need to format or parse arises frequently in programming, in a variety of situations. Hence, .NET provides a variety of mechanisms: ToString and Parse These methods provide default functionality for many types. Format providers These manifest as additional ToString (and Parse) methods that accept a format string and/or a format provider. Format providers are highly flexible and culture-aware. .NET includes format providers for the numeric types and DateTime/DateTimeOffset. XmlConvert This is a static class with methods that format and parse while honoring XML standards. XmlConvert is also useful for general-purpose conversion when you need culture independence or you want to preempt misparsing. XmlConvert supports the numeric types, bool, DateTime, DateTimeOffset, TimeSpan, and Guid. Type converters These target designers and XAML parsers. In this section, we discuss the first two mechanisms, focusing particularly on format providers. We then describe XmlConvert, type converters, and other conversion mechanisms. ToString and Parse The simplest formatting mechanism is the ToString method. It gives meaningful output on all simple value types (bool, DateTime, DateTimeOffset, TimeSpan, Guid, and all the numeric types). For the reverse operation, each of these types defines a static Parse method: Formatting and Parsing | 297 .NET Fundamentals
- 316. string s = true.ToString(); // s = "True" bool b = bool.Parse (s); // b = true If the parsing fails, a FormatException is thrown. Many types also define a Try Parse method, which returns false if the conversion fails rather than throwing an exception: bool failure = int.TryParse ("qwerty", out int i1); bool success = int.TryParse ("123", out int i2); If you don’t care about the output and want to test only whether parsing would succeed, you can use a discard: bool success = int.TryParse ("123", out int _); If you anticipate an error, calling TryParse is faster and more elegant than calling Parse in an exception handling block. The Parse and TryParse methods on DateTime(Offset) and the numeric types respect local culture settings; you can change this by specifying a CultureInfo object. Specifying invariant culture is often a good idea. For instance, parsing “1.234” into a double gives us 1234 in Germany: Console.WriteLine (double.Parse ("1.234")); // 1234 (In Germany) This is because in Germany, the period indicates a thousands separator rather than a decimal point. Specifying invariant culture fixes this: double x = double.Parse ("1.234", CultureInfo.InvariantCulture); The same applies when calling ToString(): string x = 1.234.ToString (CultureInfo.InvariantCulture); Format Providers Sometimes, you need more control over how formatting and parsing take place. There are dozens of ways to format a DateTime(Offset), for instance. Format providers allow extensive control over formatting and parsing, and are supported for numeric types and date/times. Format providers are also used by user interface controls for formatting and parsing. The gateway to using a format provider is IFormattable. All numeric types—and DateTime(Offset)—implement this interface: public interface IFormattable { string ToString (string format, IFormatProvider formatProvider); } The first argument is the format string; the second is the format provider. The format string provides instructions; the format provider determines how the instructions are translated. For example: 298 | Chapter 6: .NET Fundamentals
- 317. NumberFormatInfo f = new NumberFormatInfo(); f.CurrencySymbol = "$$"; Console.WriteLine (3.ToString ("C", f)); // $$ 3.00 Here, "C" is a format string that indicates currency, and the NumberFormatInfo object is a format provider that determines how currency—and other numeric representations—are rendered. This mechanism allows for globalization. All format strings for numbers and dates are listed in “Stan‐ dard Format Strings and Parsing Flags” on page 303. If you specify a null format string or provider, a default is applied. The default format provider is CultureInfo.CurrentCulture, which, unless reassigned, reflects the computer’s runtime control panel settings. For example, on this computer: Console.WriteLine (10.3.ToString ("C", null)); // $10.30 For convenience, most types overload ToString such that you can omit a null provider: Console.WriteLine (10.3.ToString ("C")); // $10.30 Console.WriteLine (10.3.ToString ("F4")); // 10.3000 (Fix to 4 D.P.) Calling ToString on a DateTime(Offset) or a numeric type with no arguments is equivalent to using a default format provider, with an empty format string. .NET defines three format providers (all of which implement IFormatProvider): NumberFormatInfo DateTimeFormatInfo CultureInfo All enum types are also formattable, though there’s no special IFormatProvider class. Format providers and CultureInfo Within the context of format providers, CultureInfo acts as an indirection mecha‐ nism for the other two format providers, returning a NumberFormatInfo or Date TimeFormatInfo object applicable to the culture’s regional settings. In the following example, we request a specific culture (english language in Great Britain): CultureInfo uk = CultureInfo.GetCultureInfo ("en-GB"); Console.WriteLine (3.ToString ("C", uk)); // £3.00 This executes using the default NumberFormatInfo object applicable to the en-GB culture. The next example formats a DateTime with invariant culture. Invariant culture is always the same, regardless of the computer’s settings: Formatting and Parsing | 299 .NET Fundamentals
- 318. DateTime dt = new DateTime (2000, 1, 2); CultureInfo iv = CultureInfo.InvariantCulture; Console.WriteLine (dt.ToString (iv)); // 01/02/2000 00:00:00 Console.WriteLine (dt.ToString ("d", iv)); // 01/02/2000 Invariant culture is based on American culture, with the fol‐ lowing differences: • The currency symbol is ☼ instead of $. • • Dates and times are formatted with leading zeros • (though still with the month first). • Time uses the 24-hour format rather than an AM/PM • designator. Using NumberFormatInfo or DateTimeFormatInfo In the next example, we instantiate a NumberFormatInfo and change the group separator from a comma to a space. We then use it to format a number to three decimal places: NumberFormatInfo f = new NumberFormatInfo (); f.NumberGroupSeparator = " "; Console.WriteLine (12345.6789.ToString ("N3", f)); // 12 345.679 The initial settings for a NumberFormatInfo or DateTimeFormatInfo are based on the invariant culture. Sometimes, however, it’s more useful to choose a different starting point. To do this, you can Clone an existing format provider: NumberFormatInfo f = (NumberFormatInfo) CultureInfo.CurrentCulture.NumberFormat.Clone(); A cloned format provider is always writable—even if the original was read-only. Composite formatting Composite format strings allow you to combine variable substitution with format strings. The static string.Format method accepts a composite format string (we illustrated this in “String.Format and composite format strings” on page 276): string composite = "Credit={0:C}"; Console.WriteLine (string.Format (composite, 500)); // Credit=$500.00 The Console class itself overloads its Write and WriteLine methods to accept composite format strings, allowing us to shorten this example slightly: Console.WriteLine ("Credit={0:C}", 500); // Credit=$500.00 You can also append a composite format string to a StringBuilder (via Append Format), and to a TextWriter for I/O (see Chapter 15). string.Format accepts an optional format provider. A simple application for this is to call ToString on an arbitrary object while passing in a format provider: 300 | Chapter 6: .NET Fundamentals
- 319. string s = string.Format (CultureInfo.InvariantCulture, "{0}", someObject); This is equivalent to the following: string s; if (someObject is IFormattable) s = ((IFormattable)someObject).ToString (null, CultureInfo.InvariantCulture); else if (someObject == null) s = ""; else s = someObject.ToString(); Parsing with format providers There’s no standard interface for parsing through a format provider. Instead, each participating type overloads its static Parse (and TryParse) method to accept a format provider and, optionally, a NumberStyles or DateTimeStyles enum. NumberStyles and DateTimeStyles control how parsing works: they let you specify such things as whether parentheses or a currency symbol can appear in the input string (by default, the answer to both questions is no). For example: int error = int.Parse ("(2)"); // Exception thrown int minusTwo = int.Parse ("(2)", NumberStyles.Integer | NumberStyles.AllowParentheses); // OK decimal fivePointTwo = decimal.Parse ("£5.20", NumberStyles.Currency, CultureInfo.GetCultureInfo ("en-GB")); The next section lists all NumberStyles and DateTimeStyles members as well as the default parsing rules for each type. IFormatProvider and ICustomFormatter All format providers implement IFormatProvider: public interface IFormatProvider { object GetFormat (Type formatType); } The purpose of this method is to provide indirection—this is what allows Culture Info to defer to an appropriate NumberFormatInfo or DateTimeInfo object to do the work. By implementing IFormatProvider—along with ICustomFormatter—you can also write your own format provider that works in conjunction with existing types. ICustomFormatter defines a single method, as follows: string Format (string format, object arg, IFormatProvider formatProvider); The following custom format provider writes numbers as words: public class WordyFormatProvider : IFormatProvider, ICustomFormatter { static readonly string[] _numberWords = Formatting and Parsing | 301 .NET Fundamentals
- 320. "zero one two three four five six seven eight nine minus point".Split(); IFormatProvider _parent; // Allows consumers to chain format providers public WordyFormatProvider () : this (CultureInfo.CurrentCulture) { } public WordyFormatProvider (IFormatProvider parent) => _parent = parent; public object GetFormat (Type formatType) { if (formatType == typeof (ICustomFormatter)) return this; return null; } public string Format (string format, object arg, IFormatProvider prov) { // If it's not our format string, defer to the parent provider: if (arg == null || format != "W") return string.Format (_parent, "{0:" + format + "}", arg); StringBuilder result = new StringBuilder(); string digitList = string.Format (CultureInfo.InvariantCulture, "{0}", arg); foreach (char digit in digitList) { int i = "0123456789-.".IndexOf (digit), StringComparison.InvariantCulture); if (i == -1) continue; if (result.Length > 0) result.Append (' '); result.Append (_numberWords[i]); } return result.ToString(); } } Notice that in the Format method, we used string.Format—with Invariant Culture—to convert the input number to a string. It would have been simpler just to call ToString() on arg, but then CurrentCulture would have been used instead. The reason for needing the invariant culture is evident a few lines later: int i = "0123456789-.".IndexOf (digit), StringComparison.InvariantCulture); It’s critical here that the number string comprises only the characters 0123456789-. and not any internationalized versions of these. Here’s an example of using WordyFormatProvider: double n = -123.45; IFormatProvider fp = new WordyFormatProvider(); Console.WriteLine (string.Format (fp, "{0:C} in words is {0:W}", n)); // -$123.45 in words is minus one two three point four five You can use custom format providers only in composite format strings. 302 | Chapter 6: .NET Fundamentals
- 321. Standard Format Strings and Parsing Flags The standard format strings control how a numeric type or DateTime/DateTime Offset is converted to a string. There are two kinds of format strings: Standard format strings With these, you provide general guidance. A standard format string consists of a single letter, followed, optionally, by a digit (whose meaning depends on the letter). An example is "C" or "F2". Custom format strings With these, you micromanage every character with a template. An example is "0:#.000E+00". Custom format strings are unrelated to custom format providers. Numeric Format Strings Table 6-2 lists all standard numeric format strings. Table 6-2. Standard numeric format strings Letter Meaning Sample input Result Notes G or g “General” 1.2345, "G" 0.00001, "G" 0.00001, "g" 1.2345, "G3" 12345, "G3" 1.2345 1E-05 1e-05 1.23 1.23E04 Switches to exponential notation for small or large numbers. G3 limits precision to three digits in total (before + after point). F Fixed point 2345.678, "F2" 2345.6, "F2" 2345.68 2345.60 F2 rounds to two decimal places. N Fixed point with group separator (“Numeric”) 2345.678, "N2" 2345.6, "N2" 2,345.68 2,345.60 As above, with group (1,000s) separator (details from format provider). D Pad with leading zeros 123, "D5" 123, "D1" 00123 123 For integral types only. D5 pads left to five digits; does not truncate. E or e Force exponential notation 56789, "E" 56789, "e" 56789, "E2" 5.678900E+004 5.678900e+004 5.68E+004 Six-digit default precision. C Currency 1.2, "C" 1.2, "C4" $1.20 $1.2000 C with no digit uses default number of D.P. from format provider. P Percent .503, "P" .503, "P0" 50.30% 50% Uses symbol and layout from format provider. Decimal places can optionally be overridden. Standard Format Strings and Parsing Flags | 303 .NET Fundamentals
- 322. Letter Meaning Sample input Result Notes X or x Hexadecimal 47, "X" 47, "x" 47, "X4" 2F 2f 002F X for uppercase hex digits; x for lowercase hex digits. Integrals only. R or G9/G17 Round-trip 1f / 3f, "R" 0.333333343 Use R for BigInteger, G17 for double, or G9 for float. Supplying no numeric format string (or a null or blank string) is equivalent to using the "G" standard format string followed by no digit. This exhibits the following behavior: • Numbers smaller than 10−4 or larger than the type’s precision are expressed in • exponential (scientific) notation. • The two decimal places at the limit of float or double’s precision are rounded • away to mask the inaccuracies inherent in conversion to decimal from their underlying binary form. The automatic rounding just described is usually beneficial and goes unnoticed. However, it can cause trouble if you need to round-trip a number; in other words, convert it to a string and back again (maybe repeatedly) while preserving value equality. For this reason, the R, G17, and G9 format strings exist to circumvent this implicit rounding. Table 6-3 lists custom numeric format strings. Table 6-3. Custom numeric format strings Specifier Meaning Sample input Result Notes # Digit placeholder 12.345, ".##" 12.345, ".####" 12.35 12.345 Limits digits after D.P. 0 Zero placeholder 12.345, ".00" 12.345, ".0000" 99, "000.00" 12.35 12.3450 099.00 As above, but also pads with zeros before and after D.P. . Decimal point Indicates D.P. Actual symbol comes from NumberFormatInfo. , Group separator 1234, "#,###,###" 1234, "0,000,000" 1,234 0,001,234 Symbol comes from Number FormatInfo. , (as above) Multiplier 1000000, "#," 1000000, "#,, 1000 1 If comma is at end or before D.P., it acts as a multiplier—dividing result by 1,000, 1,000,000, etc. 304 | Chapter 6: .NET Fundamentals
- 323. Specifier Meaning Sample input Result Notes % Percent notation 0.6, "00%" 60% First multiplies by 100 and then substitutes percent symbol obtained from NumberFormatInfo. E0, e0, E+0, e+0 E-0, e-0 Exponent notation 1234, "0E0" 1234, "0E+0" 1234, "0.00E00" 1234, "0.00e00" 1E3 1E+3 1.23E03 1.23e03 Literal character quote 50, @"#0" #50 Use in conjunction with an @ prefix on the string—or use . 'xx''xx' Literal string quote 50, "0 '...'" 50 ... ; Section separator 15, "#;(#);zero" 15 (If positive.) -5, "#;(#);zero" (5) (If negative.) 0, "#;(#);zero" zero (If zero.) Any other char Literal 35.2, "$0 . 00c" $35 . 20c NumberStyles Each numeric type defines a static Parse method that accepts a NumberStyles argument. NumberStyles is a flags enum that lets you determine how the string is read as it’s converted to a numeric type. It has the following combinable members: AllowLeadingWhite AllowTrailingWhite AllowLeadingSign AllowTrailingSign AllowParentheses AllowDecimalPoint AllowThousands AllowExponent AllowCurrencySymbol AllowHexSpecifier NumberStyles also defines these composite members: None Integer Float Number HexNumber Currency Any Except for None, all composite values include AllowLeadingWhite and Allow TrailingWhite. Figure 6-1 shows their remaining makeup, with the most useful three emphasized. Standard Format Strings and Parsing Flags | 305 .NET Fundamentals
- 324. Figure 6-1. Composite NumberStyles When you call Parse without specifying any flags, the defaults illustrated in Fig‐ ure 6-2 are applied. Figure 6-2. Default parsing flags for numeric types If you don’t want the defaults shown in Figure 6-2, you must explicitly specify NumberStyles: int thousand = int.Parse ("3E8", NumberStyles.HexNumber); int minusTwo = int.Parse ("(2)", NumberStyles.Integer | NumberStyles.AllowParentheses); double aMillion = double.Parse ("1,000,000", NumberStyles.Any); decimal threeMillion = decimal.Parse ("3e6", NumberStyles.Any); decimal fivePointTwo = decimal.Parse ("$5.20", NumberStyles.Currency); Because we didn’t specify a format provider, this example works with your local currency symbol, group separator, decimal point, and so on. The next example is hardcoded to work with the euro sign and a blank group separator for currencies: 306 | Chapter 6: .NET Fundamentals
- 325. NumberFormatInfo ni = new NumberFormatInfo(); ni.CurrencySymbol = "€"; ni.CurrencyGroupSeparator = " "; double million = double.Parse ("€1 000 000", NumberStyles.Currency, ni); Date/Time Format Strings Format strings for DateTime/DateTimeOffset can be divided into two groups based on whether they honor culture and format provider settings. Table 6-4 lists those that do; Table 6-5 lists those that don’t. The sample output comes from formatting the following DateTime (with invariant culture, in the case of Table 6-4): new DateTime (2000, 1, 2, 17, 18, 19); Table 6-4. Culture-sensitive date/time format strings Format string Meaning Sample output d Short date 01/02/2000 D Long date Sunday, 02 January 2000 t Short time 17:18 T Long time 17:18:19 f Long date + short time Sunday, 02 January 2000 17:18 F Long date + long time Sunday, 02 January 2000 17:18:19 g Short date + short time 01/02/2000 17:18 G (default) Short date + long time 01/02/2000 17:18:19 m, M Month and day 02 January y, Y Year and month January 2000 Table 6-5. Culture-insensitive date/time format strings Format string Meaning Sample output Notes o Round- trippable 2000-01-02T17:18:19.0000000 Will append time zone information unless DateTimeKind is Unspecified. r, R RFC 1123 standard Sun, 02 Jan 2000 17:18:19 GMT You must explicitly convert to UTC with DateTime.ToUniversalTime. s Sortable; ISO 8601 2000-01-02T17:18:19 Compatible with text-based sorting. u “Universal” sortable 2000-01-02 17:18:19Z Similar to above; must explicitly convert to UTC. U UTC Sunday, 02 January 2000 17:18:19 Long date + short time, converted to UTC. Standard Format Strings and Parsing Flags | 307 .NET Fundamentals
- 326. The format strings "r", "R", and "u" emit a suffix that implies UTC; yet they don’t automatically convert a local to a UTC DateTime (so you must do the conversion yourself). Ironically, "U" automatically converts to UTC but doesn’t write a time zone suffix! In fact, "o" is the only format specifier in the group that can write an unambiguous DateTime without intervention. DateTimeFormatInfo also supports custom format strings: these are analogous to numeric custom format strings. The list is extensive and is available online in Microsoft’s documentation. Here’s an example of a custom format string: yyyy-MM-dd HH:mm:ss Parsing and misparsing DateTimes Strings that put the month or day first are ambiguous and can easily be misparsed— particularly if you have global customers. This is not a problem in user interface controls, because the same settings are in force when parsing as when formatting. But when writing to a file, for instance, day/month misparsing can be a real prob‐ lem. There are two solutions: • Always state the same explicit culture when formatting and parsing (e.g., invar‐ • iant culture). • Format DateTime and DateTimeOffsets in a manner independent of culture. • The second approach is more robust—particularly if you choose a format that puts the four-digit year first: such strings are much more difficult to misparse by another party. Further, strings formatted with a standards-compliant year-first format (such as "o") can parse correctly alongside locally formatted strings—rather like a “universal donor.” (Dates formatted with "s" or "u" have the further benefit of being sortable.) To illustrate, suppose that we generate a culture-insensitive DateTime string s as follows: string s = DateTime.Now.ToString ("o"); The "o" format string includes milliseconds in the output. The following custom format string gives the same result as "o" but without milliseconds: yyyy-MM-ddTHH:mm:ss K We can reparse this in two ways. ParseExact demands strict compliance with the specified format string: DateTime dt1 = DateTime.ParseExact (s, "o", null); (You can achieve a similar result with XmlConvert’s ToString and ToDateTime methods.) 308 | Chapter 6: .NET Fundamentals
- 327. Parse, however, implicitly accepts both the "o" format and the CurrentCulture format: DateTime dt2 = DateTime.Parse (s); This works with both DateTime and DateTimeOffset. ParseExact is usually preferable if you know the format of the string that you’re parsing. It means that if the string is incorrectly formatted, an exception will be thrown—which is usually better than risking a misparsed date. DateTimeStyles DateTimeStyles is a flags enum that provides additional instructions when calling Parse on a DateTime(Offset). Here are its members: None, AllowLeadingWhite, AllowTrailingWhite, AllowInnerWhite, AssumeLocal, AssumeUniversal, AdjustToUniversal, NoCurrentDateDefault, RoundTripKind There is also a composite member, AllowWhiteSpaces: AllowWhiteSpaces = AllowLeadingWhite | AllowTrailingWhite | AllowInnerWhite The default is None. This means that extra whitespace is normally prohibited (white‐ space that’s part of a standard DateTime pattern is exempt). AssumeLocal and AssumeUniversal apply if the string doesn’t have a time zone suffix (such as Z or +9:00). AdjustToUniversal still honors time zone suffixes but then converts to UTC using the current regional settings. If you parse a string comprising a time but no date, today’s date is applied by default. If you apply the NoCurrentDateDefault flag, however, it instead uses 1st January 0001. Enum Format Strings In “Enums” on page 140, we described formatting and parsing enum values. Table 6-6 lists each format string and the result of applying it to the following expression: Console.WriteLine (System.ConsoleColor.Red.ToString (formatString)); Standard Format Strings and Parsing Flags | 309 .NET Fundamentals
- 328. Table 6-6. Enum format strings Format string Meaning Sample output Notes G or g “General” Red Default F or f Treat as though Flags attribute were present Red Works on combined members even if enum has no Flags attribute D or d Decimal value 12 Retrieves underlying integral value X or x Hexadecimal value 0000000C Retrieves underlying integral value Other Conversion Mechanisms In the previous two sections, we covered format providers—.NET’s primary mech‐ anism for formatting and parsing. Other important conversion mechanisms are scattered through various types and namespaces. Some convert to and from string, and some do other kinds of conversions. In this section, we discuss the following topics: • The Convert class and its functions: • — Real to integral conversions that round rather than truncate — — Parsing numbers in base 2, 8, and 16 — — Dynamic conversions — — Base-64 translations — • XmlConvert and its role in formatting and parsing for XML • • Type converters and their role in formatting and parsing for designers and • XAML • BitConverter, for binary conversions • Convert .NET calls the following types base types: • bool, char, string, System.DateTime, and System.DateTimeOffset • • All the C# numeric types • The static Convert class defines methods for converting every base type to every other base type. Unfortunately, most of these methods are useless: either they throw exceptions or they are redundant alongside implicit casts. Among the clutter, however, are some useful methods, listed in the following sections. 310 | Chapter 6: .NET Fundamentals
- 329. All base types (explicitly) implement IConvertible, which defines methods for converting to every other base type. In most cases, the implementation of each of these methods simply calls a method in Convert. On rare occasions, it can be useful to write a method that accepts an argument of type IConvertible. Rounding real to integral conversions In Chapter 2, we saw how implicit and explicit casts allow you to convert between numeric types. In summary: • Implicit casts work for nonlossy conversions (e.g., int to double). • • Explicit casts are required for lossy conversions (e.g., double to int). • Casts are optimized for efficiency; hence, they truncate data that won’t fit. This can be a problem when converting from a real number to an integer, because often you want to round rather than truncate. Convert’s numerical conversion methods address just this issue—they always round: double d = 3.9; int i = Convert.ToInt32 (d); // i == 4 Convert uses banker’s rounding, which snaps midpoint values to even integers (this prevents positive or negative bias). If banker’s rounding is a problem, first call Math.Round on the real number: this accepts an additional argument that allows you to control midpoint rounding. Parsing numbers in base 2, 8, and 16 Hidden among the To(integral-type) methods are overloads that parse numbers in another base: int thirty = Convert.ToInt32 ("1E", 16); // Parse in hexadecimal uint five = Convert.ToUInt32 ("101", 2); // Parse in binary The second argument specifies the base. It can be any base you like—as long as it’s 2, 8, 10, or 16! Dynamic conversions Occasionally, you need to convert from one type to another, but you don’t know what the types are until runtime. For this, the Convert class provides a ChangeType method: public static object ChangeType (object value, Type conversionType); The source and target types must be one of the “base” types. ChangeType also accepts an optional IFormatProvider argument. Here’s an example: Type targetType = typeof (int); object source = "42"; Other Conversion Mechanisms | 311 .NET Fundamentals
- 330. object result = Convert.ChangeType (source, targetType); Console.WriteLine (result); // 42 Console.WriteLine (result.GetType()); // System.Int32 An example of when this might be useful is in writing a deserializer that can work with multiple types. It can also convert any enum to its integral type (see “Enums” on page 140). A limitation of ChangeType is that you cannot specify a format string or parsing flag. Base-64 conversions Sometimes, you need to include binary data such as a bitmap within a text docu‐ ment, such as an XML file or email message. Base 64 is a ubiquitous means of encoding binary data as readable characters, using 64 characters from the ASCII set. Convert’s ToBase64String method converts from a byte array to base 64; From Base64String does the reverse. XmlConvert If you’re dealing with data that’s originated from or destined for an XML file, XmlConvert (in the System.Xml namespace) provides the most suitable methods for formatting and parsing. The methods in XmlConvert handle the nuances of XML formatting without needing special format strings. For instance, true in XML is “true” and not “True.” The .NET BCL internally uses XmlConvert extensively. XmlConvert is also good for general-purpose culture-independent serialization. The formatting methods in XmlConvert are all provided as overloaded ToString methods; the parsing methods are called ToBoolean, ToDateTime, and so on: string s = XmlConvert.ToString (true); // s = "true" bool isTrue = XmlConvert.ToBoolean (s); The methods that convert to and from DateTime accept an XmlDateTimeSerializa tionMode argument. This is an enum with the following values: Unspecified, Local, Utc, RoundtripKind Local and Utc cause a conversion to take place when formatting (if the DateTime is not already in that time zone). The time zone is then appended to the string: 2010-02-22T14:08:30.9375 // Unspecified 2010-02-22T14:07:30.9375+09:00 // Local 2010-02-22T05:08:30.9375Z // Utc Unspecified strips away any time-zone information embedded in the DateTime (i.e., DateTimeKind) before formatting. RoundtripKind honors the DateTime’s Date TimeKind—so when it’s reparsed, the resultant DateTime struct will be exactly as it was originally. 312 | Chapter 6: .NET Fundamentals
- 331. Type Converters Type converters are designed to format and parse in design-time environments. They also parse values in Extensible Application Markup Language (XAML) docu‐ ments—as used in Windows Presentation Foundation (WPF). In .NET, there are more than 100 type converters—covering such things as colors, images, and URIs. In contrast, format providers are implemented for only a handful of simple value types. Type converters typically parse strings in a variety of ways—without needing hints. For instance, in a WPF application in Visual Studio, if you assign a control a background color by typing "Beige" into the appropriate property window, Color’s type converter figures out that you’re referring to a color name and not an RGB string or system color. This flexibility can sometimes make type converters useful in contexts outside of designers and XAML documents. All type converters subclass TypeConverter in System.ComponentModel. To obtain a TypeConverter, call TypeDescriptor.GetConverter. The following obtains a Type Converter for the Color type (in the System.Drawing namespace): TypeConverter cc = TypeDescriptor.GetConverter (typeof (Color)); Among many other methods, TypeConverter defines methods to ConvertToString and ConvertFromString. We can call these as follows: Color beige = (Color) cc.ConvertFromString ("Beige"); Color purple = (Color) cc.ConvertFromString ("#800080"); Color window = (Color) cc.ConvertFromString ("Window"); By convention, type converters have names ending in Converter and are usually in the same namespace as the type they’re converting. A type links to its converter via a TypeConverterAttribute, allowing designers to pick up converters automatically. Type converters can also provide design-time services such as generating standard value lists for populating a drop-down list in a designer or assisting with code serialization. BitConverter Most base types can be converted to a byte array, by calling BitConverter.GetBytes: foreach (byte b in BitConverter.GetBytes (3.5)) Console.Write (b + " "); // 0 0 0 0 0 0 12 64 BitConverter also provides methods, such as ToDouble, for converting in the other direction. The decimal and DateTime(Offset) types are not supported by BitConverter. You can, however, convert a decimal to an int array by calling decimal.GetBits. To go the other way around, decimal provides a constructor that accepts an int array. Other Conversion Mechanisms | 313 .NET Fundamentals
- 332. In the case of DateTime, you can call ToBinary on an instance—this returns a long (upon which you can then use BitConverter). The static DateTime.FromBinary method does the reverse. Globalization There are two aspects to internationalizing an application: globalization and localization. Globalization is concerned with three tasks (in decreasing order of importance): 1. Making sure that your program doesn’t break when run in another culture 1. 2. Respecting a local culture’s formatting rules; for instance, when displaying dates 2. 3. Designing your program so that it picks up culture-specific data and strings 3. from satellite assemblies that you can later write and deploy Localization means concluding that last task by writing satellite assemblies for specific cultures. You can do this after writing your program (we cover the details in “Resources and Satellite Assemblies” on page 752). .NET helps you with the second task by applying culture-specific rules by default. We’ve already seen how calling ToString on a DateTime or number respects local formatting rules. Unfortunately, this makes it easy to fail the first task and have your program break because you’re expecting dates or numbers to be formatted according to an assumed culture. The solution, as we’ve seen, is either to specify a culture (such as the invariant culture) when formatting and parsing or to use culture-independent methods such as those in XmlConvert. Globalization Checklist We’ve already covered the important points in this chapter. Here’s a summary of the essential work required: • Understand Unicode and text encodings (see “Text Encodings and Unicode” on • page 281). • Be mindful that methods such as ToUpper and ToLower on char and string are • culture sensitive: use ToUpperInvariant/ToLowerInvariant unless you want culture sensitivity. • Favor culture-independent formatting and parsing mechanisms for DateTime • and DateTimeOffsets such as ToString("o") and XmlConvert. • Otherwise, specify a culture when formatting/parsing numbers or date/times • (unless you want local-culture behavior). 314 | Chapter 6: .NET Fundamentals
- 333. Testing You can test against different cultures by reassigning Thread’s CurrentCulture property (in System.Threading). The following changes the current culture to Turkey: Thread.CurrentThread.CurrentCulture = CultureInfo.GetCultureInfo ("tr-TR"); Turkey is a particularly good test case because: • "i".ToUpper() != "I" and "I".ToLower() != "i". • • Dates are formatted as day.month.year (note the period separator). • • The decimal point indicator is a comma instead of a period. • You can also experiment by changing the number and date formatting settings in the Windows Control Panel: these are reflected in the default culture (Culture Info.CurrentCulture). CultureInfo.GetCultures() returns an array of all available cultures. Thread and CultureInfo also support a CurrentUICulture property. This is concerned more with localization, which we cover in Chapter 17. Working with Numbers Conversions We covered numeric conversions in previous chapters and sections; Table 6-7 sum‐ marizes all of the options. Table 6-7. Summary of numeric conversions Task Functions Examples e Parse TryParse double d = double.Parse ("3.5"); int i; bool ok = int.TryParse ("3", out i); Parsing from base 2, 8, or 16 Convert.ToIntegral int i = Convert.ToInt32 ("1E", 16); Formatting to hexadecimal ToString ("X") string hex = 45.ToString ("X"); Lossless numeric conversion Implicit cast int i = 23; double d = i; Truncating numeric conversion Explicit cast double d = 23.5; int i = (int) d; Working with Numbers | 315 .NET Fundamentals
- 334. Task Functions Examples Rounding numeric conversion (real to integral) Convert.ToIntegral double d = 23.5; int i = Convert.ToInt32 (d); Math Table 6-8 lists the key members of the static Math class. The trigonometric functions accept arguments of type double; other methods such as Max are overloaded to operate on all numeric types. The Math class also defines the mathematical constants E (e) and PI. Table 6-8. Methods in the static Math class Category Methods Rounding Round, Truncate, Floor, Ceiling Maximum/minimum Max, Min Absolute value and sign Abs, Sign Square root Sqrt Raising to a power Pow, Exp Logarithm Log, Log10 Trigonometric Sin, Cos, Tan, Sinh, Cosh, Tanh, Asin, Acos, Atan The Round method lets you specify the number of decimal places with which to round as well as how to handle midpoints (away from zero, or with banker’s rounding). Floor and Ceiling round to the nearest integer: Floor always rounds down, and Ceiling always rounds up—even with negative numbers. Max and Min accept only two arguments. If you have an array or sequence of numbers, use the Max and Min extension methods in System.Linq.Enumerable. BigInteger The BigInteger struct is a specialized numeric type. It resides in the System.Numer ics namespace and allows you to represent an arbitrarily large integer without any loss of precision. C# doesn’t provide native support for BigInteger, so there’s no way to represent BigInteger literals. You can, however, implicitly convert from any other integral type to a BigInteger: BigInteger twentyFive = 25; // implicit conversion from integer 316 | Chapter 6: .NET Fundamentals
- 335. To represent a bigger number, such as one googol (10100 ), you can use one of BigInteger’s static methods, such as Pow (raise to the power): BigInteger googol = BigInteger.Pow (10, 100); Alternatively, you can Parse a string: BigInteger googol = BigInteger.Parse ("1".PadRight (101, '0')); Calling ToString() on this prints every digit: Console.WriteLine (googol.ToString()); // 10000000000000000000000000000 00000000000000000000000000000000000000000000000000000000000000000000000 You can perform potentially lossy conversions between BigInteger and the stan‐ dard numeric types by using the explicit cast operator: double g2 = (double) googol; // Explicit cast BigInteger g3 = (BigInteger) g2; // Explicit cast Console.WriteLine (g3); The output from this demonstrates the loss of precision: 9999999999999999673361688041166912... BigInteger overloads all the arithmetic operators including remainder (%) as well as the comparison and equality operators. You can also construct a BigInteger from a byte array. The following code gener‐ ates a 32-byte random number suitable for cryptography and then assigns it to a BigInteger: // This uses the System.Security.Cryptography namespace: RandomNumberGenerator rand = RandomNumberGenerator.Create(); byte[] bytes = new byte [32]; rand.GetBytes (bytes); var bigRandomNumber = new BigInteger (bytes); // Convert to BigInteger The advantage of storing such a number in a BigInteger over a byte array is that you get value-type semantics. Calling ToByteArray converts a BigInteger back to a byte array. Half The Half struct is a 16-bit floating point type and was introduced with .NET 5. Half is intended mainly for interoperating with graphics card processors and does not have native support in most CPUs. You can convert between Half and float or double via an explicit cast: Half h = (Half) 123.456; Console.WriteLine (h); // 123.44 (note loss of precision) There are no arithmetic operations defined for this type, so you must convert to another type such as float or double in order to perform calculations. Working with Numbers | 317 .NET Fundamentals
- 336. Half has a range of –65500 to 65500: Console.WriteLine (Half.MinValue); // -65500 Console.WriteLine (Half.MaxValue); // 65500 Note the loss of precision at the maximum range: Console.WriteLine ((Half)65500); // 65500 Console.WriteLine ((Half)65490); // 65500 Console.WriteLine ((Half)65480); // 65470 Complex The Complex struct is another specialized numeric type that represents complex numbers with real and imaginary components of type double. Complex resides in the namespace (along with BigInteger). To use Complex, instantiate the struct, specifying the real and imaginary values: var c1 = new Complex (2, 3.5); var c2 = new Complex (3, 0); There are also implicit conversions from the standard numeric types. The Complex struct exposes properties for the real and imaginary values as well as the phase and magnitude: Console.WriteLine (c1.Real); // 2 Console.WriteLine (c1.Imaginary); // 3.5 Console.WriteLine (c1.Phase); // 1.05165021254837 Console.WriteLine (c1.Magnitude); // 4.03112887414927 You can also construct a Complex number by specifying magnitude and phase: Complex c3 = Complex.FromPolarCoordinates (1.3, 5); The standard arithmetic operators are overloaded to work on Complex numbers: Console.WriteLine (c1 + c2); // (5, 3.5) Console.WriteLine (c1 * c2); // (6, 10.5) The Complex struct exposes static methods for more advanced functions, including the following: • Trigonometric (Sin, Asin, Sinh, Tan, etc.) • • Logarithms and exponentiations • • Conjugate • 318 | Chapter 6: .NET Fundamentals
- 337. Random The Random class generates a pseudorandom sequence of random bytes, integers, or doubles. To use Random, you first instantiate it, optionally providing a seed to initiate the random number series. Using the same seed guarantees the same series of numbers (if run under the same CLR version), which is sometimes useful when you want reproducibility: Random r1 = new Random (1); Random r2 = new Random (1); Console.WriteLine (r1.Next (100) + ", " + r1.Next (100)); // 24, 11 Console.WriteLine (r2.Next (100) + ", " + r2.Next (100)); // 24, 11 If you don’t want reproducibility, you can construct Random with no seed; in that case, it uses the current system time to make one up. Because the system clock has limited granularity, two Random instances created close together (typically within 10 ms) will yield the same sequence of values. A common trap is to instantiate a new Random object every time you need a random number rather than reusing the same object. A good pattern is to declare a single static Random instance. In multithreaded scenarios, however, this can cause trouble because Random objects are not thread-safe. We describe a workaround in “Thread-Local Storage” on page 898. Calling Next(n) generates a random integer between 0 and n−1. NextDouble gener‐ ates a random double between 0 and 1. NextBytes fills a byte array with random values. Random is not considered random enough for high-security applications such as cryptography. For this, .NET provides a cryptographically strong random number generator, in the System.Security.Cryptography namespace. Here’s how to use it: var rand = System.Security.Cryptography.RandomNumberGenerator.Create(); byte[] bytes = new byte [32]; rand.GetBytes (bytes); // Fill the byte array with random numbers. The downside is that it’s less flexible: filling a byte array is the only means of obtaining random numbers. To obtain an integer, you must use BitConverter: byte[] bytes = new byte [4]; rand.GetBytes (bytes); int i = BitConverter.ToInt32 (bytes, 0); Working with Numbers | 319 .NET Fundamentals
- 338. BitOperations The System.Numerics.BitOperations class (from .NET 6) exposes the following methods to help with base-2 operations: IsPow2 Returns true if a number is a power of 2 LeadingZeroCount/TrailingZeroCount Returns the number of leading zeros, when formatted as a base-2 32-bit or 64-bit unsigned integer Log2 Returns the integer base-2 log of an unsigned integer PopCount Returns the number of bits set to 1 in an unsigned integer RotateLeft/RotateRight Performs a bitwise left/right rotation RoundUpToPowerOf2 Rounds an unsigned integer up to the closest power of 2 Enums In Chapter 3, we described C#’s enum type and showed how to combine members, test equality, use logical operators, and perform conversions. .NET extends C#’s support for enums through the System.Enum type. This type has two roles: • Providing type unification for all enum types • • Defining static utility methods • Type unification means that you can implicitly cast any enum member to a System.Enum instance: Display (Nut.Macadamia); // Nut.Macadamia Display (Size.Large); // Size.Large void Display (Enum value) { Console.WriteLine (value.GetType().Name + "." + value.ToString()); } enum Nut { Walnut, Hazelnut, Macadamia } enum Size { Small, Medium, Large } The static utility methods on System.Enum are primarily related to performing conversions and obtaining lists of members. 320 | Chapter 6: .NET Fundamentals
- 339. Enum Conversions There are three ways to represent an enum value: • As an enum member • • As its underlying integral value • • As a string • In this section, we describe how to convert between each. Enum to integral conversions Recall that an explicit cast converts between an enum member and its integral value. An explicit cast is the correct approach if you know the enum type at compile time: [Flags] public enum BorderSides { Left=1, Right=2, Top=4, Bottom=8 } ... int i = (int) BorderSides.Top; // i == 4 BorderSides side = (BorderSides) i; // side == BorderSides.Top You can cast a System.Enum instance to its integral type in the same way. The trick is to first cast to an object and then the integral type: static int GetIntegralValue (Enum anyEnum) { return (int) (object) anyEnum; } This relies on you knowing the integral type: the method we just wrote would crash if passed an enum whose integral type was long. To write a method that works with an enum of any integral type, you can take one of three approaches. The first is to call Convert.ToDecimal: static decimal GetAnyIntegralValue (Enum anyEnum) { return Convert.ToDecimal (anyEnum); } This works because every integral type (including ulong) can be converted to decimal without loss of information. The second approach is to call Enum.GetUnder lyingType in order to obtain the enum’s integral type, and then call Convert.Change Type: static object GetBoxedIntegralValue (Enum anyEnum) { Type integralType = Enum.GetUnderlyingType (anyEnum.GetType()); return Convert.ChangeType (anyEnum, integralType); } This preserves the original integral type, as the following example shows: Enums | 321 .NET Fundamentals
- 340. object result = GetBoxedIntegralValue (BorderSides.Top); Console.WriteLine (result); // 4 Console.WriteLine (result.GetType()); // System.Int32 Our GetBoxedIntegralType method in fact performs no value conversion; rather, it reboxes the same value in another type. It translates an integral value in enum-type clothing to an inte‐ gral value in integral-type clothing. We describe this further in “How Enums Work” on page 323. The third approach is to call Format or ToString specifying the "d" or "D" format string. This gives you the enum’s integral value as a string, and it is useful when writing custom serialization formatters: static string GetIntegralValueAsString (Enum anyEnum) { return anyEnum.ToString ("D"); // returns something like "4" } Integral to enum conversions Enum.ToObject converts an integral value to an enum instance of the given type: object bs = Enum.ToObject (typeof (BorderSides), 3); Console.WriteLine (bs); // Left, Right This is the dynamic equivalent of the following: BorderSides bs = (BorderSides) 3; ToObject is overloaded to accept all integral types as well as object. (The latter works with any boxed integral type.) String conversions To convert an enum to a string, you can either call the static Enum.Format method or call ToString on the instance. Each method accepts a format string, which can be "G" for default formatting behavior, "D" to emit the underlying integral value as a string, "X" for the same in hexadecimal, or "F" to format combined members of an enum without the Flags attribute. We listed examples of these in “Standard Format Strings and Parsing Flags” on page 303. Enum.Parse converts a string to an enum. It accepts the enum type and a string that can include multiple members: BorderSides leftRight = (BorderSides) Enum.Parse (typeof (BorderSides), "Left, Right"); An optional third argument lets you perform case-insensitive parsing. An Argument Exception is thrown if the member is not found. Enumerating Enum Values Enum.GetValues returns an array comprising all members of a particular enum type: 322 | Chapter 6: .NET Fundamentals
- 341. foreach (Enum value in Enum.GetValues (typeof (BorderSides))) Console.WriteLine (value); Composite members such as LeftRight = Left | Right are included, too. Enum.GetNames performs the same function but returns an array of strings. Internally, the CLR implements GetValues and GetNames by reflecting over the fields in the enum’s type. The results are cached for efficiency. How Enums Work The semantics of enums are enforced largely by the compiler. In the CLR, there’s no runtime difference between an enum instance (when unboxed) and its underly‐ ing integral value. Further, an enum definition in the CLR is merely a subtype of System.Enum with static integral-type fields for each member. This makes the ordinary use of an enum highly efficient, with a runtime cost matching that of integral constants. The downside of this strategy is that enums can provide static but not strong type safety. We saw an example of this in Chapter 3: [Flags] public enum BorderSides { Left=1, Right=2, Top=4, Bottom=8 } ... BorderSides b = BorderSides.Left; b += 1234; // No error! When the compiler is unable to perform validation (as in this example), there’s no backup from the runtime to throw an exception. What we said about there being no runtime difference between an enum instance and its integral value might seem at odds with the following: [Flags] public enum BorderSides { Left=1, Right=2, Top=4, Bottom=8 } ... Console.WriteLine (BorderSides.Right.ToString()); // Right Console.WriteLine (BorderSides.Right.GetType().Name); // BorderSides Given the nature of an enum instance at runtime, you’d expect this to print 2 and Int32! The reason for its behavior is down to some more compile-time trickery. C# explicitly boxes an enum instance before calling its virtual methods—such as ToString or GetType. And when an enum instance is boxed, it gains a runtime wrapping that references its enum type. The Guid Struct The Guid struct represents a globally unique identifier: a 16-byte value that, when generated, is almost certainly unique in the world. Guids are often used for keys of various sorts, in applications and databases. There are 2128 or 3.4 × 1038 unique Guids. The Guid Struct | 323 .NET Fundamentals
- 342. The static Guid.NewGuid method generates a unique Guid: Guid g = Guid.NewGuid (); Console.WriteLine (g.ToString()); // 0d57629c-7d6e-4847-97cb-9e2fc25083fe To instantiate an existing value, you use one of the constructors. The two most useful constructors are: public Guid (byte[] b); // Accepts a 16-byte array public Guid (string g); // Accepts a formatted string When represented as a string, a Guid is formatted as a 32-digit hexadecimal number, with optional hyphens after the 8th, 12th, 16th, and 20th digits. The whole string can also be optionally wrapped in brackets or braces: Guid g1 = new Guid ("{0d57629c-7d6e-4847-97cb-9e2fc25083fe}"); Guid g2 = new Guid ("0d57629c7d6e484797cb9e2fc25083fe"); Console.WriteLine (g1 == g2); // True Being a struct, a Guid honors value-type semantics; hence, the equality operator works in the preceding example. The ToByteArray method converts a Guid to a byte array. The static Guid.Empty property returns an empty Guid (all zeros). This is often used in place of null. Equality Comparison Until now, we’ve assumed that the == and != operators are all there is to equality comparison. The issue of equality, however, is more complex and subtler, some‐ times requiring the use of additional methods and interfaces. This section explores the standard C# and .NET protocols for equality, focusing particularly on two questions: • When are == and != adequate—and inadequate—for equality comparison, and • what are the alternatives? • How and when should you customize a type’s equality logic? • But before exploring the details of equality protocols and how to customize them, we first must look at the preliminary concept of value versus referential equality. Value Versus Referential Equality There are two kinds of equality: Value equality Two values are equivalent in some sense. Referential equality Two references refer to exactly the same object. 324 | Chapter 6: .NET Fundamentals
- 343. Unless overridden: • Value types use value equality. • • Reference types use referential equality. (This is overridden with anonymous • types and records.) Value types, in fact, can use only value equality (unless boxed). A simple demonstra‐ tion of value equality is to compare two numbers: int x = 5, y = 5; Console.WriteLine (x == y); // True (by virtue of value equality) A more elaborate demonstration is to compare two DateTimeOffset structs. The following prints True because the two DateTimeOffsets refer to the same point in time and so are considered equivalent: var dt1 = new DateTimeOffset (2010, 1, 1, 1, 1, 1, TimeSpan.FromHours(8)); var dt2 = new DateTimeOffset (2010, 1, 1, 2, 1, 1, TimeSpan.FromHours(9)); Console.WriteLine (dt1 == dt2); // True DateTimeOffset is a struct whose equality semantics have been tweaked. By default, structs exhibit a special kind of value equality called structural equality in which two values are considered equal if all of their members are equal. (You can see this by creating a struct and calling its Equals method; more on this later.) Reference types exhibit referential equality by default. In the following example, f1 and f2 are not equal, despite their objects having identical content: class Foo { public int X; } ... Foo f1 = new Foo { X = 5 }; Foo f2 = new Foo { X = 5 }; Console.WriteLine (f1 == f2); // False In contrast, f3 and f1 are equal because they reference the same object: Foo f3 = f1; Console.WriteLine (f1 == f3); // True Later in this section, we explain how you can customize reference types to exhibit value equality. An example of this is the Uri class in the System namespace: Uri uri1 = new Uri ("http://www.linqpad.net"); Uri uri2 = new Uri ("http://www.linqpad.net"); Console.WriteLine (uri1 == uri2); // True The string class exhibits similar behavior: var s1 = "http://www.linqpad.net"; var s2 = "http://" + "www.linqpad.net"; Console.WriteLine (s1 == s2); // True Equality Comparison | 325 .NET Fundamentals
- 344. Standard Equality Protocols There are three standard protocols that types can implement for equality comparison: • The == and != operators • • The virtual Equals method in object • • The IEquatable<T> interface • In addition, there are the pluggable protocols and the IStructuralEquatable inter‐ face, which we describe in Chapter 7. == and != We’ve already seen in many examples how the standard == and != operators per‐ form equality/inequality comparisons. The subtleties with == and != arise because they are operators; thus, they are statically resolved (in fact, they are implemented as static functions). So, when you use == or !=, C# makes a compile-time decision as to which type will perform the comparison, and no virtual behavior comes into play. This is normally desirable. In the following example, the compiler hardwires == to the int type because x and y are both int: int x = 5; int y = 5; Console.WriteLine (x == y); // True But in the next example, the compiler wires the == operator to the object type: object x = 5; object y = 5; Console.WriteLine (x == y); // False Because object is a class (and so a reference type), object’s == operator uses referential equality to compare x and y. The result is false because x and y each refer to different boxed objects on the heap. The virtual Object.Equals method To correctly equate x and y in the preceding example, we can use the virtual Equals method. Equals is defined in System.Object and so is available to all types: object x = 5; object y = 5; Console.WriteLine (x.Equals (y)); // True Equals is resolved at runtime—according to the object’s actual type. In this case, it calls Int32’s Equals method, which applies value equality to the operands, returning true. With reference types, Equals performs referential equality comparison by default; with structs, Equals performs structural comparison by calling Equals on each of its fields. 326 | Chapter 6: .NET Fundamentals
- 345. Why the Complexity? You might wonder why the designers of C# didn’t avoid the problem by making == virtual and thus functionally identical to Equals. There are three reasons for this: • If the first operand is null, Equals fails with a NullReferenceException; a static • operator does not. • Because the == operator is statically resolved, it executes extremely quickly. • This means that you can write computationally intensive code without pen‐ alty—and without needing to learn another language such as C++. • Sometimes it can be useful to have == and Equals apply different definitions of • equality. We describe this scenario later in this section. Essentially, the complexity of the design reflects the complexity of the situation: the concept of equality covers a multitude of scenarios. Hence, Equals is suitable for equating two objects in a type-agnostic fashion. The following method equates two objects of any type: public static bool AreEqual (object obj1, object obj2) => obj1.Equals (obj2); There is one case, however, in which this fails. If the first argument is null, you get a NullReferenceException. Here’s the fix: public static bool AreEqual (object obj1, object obj2) { if (obj1 == null) return obj2 == null; return obj1.Equals (obj2); } Or, more succinctly: public static bool AreEqual (object obj1, object obj2) => obj1 == null ? obj2 == null : obj1.Equals (obj2); The static object.Equals method The object class provides a static helper method that does the work of AreEqual in the preceding example. Its name is Equals—just like the virtual method—but there’s no conflict because it accepts two arguments: public static bool Equals (object objA, object objB) This provides a null-safe equality comparison algorithm for when the types are unknown at compile time: Equality Comparison | 327 .NET Fundamentals
- 346. object x = 3, y = 3; Console.WriteLine (object.Equals (x, y)); // True x = null; Console.WriteLine (object.Equals (x, y)); // False y = null; Console.WriteLine (object.Equals (x, y)); // True A useful application is when writing generic types. The following code will not compile if object.Equals is replaced with the == or != operator: class Test <T> { T _value; public void SetValue (T newValue) { if (!object.Equals (newValue, _value)) { _value = newValue; OnValueChanged(); } } protected virtual void OnValueChanged() { ... } } Operators are prohibited here because the compiler cannot bind to the static method of an unknown type. A more elaborate way to implement this comparison is with the EqualityComparer<T> class. This has the advantage of preventing boxing: if (!EqualityComparer<T>.Default.Equals (newValue, _value)) We discuss EqualityComparer<T> in more detail in Chapter 7 (see “Plugging in Equality and Order” on page 388). The static object.ReferenceEquals method Occasionally, you need to force referential equality comparison. The static object.ReferenceEquals method does just that: Widget w1 = new Widget(); Widget w2 = new Widget(); Console.WriteLine (object.ReferenceEquals (w1, w2)); // False class Widget { ... } You might want to do this because it’s possible for Widget to override the virtual Equals method such that w1.Equals(w2) would return true. Further, it’s possible for Widget to overload the == operator so that w1==w2 would also return true. In such cases, calling object.ReferenceEquals guarantees normal referential equality semantics. 328 | Chapter 6: .NET Fundamentals
- 347. Another way to force referential equality comparison is to cast the values to object and then apply the == operator. The IEquatable<T> interface A consequence of calling object.Equals is that it forces boxing on value types. This is undesirable in highly performance-sensitive scenarios because boxing is relatively expensive compared to the actual comparison. A solution was introduced in C# 2.0, with the IEquatable<T> interface: public interface IEquatable<T> { bool Equals (T other); } The idea is that IEquatable<T>, when implemented, gives the same result as calling object’s virtual Equals method—but more quickly. Most basic .NET types imple‐ ment IEquatable<T>. You can use IEquatable<T> as a constraint in a generic type: class Test<T> where T : IEquatable<T> { public bool IsEqual (T a, T b) { return a.Equals (b); // No boxing with generic T } } If we remove the generic constraint, the class would still compile, but a.Equals(b) would instead bind to the slower object.Equals (slower assuming T was a value type). When Equals and == are not equal We said earlier that it’s sometimes useful for == and Equals to apply different definitions of equality. For example: double x = double.NaN; Console.WriteLine (x == x); // False Console.WriteLine (x.Equals (x)); // True The double type’s == operator enforces that one NaN can never equal anything else—even another NaN. This is most natural from a mathematical perspective, and it reflects the underlying CPU behavior. The Equals method, however, is obliged to apply reflexive equality; in other words: x.Equals (x) must always return true. Collections and dictionaries rely on Equals behaving this way; otherwise, they could not find an item they previously stored. Having Equals and == apply different definitions of equality is actually quite rare with value types. A more common scenario is with reference types; this happens Equality Comparison | 329 .NET Fundamentals
- 348. when the author customizes Equals so that it performs value equality while leaving == to perform (default) referential equality. The StringBuilder class does exactly that: var sb1 = new StringBuilder ("foo"); var sb2 = new StringBuilder ("foo"); Console.WriteLine (sb1 == sb2); // False (referential equality) Console.WriteLine (sb1.Equals (sb2)); // True (value equality) Let’s now look at how to customize equality. Equality and Custom Types Recall default equality comparison behavior: • Value types use value equality. • • Reference types use referential equality unless overridden (as is the case with • anonymous types and records). Further: • A struct’s Equals method applies structural value equality by default (i.e., it • compares each field in the struct). Sometimes, it makes sense to override this behavior when writing a type. There are two cases for doing so: • To change the meaning of equality • • To speed up equality comparisons for structs • Changing the meaning of equality Changing the meaning of equality makes sense when the default behavior of == and Equals is unnatural for your type and is not what a consumer would expect. An example is DateTimeOffset, a struct with two private fields: a UTC DateTime and a numeric integer offset. If you were writing this type, you’d probably want to ensure that equality comparisons considered only the UTC DateTime field and not the offset field. Another example is numeric types that support NaN values such as float and double. If you were implementing such types yourself, you’d want to ensure that NaN-comparison logic was supported in equality comparisons. With classes, it’s sometimes more natural to offer value equality as the default instead of referential equality. This is often the case with small classes that hold a simple piece of data, such as System.Uri (or System.String). With records, the compiler automatically implements structural equality (by com‐ paring each field). Sometimes, however, this will include fields that you don’t want to compare, or objects that require special comparison logic, such as collections. 330 | Chapter 6: .NET Fundamentals
- 349. The process of overriding equality with records is slightly different because records follow a special pattern that’s designed to play well with its rules for inheritance. Speeding up equality comparisons with structs The default structural equality comparison algorithm for structs is relatively slow. Taking over this process by overriding Equals can improve performance by a factor of five. Overloading the == operator and implementing IEquatable<T> allows unboxed equality comparisons, and this can speed things up by a factor of five again. Overriding equality semantics for reference types doesn’t ben‐ efit performance. The default algorithm for referential equal‐ ity comparison is already very fast because it simply compares two 32- or 64-bit references. There’s another, rather peculiar case for customizing equality, and that’s to improve a struct’s hashing algorithm for better performance in a hashtable. This comes as a result of the fact that equality comparison and hashing are joined at the hip. We examine hashing in a moment. How to override equality semantics To override equality with classes or structs, here are the steps: 1. Override GetHashCode() and Equals(). 1. 2. (Optionally) overload != and ==. 2. 3. (Optionally) implement IEquatable<T>. 3. The process is different (and simpler) with records because the compiler already overrides the equality methods and operators in line with its own special pattern. If you want to intervene, you must conform to this pattern, which means writing an Equals method with a signature like this: record Test (int X, int Y) { public virtual bool Equals (Test t) => t != null && t.X == X && t.Y == Y; } Notice that Equals is virtual (not override) and accepts the actual record type (Test in this case, and not object). The compiler will recognize that your method has the “correct” signature and will patch it in. You must also override GetHashCode(), just as you would with classes or structs. You don’t need to (and shouldn’t) overload != and ==, or implement IEquatable<T>, because this is already done for you. Equality Comparison | 331 .NET Fundamentals
- 350. Overriding GetHashCode It might seem odd that System.Object—with its small footprint of members— defines a method with a specialized and narrow purpose. GetHashCode is a virtual method in Object that fits this description; it exists primarily for the benefit of just the following two types: System.Collections.Hashtable System.Collections.Generic.Dictionary<TKey,TValue> These are hashtables—collections for which each element has a key used for storage and retrieval. A hashtable applies a very specific strategy for efficiently allocating elements based on their key. This requires that each key have an Int32 number, or hash code. The hash code need not be unique for each key but should be as varied as possible for good hashtable performance. Hashtables are considered important enough that GetHashCode is defined in System.Object—so that every type can emit a hash code. We describe hashtables in detail in Chapter 7. Both reference and value types have default implementations of GetHashCode, meaning that you don’t need to override this method—unless you override Equals. (And if you override GetHashCode, you will almost certainly want to also override Equals.) Here are the other rules for overriding object.GetHashCode: • It must return the same value on two objects for which Equals returns true • (hence, GetHashCode and Equals are overridden together). • It must not throw exceptions. • • It must return the same value if called repeatedly on the same object (unless the • object has changed). For maximum performance in hashtables, you should write GetHashCode so as to minimize the likelihood of two different values returning the same hashcode. This gives rise to the third reason for overriding Equals and GetHashCode on structs, which is to provide a more efficient hashing algorithm than the default. The default implementation for structs is at the discretion of the runtime and can be based on every field in the struct. In contrast, the default GetHashCode implementation for classes is based on an internal object token, which is unique for each instance in the CLR’s current implementation. 332 | Chapter 6: .NET Fundamentals
- 351. If an object’s hashcode changes after it’s been added as a key to a dictionary, the object will no longer be accessible in the dic‐ tionary. You can preempt this by basing hashcode calculations on immutable fields. We provide a complete example illustrating how to override GetHashCode shortly. Overriding Equals The axioms for object.Equals are as follows: • An object cannot equal null (unless it’s a nullable type). • • Equality is reflexive (an object equals itself). • • Equality is commutative (if a.Equals(b), then b.Equals(a)). • • Equality is transitive (if a.Equals(b) and b.Equals(c), then a.Equals(c)). • • Equality operations are repeatable and reliable (they don’t throw exceptions). • Overloading == and != In addition to overriding Equals, you can optionally overload the equality and inequality operators. This is nearly always done with structs because the conse‐ quence of not doing so is that the == and != operators will simply not work on your type. With classes, there are two ways to proceed: • Leave == and != alone—so that they apply referential equality. • • Overload == and != in line with Equals. • The first approach is most common with custom types—especially mutable types. It ensures that your type follows the expectation that == and != should exhibit referential equality with reference types, and this prevents confusing consumers. We saw an example earlier: var sb1 = new StringBuilder ("foo"); var sb2 = new StringBuilder ("foo"); Console.WriteLine (sb1 == sb2); // False (referential equality) Console.WriteLine (sb1.Equals (sb2)); // True (value equality) The second approach makes sense with types for which a consumer would never want referential equality. These are typically immutable—such as the string and System.Uri classes—and are sometimes good candidates for structs. Although it’s possible to overload != such that it means some‐ thing other than !(==), this is almost never done in practice, except in cases such as comparing float.NaN. Equality Comparison | 333 .NET Fundamentals
- 352. Implementing IEquatable<T> For completeness, it’s also good to implement IEquatable<T> when overriding Equals. Its results should always match those of the overridden object’s Equals method. Implementing IEquatable<T> comes at no programming cost if you struc‐ ture your Equals method implementation as in the example that follows in a moment. An example: the Area struct Imagine that we need a struct to represent an area whose width and height are interchangeable. In other words, 5 × 10 is equal to 10 × 5. (Such a type would be suitable in an algorithm that arranges rectangular shapes.) Here’s the complete code: public struct Area : IEquatable <Area> { public readonly int Measure1; public readonly int Measure2; public Area (int m1, int m2) { Measure1 = Math.Min (m1, m2); Measure2 = Math.Max (m1, m2); } public override bool Equals (object other) => other is Area a && Equals (a); // Calls method below public bool Equals (Area other) // Implements IEquatable<Area> => Measure1 == other.Measure1 && Measure2 == other.Measure2; public override int GetHashCode() => HashCode.Combine (Measure1, Measure2); public static bool operator == (Area a1, Area a2) => a1.Equals (a2); public static bool operator != (Area a1, Area a2) => !a1.Equals (a2); } From C# 10, you can shortcut the process with records. By declaring this as a record struct, you can remove all the code following the constructor. In implementing GetHashCode, we used .NET’s HashCode.Combine function to pro‐ duce a composite hashcode. (Before that function existed, a popular approach was to multiply each value by some prime number and then add them together.) Here’s a demonstration of the Area struct: Area a1 = new Area (5, 10); Area a2 = new Area (10, 5); 334 | Chapter 6: .NET Fundamentals
- 353. Console.WriteLine (a1.Equals (a2)); // True Console.WriteLine (a1 == a2); // True Pluggable equality comparers If you want a type to take on different equality semantics just for a specific scenario, you can use a pluggable IEqualityComparer. This is particularly useful in conjunc‐ tion with the standard collection classes, and we describe it in the following chapter, in “Plugging in Equality and Order” on page 388. Order Comparison As well as defining standard protocols for equality, C# and .NET define two stan‐ dard protocols for determining the order of one object relative to another: • The IComparable interfaces (IComparable and IComparable<T>) • • The > and < operators • The IComparable interfaces are used by general-purpose sorting algorithms. In the following example, the static Array.Sort method works because System.String implements the IComparable interfaces: string[] colors = { "Green", "Red", "Blue" }; Array.Sort (colors); foreach (string c in colors) Console.Write (c + " "); // Blue Green Red The < and > operators are more specialized, and they are intended mostly for numeric types. Because they are statically resolved, they can translate to highly efficient bytecode, suitable for computationally intensive algorithms. .NET also provides pluggable ordering protocols, via the IComparer interfaces. We describe these in the final section of Chapter 7. IComparable The IComparable interfaces are defined as follows: public interface IComparable { int CompareTo (object other); } public interface IComparable<in T> { int CompareTo (T other); } The two interfaces represent the same functionality. With value types, the generic type-safe interface is faster than the nongeneric interface. In both cases, the Compare To method works as follows: • If a comes after b, a.CompareTo(b) returns a positive number. • • If a is the same as b, a.CompareTo(b) returns 0. • • If a comes before b, a.CompareTo(b) returns a negative number. • For example: Order Comparison | 335 .NET Fundamentals
- 354. Console.WriteLine ("Beck".CompareTo ("Anne")); // 1 Console.WriteLine ("Beck".CompareTo ("Beck")); // 0 Console.WriteLine ("Beck".CompareTo ("Chris")); // -1 Most of the base types implement both IComparable interfaces. These interfaces are also sometimes implemented when writing custom types. We provide an example shortly. IComparable versus Equals Consider a type that both overrides Equals and implements the IComparable inter‐ faces. You’d expect that when Equals returns true, CompareTo should return 0. And you’d be right. But here’s the catch: When Equals returns false, CompareTo can return what it likes (as long as it’s internally consistent)! In other words, equality can be “fussier” than comparison, but not vice versa (violate this and sorting algorithms will break). So, CompareTo can say, “All objects are equal,” whereas Equals says, “But some are more equal than others!” A great example of this is System.String. String’s Equals method and == operator use ordinal comparison, which compares the Unicode point values of each charac‐ ter. Its CompareTo method, however, uses a less fussy culture-dependent comparison. On most computers, for instance, the strings “ṻ” and “ǖ” are different according to Equals but the same according to CompareTo. In Chapter 7, we discuss the pluggable ordering protocol, IComparer, which allows you to specify an alternative ordering algorithm when sorting or instantiating a sor‐ ted collection. A custom IComparer can further extend the gap between CompareTo and Equals—a case-insensitive string comparer, for instance, will return 0 when comparing "A" and "a". The reverse rule still applies, however: CompareTo can never be fussier than Equals. When implementing the IComparable interfaces in a custom type, you can avoid running afoul of this rule by writing the first line of CompareTo as follows: if (Equals (other)) return 0; After that, it can return what it likes, as long as it’s consistent! < and > Some types define < and > operators. For instance: bool after2010 = DateTime.Now > new DateTime (2010, 1, 1); You can expect the < and > operators, when implemented, to be functionally consis‐ tent with the IComparable interfaces. This is standard practice across .NET. It’s also standard practice to implement the IComparable interfaces whenever < and > are overloaded, although the reverse is not true. In fact, most .NET types that 336 | Chapter 6: .NET Fundamentals
- 355. implement IComparable do not overload < and >. This differs from the situation with equality for which it’s normal to overload == when overriding Equals. Typically, > and < are overloaded only when: • A type has a strong intrinsic concept of “greater than” and “less than” (versus • IComparable’s broader concepts of “comes before” and “comes after”). • There is only one way, or context, in which to perform the comparison. • • The result is invariant across cultures. • System.String doesn’t satisfy the last point: the results of string comparisons can vary according to language. Hence, string doesn’t support the > and < operators: bool error = "Beck" > "Anne"; // Compile-time error Implementing the IComparable Interfaces In the following struct representing a musical note, we implement the IComparable interfaces as well as overloading the < and > operators. For completeness, we also override Equals/GetHashCode and overload == and !=: public struct Note : IComparable<Note>, IEquatable<Note>, IComparable { int _semitonesFromA; public int SemitonesFromA { get { return _semitonesFromA; } } public Note (int semitonesFromA) { _semitonesFromA = semitonesFromA; } public int CompareTo (Note other) // Generic IComparable<T> { if (Equals (other)) return 0; // Fail-safe check return _semitonesFromA.CompareTo (other._semitonesFromA); } int IComparable.CompareTo (object other) // Nongeneric IComparable { if (!(other is Note)) throw new InvalidOperationException ("CompareTo: Not a note"); return CompareTo ((Note) other); } public static bool operator < (Note n1, Note n2) => n1.CompareTo (n2) < 0; public static bool operator > (Note n1, Note n2) => n1.CompareTo (n2) > 0; public bool Equals (Note other) // for IEquatable<Note> => _semitonesFromA == other._semitonesFromA; Order Comparison | 337 .NET Fundamentals
- 356. public override bool Equals (object other) { if (!(other is Note)) return false; return Equals ((Note) other); } public override int GetHashCode() => _semitonesFromA.GetHashCode(); public static bool operator == (Note n1, Note n2) => n1.Equals (n2); public static bool operator != (Note n1, Note n2) => !(n1 == n2); } Utility Classes Console The static Console class handles standard input/output for console-based applica‐ tions. In a command-line (console) application, the input comes from the keyboard via Read, ReadKey, and ReadLine, and the output goes to the text window via Write and WriteLine. You can control the window’s position and dimensions with the properties WindowLeft, WindowTop, WindowHeight, and WindowWidth. You can also change the BackgroundColor and ForegroundColor properties and manipulate the cursor with the CursorLeft, CursorTop, and CursorSize properties: Console.WindowWidth = Console.LargestWindowWidth; Console.ForegroundColor = ConsoleColor.Green; Console.Write ("test... 50%"); Console.CursorLeft -= 3; Console.Write ("90%"); // test... 90% The Write and WriteLine methods are overloaded to accept a composite format string (see String.Format in “String and Text Handling” on page 271). How‐ ever, neither method accepts a format provider, so you’re stuck with CultureInfo .CurrentCulture. (The workaround, of course, is to explicitly call string.Format.) The Console.Out property returns a TextWriter. Passing Console.Out to a method that expects a TextWriter is a useful way to get that method to write to the Console for diagnostic purposes. You can also redirect the Console’s input and output streams via the SetIn and SetOut methods: // First save existing output writer: System.IO.TextWriter oldOut = Console.Out; // Redirect the console's output to a file: using (System.IO.TextWriter w = System.IO.File.CreateText ("e:output.txt")) { Console.SetOut (w); 338 | Chapter 6: .NET Fundamentals
- 357. Console.WriteLine ("Hello world"); } // Restore standard console output Console.SetOut (oldOut); In Chapter 15, we describe how streams and text writers work. When running WPF or Windows Forms applications under Visual Studio, the Console’s output is automatically redirected to Visual Studio’s output window (in debug mode). This can make Console.Write useful for diagnostic purposes; although in most cases, the Debug and Trace classes in the System.Diagnostics namespace are more appropriate (see Chapter 13). Environment The static System.Environment class provides a range of useful properties: Files and folders CurrentDirectory, SystemDirectory, CommandLine Computer and operating system MachineName, ProcessorCount, OSVersion, NewLine User logon UserName, UserInteractive, UserDomainName Diagnostics TickCount, StackTrace, WorkingSet, Version You can obtain additional folders by calling GetFolderPath; we describe this in “File and Directory Operations” on page 699. You can access OS environment variables (what you see when you type set at the command prompt) with the following three methods: GetEnvironmentVariable, GetEnvironmentVariables, and SetEnvironmentVariable. The ExitCode property lets you set the return code—for when your program is called from a command or batch file—and the FailFast method terminates a program immediately, without performing cleanup. The Environment class available to Windows Store apps offers just a limited number of members (ProcessorCount, NewLine, and FailFast). Process The Process class in System.Diagnostics allows you to launch a new process. (In Chapter 13, we describe how you can also use it to interact with other processes running on the computer.) Utility Classes | 339 .NET Fundamentals
- 358. For security reasons, the Process class is not available to Windows Store apps, and you cannot start arbitrary processes. Instead, you must use the Windows.System.Launcher class to “launch” a URI or file to which you have access. For example, Launcher.LaunchUriAsync (new Uri ("http://albahari.com")); var file = await KnownFolders.DocumentsLibrary .GetFileAsync ("foo.txt"); Launcher.LaunchFileAsync (file); opens the URI or file, using whatever program is associated with the URI scheme or file extension. Your program must be in the foreground for this to work. The static Process.Start method has several overloads; the simplest accepts a simple filename with optional arguments: Process.Start ("notepad.exe"); Process.Start ("notepad.exe", "e:file.txt"); The most flexible overload accepts a ProcessStartInfo instance. With this, you can capture and redirect the launched process’s input, output, and error output (if you leave UseShellExecute as false). The following captures the output of calling ipconfig: ProcessStartInfo psi = new ProcessStartInfo { FileName = "cmd.exe", Arguments = "/c ipconfig /all", RedirectStandardOutput = true, UseShellExecute = false }; Process p = Process.Start (psi); string result = p.StandardOutput.ReadToEnd(); Console.WriteLine (result); If you don’t redirect output, Process.Start executes the program in parallel to the caller. If you want to wait for the new process to complete, you can call WaitForExit on the Process object, with an optional timeout. Redirecting output and error streams With UseShellExecute false (the default in .NET), you can capture the standard input, output, and error streams and then write/read these streams via the Standard Input, StandardOutput, and StandardError properties. A difficulty arises when you need to redirect both the standard output and standard error streams, in that you can’t usually know in which order to read data from each (because you don’t know in advance how the data will be interleaved). The solution is to read from both streams at once, which you can accomplish by reading from (at least) one of the streams asynchronously. Here’s how to do this: 340 | Chapter 6: .NET Fundamentals
- 359. • Handle the OutputDataReceived and/or ErrorDataReceived events. These • events fire when output/error data is received. • Call BeginOutputReadLine and/or BeginErrorReadLine. This enables the • aforementioned events. The following method runs an executable while capturing both the output and error streams: (string output, string errors) Run (string exePath, string args = "") { using var p = Process.Start (new ProcessStartInfo (exePath, args) { RedirectStandardOutput = true, RedirectStandardError = true, UseShellExecute = false, }); var errors = new StringBuilder (); // Read from the error stream asynchronously... p.ErrorDataReceived += (sender, errorArgs) => { if (errorArgs.Data != null) errors.AppendLine (errorArgs.Data); }; p.BeginErrorReadLine (); // ...while we read from the output stream synchronously: string output = p.StandardOutput.ReadToEnd(); p.WaitForExit(); return (output, errors.ToString()); } UseShellExecute In .NET 5+ (and .NET Core), the default for UseShellEx ecute is false, whereas in .NET Framework, it was true. Because this is a breaking change, it’s worth checking all calls to Process.Start when porting code from .NET Framework. The UseShellExecute flag changes how the CLR starts the process. With UseShell Execute true, you can do the following: • Specify a path to a file or document rather than an executable (resulting in the • operating system opening the file or document with its associated application) • Specify a URL (resulting in the operating system navigating to that URL in the • default web browser) • (Windows only) Specify a Verb (such as “runas”, to run the process with admin‐ • istrative elevation) Utility Classes | 341 .NET Fundamentals
- 360. The drawback is that you cannot redirect the input or output streams. Should you need to do so—while launching a file or document—a workaround is to set UseShellExecute to false and invoke the command-line process (cmd.exe) with the “/c” switch, as we did earlier when calling ipconfig. Under Windows, UseShellExecute instructs the CLR to use the Windows ShellExe‐ cute function instead of the CreateProcess function. Under Linux, UseShellExecute instructs the CLR to call xdg-open, gnome-open, or kfmclient. AppContext The static System.AppContext class exposes two useful properties: • BaseDirectory returns the folder in which the application started. This folder • is important for assembly resolution (finding and loading dependencies) and locating configuration files (such as appsettings.json). • TargetFrameworkName tells you the name and version of the .NET runtime that • the application targets (as specified in its .runtimeconfig.json file). This might be older than the runtime actually in use. In addition, the AppContext class manages a global string-keyed dictionary of Boolean values, intended to offer library writers a standard mechanism for allowing consumers to switch new features on or off. This untyped approach makes sense with experimental features that you want to keep undocumented to the majority of users. The consumer of a library requests that you enable a feature as follows: AppContext.SetSwitch ("MyLibrary.SomeBreakingChange", true); Code within that library can then check for that switch as follows: bool isDefined, switchValue; isDefined = AppContext.TryGetSwitch ("MyLibrary.SomeBreakingChange", out switchValue); TryGetSwitch returns false if the switch is undefined; this lets you distinguish an undefined switch from one whose value is set to false, should this be necessary. Ironically, the design of TryGetSwitch illustrates how not to write APIs. The out parameter is unnecessary, and the method should instead return a nullable bool whose value is true, false, or null for undefined. This would then enable the fol‐ lowing use: bool switchValue = AppContext.GetSwitch ("...") ?? false; 342 | Chapter 6: .NET Fundamentals
- 361. 7 Collections .NET provides a standard set of types for storing and managing collections of objects. These include resizable lists, linked lists, sorted and unsorted dictionaries, and arrays. Of these, only arrays form part of the C# language; the remaining collections are just classes you instantiate like any other. We can divide the types in the .NET BCL for collections into the following categories: • Interfaces that define standard collection protocols • • Ready-to-use collection classes (lists, dictionaries, etc.) • • Base classes for writing application-specific collections • This chapter covers each of these categories, with an additional section on the types used in determining element equality and order. The collection namespaces are as follows: Namespace Contains System.Collections Nongeneric collection classes and interfaces System.Collections.Specialized Strongly typed nongeneric collection classes System.Collections.Generic Generic collection classes and interfaces System.Collections.ObjectModel Proxies and bases for custom collections System.Collections.Concurrent Thread-safe collections (see Chapter 22) Enumeration In computing, there are many different kinds of collections, ranging from simple data structures, such as arrays or linked lists, to more complex ones, such as red/black trees and hashtables. Although the internal implementation and external 343
- 362. characteristics of these data structures vary widely, the ability to traverse the con‐ tents of the collection is an almost universal need. The .NET BCL supports this need via a pair of interfaces (IEnumerable and IEnumerator, and their generic counterparts) that allow different data structures to expose a common traversal API. These are part of a larger set of collection interfaces illustrated in Figure 7-1. Figure 7-1. Collection interfaces IEnumerable and IEnumerator The IEnumerator interface defines the basic low-level protocol by which elements in a collection are traversed—or enumerated—in a forward-only manner. Its decla‐ ration is as follows: public interface IEnumerator { bool MoveNext(); object Current { get; } void Reset(); } MoveNext advances the current element or “cursor” to the next position, returning false if there are no more elements in the collection. Current returns the element at the current position (usually cast from object to a more specific type). MoveNext must be called before retrieving the first element—this is to allow for an empty collection. The Reset method, if implemented, moves back to the start, allowing the collection to be enumerated again. Reset exists mainly for Component Object Model (COM) interoperability; calling it directly is generally avoided because it’s not universally supported (and is unnecessary in that it’s usually just as easy to instantiate a new enumerator). Collections do not usually implement enumerators; instead, they provide enumera‐ tors, via the interface IEnumerable: 344 | Chapter 7: Collections
- 363. public interface IEnumerable { IEnumerator GetEnumerator(); } By defining a single method retuning an enumerator, IEnumerable provides flexibil‐ ity in that the iteration logic can be farmed out to another class. Moreover, it means that several consumers can enumerate the collection at once without interfering with one another. You can think of IEnumerable as “IEnumeratorProvider,” and it is the most basic interface that collection classes implement. The following example illustrates low-level use of IEnumerable and IEnumerator: string s = "Hello"; // Because string implements IEnumerable, we can call GetEnumerator(): IEnumerator rator = s.GetEnumerator(); while (rator.MoveNext()) { char c = (char) rator.Current; Console.Write (c + "."); } // Output: H.e.l.l.o. However, it’s rare to call methods on enumerators directly in this manner because C# provides a syntactic shortcut: the foreach statement. Here’s the same example rewritten using foreach: string s = "Hello"; // The String class implements IEnumerable foreach (char c in s) Console.Write (c + "."); IEnumerable<T> and IEnumerator<T> IEnumerator and IEnumerable are nearly always implemented in conjunction with their extended generic versions: public interface IEnumerator<T> : IEnumerator, IDisposable { T Current { get; } } public interface IEnumerable<T> : IEnumerable { IEnumerator<T> GetEnumerator(); } By defining a typed version of Current and GetEnumerator, these interfaces strengthen static type safety, avoid the overhead of boxing with value-type elements, and are more convenient to the consumer. Arrays automatically implement IEnumer able<T> (where T is the member type of the array). Enumeration | 345 Collections
- 364. Thanks to the improved static type safety, calling the following method with an array of characters will generate a compile-time error: void Test (IEnumerable<int> numbers) { ... } It’s a standard practice for collection classes to publicly expose IEnumerable<T> while “hiding” the nongeneric IEnumerable through explicit interface implementa‐ tion. This is so that if you directly call GetEnumerator(), you get back the type-safe generic IEnumerator<T>. There are times, though, when this rule is broken for reasons of backward compatibility (generics did not exist prior to C# 2.0). A good example is arrays—these must return the nongeneric (the nice way of putting it is “classic”) IEnumerator to prevent breaking earlier code. To get a generic IEnumera tor<T>, you must cast to expose the explicit interface: int[] data = { 1, 2, 3 }; var rator = ((IEnumerable <int>)data).GetEnumerator(); Fortunately, you rarely need to write this sort of code, thanks to the foreach statement. IEnumerable<T> and IDisposable IEnumerator<T> inherits from IDisposable. This allows enumerators to hold refer‐ ences to resources such as database connections—and ensure that those resources are released when enumeration is complete (or abandoned partway through). The foreach statement recognizes this detail and translates the following: foreach (var element in somethingEnumerable) { ... } into the logical equivalent of this: using (var rator = somethingEnumerable.GetEnumerator()) while (rator.MoveNext()) { var element = rator.Current; ... } When to Use the Nongeneric Interfaces Given the extra type safety of the generic collection interfaces such as IEnumera ble<T>, the question arises: do you ever need to use the nongeneric IEnumerable (or ICollection or IList)? In the case of IEnumerable, you must implement this interface in conjunction with IEnumerable<T>—because the latter derives from the former. However, it’s very rare that you actually implement these interfaces from scratch: in nearly all cases, you can take the higher-level approach of using iterator methods, Collection<T>, and LINQ. 346 | Chapter 7: Collections
- 365. So, what about as a consumer? In nearly all cases, you can manage entirely with the generic interfaces. The nongeneric interfaces are still occasionally useful, though, in their ability to provide type unification for collections across all element types. The following method, for instance, counts elements in any collection recursively: public static int Count (IEnumerable e) { int count = 0; foreach (object element in e) { var subCollection = element as IEnumerable; if (subCollection != null) count += Count (subCollection); else count++; } return count; } Because C# offers covariance with generic interfaces, it might seem valid to have this method instead accept IEnumerable<object>. This, however, would fail with value-type elements and with legacy collections that don’t implement IEnumera ble<T>—an example is ControlCollection in Windows Forms. (On a slight tangent, you might have noticed a potential bug in our example: cyclic references will cause infinite recursion and crash the method. We could fix this most easily with the use of a HashSet (see “HashSet<T> and SortedSet<T>” on page 370). The using block ensures disposal—more on IDisposable in Chapter 12. Implementing the Enumeration Interfaces You might want to implement IEnumerable or IEnumerable<T> for one or more of the following reasons: • To support the foreach statement • • To interoperate with anything expecting a standard collection • • To meet the requirements of a more sophisticated collection interface • • To support collection initializers • To implement IEnumerable/IEnumerable<T>, you must provide an enumerator. You can do this in one of three ways: • If the class is “wrapping” another collection, by returning the wrapped collec‐ • tion’s enumerator • Via an iterator using yield return • • By instantiating your own IEnumerator/IEnumerator<T> implementation • Enumeration | 347 Collections
- 366. You can also subclass an existing collection: Collection<T> is designed just for this purpose (see “Customizable Collections and Proxies” on page 378). Yet another approach is to use the LINQ query operators, which we cover in Chapter 8. Returning another collection’s enumerator is just a matter of calling GetEnumerator on the inner collection. However, this is viable only in the simplest scenarios in which the items in the inner collection are exactly what are required. A more flexible approach is to write an iterator, using C#’s yield return statement. An iterator is a C# language feature that assists in writing collections, in the same way the foreach statement assists in consuming collections. An iterator automatically handles the implementation of IEnumerable and IEnumerator—or their generic versions. Here’s a simple example: public class MyCollection : IEnumerable { int[] data = { 1, 2, 3 }; public IEnumerator GetEnumerator() { foreach (int i in data) yield return i; } } Notice the “black magic”: GetEnumerator doesn’t appear to return an enumerator at all! Upon parsing the yield return statement, the compiler writes a hidden nested enumerator class behind the scenes and then refactors GetEnumerator to instantiate and return that class. Iterators are powerful and simple (and are used extensively in the implementation of LINQ-to-Object’s standard query operators). Keeping with this approach, we can also implement the generic interface IEnumera ble<T>: public class MyGenCollection : IEnumerable<int> { int[] data = { 1, 2, 3 }; public IEnumerator<int> GetEnumerator() { foreach (int i in data) yield return i; } // Explicit implementation keeps it hidden: IEnumerator IEnumerable.GetEnumerator() => GetEnumerator(); } Because IEnumerable<T> inherits from IEnumerable, we must implement both the generic and the nongeneric versions of GetEnumerator. In accordance with standard practice, we’ve implemented the nongeneric version explicitly. It can 348 | Chapter 7: Collections
- 367. simply call the generic GetEnumerator because IEnumerator<T> inherits from IEnumerator. The class we’ve just written would be suitable as a basis from which to write a more sophisticated collection. However, if we need nothing above a simple IEnumer able<T> implementation, the yield return statement allows for an easier variation. Rather than writing a class, you can move the iteration logic into a method return‐ ing a generic IEnumerable<T> and let the compiler take care of the rest. Here’s an example: public static IEnumerable <int> GetSomeIntegers() { yield return 1; yield return 2; yield return 3; } Here’s our method in use: foreach (int i in Test.GetSomeIntegers()) Console.WriteLine (i); The final approach in writing GetEnumerator is to write a class that implements IEnumerator directly. This is exactly what the compiler does behind the scenes, in resolving iterators. (Fortunately, it’s rare that you’ll need to go this far yourself.) The following example defines a collection that’s hardcoded to contain the integers 1, 2, and 3: public class MyIntList : IEnumerable { int[] data = { 1, 2, 3 }; public IEnumerator GetEnumerator() => new Enumerator (this); class Enumerator : IEnumerator // Define an inner class { // for the enumerator. MyIntList collection; int currentIndex = -1; public Enumerator (MyIntList items) => this.collection = items; public object Current { get { if (currentIndex == -1) throw new InvalidOperationException ("Enumeration not started!"); if (currentIndex == collection.data.Length) throw new InvalidOperationException ("Past end of list!"); return collection.data [currentIndex]; } } public bool MoveNext() Enumeration | 349 Collections
- 368. { if (currentIndex >= collection.data.Length - 1) return false; return ++currentIndex < collection.data.Length; } public void Reset() => currentIndex = -1; } } Implementing Reset is optional—you can instead throw a NotSupportedException. Note that the first call to MoveNext should move to the first (and not the second) item in the list. To get on par with an iterator in functionality, we must also implement IEnumera tor<T>. Here’s an example with bounds checking omitted for brevity: class MyIntList : IEnumerable<int> { int[] data = { 1, 2, 3 }; // The generic enumerator is compatible with both IEnumerable and // IEnumerable<T>. We implement the nongeneric GetEnumerator method // explicitly to avoid a naming conflict. public IEnumerator<int> GetEnumerator() => new Enumerator(this); IEnumerator IEnumerable.GetEnumerator() => new Enumerator(this); class Enumerator : IEnumerator<int> { int currentIndex = -1; MyIntList collection; public Enumerator (MyIntList items) => this.items = items; public int Current => collection.data [currentIndex]; object IEnumerator.Current => Current; public bool MoveNext() => ++currentIndex < collection.data.Length; public void Reset() => currentIndex = -1; // Given we don't need a Dispose method, it's good practice to // implement it explicitly, so it's hidden from the public interface. void IDisposable.Dispose() {} } } The example with generics is faster because IEnumerator<int>.Current doesn’t require casting from int to object and so avoids the overhead of boxing. 350 | Chapter 7: Collections
- 369. The ICollection and IList Interfaces Although the enumeration interfaces provide a protocol for forward-only iteration over a collection, they don’t provide a mechanism to determine the size of the collection, access a member by index, search, or modify the collection. For such functionality, .NET defines the ICollection, IList, and IDictionary interfaces. Each comes in both generic and nongeneric versions; however, the nongeneric versions exist mostly for legacy support. Figure 7-1 showed the inheritance hierarchy for these interfaces. The easiest way to summarize them is as follows: IEnumerable<T> (and IEnumerable) Provides minimum functionality (enumeration only) ICollection<T> (and ICollection) Provides medium functionality (e.g., the Count property) IList<T>/IDictionary<K,V> and their nongeneric versions Provide maximum functionality (including “random” access by index/key) It’s rare that you’ll need to implement any of these inter‐ faces. In nearly all cases when you need to write a collec‐ tion class, you can instead subclass Collection<T> (see “Customizable Collections and Proxies” on page 378). LINQ provides yet another option that covers many scenarios. The generic and nongeneric versions differ in ways over and above what you might expect, particularly in the case of ICollection. The reasons for this are mostly historical: because generics came later, the generic interfaces were developed with the benefit of hindsight, leading to a different (and better) choice of members. For this reason, ICollection<T> does not extend ICollection, IList<T> does not extend IList, and IDictionary<TKey, TValue> does not extend IDictionary. Of course, a collection class itself is free to implement both versions of an interface if beneficial (which it often is). Another, subtler reason for IList<T> not extending IList is that casting to IList<T> would then return an interface with both Add(T) and Add(object) members. This would effectively defeat static type safety because you could call Add with an object of any type. This section covers ICollection<T>, IList<T>, and their nongeneric versions; “Dictionaries” on page 372 covers the dictionary interfaces. The ICollection and IList Interfaces | 351 Collections
- 370. There is no consistent rationale in the way the words col‐ lection and list are applied throughout the .NET libraries. For instance, because IList<T> is a more functional version of ICollection<T>, you might expect the class List<T> to be correspondingly more functional than the class Collec tion<T>. This is not the case. It’s best to consider the terms collection and list as broadly synonymous, except when a spe‐ cific type is involved. ICollection<T> and ICollection ICollection<T> is the standard interface for countable collections of objects. It provides the ability to determine the size of a collection (Count), determine whether an item exists in the collection (Contains), copy the collection into an array (ToArray), and determine whether the collection is read-only (IsReadOnly). For writable collections, you can also Add, Remove, and Clear items from the collection. And because it extends IEnumerable<T>, it can also be traversed via the foreach statement: public interface ICollection<T> : IEnumerable<T>, IEnumerable { int Count { get; } bool Contains (T item); void CopyTo (T[] array, int arrayIndex); bool IsReadOnly { get; } void Add(T item); bool Remove (T item); void Clear(); } The nongeneric ICollection is similar in providing a countable collection, but it doesn’t provide functionality for altering the list or checking for element membership: public interface ICollection : IEnumerable { int Count { get; } bool IsSynchronized { get; } object SyncRoot { get; } void CopyTo (Array array, int index); } The nongeneric interface also defines properties to assist with synchronization (Chapter 14)—these were dumped in the generic version because thread safety is no longer considered intrinsic to the collection. Both interfaces are fairly straightforward to implement. If implementing a read-only ICollection<T>, the Add, Remove, and Clear methods should throw a NotSuppor tedException. 352 | Chapter 7: Collections
- 371. These interfaces are usually implemented in conjunction with either the IList or the IDictionary interface. IList<T> and IList IList<T> is the standard interface for collections indexable by position. In addi‐ tion to the functionality inherited from ICollection<T> and IEnumerable<T>, it provides the ability to read or write an element by position (via an indexer) and insert/remove by position: public interface IList<T> : ICollection<T>, IEnumerable<T>, IEnumerable { T this [int index] { get; set; } int IndexOf (T item); void Insert (int index, T item); void RemoveAt (int index); } The IndexOf methods perform a linear search on the list, returning −1 if the specified item is not found. The nongeneric version of IList has more members because it inherits less from ICollection: public interface IList : ICollection, IEnumerable { object this [int index] { get; set } bool IsFixedSize { get; } bool IsReadOnly { get; } int Add (object value); void Clear(); bool Contains (object value); int IndexOf (object value); void Insert (int index, object value); void Remove (object value); void RemoveAt (int index); } The Add method on the nongeneric IList interface returns an integer—this is the index of the newly added item. In contrast, the Add method on ICollection<T> has a void return type. The general-purpose List<T> class is the quintessential implementation of both IList<T> and IList. C# arrays also implement both the generic and nongeneric ILists (although the methods that add or remove elements are hidden via explicit interface implementation and throw a NotSupportedException if called). The ICollection and IList Interfaces | 353 Collections
- 372. An ArgumentException is thrown if you try to access a mul‐ tidimensional array via IList’s indexer. This is a trap when writing methods such as the following: public object FirstOrNull (IList list) { if (list == null || list.Count == 0) return null; return list[0]; } This might appear bulletproof, but it will throw an exception if called with a multidimensional array. You can test for a multidimensional array at runtime with this expression (more on this in Chapter 19): list.GetType().IsArray && list.GetType().GetArrayRank()>1 IReadOnlyCollection<T> and IReadOnlyList<T> .NET also defines collection and list interfaces that expose just the members required for read-only operations: public interface IReadOnlyCollection<out T> : IEnumerable<T>, IEnumerable { int Count { get; } } public interface IReadOnlyList<out T> : IReadOnlyCollection<T>, IEnumerable<T>, IEnumerable { T this[int index] { get; } } Because the type parameter for these interfaces is used only in output positions, it’s marked as covariant. This allows a list of cats, for instance, to be treated as a read- only list of animals. In contrast, T is not marked as covariant with ICollection<T> and IList<T>, because T is used in both input and output positions. These interfaces represent a read-only view of a collection or list; the underlying implementation might still be writable. Most of the writable (mutable) collections implement both the read-only and read/write interfaces. In addition to letting you work with collections covariantly, the read-only interfaces allow a class to publicly expose a read-only view of a private writable collection. We demonstrate this—along with a better solution—in “ReadOnlyCollection<T>” on page 383. IReadOnlyList<T> maps to the Windows Runtime type IVectorView<T>. 354 | Chapter 7: Collections
- 373. The Array Class The Array class is the implicit base class for all single and multidimensional arrays, and it is one of the most fundamental types implementing the standard collection interfaces. The Array class provides type unification, so a common set of methods is available to all arrays, regardless of their declaration or underlying element type. Because arrays are so fundamental, C# provides explicit syntax for their decla‐ ration and initialization, which we described in Chapters 2 and 3. When an array is declared using C#’s syntax, the CLR implicitly subtypes the Array class —synthesizing a pseudotype appropriate to the array’s dimensions and element types. This pseudotype implements the typed generic collection interfaces, such as IList<string>. The CLR also treats array types specially upon construction, assigning them a contiguous space in memory. This makes indexing into arrays highly efficient, but prevents them from being resized later on. Array implements the collection interfaces up to IList<T> in both their generic and nongeneric forms. IList<T> itself is implemented explicitly, though, to keep Array’s public interface clean of methods such as Add or Remove, which throw an exception on fixed-length collections such as arrays. The Array class does actually offer a static Resize method, although this works by creating a new array and then copying over each element. As well as being inefficient, references to the array elsewhere in the program will still point to the original version. A better solution for resizable collections is to use the List<T> class (described in the following section). An array can contain value-type or reference-type elements. Value-type elements are stored in place in the array, so an array of three long integers (each 8 bytes) will occupy 24 bytes of contiguous memory. A reference type element, however, occu‐ pies only as much space in the array as a reference (4 bytes in a 32-bit environment or 8 bytes in a 64-bit environment). Figure 7-2 illustrates the effect, in memory, of the following program: StringBuilder[] builders = new StringBuilder [5]; builders [0] = new StringBuilder ("builder1"); builders [1] = new StringBuilder ("builder2"); builders [2] = new StringBuilder ("builder3"); long[] numbers = new long [3]; numbers [0] = 12345; numbers [1] = 54321; The Array Class | 355 Collections
- 374. Figure 7-2. Arrays in memory Because Array is a class, arrays are always (themselves) reference types—regardless of the array’s element type. This means that the statement arrayB = arrayA results in two variables that reference the same array. Similarly, two distinct arrays will always fail an equality test, unless you employ a structural equality comparer, which compares every element of the array: object[] a1 = { "string", 123, true }; object[] a2 = { "string", 123, true }; Console.WriteLine (a1 == a2); // False Console.WriteLine (a1.Equals (a2)); // False IStructuralEquatable se1 = a1; Console.WriteLine (se1.Equals (a2, StructuralComparisons.StructuralEqualityComparer)); // True Arrays can be duplicated by calling the Clone method: arrayB = arrayA.Clone(). However, this results in a shallow clone, meaning that only the memory represented by the array itself is copied. If the array contains value-type objects, the values them‐ selves are copied; if the array contains reference type objects, just the references are copied (resulting in two arrays whose members reference the same objects). Figure 7-3 demonstrates the effect of adding the following code to our example: StringBuilder[] builders2 = builders; StringBuilder[] shallowClone = (StringBuilder[]) builders.Clone(); 356 | Chapter 7: Collections
- 375. Figure 7-3. Shallow-cloning an array To create a deep copy—for which reference type subobjects are duplicated—you must loop through the array and clone each element manually. The same rules apply to other .NET collection types. Although Array is designed primarily for use with 32-bit indexers, it also has limited support for 64-bit indexers (allowing an array to theoretically address up to 264 elements) via several methods that accept both Int32 and Int64 parameters. These overloads are useless in practice because the CLR does not permit any object—including arrays—to exceed two gigabytes in size (whether running on a 32- or 64-bit environment). Many of the methods on the Array class that you expect to be instance methods are in fact static methods. This is an odd design decision, and means that you should check for both static and instance methods when looking for a method on Array. Construction and Indexing The easiest way to create and index arrays is through C#’s language constructs: int[] myArray = { 1, 2, 3 }; int first = myArray [0]; int last = myArray [myArray.Length - 1]; Alternatively, you can instantiate an array dynamically by calling Array.Create Instance. This allows you to specify element type and rank (number of dimen‐ sions) at runtime as well as allowing nonzero-based arrays through specifying a lower bound. Nonzero-based arrays are not compatible with the .NET Common The Array Class | 357 Collections
- 376. Language Specification (CLS) and should not be exposed as public members in a library that might be consumed by a program written in F# or Visual Basic. The GetValue and SetValue methods let you access elements in a dynamically created array (they also work on ordinary arrays): // Create a string array 2 elements in length: Array a = Array.CreateInstance (typeof(string), 2); a.SetValue ("hi", 0); // → a[0] = "hi"; a.SetValue ("there", 1); // → a[1] = "there"; string s = (string) a.GetValue (0); // → s = a[0]; // We can also cast to a C# array as follows: string[] cSharpArray = (string[]) a; string s2 = cSharpArray [0]; Zero-indexed arrays created dynamically can be cast to a C# array of a matching or compatible type (compatible by standard array-variance rules). For example, if Apple subclasses Fruit, Apple[] can be cast to Fruit[]. This leads to the issue of why object[] was not used as the unifying array type rather than the Array class. The answer is that object[] is incompatible with both multidimensional and value-type arrays (and nonzero-based arrays). An int[] array cannot be cast to object[]. Hence, we require the Array class for full type unification. GetValue and SetValue also work on compiler-created arrays, and they are useful when writing methods that can deal with an array of any type and rank. For multidimensional arrays, they accept an array of indexers: public object GetValue (params int[] indices) public void SetValue (object value, params int[] indices) The following method prints the first element of any array, regardless of rank: void WriteFirstValue (Array a) { Console.Write (a.Rank + "-dimensional; "); // The indexers array will automatically initialize to all zeros, so // passing it into GetValue or SetValue will get/set the zero-based // (i.e., first) element in the array. int[] indexers = new int[a.Rank]; Console.WriteLine ("First value is " + a.GetValue (indexers)); } void Demo() { int[] oneD = { 1, 2, 3 }; int[,] twoD = { {5,6}, {8,9} }; WriteFirstValue (oneD); // 1-dimensional; first value is 1 WriteFirstValue (twoD); // 2-dimensional; first value is 5 } 358 | Chapter 7: Collections
- 377. For working with arrays of unknown type but known rank, generics provide an easier and more efficient solution: void WriteFirstValue<T> (T[] array) { Console.WriteLine (array[0]); } SetValue throws an exception if the element is of an incompatible type for the array. When an array is instantiated, whether via language syntax or Array.CreateIn stance, its elements are automatically initialized. For arrays with reference type elements, this means writing nulls; for arrays with value-type elements, this means calling the value-type’s default constructor (effectively “zeroing” the members). The Array class also provides this functionality on demand via the Clear method: public static void Clear (Array array, int index, int length); This method doesn’t affect the size of the array. This is in contrast to the usual use of Clear (such as in ICollection<T>.Clear) whereby the collection is reduced to zero elements. Enumeration Arrays are easily enumerated with a foreach statement: int[] myArray = { 1, 2, 3}; foreach (int val in myArray) Console.WriteLine (val); You can also enumerate using the static Array.ForEach method, defined as follows: public static void ForEach<T> (T[] array, Action<T> action); This uses an Action delegate, with this signature: public delegate void Action<T> (T obj); Here’s the first example rewritten with Array.ForEach: Array.ForEach (new[] { 1, 2, 3 }, Console.WriteLine); Length and Rank Array provides the following methods and properties for querying length and rank: public int GetLength (int dimension); public long GetLongLength (int dimension); public int Length { get; } public long LongLength { get; } public int GetLowerBound (int dimension); public int GetUpperBound (int dimension); public int Rank { get; } // Returns number of dimensions in array The Array Class | 359 Collections
- 378. GetLength and GetLongLength return the length for a given dimension (0 for a single-dimensional array), and Length and LongLength return the total number of elements in the array—all dimensions included. GetLowerBound and GetUpperBound are useful with nonzero indexed arrays. Get UpperBound returns the same result as adding GetLowerBound to GetLength for any given dimension. Searching The Array class offers a range of methods for finding elements within a one- dimensional array: BinarySearch methods For rapidly searching a sorted array for a particular item IndexOf/LastIndex methods For searching unsorted arrays for a particular item Find/FindLast/FindIndex/FindLastIndex/FindAll/Exists/TrueForAll For searching unsorted arrays for item(s) that satisfy a given Predicate<T> None of the array searching methods throws an exception if the specified value is not found. Instead, if an item is not found, methods returning an integer return −1 (assuming a zero-indexed array), and methods returning a generic type return the type’s default value (e.g., 0 for an int, or null for a string). The binary search methods are fast, but they work only on sorted arrays and require that the elements be compared for order rather than simply equality. To this effect, the binary search methods can accept an IComparer or IComparer<T> object to arbitrate on ordering decisions (see “Plugging in Equality and Order” on page 388). This must be consistent with any comparer used in originally sorting the array. If no comparer is provided, the type’s default ordering algorithm will be applied based on its implementation of IComparable / IComparable<T>. The IndexOf and LastIndexOf methods perform a simple enumeration over the array, returning the position of the first (or last) element that matches the given value. The predicate-based searching methods allow a method delegate or lambda expres‐ sion to arbitrate on whether a given element is a “match.” A predicate is simply a delegate accepting an object and returning true or false: public delegate bool Predicate<T> (T object); In the following example, we search an array of strings for a name containing the letter “a”: string[] names = { "Rodney", "Jack", "Jill" }; string match = Array.Find (names, ContainsA); Console.WriteLine (match); // Jack 360 | Chapter 7: Collections
- 379. ContainsA (string name) { return name.Contains ("a"); } Here’s the same code shortened with a lambda expression: string[] names = { "Rodney", "Jack", "Jill" }; string match = Array.Find (names, n => n.Contains ("a")); // Jack FindAll returns an array of all items satisfying the predicate. In fact, it’s equivalent to Enumerable.Where in the System.Linq namespace, except that FindAll returns an array of matching items rather than an IEnumerable<T> of the same. Exists returns true if any array member satisfies the given predicate, and is equiva‐ lent to Any in System.Linq.Enumerable. TrueForAll returns true if all items satisfy the predicate, and is equivalent to All in System.Linq.Enumerable. Sorting Array has the following built-in sorting methods: // For sorting a single array: public static void Sort<T> (T[] array); public static void Sort (Array array); // For sorting a pair of arrays: public static void Sort<TKey,TValue> (TKey[] keys, TValue[] items); public static void Sort (Array keys, Array items); Each of these methods is additionally overloaded to also accept the following: int index // Starting index at which to begin sorting int length // Number of elements to sort IComparer<T> comparer // Object making ordering decisions Comparison<T> comparison // Delegate making ordering decisions The following illustrates the simplest use of Sort: int[] numbers = { 3, 2, 1 }; Array.Sort (numbers); // Array is now { 1, 2, 3 } The methods accepting a pair of arrays work by rearranging the items of each array in tandem, basing the ordering decisions on the first array. In the next example, both the numbers and their corresponding words are sorted into numerical order: int[] numbers = { 3, 2, 1 }; string[] words = { "three", "two", "one" }; Array.Sort (numbers, words); // numbers array is now { 1, 2, 3 } // words array is now { "one", "two", "three" } The Array Class | 361 Collections
- 380. Array.Sort requires that the elements in the array implement IComparable (see “Order Comparison” on page 335). This means that most built-in C# types (such as integers, as in the preceding example) can be sorted. If the elements are not intrins‐ ically comparable or you want to override the default ordering, you must provide Sort with a custom comparison provider that reports on the relative position of two elements. There are ways to do this: • Via a helper object that implements IComparer /IComparer<T> (see “Plugging • in Equality and Order” on page 388) • Via a Comparison delegate: • public delegate int Comparison<T> (T x, T y); The Comparison delegate follows the same semantics as IComparer<T>.CompareTo: if x comes before y, a negative integer is returned; if x comes after y, a positive integer is returned; if x and y have the same sorting position, 0 is returned. In the following example, we sort an array of integers such that the odd numbers come first: int[] numbers = { 1, 2, 3, 4, 5 }; Array.Sort (numbers, (x, y) => x % 2 == y % 2 ? 0 : x % 2 == 1 ? -1 : 1); // numbers array is now { 1, 3, 5, 2, 4 } As an alternative to calling Sort, you can use LINQ’s OrderBy and ThenBy operators. Unlike Array.Sort, the LINQ opera‐ tors don’t alter the original array, instead emitting the sorted result in a fresh IEnumerable<T> sequence. Reversing Elements The following Array methods reverse the order of all—or a portion of—elements in an array: public static void Reverse (Array array); public static void Reverse (Array array, int index, int length); Copying Array provides four methods to perform shallow copying: Clone, CopyTo, Copy, and ConstrainedCopy. The former two are instance methods; the latter two are static methods. The Clone method returns a whole new (shallow-copied) array. The CopyTo and Copy methods copy a contiguous subset of the array. Copying a multidimensional rectangular array requires you to map the multidimensional index to a linear index. For example, the middle square (position[1,1]) in a 3 × 3 array is represented with the index 4, from the calculation: 1 * 3 + 1. The source and destination ranges can overlap without causing a problem. 362 | Chapter 7: Collections
- 381. ConstrainedCopy performs an atomic operation: if all of the requested elements cannot be successfully copied (due to a type error, for instance), the operation is rolled back. Array also provides an AsReadOnly method that returns a wrapper that prevents elements from being reassigned. Converting and Resizing Array.ConvertAll creates and returns a new array of element type TOutput, calling the supplied Converter delegate to copy over the elements. Converter is defined as follows: public delegate TOutput Converter<TInput,TOutput> (TInput input) The following converts an array of floats to an array of integers: float[] reals = { 1.3f, 1.5f, 1.8f }; int[] wholes = Array.ConvertAll (reals, r => Convert.ToInt32 (r)); // wholes array is { 1, 2, 2 } The Resize method works by creating a new array and copying over the elements, returning the new array via the reference parameter. However, any references to the original array in other objects will remain unchanged. The System.Linq namespace offers an additional buffet of extension methods suitable for array conversion. These meth‐ ods return an IEnumerable<T>, which you can convert back to an array via Enumerable’s ToArray method. Lists, Queues, Stacks, and Sets .NET provides a basic set of concrete collection classes that implement the inter‐ faces described in this chapter. This section concentrates on the list-like collections (versus the dictionary-like collections, which we cover in “Dictionaries” on page 372). As with the interfaces we discussed previously, you usually have a choice of generic or nongeneric versions of each type. In terms of flexibility and performance, the generic classes win, making their nongeneric counterparts redundant except for backward compatibility. This differs from the situation with collection interfaces, for which the nongeneric versions are still occasionally useful. Of the classes described in this section, the generic List class is the most commonly used. List<T> and ArrayList The generic List and nongeneric ArrayList classes provide a dynamically sized array of objects and are among the most commonly used of the collection classes. ArrayList implements IList, whereas List<T> implements both IList and IList<T> (and the read-only version, IReadOnlyList<T>). Unlike with arrays, Lists, Queues, Stacks, and Sets | 363 Collections
- 382. all interfaces are implemented publicly, and methods such as Add and Remove are exposed and work as you would expect. Internally, List<T> and ArrayList work by maintaining an internal array of objects, replaced with a larger array upon reaching capacity. Appending elements is efficient (because there is usually a free slot at the end), but inserting elements can be slow (because all elements after the insertion point must be shifted to make a free slot), as can removing elements (especially near the start). As with arrays, searching is efficient if the BinarySearch method is used on a list that has been sorted, but it is otherwise inefficient because each item must be individually checked. List<T> is up to several times faster than ArrayList if T is a value type, because List<T> avoids the overhead of boxing and unboxing elements. List<T> and ArrayList provide constructors that accept an existing collection of elements: these copy each element from the existing collection into the new List<T> or ArrayList: public class List<T> : IList<T>, IReadOnlyList<T> { public List (); public List (IEnumerable<T> collection); public List (int capacity); // Add+Insert public void Add (T item); public void AddRange (IEnumerable<T> collection); public void Insert (int index, T item); public void InsertRange (int index, IEnumerable<T> collection); // Remove public bool Remove (T item); public void RemoveAt (int index); public void RemoveRange (int index, int count); public int RemoveAll (Predicate<T> match); // Indexing public T this [int index] { get; set; } public List<T> GetRange (int index, int count); public Enumerator<T> GetEnumerator(); // Exporting, copying and converting: public T[] ToArray(); public void CopyTo (T[] array); public void CopyTo (T[] array, int arrayIndex); public void CopyTo (int index, T[] array, int arrayIndex, int count); public ReadOnlyCollection<T> AsReadOnly(); public List<TOutput> ConvertAll<TOutput> (Converter <T,TOutput> converter); // Other: public void Reverse(); // Reverses order of elements in list. public int Capacity { get;set; } // Forces expansion of internal array. 364 | Chapter 7: Collections
- 383. public void TrimExcess(); // Trims internal array back to size. public void Clear(); // Removes all elements, so Count=0. } public delegate TOutput Converter <TInput, TOutput> (TInput input); In addition to these members, List<T> provides instance versions of all of Array’s searching and sorting methods. The following code demonstrates List’s properties and methods (for examples on searching and sorting, see “The Array Class” on page 355): var words = new List<string>(); // New string-typed list words.Add ("melon"); words.Add ("avocado"); words.AddRange (new[] { "banana", "plum" } ); words.Insert (0, "lemon"); // Insert at start words.InsertRange (0, new[] { "peach", "nashi" }); // Insert at start words.Remove ("melon"); words.RemoveAt (3); // Remove the 4th element words.RemoveRange (0, 2); // Remove first 2 elements // Remove all strings starting in 'n': words.RemoveAll (s => s.StartsWith ("n")); Console.WriteLine (words [0]); // first word Console.WriteLine (words [words.Count - 1]); // last word foreach (string s in words) Console.WriteLine (s); // all words List<string> subset = words.GetRange (1, 2); // 2nd->3rd words string[] wordsArray = words.ToArray(); // Creates a new typed array // Copy first two elements to the end of an existing array: string[] existing = new string [1000]; words.CopyTo (0, existing, 998, 2); List<string> upperCaseWords = words.ConvertAll (s => s.ToUpper()); List<int> lengths = words.ConvertAll (s => s.Length); The nongeneric ArrayList class requires clumsy casts—as the following example demonstrates: ArrayList al = new ArrayList(); al.Add ("hello"); string first = (string) al [0]; string[] strArr = (string[]) al.ToArray (typeof (string)); Such casts cannot be verified by the compiler; the following compiles successfully but then fails at runtime: int first = (int) al [0]; // Runtime exception Lists, Queues, Stacks, and Sets | 365 Collections
- 384. An ArrayList is functionally similar to List<object>. Both are useful when you need a list of mixed-type elements that share no common base type (other than object). A possi‐ ble advantage of choosing an ArrayList, in this case, would be if you need to deal with the list using reflection (Chap‐ ter 19). Reflection is easier with a nongeneric ArrayList than a List<object>. If you import the System.Linq namespace, you can convert an ArrayList to a generic List by calling Cast and then ToList: ArrayList al = new ArrayList(); al.AddRange (new[] { 1, 5, 9 } ); List<int> list = al.Cast<int>().ToList(); Cast and ToList are extension methods in the System.Linq.Enumerable class. LinkedList<T> LinkedList<T> is a generic doubly linked list (see Figure 7-4). A doubly linked list is a chain of nodes in which each references the node before, the node after, and the actual element. Its main benefit is that an element can always be inserted efficiently anywhere in the list because it just involves creating a new node and updating a few references. However, finding where to insert the node in the first place can be slow because there’s no intrinsic mechanism to index directly into a linked list; each node must be traversed, and binary-chop searches are not possible. Figure 7-4. LinkedList<T> LinkedList<T> implements IEnumerable<T> and ICollection<T> (and their non‐ generic versions), but not IList<T> because access by index is not supported. List nodes are implemented via the following class: 366 | Chapter 7: Collections
- 385. public sealed class LinkedListNode<T> { public LinkedList<T> List { get; } public LinkedListNode<T> Next { get; } public LinkedListNode<T> Previous { get; } public T Value { get; set; } } When adding a node, you can specify its position either relative to another node or at the start/end of the list. LinkedList<T> provides the following methods for this: public void AddFirst(LinkedListNode<T> node); public LinkedListNode<T> AddFirst (T value); public void AddLast (LinkedListNode<T> node); public LinkedListNode<T> AddLast (T value); public void AddAfter (LinkedListNode<T> node, LinkedListNode<T> newNode); public LinkedListNode<T> AddAfter (LinkedListNode<T> node, T value); public void AddBefore (LinkedListNode<T> node, LinkedListNode<T> newNode); public LinkedListNode<T> AddBefore (LinkedListNode<T> node, T value); Similar methods are provided to remove elements: public void Clear(); public void RemoveFirst(); public void RemoveLast(); public bool Remove (T value); public void Remove (LinkedListNode<T> node); LinkedList<T> has internal fields to keep track of the number of elements in the list as well as the head and tail of the list. These are exposed in the following public properties: public int Count { get; } // Fast public LinkedListNode<T> First { get; } // Fast public LinkedListNode<T> Last { get; } // Fast LinkedList<T> also supports the following searching methods (each requiring that the list be internally enumerated): public bool Contains (T value); public LinkedListNode<T> Find (T value); public LinkedListNode<T> FindLast (T value); Finally, LinkedList<T> supports copying to an array for indexed processing and obtaining an enumerator to support the foreach statement: public void CopyTo (T[] array, int index); public Enumerator<T> GetEnumerator(); Lists, Queues, Stacks, and Sets | 367 Collections
- 386. Here’s a demonstration on the use of LinkedList<string>: var tune = new LinkedList<string>(); tune.AddFirst ("do"); // do tune.AddLast ("so"); // do - so tune.AddAfter (tune.First, "re"); // do - re- so tune.AddAfter (tune.First.Next, "mi"); // do - re - mi- so tune.AddBefore (tune.Last, "fa"); // do - re - mi - fa- so tune.RemoveFirst(); // re - mi - fa - so tune.RemoveLast(); // re - mi - fa LinkedListNode<string> miNode = tune.Find ("mi"); tune.Remove (miNode); // re - fa tune.AddFirst (miNode); // mi- re - fa foreach (string s in tune) Console.WriteLine (s); Queue<T> and Queue Queue<T> and Queue are first-in, first-out (FIFO) data structures, providing meth‐ ods to Enqueue (add an item to the tail of the queue) and Dequeue (retrieve and remove the item at the head of the queue). A Peek method is also provided to return the element at the head of the queue without removing it, as well as a Count property (useful in checking that elements are present before dequeuing). Although queues are enumerable, they do not implement IList<T>/IList, because members cannot be accessed directly by index. A ToArray method is provided, however, for copying the elements to an array from which they can be randomly accessed: public class Queue<T> : IEnumerable<T>, ICollection, IEnumerable { public Queue(); public Queue (IEnumerable<T> collection); // Copies existing elements public Queue (int capacity); // To lessen auto-resizing public void Clear(); public bool Contains (T item); public void CopyTo (T[] array, int arrayIndex); public int Count { get; } public T Dequeue(); public void Enqueue (T item); public Enumerator<T> GetEnumerator(); // To support foreach public T Peek(); public T[] ToArray(); public void TrimExcess(); } The following is an example of using Queue<int>: var q = new Queue<int>(); q.Enqueue (10); q.Enqueue (20); 368 | Chapter 7: Collections
- 387. int[] data = q.ToArray(); // Exports to an array Console.WriteLine (q.Count); // "2" Console.WriteLine (q.Peek()); // "10" Console.WriteLine (q.Dequeue()); // "10" Console.WriteLine (q.Dequeue()); // "20" Console.WriteLine (q.Dequeue()); // throws an exception (queue empty) Queues are implemented internally using an array that’s resized as required—much like the generic List class. The queue maintains indexes that point directly to the head and tail elements; therefore, enqueuing and dequeuing are extremely quick operations (except when an internal resize is required). Stack<T> and Stack Stack<T> and Stack are last-in, first-out (LIFO) data structures, providing methods to Push (add an item to the top of the stack) and Pop (retrieve and remove an element from the top of the stack). A nondestructive Peek method is also provided, as is a Count property and a ToArray method for exporting the data for random access: public class Stack<T> : IEnumerable<T>, ICollection, IEnumerable { public Stack(); public Stack (IEnumerable<T> collection); // Copies existing elements public Stack (int capacity); // Lessens auto-resizing public void Clear(); public bool Contains (T item); public void CopyTo (T[] array, int arrayIndex); public int Count { get; } public Enumerator<T> GetEnumerator(); // To support foreach public T Peek(); public T Pop(); public void Push (T item); public T[] ToArray(); public void TrimExcess(); } The following example demonstrates Stack<int>: var s = new Stack<int>(); s.Push (1); // Stack = 1 s.Push (2); // Stack = 1,2 s.Push (3); // Stack = 1,2,3 Console.WriteLine (s.Count); // Prints 3 Console.WriteLine (s.Peek()); // Prints 3, Stack = 1,2,3 Console.WriteLine (s.Pop()); // Prints 3, Stack = 1,2 Console.WriteLine (s.Pop()); // Prints 2, Stack = 1 Console.WriteLine (s.Pop()); // Prints 1, Stack = <empty> Console.WriteLine (s.Pop()); // throws exception Stacks are implemented internally with an array that’s resized as required, as with Queue<T> and List<T>. Lists, Queues, Stacks, and Sets | 369 Collections
- 388. BitArray A BitArray is a dynamically sized collection of compacted bool values. It is more memory efficient than both a simple array of bool and a generic List of bool because it uses only one bit for each value, whereas the bool type otherwise occu‐ pies one byte for each value. BitArray’s indexer reads and writes individual bits: var bits = new BitArray(2); bits[1] = true; There are four bitwise operator methods (And, Or, Xor, and Not). All but the last accept another BitArray: bits.Xor (bits); // Bitwise exclusive-OR bits with itself Console.WriteLine (bits[1]); // False HashSet<T> and SortedSet<T> HashSet<T> and SortedSet<T> have the following distinguishing features: • Their Contains methods execute quickly using a hash-based lookup. • • They do not store duplicate elements and silently ignore requests to add • duplicates. • You cannot access an element by position. • SortedSet<T> keeps elements in order, whereas HashSet<T> does not. The commonality of the HashSet<T> and SortedSet<T> types is captured by the interface ISet<T>. From .NET 5, these classes also implement an interface called IReadOnlySet<T>, which is also implemented by the immutable set types (see “Immutable Collections” on page 384). HashSet<T> is implemented with a hashtable that stores just keys; SortedSet<T> is implemented with a red/black tree. Both collections implement ICollection<T> and offer methods that you would expect, such as Contains, Add, and Remove. In addition, there’s a predicate-based removal method called RemoveWhere. The following constructs a HashSet<char> from an existing collection, tests for membership, and then enumerates the collection (notice the absence of duplicates): var letters = new HashSet<char> ("the quick brown fox"); Console.WriteLine (letters.Contains ('t')); // true Console.WriteLine (letters.Contains ('j')); // false foreach (char c in letters) Console.Write (c); // the quickbrownfx 370 | Chapter 7: Collections
- 389. (The reason we can pass a string into HashSet<char>’s constructor is that string implements IEnumerable<char>.) The really interesting methods are the set operations. The following set operations are destructive in that they modify the set: public void UnionWith (IEnumerable<T> other); // Adds public void IntersectWith (IEnumerable<T> other); // Removes public void ExceptWith (IEnumerable<T> other); // Removes public void SymmetricExceptWith (IEnumerable<T> other); // Removes whereas the following methods simply query the set and so are nondestructive: public bool IsSubsetOf (IEnumerable<T> other); public bool IsProperSubsetOf (IEnumerable<T> other); public bool IsSupersetOf (IEnumerable<T> other); public bool IsProperSupersetOf (IEnumerable<T> other); public bool Overlaps (IEnumerable<T> other); public bool SetEquals (IEnumerable<T> other); UnionWith adds all the elements in the second set to the original set (excluding duplicates). IntersectWith removes the elements that are not in both sets. We can extract all of the vowels from our set of characters as follows: var letters = new HashSet<char> ("the quick brown fox"); letters.IntersectWith ("aeiou"); foreach (char c in letters) Console.Write (c); // euio ExceptWith removes the specified elements from the source set. Here, we strip all vowels from the set: var letters = new HashSet<char> ("the quick brown fox"); letters.ExceptWith ("aeiou"); foreach (char c in letters) Console.Write (c); // th qckbrwnfx SymmetricExceptWith removes all but the elements that are unique to one set or the other: var letters = new HashSet<char> ("the quick brown fox"); letters.SymmetricExceptWith ("the lazy brown fox"); foreach (char c in letters) Console.Write (c); // quicklazy Note that because HashSet<T> and SortedSet<T> implement IEnumerable<T>, you can use another type of set (or collection) as the argument to any of the set operation methods. SortedSet<T> offers all the members of HashSet<T>, plus the following: public virtual SortedSet<T> GetViewBetween (T lowerValue, T upperValue) public IEnumerable<T> Reverse() public T Min { get; } public T Max { get; } SortedSet<T> also accepts an optional IComparer<T> in its constructor (rather than an equality comparer). Lists, Queues, Stacks, and Sets | 371 Collections
- 390. Here’s an example of loading the same letters into a SortedSet<char>: var letters = new SortedSet<char> ("the quick brown fox"); foreach (char c in letters) Console.Write (c); // bcefhiknoqrtuwx Following on from this, we can obtain the letters in the set between f and i as follows: foreach (char c in letters.GetViewBetween ('f', 'i')) Console.Write (c); // fhi Dictionaries A dictionary is a collection in which each element is a key/value pair. Dictionaries are most commonly used for lookups and sorted lists. .NET defines a standard protocol for dictionaries, via the interfaces IDictionary and IDictionary<TKey,TValue>, as well as a set of general-purpose dictionary classes. The classes each differ in the following regard: • Whether items are stored in sorted sequence • • Whether items can be accessed by position (index) as well as by key • • Whether generic or nongeneric • • Whether it’s fast or slow to retrieve items by key from a large dictionary • Table 7-1 summarizes each of the dictionary classes and how they differ in these respects. The performance times are in milliseconds and based on performing 50,000 oper‐ ations on a dictionary with integer keys and values on a 1.5 GHz PC. (The differ‐ ences in performance between generic and nongeneric counterparts using the same underlying collection structure are due to boxing, and show up only with value-type elements.) Table 7-1. Dictionary classes Type Internal structure Retrieve by index? Memory overhead (avg. bytes per item) Speed: random insertion Speed: sequential insertion Speed: retrieval by key Unsorted Dictionary <K,V> Hashtable No 22 30 30 20 Hashtable Hashtable No 38 50 50 30 ListDictionary Linked list No 36 50,000 50,000 50,000 OrderedDictionary Hashtable + array Yes 59 70 70 40 372 | Chapter 7: Collections
- 391. Type Internal structure Retrieve by index? Memory overhead (avg. bytes per item) Speed: random insertion Speed: sequential insertion Speed: retrieval by key Sorted SortedDictionary <K,V> Red/black tree No 20 130 100 120 SortedList <K,V> 2xArray Yes 2 3,300 30 40 SortedList 2xArray Yes 27 4,500 100 180 In Big-O notation, retrieval time by key is as follows: • O(1) for Hashtable, Dictionary, and OrderedDictionary • • O(log n) for SortedDictionary and SortedList • • O(n) for ListDictionary (and nondictionary types such as List<T>) • n is the number of elements in the collection. IDictionary<TKey,TValue> IDictionary<TKey,TValue> defines the standard protocol for all key/value-based collections. It extends ICollection<T> by adding methods and properties to access elements based on a key of arbitrary type: public interface IDictionary <TKey, TValue> : ICollection <KeyValuePair <TKey, TValue>>, IEnumerable { bool ContainsKey (TKey key); bool TryGetValue (TKey key, out TValue value); void Add (TKey key, TValue value); bool Remove (TKey key); TValue this [TKey key] { get; set; } // Main indexer - by key ICollection <TKey> Keys { get; } // Returns just keys ICollection <TValue> Values { get; } // Returns just values } There’s also an interface called IReadOnlyDiction ary<TKey,TValue>, which defines the read-only subset of dic‐ tionary members. To add an item to a dictionary, you either call Add or use the index’s set accessor— the latter adds an item to the dictionary if the key is not already present (or updates the item if it is present). Duplicate keys are forbidden in all dictionary implementations, so calling Add twice with the same key throws an exception. Dictionaries | 373 Collections
- 392. To retrieve an item from a dictionary, use either the indexer or the TryGetValue method. If the key doesn’t exist, the indexer throws an exception, whereas TryGet Value returns false. You can test for membership explicitly by calling Contains Key; however, this incurs the cost of two lookups if you then subsequently retrieve the item. Enumerating directly over an IDictionary<TKey,TValue> returns a sequence of KeyValuePair structs: public struct KeyValuePair <TKey, TValue> { public TKey Key { get; } public TValue Value { get; } } You can enumerate over just the keys or values via the dictionary’s Keys/Values properties. We demonstrate the use of this interface with the generic Dictionary class in the following section. IDictionary The nongeneric IDictionary interface is the same in principle as IDiction ary<TKey,TValue>, apart from two important functional differences. It’s important to be aware of these differences, because IDictionary appears in legacy code (including the .NET BCL itself in places): • Retrieving a nonexistent key via the indexer returns null (rather than throwing • an exception). • Contains tests for membership rather than ContainsKey. • Enumerating over a nongeneric IDictionary returns a sequence of Dictionary Entry structs: public struct DictionaryEntry { public object Key { get; set; } public object Value { get; set; } } Dictionary<TKey,TValue> and Hashtable The generic Dictionary class is one of the most commonly used collections (along with the List<T> collection). It uses a hashtable data structure to store keys and values, and it is fast and efficient. The nongeneric version of Dictionary<TKey,TValue> is called Hashtable; there is no nongeneric class called Dictio nary. When we refer simply to Dictionary, we mean the generic Dictionary<TKey,TValue> class. 374 | Chapter 7: Collections
- 393. Dictionary implements both the generic and nongeneric IDictionary interfaces, the generic IDictionary being exposed publicly. Dictionary is, in fact, a “textbook” implementation of the generic IDictionary. Here’s how to use it: var d = new Dictionary<string, int>(); d.Add("One", 1); d["Two"] = 2; // adds to dictionary because "two" not already present d["Two"] = 22; // updates dictionary because "two" is now present d["Three"] = 3; Console.WriteLine (d["Two"]); // Prints "22" Console.WriteLine (d.ContainsKey ("One")); // true (fast operation) Console.WriteLine (d.ContainsValue (3)); // true (slow operation) int val = 0; if (!d.TryGetValue ("onE", out val)) Console.WriteLine ("No val"); // "No val" (case sensitive) // Three different ways to enumerate the dictionary: foreach (KeyValuePair<string, int> kv in d) // One; 1 Console.WriteLine (kv.Key + "; " + kv.Value); // Two; 22 // Three; 3 foreach (string s in d.Keys) Console.Write (s); // OneTwoThree Console.WriteLine(); foreach (int i in d.Values) Console.Write (i); // 1223 Its underlying hashtable works by converting each element’s key into an integer hashcode—a pseudo-unique value—and then applying an algorithm to convert the hashcode into a hash key. This hash key is used internally to determine which “bucket” an entry belongs to. If the bucket contains more than one value, a linear search is performed on the bucket. A good hash function does not strive to return strictly unique hashcodes (which would usually be impossible); it strives to return hashcodes that are evenly distributed across the 32-bit integer space. This prevents the scenario of ending up with a few very large (and inefficient) buckets. A dictionary can work with keys of any type, providing it’s able to determine equality between keys and obtain hashcodes. By default, equality is determined via the key’s object.Equals method, and the pseudo-unique hashcode is obtained via the key’s GetHashCode method. You can change this behavior either by overriding these methods or by providing an IEqualityComparer object when constructing the dictionary. A common application of this is to specify a case-insensitive equality comparer when using string keys: var d = new Dictionary<string, int> (StringComparer.OrdinalIgnoreCase); We discuss this further in “Plugging in Equality and Order” on page 388. Dictionaries | 375 Collections
- 394. As with many other types of collections, you can improve the performance of a dictionary slightly by specifying the collection’s expected size in the constructor, avoiding or lessening the need for internal resizing operations. The nongeneric version is named Hashtable and is functionally similar apart from differences stemming from it exposing the nongeneric IDictionary interface dis‐ cussed previously. The downside to Dictionary and Hashtable is that the items are not sorted. Furthermore, the original order in which the items were added is not retained. As with all dictionaries, duplicate keys are not allowed. When the generic collections were introduced back in 2005, the CLR team chose to name them according to what they represent (Dictionary, List) rather than how they are inter‐ nally implemented (Hashtable, ArrayList). Although this is good because it gives them the freedom to later change the implementation, it also means that the performance contract (often the most important criteria in choosing one kind of collection over another) is no longer captured in the name. OrderedDictionary An OrderedDictionary is a nongeneric dictionary that maintains elements in the same order that they were added. With an OrderedDictionary, you can access elements both by index and by key. An OrderedDictionary is not a sorted dictionary. An OrderedDictionary is a combination of a Hashtable and an ArrayList. This means that it has all the functionality of a Hashtable, plus functions such as RemoveAt and an integer indexer. It also exposes Keys and Values properties that return elements in their original order. This class was introduced in .NET 2.0, yet, peculiarly, there’s no generic version. ListDictionary and HybridDictionary ListDictionary uses a singly linked list to store the underlying data. It doesn’t provide sorting, although it does preserve the original entry order of the items. ListDictionary is extremely slow with large lists. Its only real “claim to fame” is its efficiency with very small lists (fewer than 10 items). HybridDictionary is a ListDictionary that automatically converts to a Hashtable upon reaching a certain size, to address ListDictionary’s problems with perfor‐ mance. The idea is to get a low memory footprint when the dictionary is small and good performance when the dictionary is large. However, given the overhead in converting from one to the other—and the fact that a Dictionary is not excessively 376 | Chapter 7: Collections
- 395. 1 There’s also a functionally identical nongeneric version of this called SortedList. heavy or slow in either scenario—you wouldn’t suffer unreasonably by using a Dictionary to begin with. Both classes come only in nongeneric form. Sorted Dictionaries The .NET BCL provides two dictionary classes internally structured such that their content is always sorted by key: • SortedDictionary<TKey,TValue> • • SortedList<TKey,TValue>1 • (In this section, we abbreviate <TKey,TValue> to <,>.) SortedDictionary<,> uses a red/black tree: a data structure designed to perform consistently well in any insertion or retrieval scenario. SortedList<,> is implemented internally with an ordered array pair, providing fast retrieval (via a binary-chop search) but poor insertion performance (because existing values need to be shifted to make room for a new entry). SortedDictionary<,> is much faster than SortedList<,> at inserting elements in a random sequence (particularly with large lists). SortedList<,>, however, has an extra ability: to access items by index as well as by key. With a sorted list, you can go directly to the nth element in the sorting sequence (via the indexer on the Keys/Values properties). To do the same with a SortedDictionary<,>, you must manually enumerate over n items. (Alternatively, you could write a class that combines a sorted dictionary with a list class.) None of the three collections allows duplicate keys (as is the case with all dictionaries). The following example uses reflection to load all of the methods defined in System.Object into a sorted list keyed by name and then enumerates their keys and values: // MethodInfo is in the System.Reflection namespace var sorted = new SortedList <string, MethodInfo>(); foreach (MethodInfo m in typeof (object).GetMethods()) sorted [m.Name] = m; foreach (string name in sorted.Keys) Console.WriteLine (name); Dictionaries | 377 Collections
- 396. foreach (MethodInfo m in sorted.Values) Console.WriteLine (m.Name + " returns a " + m.ReturnType); Here’s the result of the first enumeration: Equals GetHashCode GetType ReferenceEquals ToString Here’s the result of the second enumeration: Equals returns a System.Boolean GetHashCode returns a System.Int32 GetType returns a System.Type ReferenceEquals returns a System.Boolean ToString returns a System.String Notice that we populated the dictionary through its indexer. If we instead used the Add method, it would throw an exception because the object class upon which we’re reflecting overloads the Equals method, and you can’t add the same key twice to a dictionary. By using the indexer, the later entry overwrites the earlier entry, preventing this error. You can store multiple members of the same key by making each value element a list: SortedList <string, List<MethodInfo>> Extending our example, the following retrieves the MethodInfo whose key is "GetHashCode", just as with an ordinary dictionary: Console.WriteLine (sorted ["GetHashCode"]); // Int32 GetHashCode() So far, everything we’ve done would also work with a SortedDictionary<,>. The following two lines, however, which retrieve the last key and value, work only with a sorted list: Console.WriteLine (sorted.Keys [sorted.Count - 1]); // ToString Console.WriteLine (sorted.Values[sorted.Count - 1].IsVirtual); // True Customizable Collections and Proxies The collection classes discussed in previous sections are convenient in that you can directly instantiate them, but they don’t allow you to control what happens when an item is added to or removed from the collection. With strongly typed collections in an application, you sometimes need this control; for instance: • To fire an event when an item is added or removed • • To update properties because of the added or removed item • 378 | Chapter 7: Collections
- 397. • To detect an “illegal” add/remove operation and throw an exception (for exam‐ • ple, if the operation violates a business rule) The .NET BCL provides collection classes for this exact purpose, in the System.Col lections.ObjectModel namespace. These are essentially proxies or wrappers that implement IList<T> or IDictionary<,> by forwarding the methods through to an underlying collection. Each Add, Remove, or Clear operation is routed via a virtual method that acts as a “gateway” when overridden. Customizable collection classes are commonly used for publicly exposed col‐ lections; for instance, a collection of controls exposed publicly on a System .Windows.Form class. Collection<T> and CollectionBase The Collection<T> class is a customizable wrapper for List<T>. As well as implementing IList<T> and IList, it defines four additional virtual methods and a protected property as follows: public class Collection<T> : IList<T>, ICollection<T>, IEnumerable<T>, IList, ICollection, IEnumerable { // ... protected virtual void ClearItems(); protected virtual void InsertItem (int index, T item); protected virtual void RemoveItem (int index); protected virtual void SetItem (int index, T item); protected IList<T> Items { get; } } The virtual methods provide the gateway by which you can “hook in” to change or enhance the list’s normal behavior. The protected Items property allows the imple‐ menter to directly access the “inner list”—this is used to make changes internally without the virtual methods firing. The virtual methods need not be overridden; they can be left alone until there’s a requirement to alter the list’s default behavior. The following example demonstrates the typical “skeleton” use of Collection<T>: Zoo zoo = new Zoo(); zoo.Animals.Add (new Animal ("Kangaroo", 10)); zoo.Animals.Add (new Animal ("Mr Sea Lion", 20)); foreach (Animal a in zoo.Animals) Console.WriteLine (a.Name); public class Animal { public string Name; public int Popularity; public Animal (string name, int popularity) Customizable Collections and Proxies | 379 Collections
- 398. { Name = name; Popularity = popularity; } } public class AnimalCollection : Collection <Animal> { // AnimalCollection is already a fully functioning list of animals. // No extra code is required. } public class Zoo // The class that will expose AnimalCollection. { // This would typically have additional members. public readonly AnimalCollection Animals = new AnimalCollection(); } As it stands, AnimalCollection is no more functional than a simple List<Animal>; its role is to provide a base for future extension. To illustrate, let’s now add a Zoo property to Animal so that it can reference the Zoo in which it lives and override each of the virtual methods in Collection<Animal> to maintain that property automatically: public class Animal { public string Name; public int Popularity; public Zoo Zoo { get; internal set; } public Animal(string name, int popularity) { Name = name; Popularity = popularity; } } public class AnimalCollection : Collection <Animal> { Zoo zoo; public AnimalCollection (Zoo zoo) { this.zoo = zoo; } protected override void InsertItem (int index, Animal item) { base.InsertItem (index, item); item.Zoo = zoo; } protected override void SetItem (int index, Animal item) { base.SetItem (index, item); item.Zoo = zoo; } protected override void RemoveItem (int index) { this [index].Zoo = null; base.RemoveItem (index); } protected override void ClearItems() 380 | Chapter 7: Collections
- 399. { foreach (Animal a in this) a.Zoo = null; base.ClearItems(); } } public class Zoo { public readonly AnimalCollection Animals; public Zoo() { Animals = new AnimalCollection (this); } } Collection<T> also has a constructor accepting an existing IList<T>. Unlike with other collection classes, the supplied list is proxied rather than copied, meaning that subsequent changes will be reflected in the wrapping Collection<T> (although without Collection<T>’s virtual methods firing). Conversely, changes made via the Collection<T> will change the underlying list. CollectionBase CollectionBase is the nongeneric version of Collection<T>. This provides most of the same features as Collection<T> but is clumsier to use. Instead of the template methods InsertItem, RemoveItem, SetItem, and ClearItem, Collection Base has “hook” methods that double the number of methods required: OnInsert, OnInsertComplete, OnSet, OnSetComplete, OnRemove, OnRemoveComplete, OnClear, and OnClearComplete. Because CollectionBase is nongeneric, you must also implement typed methods when subclassing it—at a minimum, a typed indexer and Add method. KeyedCollection<TKey,TItem> and DictionaryBase KeyedCollection<TKey,TItem> subclasses Collection<TItem>. It both adds and subtracts functionality. What it adds is the ability to access items by key, much like with a dictionary. What it subtracts is the ability to proxy your own inner list. A keyed collection has some resemblance to an OrderedDictionary in that it com‐ bines a linear list with a hashtable. However, unlike OrderedDictionary, it doesn’t implement IDictionary and doesn’t support the concept of a key/value pair. Keys are obtained instead from the items themselves: via the abstract GetKeyForItem method. This means that enumerating a keyed collection is just like enumerating an ordinary list. You can best think of KeyedCollection<TKey,TItem> as Collection<TItem> plus fast lookup by key. Because it subclasses Collection<>, a keyed collection inherits all of Collec tion<>’s functionality, except for the ability to specify an existing list in construc‐ tion. The additional members it defines are as follows: public abstract class KeyedCollection <TKey, TItem> : Collection <TItem> Customizable Collections and Proxies | 381 Collections
- 400. // ... protected abstract TKey GetKeyForItem(TItem item); protected void ChangeItemKey(TItem item, TKey newKey); // Fast lookup by key - this is in addition to lookup by index. public TItem this[TKey key] { get; } protected IDictionary<TKey, TItem> Dictionary { get; } } GetKeyForItem is what the implementer overrides to obtain an item’s key from the underlying object. The ChangeItemKey method must be called if the item’s key property changes, in order to update the internal dictionary. The Dictionary property returns the internal dictionary used to implement the lookup, which is created when the first item is added. This behavior can be changed by specifying a creation threshold in the constructor, delaying the internal dictionary from being created until the threshold is reached (in the interim, a linear search is performed if an item is requested by key). A good reason not to specify a creation threshold is that having a valid dictionary can be useful in obtaining an ICollection<> of keys, via the Dictionary’s Keys property. This collection can then be passed on to a public property. The most common use for KeyedCollection<,> is in providing a collection of items accessible both by index and by name. To demonstrate this, let’s revisit the zoo, this time implementing AnimalCollection as a KeyedCollection<string, Animal>: public class Animal { string name; public string Name { get { return name; } set { if (Zoo != null) Zoo.Animals.NotifyNameChange (this, value); name = value; } } public int Popularity; public Zoo Zoo { get; internal set; } public Animal (string name, int popularity) { Name = name; Popularity = popularity; } } public class AnimalCollection : KeyedCollection <string, Animal> { Zoo zoo; public AnimalCollection (Zoo zoo) { this.zoo = zoo; } 382 | Chapter 7: Collections
- 401. internal void NotifyNameChange (Animal a, string newName) => this.ChangeItemKey (a, newName); protected override string GetKeyForItem (Animal item) => item.Name; // The following methods would be implemented as in the previous example protected override void InsertItem (int index, Animal item)... protected override void SetItem (int index, Animal item)... protected override void RemoveItem (int index)... protected override void ClearItems()... } public class Zoo { public readonly AnimalCollection Animals; public Zoo() { Animals = new AnimalCollection (this); } } The following code demonstrates its use: Zoo zoo = new Zoo(); zoo.Animals.Add (new Animal ("Kangaroo", 10)); zoo.Animals.Add (new Animal ("Mr Sea Lion", 20)); Console.WriteLine (zoo.Animals [0].Popularity); // 10 Console.WriteLine (zoo.Animals ["Mr Sea Lion"].Popularity); // 20 zoo.Animals ["Kangaroo"].Name = "Mr Roo"; Console.WriteLine (zoo.Animals ["Mr Roo"].Popularity); // 10 DictionaryBase The nongeneric version of KeyedCollection is called DictionaryBase. This legacy class takes a very different approach in that it implements IDictionary and uses clumsy hook methods like CollectionBase: OnInsert, OnInsertComplete, OnSet, OnSetComplete, OnRemove, OnRemoveComplete, OnClear, and OnClearComplete (and additionally, OnGet). The primary advantage of implementing IDictionary over taking the KeyedCollection approach is that you don’t need to subclass it in order to obtain keys. But since the very purpose of DictionaryBase is to be subclassed, it’s no advantage at all. The improved model in KeyedCollection is almost certainly due to the fact that it was written some years later, with the benefit of hindsight. DictionaryBase is best considered useful for backward compatibility. ReadOnlyCollection<T> ReadOnlyCollection<T> is a wrapper, or proxy, that provides a read-only view of a collection. This is useful in allowing a class to publicly expose read-only access to a collection that the class can still update internally. A read-only collection accepts the input collection in its constructor, to which it maintains a permanent reference. It doesn’t take a static copy of the input collection, so subsequent changes to the input collection are visible through the read-only wrapper. Customizable Collections and Proxies | 383 Collections
- 402. To illustrate, suppose that your class wants to provide read-only public access to a list of strings called Names. We could do this as follows: public class Test { List<string> names = new List<string>(); public IReadOnlyList<string> Names => names; } Although Names returns a read-only interface, the consumer can still downcast at runtime to List<string> or IList<string> and then call Add, Remove, or Clear on the list. ReadOnlyCollection<T> provides a more robust solution: public class Test { List<string> names = new List<string>(); public ReadOnlyCollection<string> Names { get; private set; } public Test() => Names = new ReadOnlyCollection<string> (names); public void AddInternally() => names.Add ("test"); } Now, only members within the Test class can alter the list of names: Test t = new Test(); Console.WriteLine (t.Names.Count); // 0 t.AddInternally(); Console.WriteLine (t.Names.Count); // 1 t.Names.Add ("test"); // Compiler error ((IList<string>) t.Names).Add ("test"); // NotSupportedException Immutable Collections We just described how ReadOnlyCollection<T> creates a read-only view of a col‐ lection. Restricting the ability to write (mutate) a collection—or any other object— simplifies software and reduces bugs. The immutable collections extend this principle, by providing collections that cannot be modified at all after initialization. Should you need to add an item to an immuta‐ ble collection, you must instantiate a new collection, leaving the old one untouched. Immutability is a hallmark of functional programming and has the following benefits: • It eliminates a large class of bugs associated with changing state. • • It vastly simplifies parallelism and multithreading, by preventing most of the • thread-safety problems that we describe in Chapters 14, 22, and 23. • It makes code easier to reason about. • 384 | Chapter 7: Collections
- 403. The disadvantage of immutability is that when you need to make a change, you must create a whole new object. This incurs a performance hit, although there are mitigating strategies that we discuss in this section, including the ability to reuse portions of the original structure. The immutable collections are part of .NET (in .NET Framework, they are available via the System.Collections.Immutable NuGet package). All collections are defined in the System.Collections.Immutable namespace: Type Internal structure ImmutableArray<T> Array ImmutableList<T> AVL tree ImmutableDictionary<K,V> AVL tree ImmutableHashSet<T> AVL tree ImmutableSortedDictionary<K,V> AVL tree ImmutableSortedSet<T> AVL tree ImmutableStack<T> Linked list ImmutableQueue<T> Linked list The ImmutableArray<T> and ImmutableList<T> types are both immutable versions of List<T>. Both do the same job but with different performance characteristics that we discuss in “Immutable Collections and Performance” on page 387. The immutable collections expose a public interface similar to their mutable coun‐ terparts. The key difference is that the methods that appear to alter the collection (such as Add or Remove) don’t alter the original collection; instead they return a new collection with the requested item added or removed. Immutable collections prevent the adding and removing of items; they don’t prevent the items themselves from being mutated. To get the full benefits of immutability, you need to ensure that only immutable items end up in an immutable collection. Creating Immutable Collections Each immutable collection type offers a Create<T>() method, which accepts optional initial values and returns an initialized immutable collection: ImmutableArray<int> array = ImmutableArray.Create<int> (1, 2, 3); Each collection also offers a CreateRange<T> method, which does the same job as Create<T>; the difference is that its parameter type is IEnumerable<T> instead of params T[]. Immutable Collections | 385 Collections
- 404. You can also create an immutable collection from an existing IEnumerable<T>, using appropriate extension methods (ToImmutableArray, ToImmutableList, ToIm mutableDictionary, and so on): var list = new[] { 1, 2, 3 }.ToImmutableList(); Manipulating Immutable Collections The Add method returns a new collection containing the existing elements plus the new one: var oldList = ImmutableList.Create<int> (1, 2, 3); ImmutableList<int> newList = oldList.Add (4); Console.WriteLine (oldList.Count); // 3 (unaltered) Console.WriteLine (newList.Count); // 4 The Remove method operates in the same fashion, returning a new collection with the item removed. Repeatedly adding or removing elements in this manner is inefficient, because a new immutable collection is created for each add or remove operation. A better solution is to call AddRange (or RemoveRange), which accepts an IEnumerable<T> of items, which are all added or removed in one go: var anotherList = oldList.AddRange (new[] { 4, 5, 6 }); The immutable list and array also defines Insert and InsertRange methods to insert elements at a particular index, a RemoveAt method to remove at an index, and RemoveAll, which removes based on a predicate. Builders For more complex initialization needs, each immutable collection class defines a builder counterpart. Builders are classes that are functionally equivalent to a muta‐ ble collection, with similar performance characteristics. After the data is initialized, calling .ToImmutable() on a builder returns an immutable collection: ImmutableArray<int>.Builder builder = ImmutableArray.CreateBuilder<int>(); builder.Add (1); builder.Add (2); builder.Add (3); builder.RemoveAt (0); ImmutableArray<int> myImmutable = builder.ToImmutable(); You also can use builders to batch multiple updates to an existing immutable collection: var builder2 = myImmutable.ToBuilder(); builder2.Add (4); // Efficient builder2.Remove (2); // Efficient ... // More changes to builder... 386 | Chapter 7: Collections
- 405. // Return a new immutable collection with all the changes applied: ImmutableArray<int> myImmutable2 = builder2.ToImmutable(); Immutable Collections and Performance Most of the immutable collections use an AVL tree internally, which allows the add/ remove operations to reuse portions of the original internal structure rather than having to re-create the entire thing from scratch. This reduces the overhead of add/ remove operations from potentially huge (with large collections) to just moderately large, but it comes at the cost of making read operations slower. The end result is that most immutable collections are slower than their mutable counterparts for both reading and writing. The most seriously affected is ImmutableList<T>, which for both read and add operations is 10 to 200 times slower than List<T> (depending on the size of the list). This is why ImmutableArray<T> exists: by using an array inside, it avoids the overhead for read operations (for which it’s comparable in performance to an ordinary mutable array). The flipside is that it’s much slower than (even) Immu tableList<T> for add operations because none of the original structure can be reused. Hence, ImmutableArray<T> is desirable when you want unimpeded read- performance and don’t expect many subsequent calls to Add or Remove (without using a builder): Type Read performance Add performance ImmutableList<T> Slow Slow ImmutableArray<T> Very fast Very slow Calling Remove on an ImmutableArray is more expensive than calling Remove on a List<T>—even in the worst-case scenario of removing the first element—because allocating the new collection places additional load on the garbage collector. Although the immutable collections as a whole incur a potentially significant per‐ formance cost, it’s important to keep the overall magnitude in perspective. An Add operation on an ImmutableList with a million elements is still likely to occur in less than a microsecond on a typical laptop, and a read operation, in less than 100 nanoseconds. And, if you need to perform write-operations in a loop, you can avoid the accumulated cost with a builder. The following factors also work to mitigate the costs: • Immutability allows for easy concurrency and parallelization (Chapter 23), so • you can employ all available cores. Parallelizing with mutable state easily leads to errors and requires the use of locks or concurrent collections, both of which hurt performance. Immutable Collections | 387 Collections
- 406. • With immutability, you don’t need to “defensively copy” collections or data • structures to guard against unexpected change. This was a factor in favoring the use of immutable collections in writing recent portions of Visual Studio. • In most typical programs, few collections have enough items for the difference • to matter. In addition to Visual Studio, the well-performing Microsoft Roslyn toolchain was built with immutable collections, demonstrating how the benefits can outweigh the costs. Plugging in Equality and Order In the sections “Equality Comparison” on page 324 and “Order Comparison” on page 335, we described the standard .NET protocols that make a type equatable, hashable, and comparable. A type that implements these protocols can function correctly in a dictionary or sorted list “out of the box.” More specifically: • A type for which Equals and GetHashCode return meaningful results can be • used as a key in a Dictionary or Hashtable. • A type that implements IComparable /IComparable<T> can be used as a key in • any of the sorted dictionaries or lists. A type’s default equating or comparison implementation typically reflects what is most “natural” for that type. Sometimes, however, the default behavior is not what you want. You might need a dictionary whose string type key is treated without respect to case. Or you might want a sorted list of customers, sorted by each customer’s postcode. For this reason, .NET also defines a matching set of “plug-in” protocols. The plug-in protocols achieve two things: • They allow you to switch in alternative equating or comparison behavior. • • They allow you to use a dictionary or sorted collection with a key type that’s • not intrinsically equatable or comparable. The plug-in protocols consist of the following interfaces: IEqualityComparer and IEqualityComparer<T> • Performs plug-in equality comparison and hashing • • Recognized by Hashtable and Dictionary • IComparer and IComparer<T> • Performs plug-in order comparison • • Recognized by the sorted dictionaries and collections; also, Array.Sort • 388 | Chapter 7: Collections
- 407. Each interface comes in both generic and nongeneric forms. The IEqualityCom parer interfaces also have a default implementation in a class called Equality Comparer. In addition, there are interfaces called IStructuralEquatable and IStructuralCom parable, which allow for the option of structural comparisons on classes and arrays. IEqualityComparer and EqualityComparer An equality comparer switches in nondefault equality and hashing behavior, pri‐ marily for the Dictionary and Hashtable classes. Recall the requirements of a hashtable-based dictionary. It needs answers to two questions for any given key: • Is it the same as another? • • What is its integer hashcode? • An equality comparer answers these questions by implementing the IEquality Comparer interfaces: public interface IEqualityComparer<T> { bool Equals (T x, T y); int GetHashCode (T obj); } public interface IEqualityComparer // Nongeneric version { bool Equals (object x, object y); int GetHashCode (object obj); } To write a custom comparer, you implement one or both of these interfaces (imple‐ menting both gives maximum interoperability). Because this is somewhat tedious, an alternative is to subclass the abstract EqualityComparer class, defined as follows: public abstract class EqualityComparer<T> : IEqualityComparer, IEqualityComparer<T> { public abstract bool Equals (T x, T y); public abstract int GetHashCode (T obj); bool IEqualityComparer.Equals (object x, object y); int IEqualityComparer.GetHashCode (object obj); public static EqualityComparer<T> Default { get; } } EqualityComparer implements both interfaces; your job is simply to override the two abstract methods. Plugging in Equality and Order | 389 Collections
- 408. The semantics for Equals and GetHashCode follow the same rules for object .Equals and object.GetHashCode, described in Chapter 6. In the following exam‐ ple, we define a Customer class with two fields and then write an equality comparer that matches both the first and last names: public class Customer { public string LastName; public string FirstName; public Customer (string last, string first) { LastName = last; FirstName = first; } } public class LastFirstEqComparer : EqualityComparer <Customer> { public override bool Equals (Customer x, Customer y) => x.LastName == y.LastName && x.FirstName == y.FirstName; public override int GetHashCode (Customer obj) => (obj.LastName + ";" + obj.FirstName).GetHashCode(); } To illustrate how this works, let’s create two customers: Customer c1 = new Customer ("Bloggs", "Joe"); Customer c2 = new Customer ("Bloggs", "Joe"); Because we’ve not overridden object.Equals, normal reference type equality semantics apply: Console.WriteLine (c1 == c2); // False Console.WriteLine (c1.Equals (c2)); // False The same default equality semantics apply when using these customers in a Dictionary without specifying an equality comparer: var d = new Dictionary<Customer, string>(); d [c1] = "Joe"; Console.WriteLine (d.ContainsKey (c2)); // False Now, with the custom equality comparer: var eqComparer = new LastFirstEqComparer(); var d = new Dictionary<Customer, string> (eqComparer); d [c1] = "Joe"; Console.WriteLine (d.ContainsKey (c2)); // True In this example, we would have to be careful not to change the customer’s First Name or LastName while it was in use in the dictionary; otherwise, its hashcode would change and the Dictionary would break. 390 | Chapter 7: Collections
- 409. EqualityComparer<T>.Default Calling EqualityComparer<T>.Default returns a general-purpose equality com‐ parer that you can use as an alternative to the static object.Equals method. The advantage is that it first checks whether T implements IEquatable<T>, and if so, it calls that implementation instead, avoiding the boxing overhead. This is particularly useful in generic methods: static bool Foo<T> (T x, T y) { bool same = EqualityComparer<T>.Default.Equals (x, y); ... ReferenceEqualityComparer.Instance (.NET 5+) From .NET 5, ReferenceEqualityComparer.Instance returns an equality com‐ parer that always applies referential equality. In the case of value types, its Equals method always returns false. IComparer and Comparer Comparers are used to switch in custom ordering logic for sorted dictionaries and collections. Note that a comparer is useless to the unsorted dictionaries such as Dictionary and Hashtable—these require an IEqualityComparer to get hashcodes. Similarly, an equality comparer is useless for sorted dictionaries and collections. Here are the IComparer interface definitions: public interface IComparer { int Compare(object x, object y); } public interface IComparer <in T> { int Compare(T x, T y); } As with equality comparers, there’s an abstract class that you can subtype instead of implementing the interfaces: public abstract class Comparer<T> : IComparer, IComparer<T> { public static Comparer<T> Default { get; } public abstract int Compare (T x, T y); // Implemented by you int IComparer.Compare (object x, object y); // Implemented for you } The following example illustrates a class that describes a wish as well as a comparer that sorts wishes by priority: Plugging in Equality and Order | 391 Collections
- 410. class Wish { public string Name; public int Priority; public Wish (string name, int priority) { Name = name; Priority = priority; } } class PriorityComparer : Comparer<Wish> { public override int Compare (Wish x, Wish y) { if (object.Equals (x, y)) return 0; // Optimization if (x == null) return -1; if (y == null) return 1; return x.Priority.CompareTo (y.Priority); } } The object.Equals check ensures that we can never contradict the Equals method. Calling the static object.Equals method in this case is better than calling x.Equals because it still works if x is null! Here’s how our PriorityComparer is used to sort a List: var wishList = new List<Wish>(); wishList.Add (new Wish ("Peace", 2)); wishList.Add (new Wish ("Wealth", 3)); wishList.Add (new Wish ("Love", 2)); wishList.Add (new Wish ("3 more wishes", 1)); wishList.Sort (new PriorityComparer()); foreach (Wish w in wishList) Console.Write (w.Name + " | "); // OUTPUT: 3 more wishes | Love | Peace | Wealth | In the next example, SurnameComparer allows you to sort surname strings in an order suitable for a phonebook listing: class SurnameComparer : Comparer <string> { string Normalize (string s) { s = s.Trim().ToUpper(); if (s.StartsWith ("MC")) s = "MAC" + s.Substring (2); return s; } public override int Compare (string x, string y) => Normalize (x).CompareTo (Normalize (y)); } 392 | Chapter 7: Collections
- 411. Here’s SurnameComparer in use in a sorted dictionary: var dic = new SortedDictionary<string,string> (new SurnameComparer()); dic.Add ("MacPhail", "second!"); dic.Add ("MacWilliam", "third!"); dic.Add ("McDonald", "first!"); foreach (string s in dic.Values) Console.Write (s + " "); // first! second! third! StringComparer StringComparer is a predefined plug-in class for equating and comparing strings, allowing you to specify language and case sensitivity. StringComparer implements both IEqualityComparer and IComparer (and their generic versions), so you can use it with any type of dictionary or sorted collection. Because StringComparer is abstract, you obtain instances via its static properties. StringComparer.Ordinal mirrors the default behavior for string equality compari‐ son and StringComparer.CurrentCulture for order comparison. Here are all of its static members: public static StringComparer CurrentCulture { get; } public static StringComparer CurrentCultureIgnoreCase { get; } public static StringComparer InvariantCulture { get; } public static StringComparer InvariantCultureIgnoreCase { get; } public static StringComparer Ordinal { get; } public static StringComparer OrdinalIgnoreCase { get; } public static StringComparer Create (CultureInfo culture, bool ignoreCase); In the following example, an ordinal case-insensitive dictionary is created such that dict["Joe"] and dict["JOE"] mean the same thing: var dict = new Dictionary<string, int> (StringComparer.OrdinalIgnoreCase); In the next example, an array of names is sorted, using Australian English: string[] names = { "Tom", "HARRY", "sheila" }; CultureInfo ci = new CultureInfo ("en-AU"); Array.Sort<string> (names, StringComparer.Create (ci, false)); The final example is a culture-aware version of the SurnameComparer we wrote in the previous section (to compare names suitable for a phonebook listing): class SurnameComparer : Comparer<string> { StringComparer strCmp; public SurnameComparer (CultureInfo ci) { // Create a case-sensitive, culture-sensitive string comparer strCmp = StringComparer.Create (ci, false); } Plugging in Equality and Order | 393 Collections
- 412. string Normalize (string s) { s = s.Trim(); if (s.ToUpper().StartsWith ("MC")) s = "MAC" + s.Substring (2); return s; } public override int Compare (string x, string y) { // Directly call Compare on our culture-aware StringComparer return strCmp.Compare (Normalize (x), Normalize (y)); } } IStructuralEquatable and IStructuralComparable As we discussed in Chapter 6, structs implement structural comparison by default: two structs are equal if all of their fields are equal. Sometimes, however, structural equality and order comparison are useful as plug-in options on other types, as well—such as arrays. The following interfaces help with this: public interface IStructuralEquatable { bool Equals (object other, IEqualityComparer comparer); int GetHashCode (IEqualityComparer comparer); } public interface IStructuralComparable { int CompareTo (object other, IComparer comparer); } The IEqualityComparer/IComparer that you pass in are applied to each individual element in the composite object. We can demonstrate this by using arrays. In the following example, we compare two arrays for equality, first using the default Equals method, then using IStructuralEquatable’s version: int[] a1 = { 1, 2, 3 }; int[] a2 = { 1, 2, 3 }; IStructuralEquatable se1 = a1; Console.Write (a1.Equals (a2)); // False Console.Write (se1.Equals (a2, EqualityComparer<int>.Default)); // True Here’s another example: string[] a1 = "the quick brown fox".Split(); string[] a2 = "THE QUICK BROWN FOX".Split(); IStructuralEquatable se1 = a1; bool isTrue = se1.Equals (a2, StringComparer.InvariantCultureIgnoreCase); 394 | Chapter 7: Collections
- 413. 8 LINQ Queries LINQ, or Language Integrated Query, is a set of language and runtime features for writing structured type-safe queries over local object collections and remote data sources. LINQ enables you to query any collection implementing IEnumerable<T>, whether an array, list, or XML Document Object Model (DOM), as well as remote data sources, such as tables in an SQL Server database. LINQ offers the benefits of both compile-time type checking and dynamic query composition. This chapter describes the LINQ architecture and the fundamentals of writing quer‐ ies. All core types are defined in the System.Linq and System.Linq.Expressions namespaces. The examples in this and the following two chapters are pre‐ loaded into an interactive querying tool called LINQPad. You can download LINQPad from http://www.linqpad.net. Getting Started The basic units of data in LINQ are sequences and elements. A sequence is any object that implements IEnumerable<T>, and an element is each item in the sequence. In the following example, names is a sequence, and "Tom", "Dick", and "Harry" are elements: string[] names = { "Tom", "Dick", "Harry" }; We call this a local sequence because it represents a local collection of objects in memory. A query operator is a method that transforms a sequence. A typical query operator accepts an input sequence and emits a transformed output sequence. In the Enumera ble class in System.Linq, there are around 40 query operators—all implemented as static extension methods. These are called standard query operators. 395
- 414. Queries that operate over local sequences are called local quer‐ ies or LINQ-to-objects queries. LINQ also supports sequences that can be dynamically fed from a remote data source such as an SQL Server database. These sequences additionally implement the IQueryable<T> interface and are supported through a matching set of stan‐ dard query operators in the Queryable class. We discuss this further in “Interpreted Queries” on page 424. A query is an expression that, when enumerated, transforms sequences with query operators. The simplest query comprises one input sequence and one operator. For instance, we can apply the Where operator on a simple array to extract those strings whose length is at least four characters, as follows: string[] names = { "Tom", "Dick", "Harry" }; IEnumerable<string> filteredNames = System.Linq.Enumerable.Where (names, n => n.Length >= 4); foreach (string n in filteredNames) Console.WriteLine (n); Dick Harry Because the standard query operators are implemented as extension methods, we can call Where directly on names, as though it were an instance method: IEnumerable<string> filteredNames = names.Where (n => n.Length >= 4); For this to compile, you must import the System.Linq namespace. Here’s a com‐ plete example: using System; using System.Collections.Generic; using System.Linq; string[] names = { "Tom", "Dick", "Harry" }; IEnumerable<string> filteredNames = names.Where (n => n.Length >= 4); foreach (string name in filteredNames) Console.WriteLine (name); Dick Harry We could further shorten our code by implicitly typing filter edNames: var filteredNames = names.Where (n => n.Length >= 4); This can hinder readability, however, outside of an IDE, where there are no tool tips to help. For this reason, we make less use of implicit typing in this chapter than you might in your own projects. 396 | Chapter 8: LINQ Queries
- 415. 1 The term is based on Eric Evans and Martin Fowler’s work on fluent interfaces. Most query operators accept a lambda expression as an argument. The lambda expression helps guide and shape the query. In our example, the lambda expression is as follows: n => n.Length >= 4 The input argument corresponds to an input element. In this case, the input argu‐ ment n represents each name in the array and is of type string. The Where operator requires that the lambda expression return a bool value, which if true, indicates that the element should be included in the output sequence. Here’s its signature: public static IEnumerable<TSource> Where<TSource> (this IEnumerable<TSource> source, Func<TSource,bool> predicate) The following query extracts all names that contain the letter “a”: IEnumerable<string> filteredNames = names.Where (n => n.Contains ("a")); foreach (string name in filteredNames) Console.WriteLine (name); // Harry So far, we’ve built queries using extension methods and lambda expressions. As you’ll see shortly, this strategy is highly composable in that it allows the chaining of query operators. In this book, we refer to this as fluent syntax.1 C# also provides another syntax for writing queries, called query expression syntax. Here’s our preced‐ ing query written as a query expression: IEnumerable<string> filteredNames = from n in names where n.Contains ("a") select n; Fluent syntax and query syntax are complementary. In the following two sections, we explore each in more detail. Fluent Syntax Fluent syntax is the most flexible and fundamental. In this section, we describe how to chain query operators to form more complex queries—and show why extension methods are important to this process. We also describe how to formulate lambda expressions for a query operator and introduce several new query operators. Chaining Query Operators In the preceding section, we showed two simple queries, each comprising a single query operator. To build more complex queries, you append additional query oper‐ ators to the expression, creating a chain. To illustrate, the following query extracts all strings containing the letter “a,” sorts them by length, and then converts the results to uppercase: Fluent Syntax | 397 LINQ Queries
- 416. using System; using System.Collections.Generic; using System.Linq; string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; IEnumerable<string> query = names .Where (n => n.Contains ("a")) .OrderBy (n => n.Length) .Select (n => n.ToUpper()); foreach (string name in query) Console.WriteLine (name); JAY MARY HARRY The variable, n, in our example, is privately scoped to each of the lambda expressions. We can reuse the identifier n for the same reason that we can reuse the identifier c in the following method: void Test() { foreach (char c in "string1") Console.Write (c); foreach (char c in "string2") Console.Write (c); foreach (char c in "string3") Console.Write (c); } Where, OrderBy, and Select are standard query operators that resolve to extension methods in the Enumerable class (if you import the System.Linq namespace). We already introduced the Where operator, which emits a filtered version of the input sequence. The OrderBy operator emits a sorted version of its input sequence; the Select method emits a sequence in which each input element is transformed or projected with a given lambda expression (n.ToUpper(), in this case). Data flows from left to right through the chain of operators, so the data is first filtered, then sorted, and then projected. A query operator never alters the input sequence; instead, it returns a new sequence. This is consistent with the functional programming paradigm from which LINQ was inspired. Here are the signatures of each of these extension methods (with the OrderBy signature slightly simplified): public static IEnumerable<TSource> Where<TSource> (this IEnumerable<TSource> source, Func<TSource,bool> predicate) public static IEnumerable<TSource> OrderBy<TSource,TKey> (this IEnumerable<TSource> source, Func<TSource,TKey> keySelector) public static IEnumerable<TResult> Select<TSource,TResult> (this IEnumerable<TSource> source, Func<TSource,TResult> selector) 398 | Chapter 8: LINQ Queries
- 417. When query operators are chained as in this example, the output sequence of one operator is the input sequence of the next. The complete query resembles a production line of conveyor belts, as illustrated in Figure 8-1. Figure 8-1. Chaining query operators We can construct the identical query progressively, as follows: // You must import the System.Linq namespace for this to compile: IEnumerable<string> filtered = names .Where (n => n.Contains ("a")); IEnumerable<string> sorted = filtered.OrderBy (n => n.Length); IEnumerable<string> finalQuery = sorted .Select (n => n.ToUpper()); finalQuery is compositionally identical to the query we constructed previously. Further, each intermediate step also comprises a valid query that we can execute: foreach (string name in filtered) Console.Write (name + "|"); // Harry|Mary|Jay| Console.WriteLine(); foreach (string name in sorted) Console.Write (name + "|"); // Jay|Mary|Harry| Console.WriteLine(); foreach (string name in finalQuery) Console.Write (name + "|"); // JAY|MARY|HARRY| Why extension methods are important Instead of using extension method syntax, you can use conventional static method syntax to call the query operators: IEnumerable<string> filtered = Enumerable.Where (names, n => n.Contains ("a")); IEnumerable<string> sorted = Enumerable.OrderBy (filtered, n => n.Length); IEnumerable<string> finalQuery = Enumerable.Select (sorted, n => n.ToUpper()); This is, in fact, how the compiler translates extension method calls. Shunning extension methods comes at a cost, however, if you want to write a query in a single statement as we did earlier. Let’s revisit the single-statement query—first in extension method syntax: Fluent Syntax | 399 LINQ Queries
- 418. IEnumerable<string> query = names.Where (n => n.Contains ("a")) .OrderBy (n => n.Length) .Select (n => n.ToUpper()); Its natural linear shape reflects the left-to-right flow of data and also keeps lambda expressions alongside their query operators (infix notation). Without extension methods, the query loses its fluency: IEnumerable<string> query = Enumerable.Select ( Enumerable.OrderBy ( Enumerable.Where ( names, n => n.Contains ("a") ), n => n.Length ), n => n.ToUpper() ); Composing Lambda Expressions In previous examples, we fed the following lambda expression to the Where operator: n => n.Contains ("a") // Input type = string, return type = bool. A lambda expression that takes a value and returns a bool is called a predicate. The purpose of the lambda expression depends on the particular query operator. With the Where operator, it indicates whether an element should be included in the output sequence. In the case of the OrderBy operator, the lambda expression maps each element in the input sequence to its sorting key. With the Select operator, the lambda expression determines how each element in the input sequence is transformed before being fed to the output sequence. A lambda expression in a query operator always works on individual elements in the input sequence—not the sequence as a whole. The query operator evaluates your lambda expression upon demand, typically once per element in the input sequence. Lambda expressions allow you to feed your own logic into the query operators. This makes the query operators versatile as well as being simple under the hood. Here’s a complete implementation of Enumera ble.Where, exception handling aside: public static IEnumerable<TSource> Where<TSource> (this IEnumerable<TSource> source, Func<TSource,bool> predicate) { foreach (TSource element in source) if (predicate (element)) yield return element; } 400 | Chapter 8: LINQ Queries
- 419. Lambda expressions and Func signatures The standard query operators utilize generic Func delegates. Func is a family of general-purpose generic delegates in the System namespace, defined with the fol‐ lowing intent: The type arguments in Func appear in the same order they do in lambda expressions. Hence, Func<TSource,bool> matches a TSource=>bool lambda expression: one that accepts a TSource argument and returns a bool value. Similarly, Func<TSource,TResult> matches a TSource=>TResult lambda expression. The Func delegates are listed in “Lambda Expressions” on page 174. Lambda expressions and element typing The standard query operators use the following type parameter names: Generic type letter Meaning TSource Element type for the input sequence TResult Element type for the output sequence (if different from TSource) TKey Element type for the key used in sorting, grouping, or joining TSource is determined by the input sequence. TResult and TKey are typically inferred from your lambda expression. For example, consider the signature of the Select query operator: public static IEnumerable<TResult> Select<TSource,TResult> (this IEnumerable<TSource> source, Func<TSource,TResult> selector) Func<TSource,TResult> matches a TSource=>TResult lambda expression: one that maps an input element to an output element. TSource and TResult can be different types, so the lambda expression can change the type of each element. Further, the lambda expression determines the output sequence type. The following query uses Select to transform string type elements to integer type elements: string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; IEnumerable<int> query = names.Select (n => n.Length); foreach (int length in query) Console.Write (length + "|"); // 3|4|5|4|3| The compiler can infer the type of TResult from the return value of the lambda expression. In this case, n.Length returns an int value, so TResult is inferred to be int. Fluent Syntax | 401 LINQ Queries
- 420. The Where query operator is simpler and requires no type inference for the output because input and output elements are of the same type. This makes sense because the operator merely filters elements; it does not transform them: public static IEnumerable<TSource> Where<TSource> (this IEnumerable<TSource> source, Func<TSource,bool> predicate) Finally, consider the signature of the OrderBy operator: // Slightly simplified: public static IEnumerable<TSource> OrderBy<TSource,TKey> (this IEnumerable<TSource> source, Func<TSource,TKey> keySelector) Func<TSource,TKey> maps an input element to a sorting key. TKey is inferred from your lambda expression and is separate from the input and output element types. For instance, we could choose to sort a list of names by length (int key) or alphabetically (string key): string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; IEnumerable<string> sortedByLength, sortedAlphabetically; sortedByLength = names.OrderBy (n => n.Length); // int key sortedAlphabetically = names.OrderBy (n => n); // string key You can call the query operators in Enumerable with tra‐ ditional delegates that refer to methods instead of lambda expressions. This approach is effective in simplifying certain kinds of local queries—particularly with LINQ to XML—and is demonstrated in Chapter 10. It doesn’t work with IQuerya ble<T>-based sequences, however (e.g., when querying a data‐ base), because the operators in Queryable require lambda expressions in order to emit expression trees. We discuss this later in “Interpreted Queries” on page 424. Natural Ordering The original ordering of elements within an input sequence is significant in LINQ. Some query operators rely on this ordering, such as Take, Skip, and Reverse. The Take operator outputs the first x elements, discarding the rest: int[] numbers = { 10, 9, 8, 7, 6 }; IEnumerable<int> firstThree = numbers.Take (3); // { 10, 9, 8 } The Skip operator ignores the first x elements and outputs the rest: IEnumerable<int> lastTwo = numbers.Skip (3); // { 7, 6 } Reverse does exactly as it says: IEnumerable<int> reversed = numbers.Reverse(); // { 6, 7, 8, 9, 10 } With local queries (LINQ-to-objects), operators such as Where and Select preserve the original ordering of the input sequence (as do all other query operators, except for those that specifically change the ordering). 402 | Chapter 8: LINQ Queries
- 421. Other Operators Not all query operators return a sequence. The element operators extract one ele‐ ment from the input sequence; examples are First, Last, and ElementAt: int[] numbers = { 10, 9, 8, 7, 6 }; int firstNumber = numbers.First(); // 10 int lastNumber = numbers.Last(); // 6 int secondNumber = numbers.ElementAt(1); // 9 int secondLowest = numbers.OrderBy(n=>n).Skip(1).First(); // 7 Because these operators return a single element, you don’t usually call further query operators on their result unless that element itself is a collection. The aggregation operators return a scalar value, usually of numeric type: int count = numbers.Count(); // 5; int min = numbers.Min(); // 6; The quantifiers return a bool value: bool hasTheNumberNine = numbers.Contains (9); // true bool hasMoreThanZeroElements = numbers.Any(); // true bool hasAnOddElement = numbers.Any (n => n % 2 != 0); // true Some query operators accept two input sequences. Examples are Concat, which appends one sequence to another, and Union, which does the same but with dupli‐ cates removed: int[] seq1 = { 1, 2, 3 }; int[] seq2 = { 3, 4, 5 }; IEnumerable<int> concat = seq1.Concat (seq2); // { 1, 2, 3, 3, 4, 5 } IEnumerable<int> union = seq1.Union (seq2); // { 1, 2, 3, 4, 5 } The joining operators also fall into this category. Chapter 9 covers all of the query operators in detail. Query Expressions C# provides a syntactic shortcut for writing LINQ queries, called query expressions. Contrary to popular belief, a query expression is not a means of embedding SQL into C#. In fact, the design of query expressions was inspired primarily by list comprehensions from functional programming languages such as LISP and Haskell, although SQL had a cosmetic influence. In this book, we refer to query expression syntax simply as query syntax. Query Expressions | 403 LINQ Queries
- 422. In the preceding section, we wrote a fluent-syntax query to extract strings contain‐ ing the letter “a,” sorted by length and converted to uppercase. Here’s the same thing in query syntax: using System; using System.Collections.Generic; using System.Linq; string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; IEnumerable<string> query = from n in names where n.Contains ("a") // Filter elements orderby n.Length // Sort elements select n.ToUpper(); // Translate each element (project) foreach (string name in query) Console.WriteLine (name); JAY MARY HARRY Query expressions always start with a from clause and end with either a select or group clause. The from clause declares a range variable (in this case, n), which you can think of as traversing the input sequence—rather like foreach. Figure 8-2 illustrates the complete syntax as a railroad diagram. To read this diagram, start at the left and then proceed along the track as if you were a train. For instance, after the manda‐ tory from clause, you can optionally include an orderby, where, let, or join clause. After that, you can either continue with a select or group clause, or go back and include another from, orderby, where, let, or join clause. 404 | Chapter 8: LINQ Queries
- 423. Figure 8-2. Query syntax The compiler processes a query expression by translating it into fluent syntax. It does this in a fairly mechanical fashion—much like it translates foreach statements into calls to GetEnumerator and MoveNext. This means that anything you can write in query syntax you can also write in fluent syntax. The compiler (initially) translates our example query into the following: IEnumerable<string> query = names.Where (n => n.Contains ("a")) .OrderBy (n => n.Length) .Select (n => n.ToUpper()); The Where, OrderBy, and Select operators then resolve using the same rules that would apply if the query were written in fluent syntax. In this case, they bind to extension methods in the Enumerable class because the System.Linq namespace is imported and names implements IEnumerable<string>. The compiler doesn’t specifically favor the Enumerable class, however, when translating query expres‐ sions. You can think of the compiler as mechanically injecting the words “Where,” Query Expressions | 405 LINQ Queries
- 424. “OrderBy,” and “Select” into the statement and then compiling it as though you had typed the method names yourself. This offers flexibility in how they resolve. The operators in the database queries that we write in later sections, for instance, will bind instead to extension methods in Queryable. If we remove the using System.Linq directive from our pro‐ gram, the query would not compile, since the Where, OrderBy, and Select methods would have nowhere to bind. Query expressions cannot compile unless you import System.Linq or another namespace with an implementation of these query methods. Range Variables The identifier immediately following the from keyword syntax is called the range variable. A range variable refers to the current element in the sequence on which the operation is to be performed. In our examples, the range variable n appears in every clause in the query. And yet, the variable actually enumerates over a different sequence with each clause: from n in names // n is our range variable where n.Contains ("a") // n = directly from the array orderby n.Length // n = subsequent to being filtered select n.ToUpper() // n = subsequent to being sorted This becomes clear when we examine the compiler’s mechanical translation to fluent syntax: names.Where (n => n.Contains ("a")) // Locally scoped n .OrderBy (n => n.Length) // Locally scoped n .Select (n => n.ToUpper()) // Locally scoped n As you can see, each instance of n is scoped privately to its own lambda expression. Query expressions also let you introduce new range variables via the following clauses: • let • • into • • An additional from clause • • join • We cover these later in this chapter in “Composition Strategies” on page 418 as well as in Chapter 9, in “Projecting” on page 449 and “Joining” on page 449. Query Syntax Versus SQL Syntax Query expressions look superficially like SQL, yet the two are very different. A LINQ query boils down to a C# expression, and so follows standard C# rules. For 406 | Chapter 8: LINQ Queries
- 425. example, with LINQ, you cannot use a variable before you declare it. In SQL, you can reference a table alias in the SELECT clause before defining it in a FROM clause. A subquery in LINQ is just another C# expression and so requires no special syntax. Subqueries in SQL are subject to special rules. With LINQ, data logically flows from left to right through the query. With SQL, the order is less well structured with regard to data flow. A LINQ query comprises a conveyor belt or pipeline of operators that accept and emit sequences whose element order can matter. An SQL query comprises a network of clauses that work mostly with unordered sets. Query Syntax Versus Fluent Syntax Query and fluent syntax each have advantages. Query syntax is simpler for queries that involve any of the following: • A let clause for introducing a new variable alongside the range variable • • SelectMany, Join, or GroupJoin, followed by an outer range variable reference • (We describe the let clause in “Composition Strategies” on page 418; we describe SelectMany, Join, and GroupJoin in Chapter 9.) The middle ground is queries that involve the simple use of Where, OrderBy, and Select. Either syntax works well; the choice here is largely personal. For queries that comprise a single operator, fluent syntax is shorter and less cluttered. Finally, there are many operators that have no keyword in query syntax. These require that you use fluent syntax—at least in part. This means any operator outside of the following: Where, Select, SelectMany OrderBy, ThenBy, OrderByDescending, ThenByDescending GroupBy, Join, GroupJoin Mixed-Syntax Queries If a query operator has no query-syntax support, you can mix query syntax and fluent syntax. The only restriction is that each query-syntax component must be complete (i.e., start with a from clause and end with a select or group clause). Assuming this array declaration string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; the following example counts the number of names containing the letter “a”: int matches = (from n in names where n.Contains ("a") select n).Count(); // 3 Query Expressions | 407 LINQ Queries
- 426. The next query obtains the first name in alphabetical order: string first = (from n in names orderby n select n).First(); // Dick The mixed-syntax approach is sometimes beneficial in more complex queries. With these simple examples, however, we could stick to fluent syntax throughout without penalty: int matches = names.Where (n => n.Contains ("a")).Count(); // 3 string first = names.OrderBy (n => n).First(); // Dick There are times when mixed-syntax queries offer by far the highest “bang for the buck” in terms of function and simplic‐ ity. It’s important not to unilaterally favor either query or flu‐ ent syntax; otherwise, you’ll be unable to write mixed-syntax queries when they are the best option. Where applicable, the remainder of this chapter shows key concepts in both fluent and query syntax. Deferred Execution An important feature of most query operators is that they execute not when con‐ structed but when enumerated (in other words, when MoveNext is called on its enumerator). Consider the following query: var numbers = new List<int> { 1 }; IEnumerable<int> query = numbers.Select (n => n * 10); // Build query numbers.Add (2); // Sneak in an extra element foreach (int n in query) Console.Write (n + "|"); // 10|20| The extra number that we sneaked into the list after constructing the query is included in the result because it’s not until the foreach statement runs that any filtering or sorting takes place. This is called deferred or lazy execution and is the same as what happens with delegates: Action a = () => Console.WriteLine ("Foo"); // We’ve not written anything to the Console yet. Now let’s run it: a(); // Deferred execution! All standard query operators provide deferred execution, with the following exceptions: • Operators that return a single element or scalar value, such as First or Count • • The following conversion operators: • ToArray, ToList, ToDictionary, ToLookup, ToHashSet 408 | Chapter 8: LINQ Queries
- 427. These operators cause immediate query execution because their result types have no mechanism to provide deferred execution. The Count method, for instance, returns a simple integer, which doesn’t then get enumerated. The following query is executed immediately: int matches = numbers.Where (n => n <= 2).Count(); // 1 Deferred execution is important because it decouples query construction from query execution. This allows you to construct a query in several steps and also makes database queries possible. Subqueries provide another level of indirection. Everything in a subquery is subject to deferred execution, including aggre‐ gation and conversion methods. We describe this in “Subquer‐ ies” on page 414. Reevaluation Deferred execution has another consequence: a deferred execution query is reevalu‐ ated when you reenumerate: var numbers = new List<int>() { 1, 2 }; IEnumerable<int> query = numbers.Select (n => n * 10); foreach (int n in query) Console.Write (n + "|"); // 10|20| numbers.Clear(); foreach (int n in query) Console.Write (n + "|"); // <nothing> There are a couple of reasons why reevaluation is sometimes disadvantageous: • Sometimes, you want to “freeze” or cache the results at a certain point in time. • • Some queries are computationally intensive (or rely on querying a remote • database), so you don’t want to unnecessarily repeat them. You can defeat reevaluation by calling a conversion operator such as ToArray or ToList. ToArray copies the output of a query to an array; ToList copies to a generic List<T>: var numbers = new List<int>() { 1, 2 }; List<int> timesTen = numbers .Select (n => n * 10) .ToList(); // Executes immediately into a List<int> numbers.Clear(); Console.WriteLine (timesTen.Count); // Still 2 Deferred Execution | 409 LINQ Queries
- 428. Captured Variables If your query’s lambda expressions capture outer variables, the query will honor the value of those variables at the time the query runs: int[] numbers = { 1, 2 }; int factor = 10; IEnumerable<int> query = numbers.Select (n => n * factor); factor = 20; foreach (int n in query) Console.Write (n + "|"); // 20|40| This can be a trap when building up a query within a for loop. For example, suppose that we want to remove all vowels from a string. The following, although inefficient, gives the correct result: IEnumerable<char> query = "Not what you might expect"; query = query.Where (c => c != 'a'); query = query.Where (c => c != 'e'); query = query.Where (c => c != 'i'); query = query.Where (c => c != 'o'); query = query.Where (c => c != 'u'); foreach (char c in query) Console.Write (c); // Nt wht y mght xpct Now watch what happens when we refactor this with a for loop: IEnumerable<char> query = "Not what you might expect"; string vowels = "aeiou"; for (int i = 0; i < vowels.Length; i++) query = query.Where (c => c != vowels[i]); foreach (char c in query) Console.Write (c); An IndexOutOfRangeException is thrown upon enumerating the query because, as we saw in Chapter 4 (see “Capturing Outer Variables” on page 176), the compiler scopes the iteration variable in the for loop as if it were declared outside the loop. Hence, each closure captures the same variable (i) whose value is 5 when the query is actually enumerated. To solve this, you must assign the loop variable to another variable declared inside the statement block: for (int i = 0; i < vowels.Length; i++) { char vowel = vowels[i]; query = query.Where (c => c != vowel); } This forces a fresh local variable to be captured on each loop iteration. Another way to solve the problem is to replace the for loop with a foreach loop: foreach (char vowel in vowels) query = query.Where (c => c != vowel); 410 | Chapter 8: LINQ Queries
- 429. How Deferred Execution Works Query operators provide deferred execution by returning decorator sequences. Unlike a traditional collection class such as an array or linked list, a decorator sequence (in general) has no backing structure of its own to store elements. Instead, it wraps another sequence that you supply at runtime, to which it maintains a permanent dependency. Whenever you request data from a decorator, it in turn must request data from the wrapped input sequence. The query operator’s transformation constitutes the “decora‐ tion.” If the output sequence performed no transformation, it would be a proxy rather than a decorator. Calling Where merely constructs the decorator wrapper sequence, which holds a reference to the input sequence, the lambda expression, and any other arguments supplied. The input sequence is enumerated only when the decorator is enumerated. Figure 8-3 illustrates the composition of the following query: IEnumerable<int> lessThanTen = new int[] { 5, 12, 3 }.Where (n => n < 10); Figure 8-3. Decorator sequence When you enumerate lessThanTen, you are, in effect, querying the array through the Where decorator. The good news—should you ever want to write your own query operator—is that implementing a decorator sequence is easy with a C# iterator. Here’s how you can write your own Select method: public static IEnumerable<TResult> MySelect<TSource,TResult> (this IEnumerable<TSource> source, Func<TSource,TResult> selector) { foreach (TSource element in source) yield return selector (element); } Deferred Execution | 411 LINQ Queries
- 430. This method is an iterator by virtue of the yield return statement. Functionally, it’s a shortcut for the following: public static IEnumerable<TResult> MySelect<TSource,TResult> (this IEnumerable<TSource> source, Func<TSource,TResult> selector) { return new SelectSequence (source, selector); } where SelectSequence is a (compiler-written) class whose enumerator encapsulates the logic in the iterator method. Hence, when you call an operator such as Select or Where, you’re doing nothing more than instantiating an enumerable class that decorates the input sequence. Chaining Decorators Chaining query operators creates a layering of decorators. Consider the following query: IEnumerable<int> query = new int[] { 5, 12, 3 }.Where (n => n < 10) .OrderBy (n => n) .Select (n => n * 10); Each query operator instantiates a new decorator that wraps the previous sequence (rather like a Russian nesting doll). Figure 8-4 illustrates the object model of this query. Note that this object model is fully constructed prior to any enumeration. Figure 8-4. Layered decorator sequences When you enumerate query, you’re querying the original array, transformed through a layering or chain of decorators. 412 | Chapter 8: LINQ Queries
- 431. Adding ToList onto the end of this query would cause the preceding operators to execute immediately, collapsing the whole object model into a single list. Figure 8-5 shows the same object composition in Unified Modeling Language (UML) syntax. Select’s decorator references the OrderBy decorator, which refer‐ ences Where’s decorator, which references the array. A feature of deferred execution is that you build the identical object model if you compose the query progressively: IEnumerable<int> source = new int[] { 5, 12, 3 }, filtered = source .Where (n => n < 10), sorted = filtered .OrderBy (n => n), query = sorted .Select (n => n * 10); Figure 8-5. UML decorator composition How Queries Are Executed Here are the results of enumerating the preceding query: foreach (int n in query) Console.WriteLine (n); 30 50 Behind the scenes, the foreach calls GetEnumerator on Select’s decorator (the last or outermost operator), which kicks off everything. The result is a chain of enumerators that structurally mirrors the chain of decorator sequences. Figure 8-6 illustrates the flow of execution as enumeration proceeds. Deferred Execution | 413 LINQ Queries
- 432. Figure 8-6. Execution of a local query In the first section of this chapter, we depicted a query as a production line of conveyor belts. Extending this analogy, we can say a LINQ query is a lazy produc‐ tion line, where the conveyor belts roll elements only upon demand. Constructing a query constructs a production line—with everything in place—but with nothing rolling. Then, when the consumer requests an element (enumerates over the query), the rightmost conveyor belt activates; this in turn triggers the others to roll—as and when input sequence elements are needed. LINQ follows a demand-driven pull model, rather than a supply-driven push model. This is important—as you’ll see later—in allowing LINQ to scale to querying SQL databases. Subqueries A subquery is a query contained within another query’s lambda expression. The following example uses a subquery to sort musicians by their last name: string[] musos = { "David Gilmour", "Roger Waters", "Rick Wright", "Nick Mason" }; IEnumerable<string> query = musos.OrderBy (m => m.Split().Last()); m.Split converts each string into a collection of words, upon which we then call the Last query operator. m.Split().Last is the subquery; query references the outer query. 414 | Chapter 8: LINQ Queries
- 433. Subqueries are permitted because you can put any valid C# expression on the right‐ hand side of a lambda. A subquery is simply another C# expression. This means that the rules for subqueries are a consequence of the rules for lambda expressions (and the behavior of query operators in general). The term subquery, in the general sense, has a broader mean‐ ing. For the purpose of describing LINQ, we use the term only for a query referenced from within the lambda expression of another query. In a query expression, a subquery amounts to a query referenced from an expression in any clause except the from clause. A subquery is privately scoped to the enclosing expression and can reference parameters in the outer lambda expression (or range variables in a query expression). m.Split().Last is a very simple subquery. The next query retrieves all strings in an array whose length matches that of the shortest string: string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; IEnumerable<string> outerQuery = names .Where (n => n.Length == names.OrderBy (n2 => n2.Length) .Select (n2 => n2.Length).First()); // Tom, Jay Here’s the same thing as a query expression: IEnumerable<string> outerQuery = from n in names where n.Length == (from n2 in names orderby n2.Length select n2.Length).First() select n; Because the outer range variable (n) is in scope for a subquery, we cannot reuse n as the subquery’s range variable. A subquery is executed whenever the enclosing lambda expression is evaluated. This means that a subquery is executed upon demand, at the discretion of the outer query. You could say that execution proceeds from the outside in. Local queries follow this model literally; interpreted queries (e.g., database queries) follow this model conceptually. Subqueries | 415 LINQ Queries
- 434. The subquery executes as and when required, to feed the outer query. As Figures 8-7 and 8-8 illustrate, the subquery in our example (the top conveyor belt in Figure 8-7) executes once for every outer loop iteration. Figure 8-7. Subquery composition We can express our preceding subquery more succinctly as follows: IEnumerable<string> query = from n in names where n.Length == names.OrderBy (n2 => n2.Length).First().Length select n; With the Min aggregation function, we can simplify the query further: IEnumerable<string> query = from n in names where n.Length == names.Min (n2 => n2.Length) select n; In “Interpreted Queries” on page 424, we describe how remote sources such as SQL tables can be queried. Our example makes an ideal database query because it would be processed as a unit, requiring only one round trip to the database server. This query, however, is inefficient for a local collection because the subquery is recalculated on each outer loop iteration. We can avoid this inefficiency by running the subquery separately (so that it’s no longer a subquery): int shortest = names.Min (n => n.Length); IEnumerable<string> query = from n in names where n.Length == shortest select n; 416 | Chapter 8: LINQ Queries
- 435. Figure 8-8. UML subquery composition Factoring out subqueries in this manner is nearly always desirable when querying local collections. An exception is when the subquery is correlated, meaning that it references the outer range variable. We explore correlated subqueries in “Projecting” on page 449. Subqueries and Deferred Execution An element or aggregation operator such as First or Count in a subquery doesn’t force the outer query into immediate execution—deferred execution still holds for the outer query. This is because subqueries are called indirectly—through a delegate in the case of a local query, or through an expression tree in the case of an interpreted query. An interesting case arises when you include a subquery within a Select expression. In the case of a local query, you’re actually projecting a sequence of queries—each itself subject to deferred execution. The effect is generally transparent, and it Subqueries | 417 LINQ Queries
- 436. serves to further improve efficiency. We revisit Select subqueries in some detail in Chapter 9. Composition Strategies In this section, we describe three strategies for building more complex queries: • Progressive query construction • • Using the into keyword • • Wrapping queries • All are chaining strategies and produce identical runtime queries. Progressive Query Building At the start of the chapter, we demonstrated how you could build a fluent query progressively: var filtered = names .Where (n => n.Contains ("a")); var sorted = filtered .OrderBy (n => n); var query = sorted .Select (n => n.ToUpper()); Because each of the participating query operators returns a decorator sequence, the resultant query is the same chain or layering of decorators that you would get from a single-expression query. There are a couple of potential benefits, however, to building queries progressively: • It can make queries easier to write. • • You can add query operators conditionally. For example, • if (includeFilter) query = query.Where (...) is more efficient than query = query.Where (n => !includeFilter || <expression>) because it avoids adding an extra query operator if includeFilter is false. A progressive approach is often useful in query comprehensions. Imagine that we want to remove all vowels from a list of names and then present in alphabetical order those whose length is still more than two characters. In fluent syntax, we could write this query as a single expression—by projecting before we filter: IEnumerable<string> query = names .Select (n => n.Replace ("a", "").Replace ("e", "").Replace ("i", "") .Replace ("o", "").Replace ("u", "")) .Where (n => n.Length > 2) .OrderBy (n => n); // Dck // Hrry // Mry 418 | Chapter 8: LINQ Queries
- 437. Rather than calling string’s Replace method five times, we could remove vowels from a string more efficiently with a regular expression: n => Regex.Replace (n, "[aeiou]", "") string’s Replace method has the advantage, though, of also working in database queries. Translating this directly into a query expression is troublesome because the select clause must come after the where and orderby clauses. And if we rearrange the query so as to project last, the result would be different: IEnumerable<string> query = from n in names where n.Length > 2 orderby n select n.Replace ("a", "").Replace ("e", "").Replace ("i", "") .Replace ("o", "").Replace ("u", ""); // Dck // Hrry // Jy // Mry // Tm Fortunately, there are a number of ways to get the original result in query syntax. The first is by querying progressively: IEnumerable<string> query = from n in names select n.Replace ("a", "").Replace ("e", "").Replace ("i", "") .Replace ("o", "").Replace ("u", ""); query = from n in query where n.Length > 2 orderby n select n; // Dck // Hrry // Mry The into Keyword The into keyword is interpreted in two very different ways by query expressions, depending on context. The meaning we’re describing now is for signaling query continuation (the other is for signaling a GroupJoin). The into keyword lets you “continue” a query after a projection and is a shortcut for progressively querying. With into, we can rewrite the preceding query as follows: IEnumerable<string> query = from n in names select n.Replace ("a", "").Replace ("e", "").Replace ("i", "") .Replace ("o", "").Replace ("u", "") into noVowel where noVowel.Length > 2 orderby noVowel select noVowel; Composition Strategies | 419 LINQ Queries
- 438. The only place you can use into is after a select or group clause. into “restarts” a query, allowing you to introduce fresh where, orderby, and select clauses. Although it’s easiest to think of into as restarting a query from the perspective of a query expression, it’s all one query when translated to its final fluent form. Hence, there’s no intrinsic performance hit with into. Nor do you lose any points for its use! The equivalent of into in fluent syntax is simply a longer chain of operators. Scoping rules All range variables are out of scope following an into keyword. The following will not compile: var query = from n1 in names select n1.ToUpper() into n2 // Only n2 is visible from here on. where n1.Contains ("x") // Illegal: n1 is not in scope. select n2; To see why, consider how this maps to fluent syntax: var query = names .Select (n1 => n1.ToUpper()) .Where (n2 => n1.Contains ("x")); // Error: n1 no longer in scope The original name (n1) is lost by the time the Where filter runs. Where’s input sequence contains only uppercase names, so it cannot filter based on n1. Wrapping Queries A query built progressively can be formulated into a single statement by wrapping one query around another. In general terms, var tempQuery = tempQueryExpr var finalQuery = from ... in tempQuery ... can be reformulated as: var finalQuery = from ... in (tempQueryExpr) Wrapping is semantically identical to progressive query building or using the into keyword (without the intermediate variable). The end result in all cases is a linear chain of query operators. For example, consider the following query: IEnumerable<string> query = from n in names select n.Replace ("a", "").Replace ("e", "").Replace ("i", "") .Replace ("o", "").Replace ("u", ""); query = from n in query where n.Length > 2 orderby n select n; 420 | Chapter 8: LINQ Queries
- 439. Reformulated in wrapped form, it’s the following: IEnumerable<string> query = from n1 in ( from n2 in names select n2.Replace ("a", "").Replace ("e", "").Replace ("i", "") .Replace ("o", "").Replace ("u", "") ) where n1.Length > 2 orderby n1 select n1; When converted to fluent syntax, the result is the same linear chain of operators as in previous examples: IEnumerable<string> query = names .Select (n => n.Replace ("a", "").Replace ("e", "").Replace ("i", "") .Replace ("o", "").Replace ("u", "")) .Where (n => n.Length > 2) .OrderBy (n => n); (The compiler does not emit the final .Select (n => n), because it’s redundant.) Wrapped queries can be confusing because they resemble the subqueries we wrote earlier. Both have the concept of an inner and outer query. When converted to fluent syntax, however, you can see that wrapping is simply a strategy for sequen‐ tially chaining operators. The end result bears no resemblance to a subquery, which embeds an inner query within the lambda expression of another. Returning to a previous analogy: when wrapping, the “inner” query amounts to the preceding conveyor belts. In contrast, a subquery rides above a conveyor belt and is activated upon demand through the conveyor belt’s lambda worker (as illustrated in Figure 8-7). Projection Strategies Object Initializers So far, all of our select clauses have projected scalar element types. With C# object initializers, you can project into more complex types. For example, suppose, as a first step in a query, we want to strip vowels from a list of names while still retaining the original versions alongside, for the benefit of subsequent queries. We can write the following class to assist: class TempProjectionItem { public string Original; // Original name public string Vowelless; // Vowel-stripped name } Projection Strategies | 421 LINQ Queries
- 440. We then can project into it with object initializers: string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; IEnumerable<TempProjectionItem> temp = from n in names select new TempProjectionItem { Original = n, Vowelless = n.Replace ("a", "").Replace ("e", "").Replace ("i", "") .Replace ("o", "").Replace ("u", "") }; The result is of type IEnumerable<TempProjectionItem>, which we can subse‐ quently query: IEnumerable<string> query = from item in temp where item.Vowelless.Length > 2 select item.Original; // Dick // Harry // Mary Anonymous Types Anonymous types allow you to structure your intermediate results without writing special classes. We can eliminate the TempProjectionItem class in our previous example with anonymous types: var intermediate = from n in names select new { Original = n, Vowelless = n.Replace ("a", "").Replace ("e", "").Replace ("i", "") .Replace ("o", "").Replace ("u", "") }; IEnumerable<string> query = from item in intermediate where item.Vowelless.Length > 2 select item.Original; This gives the same result as the previous example, but without needing to write a one-off class. The compiler does the job instead, generating a temporary class with fields that match the structure of our projection. This means, however, that the intermediate query has the following type: IEnumerable <random-compiler-generated-name> The only way we can declare a variable of this type is with the var keyword. In this case, var is more than just a clutter reduction device; it’s a necessity. We can write the entire query more succinctly with the into keyword: 422 | Chapter 8: LINQ Queries
- 441. var query = from n in names select new { Original = n, Vowelless = n.Replace ("a", "").Replace ("e", "").Replace ("i", "") .Replace ("o", "").Replace ("u", "") } into temp where temp.Vowelless.Length > 2 select temp.Original; Query expressions provide a shortcut for writing this kind of query: the let keyword. The let Keyword The let keyword introduces a new variable alongside the range variable. With let, we can write a query extracting strings whose length, excluding vowels, exceeds two characters, as follows: string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; IEnumerable<string> query = from n in names let vowelless = n.Replace ("a", "").Replace ("e", "").Replace ("i", "") .Replace ("o", "").Replace ("u", "") where vowelless.Length > 2 orderby vowelless select n; // Thanks to let, n is still in scope. The compiler resolves a let clause by projecting into a temporary anonymous type that contains both the range variable and the new expression variable. In other words, the compiler translates this query into the preceding example. let accomplishes two things: • It projects new elements alongside existing elements. • • It allows an expression to be used repeatedly in a query without being • rewritten. The let approach is particularly advantageous in this example because it allows the select clause to project either the original name (n) or its vowel-removed version (vowelless). You can have any number of let statements before or after a where statement (see Figure 8-2). A let statement can reference variables introduced in earlier let statements (subject to the boundaries imposed by an into clause). let reprojects all existing variables transparently. A let expression need not evaluate to a scalar type: sometimes it’s useful to have it evaluate to a subsequence, for instance. Projection Strategies | 423 LINQ Queries
- 442. Interpreted Queries LINQ provides two parallel architectures: local queries for local object collections and interpreted queries for remote data sources. So far, we’ve examined the architec‐ ture of local queries, which operate over collections implementing IEnumerable<T>. Local queries resolve to query operators in the Enumerable class (by default), which in turn resolve to chains of decorator sequences. The delegates that they accept— whether expressed in query syntax, fluent syntax, or traditional delegates—are fully local to Intermediate Language (IL) code, just like any other C# method. By contrast, interpreted queries are descriptive. They operate over sequences that implement IQueryable<T>, and they resolve to the query operators in the Querya ble class, which emit expression trees that are interpreted at runtime. These expres‐ sion trees can be translated, for instance, to SQL queries, allowing you to use LINQ to query a database. The query operators in Enumerable can actually work with IQueryable<T> sequences. The difficulty is that the resultant queries always execute locally on the client. This is why a second set of query operators is provided in the Queryable class. To write interpreted queries, you need to start with an API that exposes sequences of type IQueryable<T>. An example is Microsoft’s Entity Framework Core (EF Core), which allows you to query a variety of databases, including SQL Server, Oracle, MySQL, PostgreSQL, and SQLite. It’s also possible to generate an IQueryable<T> wrapper around an ordinary enu‐ merable collection by calling the AsQueryable method. We describe AsQueryable in “Building Query Expressions” on page 442. IQueryable<T> is an extension of IEnumerable<T> with addi‐ tional methods for constructing expression trees. Most of the time you can ignore the details of these methods; they’re called indirectly by the runtime. “Building Query Expressions” on page 442 covers IQueryable<T> in more detail. To illustrate, let’s create a simple customer table in SQL Server and populate it with a few names using the following SQL script: create table Customer ( ID int not null primary key, Name varchar(30) ) insert Customer values (1, 'Tom') insert Customer values (2, 'Dick') insert Customer values (3, 'Harry') insert Customer values (4, 'Mary') insert Customer values (5, 'Jay') 424 | Chapter 8: LINQ Queries
- 443. With this table in place, we can write an interpreted LINQ query in C# that uses EF Core to retrieve customers whose name contains the letter “a,” as follows: using System; using System.Linq; using Microsoft.EntityFrameworkCore; using var dbContext = new NutshellContext(); IQueryable<string> query = from c in dbContext.Customers where c.Name.Contains ("a") orderby c.Name.Length select c.Name.ToUpper(); foreach (string name in query) Console.WriteLine (name); public class Customer { public int ID { get; set; } public string Name { get; set; } } // We’ll explain the following class in more detail in the next section. public class NutshellContext : DbContext { public virtual DbSet<Customer> Customers { get; set; } protected override void OnConfiguring (DbContextOptionsBuilder builder) => builder.UseSqlServer ("...connection string..."); protected override void OnModelCreating (ModelBuilder modelBuilder) => modelBuilder.Entity<Customer>().ToTable ("Customer") .HasKey (c => c.ID); } EF Core translates this query into the following SQL: SELECT UPPER([c].[Name]) FROM [Customers] AS [c] WHERE CHARINDEX(N'a', [c].[Name]) > 0 ORDER BY CAST(LEN([c].[Name]) AS int) Here’s the end result: // JAY // MARY // HARRY Interpreted Queries | 425 LINQ Queries
- 444. How Interpreted Queries Work Let’s examine how the preceding query is processed. First, the compiler converts query syntax to fluent syntax. This is done exactly as with local queries: IQueryable<string> query = dbContext.customers .Where (n => n.Name.Contains ("a")) .OrderBy (n => n.Name.Length) .Select (n => n.Name.ToUpper()); Next, the compiler resolves the query operator methods. Here’s where local and interpreted queries differ—interpreted queries resolve to query operators in the Queryable class instead of the Enumerable class. To see why, we need to look at the dbContext.Customers variable, the source upon which the entire query builds. dbContext.Customers is of type DbSet<T>, which implements IQueryable<T> (a subtype of IEnumerable<T>). This means that the compiler has a choice in resolving Where: it could call the extension method in Enumerable or the following extension method in Queryable: public static IQueryable<TSource> Where<TSource> (this IQueryable<TSource> source, Expression <Func<TSource,bool>> predicate) The compiler chooses Queryable.Where because its signature is a more specific match. Queryable.Where accepts a predicate wrapped in an Expression<TDelegate> type. This instructs the compiler to translate the supplied lambda expression— in other words, n=>n.Name.Contains("a")—to an expression tree rather than a compiled delegate. An expression tree is an object model based on the types in System.Linq.Expressions that can be inspected at runtime (so that EF Core can later translate it to an SQL statement). Because Queryable.Where also returns IQueryable<T>, the same process follows with the OrderBy and Select operators. Figure 8-9 illustrates the end result. In the shaded box, there is an expression tree describing the entire query, which can be traversed at runtime. 426 | Chapter 8: LINQ Queries
- 445. Figure 8-9. Interpreted query composition Execution Interpreted queries follow a deferred execution model—just like local queries. This means that the SQL statement is not generated until you start enumerating the query. Further, enumerating the same query twice results in the database being queried twice. Under the hood, interpreted queries differ from local queries in how they execute. When you enumerate over an interpreted query, the outermost sequence runs a program that traverses the entire expression tree, processing it as a unit. In our example, EF Core translates the expression tree to an SQL statement, which it then executes, yielding the results as a sequence. To work, EF Core needs to understand the schema of the data‐ base. It does this by leveraging conventions, code attributes, and a fluent configuration API. We’ll explore this in detail later in the chapter. We said previously that a LINQ query is like a production line. However, when you enumerate an IQueryable conveyor belt, it doesn’t start up the whole production line, like with a local query. Instead, just the IQueryable belt starts up, with a special enumerator that calls upon a production manager. The manager reviews the entire production line—which consists not of compiled code but of dummies (method call expressions) with instructions pasted to their foreheads (expression trees). The manager then traverses all the expressions, in this case transcribing them to a single Interpreted Queries | 427 LINQ Queries
- 446. piece of paper (an SQL statement), which it then executes, feeding the results back to the consumer. Only one belt turns; the rest of the production line is a network of empty shells, existing just to describe what needs to be done. This has some practical implications. For instance, with local queries, you can write your own query methods (fairly easily, with iterators) and then use them to supplement the predefined set. With remote queries, this is difficult and even undesirable. If you wrote a MyWhere extension method accepting IQueryable<T>, it would be like putting your own dummy into the production line. The production manager wouldn’t know what to do with your dummy. Even if you intervened at this stage, your solution would be hardwired to a particular provider, such as EF Core, and would not work with other IQueryable implementations. Part of the benefit of having a standard set of methods in Queryable is that they define a standard vocabulary for querying any remote collection. As soon as you try to extend the vocabulary, you’re no longer interoperable. Another consequence of this model is that an IQueryable provider might be unable to cope with some queries—even if you stick to the standard methods. EF Core is limited by the capabilities of the database server; some LINQ queries have no SQL translation. If you’re familiar with SQL, you’ll have a good intuition for what these are, although at times you’ll need to experiment to see what causes a runtime error; it can be surprising what does work! Combining Interpreted and Local Queries A query can include both interpreted and local operators. A typical pattern is to have the local operators on the outside and the interpreted components on the inside; in other words, the interpreted queries feed the local queries. This pattern works well when querying a database. For instance, suppose that we write a custom extension method to pair up strings in a collection: public static IEnumerable<string> Pair (this IEnumerable<string> source) { string firstHalf = null; foreach (string element in source) if (firstHalf == null) firstHalf = element; else { yield return firstHalf + ", " + element; firstHalf = null; } } We can use this extension method in a query that mixes EF Core and local operators: using var dbContext = new NutshellContext (); IEnumerable<string> q = dbContext.Customers .Select (c => c.Name.ToUpper()) 428 | Chapter 8: LINQ Queries
- 447. .OrderBy (n => n) .Pair() // Local from this point on. .Select ((n, i) => "Pair " + i.ToString() + " = " + n); foreach (string element in q) Console.WriteLine (element); // Pair 0 = DICK, HARRY // Pair 1 = JAY, MARY Because dbContext.Customers is of a type implementing IQueryable<T>, the Select operator resolves to Queryable.Select. This returns an output sequence also of type IQueryable<T>, so the OrderBy operator similarly resolves to Queryable.OrderBy. But the next query operator, Pair, has no overload accepting IQueryable<T>—only the less specific IEnumerable<T>. So, it resolves to our local Pair method—wrapping the interpreted query in a local query. Pair also returns IEnumerable, so the Select that follows resolves to another local operator. On the EF Core side, the resulting SQL statement is equivalent to this: SELECT UPPER([c].[Name]) FROM [Customers] AS [c] ORDER BY UPPER([c].[Name]) The remaining work is done locally. In effect, we end up with a local query (on the outside) whose source is an interpreted query (the inside). AsEnumerable Enumerable.AsEnumerable is the simplest of all query operators. Here’s its complete definition: public static IEnumerable<TSource> AsEnumerable<TSource> (this IEnumerable<TSource> source) { return source; } Its purpose is to cast an IQueryable<T> sequence to IEnumerable<T>, forcing sub‐ sequent query operators to bind to Enumerable operators instead of Queryable operators. This causes the remainder of the query to execute locally. To illustrate, suppose that we had a MedicalArticles table in SQL Server and wanted to use EF Core to retrieve all articles on influenza whose abstract contained fewer than 100 words. For the latter predicate, we need a regular expression: Regex wordCounter = new Regex (@"b(w|[-'])+b"); using var dbContext = new NutshellContext (); var query = dbContext.MedicalArticles .Where (article => article.Topic == "influenza" && wordCounter.Matches (article.Abstract).Count < 100); The problem is that SQL Server doesn’t support regular expressions, so EF Core will throw an exception, complaining that the query cannot be translated to SQL. We can solve this by querying in two steps: first retrieving all articles on influenza Interpreted Queries | 429 LINQ Queries
- 448. through an EF Core query, and then filtering locally for abstracts of fewer than 100 words: Regex wordCounter = new Regex (@"b(w|[-'])+b"); using var dbContext = new NutshellContext (); IEnumerable<MedicalArticle> efQuery = dbContext.MedicalArticles .Where (article => article.Topic == "influenza"); IEnumerable<MedicalArticle> localQuery = efQuery .Where (article => wordCounter.Matches (article.Abstract).Count < 100); Because efQuery is of type IEnumerable<MedicalArticle>, the second query binds to the local query operators, forcing that part of the filtering to run on the client. With AsEnumerable, we can do the same in a single query: Regex wordCounter = new Regex (@"b(w|[-'])+b"); using var dbContext = new NutshellContext (); var query = dbContext.MedicalArticles .Where (article => article.Topic == "influenza") .AsEnumerable() .Where (article => wordCounter.Matches (article.Abstract).Count < 100); An alternative to calling AsEnumerable is to call ToArray or ToList. The advantage of AsEnumerable is that it doesn’t force immediate query execution, nor does it create any storage structure. Moving query processing from the database server to the cli‐ ent can hurt performance, especially if it means retrieving more rows. A more efficient (though more complex) way to solve our example would be to use SQL CLR integration to expose a function on the database that implemented the regular expression. We further demonstrate combined interpreted and local queries in Chapter 10. EF Core Throughout this and Chapter 9, we use EF Core to demonstrate interpreted queries. Let’s now examine the key features of this technology. EF Core Entity Classes EF Core lets you use any class to represent data, as long as it contains a public property for each column that you want to query. For instance, we could define the following entity class to query and update a Customers table in the database: 430 | Chapter 8: LINQ Queries
- 449. public class Customer { public int ID { get; set; } public string Name { get; set; } } DbContext After defining entity classes, the next step is to subclass DbContext. An instance of that class represents your sessions working with the database. Typically, your DbContext subclass will contain one DbSet<T> property for each entity in your model: public class NutshellContext : DbContext { public DbSet<Customer> Customers { get; set; } ... properties for other tables ... } A DbContext object does three things: • It acts as a factory for generating DbSet<> objects that you can query. • • It keeps track of any changes that you make to your entities so that you can • write them back (see “Change Tracking” on page 437). • It provides virtual methods that you can override to configure the connection • and model. Configuring the connection By overriding the OnConfiguring method, you can specify the database provider and connection string: public class NutshellContext : DbContext { ... protected override void OnConfiguring (DbContextOptionsBuilder optionsBuilder) => optionsBuilder.UseSqlServer (@"Server=(local);Database=Nutshell;Trusted_Connection=True"); } In this example, the connection string is specified as a string literal. Produc‐ tion applications would typically retrieve it from a configuration file such as appsettings.json. UseSqlServer is an extension method defined in an assembly that’s part of the Microsoft.EntityFramework.SqlServer NuGet package. Packages are available for other database providers, including Oracle, MySQL, PostgreSQL, and SQLite. EF Core | 431 LINQ Queries
- 450. If you’re using ASP.NET, you can allow its dependency injec‐ tion framework to preconfigure optionsBuilder; in most cases, this lets you avoid overriding OnConfiguring altogether. To enable this, define a constructor on DbContext as follows: public NutshellContext (DbContextOptions<NutshellContext> options) : base(options) { } If you do choose to override OnConfiguring (perhaps to pro‐ vide a configuration if your DbContext is used in another scenario), you can check whether options have already been configured as follows: protected override void OnConfiguring ( DbContextOptionsBuilder optionsBuilder) { if (!optionsBuilder.IsConfigured) { ... } } In the OnConfiguring method, you can enable other options, including lazy loading (see “Lazy loading” on page 440). Configuring the Model By default, EF Core is convention based, meaning that it infers the database schema from your class and property names. You can override the defaults using the fluent api by overriding OnModelCreating and calling extension methods on the ModelBuilder parameter. For example, we can explicitly specify the database table name for our Customer entity as follows: protected override void OnModelCreating (ModelBuilder modelBuilder) => modelBuilder.Entity<Customer>() .ToTable ("Customer"); // Table is called 'Customer' Without this code, EF Core would map this entity to a table named “Customers” rather than “Customer”, because we have a DbSet<Customer> property in our DbContext called Customers: public DbSet<Customer> Customers { get; set; } 432 | Chapter 8: LINQ Queries
- 451. The following code maps all of your entities to table names that match the entity class name (which is typically singular) rather than the DbSet<T> property name (which is typically plural): protected override void OnModelCreating (ModelBuilder modelBuilder) { foreach (IMutableEntityType entityType in modelBuilder.Model.GetEntityTypes()) { modelBuilder.Entity (entityType.Name) .ToTable (entityType.ClrType.Name); } } The fluent API offers an expanded syntax for configuring columns. In the next example, we use two popular methods: • HasColumnName, which maps a property to a differently named column • • IsRequired, which indicates that a column is not nullable • protected override void OnModelCreating (ModelBuilder modelBuilder) => modelBuilder.Entity<Customer> (entity => { entity.ToTable ("Customer"); entity.Property (e => e.Name) .HasColumnName ("Full Name") // Column name is 'Full Name' .IsRequired(); // Column is not nullable }); Table 8-1 lists some of the most important methods in the fluent API. Instead of using the fluent API, you can configure your model by applying special attributes to your entity classes and prop‐ erties (“data annotations”). This approach is less flexible in that the configuration must be fixed at compile-time, and is less powerful in that there are some options that can be configured only via the fluent API. Table 8-1. Fluent API model configuration methods Method Purpose Example ToTable Specify the database table name for a given entity builder .Entity<Customer>() .ToTable("Customer"); HasColumnName Specify the column name for a given property builder.Entity<Customer>() .Property(c => c.Name) .HasColumnName("Full Name"); HasKey Specify a key (usually that deviates from convention) builder.Entity<Customer>() .HasKey(c => c.CustomerNr); EF Core | 433 LINQ Queries
- 452. Method Purpose Example IsRequired Specify that the property requires a value (is not nullable) builder.Entity<Customer>() .Property(c => c.Name) .IsRequired(); HasMaxLength Specify the maximum length of a variable-length type (usually a string) whose width can vary builder.Entity<Customer>() .Property(c => c.Name) .HasMaxLength(60); HasColumnType Specify the database data type for a column builder.Entity<Purchase>() .Property(p => p.Description) .HasColumnType("varchar(80)"); Ignore Ignore a type builder.Ignore<Products>(); Ignore Ignore a property of a type builder.Entity<Customer>() .Ignore(c => c.ChatName); HasIndex Specify a property (or combination of properties) should serve in the database as an index // Compound index: builder.Entity<Purchase>() .HasIndex(p => new { p.Date, p.Price }); // Unique index on one property builder .Entity<MedicalArticle>() .HasIndex(a => a.Topic) .IsUnique(); HasOne See “Navigation Properties” on page 438 builder.Entity<Purchase>() .HasOne(p => p.Customer) .WithMany(c => c.Purchases); HasMany See “Navigation Properties” on page 438 builder.Entity<Customer>() .HasMany(c => c.Purchases) .WithOne(p => p.Customer); Creating the database EF Core supports a code-first approach, which means that you can start by defining entity classes and then ask EF Core to create the database. The easiest way to do the latter is to call the following method on a DbContext instance: dbContext.Database.EnsureCreated(); A better approach, however, is to use EF Core’s migrations feature, which not only creates the database but configures it such that EF Core can automatically update the schema in the future when your entity classes change. You can enable migrations in Visual Studio’s Package Manager Console and ask it to create the database with the following commands: 434 | Chapter 8: LINQ Queries
- 453. Install-Package Microsoft.EntityFrameworkCore.Tools Add-Migration InitialCreate Update-Database The first command installs tools to manage EF Core from within Visual Studio. The second command generates a special C# class known as a code migration that contains instructions to create the database. The final command runs those instruc‐ tions against the database connection string specified in the project’s application configuration file. Using DbContext After you’ve defined Entity classes and subclassed DbContext, you can instantiate your DbContext and query the database, as follows: using var dbContext = new NutshellContext(); Console.WriteLine (dbContext.Customers.Count()); // Executes "SELECT COUNT(*) FROM [Customer] AS [c]" You can also use your DbContext instance to write to the database. The following code inserts a row into the Customer table: using var dbContext = new NutshellContext(); Customer cust = new Customer() { Name = "Sara Wells" }; dbContext.Customers.Add (cust); dbContext.SaveChanges(); // Writes changes back to database The following queries the database for the customer that was just inserted: using var dbContext = new NutshellContext(); Customer cust = dbContext.Customers .Single (c => c.Name == "Sara Wells") The following updates that customer’s name and writes the change to the database: cust.Name = "Dr. Sara Wells"; dbContext.SaveChanges(); The Single operator is ideal for retrieving a row by primary key. Unlike First, it throws an exception if more than one element is returned. Object Tracking A DbContext instance keeps track of all the entities it instantiates, so it can feed the same ones back to you whenever you request the same rows in a table. In other words, a context in its lifetime will never emit two separate entities that refer to the same row in a table (where a row is identified by primary key). This capability is called object tracking. EF Core | 435 LINQ Queries
- 454. To illustrate, suppose the customer whose name is alphabetically first also has the lowest ID. In the following example, a and b will reference the same object: using var dbContext = new NutshellContext (); Customer a = dbContext.Customers.OrderBy (c => c.Name).First(); Customer b = dbContext.Customers.OrderBy (c => c.ID).First(); Disposing DbContext Although DbContext implements IDisposable, you can (in general) get away without disposing instances. Disposing forces the context’s connection to dispose—but this is usually unnecessary because EF Core closes connections automatically whenever you finish retrieving results from a query. Disposing a context prematurely can actually be problematic because of lazy evalua‐ tion. Consider the following: IQueryable<Customer> GetCustomers (string prefix) { using (var dbContext = new NutshellContext ()) return dbContext.Customers .Where (c => c.Name.StartsWith (prefix)); } ... foreach (Customer c in GetCustomers ("a")) Console.WriteLine (c.Name); This will fail because the query is evaluated when we enumerate it—which is after disposing its DbContext. There are some caveats, though, on not disposing contexts: • It relies on the connection object releasing all unmanaged resources on the • Close method. Even though this holds true with SqlConnection, it’s theoreti‐ cally possible for a third-party connection to keep resources open if you call Close but not Dispose (though this would arguably violate the contract defined by IDbConnection.Close). • If you manually call GetEnumerator on a query (instead of using foreach) • and then fail to either dispose the enumerator or consume the sequence, the connection will remain open. Disposing the DbContext provides a backup in such scenarios. • Some people feel that it’s tidier to dispose contexts (and all objects that imple‐ • ment IDisposable). If you want to explicitly dispose contexts, you must pass a DbContext instance into methods such as GetCustomers to avoid the problem described. In scenarios such as ASP.NET Core MVC where the context instance is provided via dependency injection (DI), the DI infrastructure will manage the context lifetime. It will be created when a unit of work (such as an HTTP request processed in the controller) begins and disposed when that unit of work ends. 436 | Chapter 8: LINQ Queries
- 455. Consider what happens when EF Core encounters the second query. It starts by querying the database—and obtaining a single row. It then reads the primary key of this row and performs a lookup in the context’s entity cache. Seeing a match, it returns the existing object without updating any values. So, if another user had just updated that customer’s Name in the database, the new value would be ignored. This is essential for avoiding unexpected side effects (the Customer object could be in use elsewhere) and also for managing concurrency. If you had altered properties on the Customer object and not yet called SaveChanges, you wouldn’t want your properties automatically overwritten. You can disable object tracking by chaining the AsNo Tracking extension method to your query or by setting ChangeTracker.QueryTrackingBehavior on the context to QueryTrackingBehavior.NoTracking. No-tracking queries are useful when data is used read-only as it improves perfor‐ mance and reduces memory use. To get fresh information from the database, you must either instantiate a new context or call the Reload method, as follows: dbContext.Entry (myCustomer).Reload(); The best practice is to use a fresh DbContext instance per unit of work so that the need to manually reload an entity is rare. Change Tracking When you change a property value in an entity loaded via DbContext, EF Core rec‐ ognizes the change and updates the database accordingly upon calling SaveChanges. To do that, it creates a snapshot of the state of entities loaded through your DbCon text subclass and compares the current state to the original one when SaveChanges is called (or when you manually query change tracking, as you’ll see in a moment). You can enumerate the tracked changes in a DbContext as follows: foreach (var e in dbContext.ChangeTracker.Entries()) { Console.WriteLine ($"{e.Entity.GetType().FullName} is {e.State}"); foreach (var m in e.Members) Console.WriteLine ( $" {m.Metadata.Name}: '{m.CurrentValue}' modified: {m.IsModified}"); } When you call SaveChanges, EF Core uses the information in the ChangeTracker to construct SQL statements that will update the database to match the changes in your objects, issuing insert statements to add new rows, update statements to modify data, and delete statements to remove rows that were removed from the object graph in your DbContext subclass. Any TransactionScope is honored; if none is present, it wraps all statements in a new transaction. You can optimize change tracking by implementing INotifyPropertyChanged and, optionally, INotifyPropertyChanging in your entities. The former allows EF Core EF Core | 437 LINQ Queries
- 456. to avoid the overhead of comparing modified with original entities; the latter allows EF Core to avoid storing the original values altogether. After implementing these interfaces, call the HasChangeTrackingStrategy method on the ModelBuilder when configuring the model in order to activate the optimized change tracking. Navigation Properties Navigation properties allow you to do the following: • Query related tables without having to manually join • • Insert, remove, and update related rows without explicitly updating foreign • keys For example, suppose that a customer can have a number of purchases. We can represent a one-to-many relationship between Customer and Purchase with the following entities: public class Customer { public int ID { get; set; } public string Name { get; set; } // Child navigation property, which must be of type ICollection<T>: public virtual List<Purchase> Purchases {get;set;} = new List<Purchase>(); } public class Purchase { public int ID { get; set; } public DateTime Date { get; set; } public string Description { get; set; } public decimal Price { get; set; } public int CustomerID? { get; set; } // Foreign key field public Customer Customer { get; set; } // Parent navigation property } EF Core is able to infer from these entities that CustomerID is a foreign key to the Customer table, because the name “CustomerID” follows a popular naming conven‐ tion. If we were to ask EF Core to create a database from these entities, it would create a foreign key constraint between Purchase.CustomerID and Customer.ID. If EF Core is unable to infer the relationship, you can config‐ ure it explicitly in the OnModelCreating method as follows: modelBuilder.Entity<Purchase>() .HasOne (e => e.Customer) .WithMany (e => e.Purchases) .HasForeignKey (e => e.CustomerID); With these navigation properties set up, we can write queries such as this: var customersWithPurchases = Customers.Where (c => c.Purchases.Any()); 438 | Chapter 8: LINQ Queries
- 457. We cover how to write such queries in detail in Chapter 9. Adding and removing entities from navigation collections When you add new entities to a collection navigation property, EF Core automati‐ cally populates the foreign keys upon calling SaveChanges: Customer cust = dbContext.Customers.Single (c => c.ID == 1); Purchase p1 = new Purchase { Description="Bike", Price=500 }; Purchase p2 = new Purchase { Description="Tools", Price=100 }; cust.Purchases.Add (p1); cust.Purchases.Add (p2); dbContext.SaveChanges(); In this example, EF Core automatically writes 1 into the CustomerID column of each of the new purchases and writes the database-generated ID for each purchase to Purchase.ID. When you remove an entity from a collection navigation property and call Save Changes, EF Core will either clear the foreign key field or delete the corresponding row from the database, depending on how the relationship has been configured or inferred. In this case, we’ve defined Purchase.CustomerID as a nullable integer (so that we can represent purchases without a customer, or cash transactions), so removing a purchase from a customer would clear its foreign key field rather than deleting it from the database. Loading navigation properties When EF Core populates an entity, it does not (by default) populate its navigation properties: using var dbContext = new NutshellContext(); var cust = dbContext.Customers.First(); Console.WriteLine (cust.Purchases.Count); // Always 0 One solution is to use the Include extension method, which instructs EF Core to eagerly load navigation properties: var cust = dbContext.Customers .Include (c => c.Purchases) .Where (c => c.ID == 2).First(); Another solution is to use a projection. This technique is particularly useful when you need to work with only some of the entity properties, because it reduces data transfer: var custInfo = dbContext.Customers .Where (c => c.ID == 2) .Select (c => new { Name = c.Name, EF Core | 439 LINQ Queries
- 458. Purchases = c.Purchases.Select (p => new { p.Description, p.Price }) }) .First(); Both of these techniques inform EF Core what data you require so that it can be fetched in a single database query. It’s also possible to manually instruct EF Core to populate a navigation property as needed: dbContext.Entry (cust).Collection (b => b.Purchases).Load(); // cust.Purchases is now populated. This is called explicit loading. Unlike the preceding approaches, this generates an extra round trip to the database. Lazy loading Another approach for loading navigation properties is called lazy loading. When enabled, EF Core populates navigation properties on demand by generating a proxy class for each of your entity classes that intercepts attempts to access unloaded navigation properties. For this to work, each navigation property must be virtual, and the class it’s defined in must be inheritable (not sealed). Also, the context must not have been disposed when the lazy load occurs, so that an additional database request can be performed. You can enable lazy loading in the OnConfiguring method of your DbContext subclass, as follows: protected override void OnConfiguring (DbContextOptionsBuilder optionsBuilder) { optionsBuilder .UseLazyLoadingProxies() ... } (You will also need to add a reference to the Microsoft.EntityFramework Core.Proxies NuGet package.) The cost of lazy loading is that EF Core must make an additional request to the database each time you access an unloaded navigation property. If you make many such requests, performance can suffer as a result of excessive round-tripping. With lazy loading enabled, the runtime type of your classes is a proxy derived from your entity class. For example: using var dbContext = new NutshellContext(); var cust = dbContext.Customers.First(); Console.WriteLine (cust.GetType()); // Castle.Proxies.CustomerProxy Deferred Execution EF Core queries are subject to deferred execution, just like local queries. This allows you to build queries progressively. There is one aspect, however, in which EF Core 440 | Chapter 8: LINQ Queries
- 459. has special deferred execution semantics, and that is when a subquery appears within a Select expression. With local queries, you get double-deferred execution, because from a functional perspective, you’re selecting a sequence of queries. So, if you enumerate the outer result sequence but never enumerate the inner sequences, the subquery will never execute. With EF Core, the subquery is executed at the same time as the main outer query. This prevents excessive round-tripping. For example, the following query executes in a single round trip upon reaching the first foreach statement: using var dbContext = new NutshellContext (); var query = from c in dbContext.Customers select from p in c.Purchases select new { c.Name, p.Price }; foreach (var customerPurchaseResults in query) foreach (var namePrice in customerPurchaseResults) Console.WriteLine ($"{ namePrice.Name} spent { namePrice.Price}"); Any navigation properties that you explicitly project are fully populated in a single round trip: var query = from c in dbContext.Customers select new { c.Name, c.Purchases }; foreach (var row in query) foreach (Purchase p in row.Purchases) // No extra round-tripping Console.WriteLine (row.Name + " spent " + p.Price); But if we enumerate a navigation property without first having either eagerly loaded or projected, deferred execution rules apply. In the following example, EF Core executes another Purchases query on each loop iteration (assuming lazy loading is enabled): foreach (Customer c in dbContext.Customers.ToArray()) foreach (Purchase p in c.Purchases) // Another SQL round-trip Console.WriteLine (c.Name + " spent " + p.Price); This model is advantageous when you want to selectively execute the inner loop, based on a test that can be performed only on the client: foreach (Customer c in dbContext.Customers.ToArray()) if (myWebService.HasBadCreditHistory (c.ID)) foreach (Purchase p in c.Purchases) // Another SQL round trip Console.WriteLine (c.Name + " spent " + p.Price); EF Core | 441 LINQ Queries
- 460. Note the use of ToArray in the previous two queries. By default, SQL Server cannot initiate a new query while the results of the current query are still being processed. Calling ToArray materializes the customers so that additional queries can be issued to retrieve purchases per customer. It is possible to configure SQL Server to allow multiple active result sets (MARS) by appending ;MultipleActiveResultSets=True to the database connection string. Use MARS with caution as it can mask a chatty database design that could be improved by eager loading and/or projecting the required data. (In Chapter 9, we explore Select subqueries in more detail, in “Projecting” on page 449.) Building Query Expressions So far in this chapter, when we’ve needed to dynamically compose queries, we’ve done so by conditionally chaining query operators. Although this is adequate in many scenarios, sometimes you need to work at a more granular level and dynami‐ cally compose the lambda expressions that feed the operators. In this section, we assume the following Product class: public class Product { public int ID { get; set; } public string Description { get; set; } public bool Discontinued { get; set; } public DateTime LastSale { get; set; } } Delegates Versus Expression Trees Recall that: • Local queries, which use Enumerable operators, take delegates. • • Interpreted queries, which use Queryable operators, take expression trees. • We can see this by comparing the signature of the Where operator in Enumerable and Queryable: public static IEnumerable<TSource> Where<TSource> (this IEnumerable<TSource> source, Func<TSource,bool> predicate) public static IQueryable<TSource> Where<TSource> (this IQueryable<TSource> source, Expression<Func<TSource,bool>> predicate) When embedded within a query, a lambda expression looks identical whether it binds to Enumerable’s operators or Queryable’s operators: IEnumerable<Product> q1 = localProducts.Where (p => !p.Discontinued); IQueryable<Product> q2 = sqlProducts.Where (p => !p.Discontinued); 442 | Chapter 8: LINQ Queries
- 461. When you assign a lambda expression to an intermediate variable, however, you must be explicit about whether to resolve to a delegate (i.e., Func<>) or an expres‐ sion tree (i.e., Expression<Func<>>). In the following example, predicate1 and predicate2 are not interchangeable: Func <Product, bool> predicate1 = p => !p.Discontinued; IEnumerable<Product> q1 = localProducts.Where (predicate1); Expression <Func <Product, bool>> predicate2 = p => !p.Discontinued; IQueryable<Product> q2 = sqlProducts.Where (predicate2); Compiling expression trees You can convert an expression tree to a delegate by calling Compile. This is of par‐ ticular value when writing methods that return reusable expressions. To illustrate, let’s add a static method to the Product class that returns a predicate evaluating to true if a product is not discontinued and has sold in the past 30 days: public class Product { public static Expression<Func<Product, bool>> IsSelling() { return p => !p.Discontinued && p.LastSale > DateTime.Now.AddDays (-30); } } The method just written can be used both in interpreted and local queries, as follows: void Test() { var dbContext = new NutshellContext(); Product[] localProducts = dbContext.Products.ToArray(); IQueryable<Product> sqlQuery = dbContext.Products.Where (Product.IsSelling()); IEnumerable<Product> localQuery = localProducts.Where (Product.IsSelling().Compile()); } .NET does not provide an API to convert in the reverse direction, from a delegate to an expression tree. This makes expression trees more versatile. AsQueryable The AsQueryable operator lets you write whole queries that can run over either local or remote sequences: IQueryable<Product> FilterSortProducts (IQueryable<Product> input) { return from p in input where ... Building Query Expressions | 443 LINQ Queries
- 462. orderby ... select p; } void Test() { var dbContext = new NutshellContext(); Product[] localProducts = dbContext.Products.ToArray(); var sqlQuery = FilterSortProducts (dbContext.Products); var localQuery = FilterSortProducts (localProducts.AsQueryable()); ... } AsQueryable wraps IQueryable<T> clothing around a local sequence so that subse‐ quent query operators resolve to expression trees. When you later enumerate over the result, the expression trees are implicitly compiled (at a small performance cost), and the local sequence enumerates as it would ordinarily. Expression Trees We said previously that an implicit conversion from a lambda expression to Expres sion<TDelegate> causes the C# compiler to emit code that builds an expression tree. With some programming effort, you can do the same thing manually at runtime—in other words, dynamically build an expression tree from scratch. The result can be cast to an Expression<TDelegate> and used in EF Core queries or compiled into an ordinary delegate by calling Compile. The Expression DOM An expression tree is a miniature code DOM. Each node in the tree is represented by a type in the System.Linq.Expressions namespace. Figure 8-10 illustrates these types. Figure 8-10. Expression types 444 | Chapter 8: LINQ Queries
- 463. The base class for all nodes is the (nongeneric) Expression class. The generic Expression<TDelegate> class actually means “typed lambda expression” and might have been named LambdaExpression<TDelegate> if it wasn’t for the clumsiness of this: LambdaExpression<Func<Customer,bool>> f = ... Expression<T>’s base type is the (nongeneric) LambdaExpression class. Lamdba Expression provides type unification for lambda expression trees: any typed Expression<T> can be cast to a LambdaExpression. The thing that distinguishes LambdaExpressions from ordinary Expressions is that lambda expressions have parameters. To create an expression tree, don’t instantiate node types directly; rather, call static methods provided on the Expression class, such as Add, And, Call, Constant, LessThan, and so on. Figure 8-11 shows the expression tree that the following assignment creates: Expression<Func<string, bool>> f = s => s.Length < 5; Figure 8-11. Expression tree We can demonstrate this as follows: Console.WriteLine (f.Body.NodeType); // LessThan Console.WriteLine (((BinaryExpression) f.Body).Right); // 5 Let’s now build this expression from scratch. The principle is that you start from the bottom of the tree and work your way up. The bottommost thing in our tree is a ParameterExpression, the lambda expression parameter called “s” of type string: ParameterExpression p = Expression.Parameter (typeof (string), "s"); Building Query Expressions | 445 LINQ Queries
- 464. The next step is to build the MemberExpression and ConstantExpression. In the former case, we need to access the Length property of our parameter, “s”: MemberExpression stringLength = Expression.Property (p, "Length"); ConstantExpression five = Expression.Constant (5); Next is the LessThan comparison: BinaryExpression comparison = Expression.LessThan (stringLength, five); The final step is to construct the lambda expression, which links an expression Body to a collection of parameters: Expression<Func<string, bool>> lambda = Expression.Lambda<Func<string, bool>> (comparison, p); A convenient way to test our lambda is by compiling it to a delegate: Func<string, bool> runnable = lambda.Compile(); Console.WriteLine (runnable ("kangaroo")); // False Console.WriteLine (runnable ("dog")); // True The easiest way to determine which expression type to use is to examine an existing lambda expression in the Visual Studio debugger. We continue this discussion online, at http://www.albahari.com/expressions. 446 | Chapter 8: LINQ Queries
- 465. 9 LINQ Operators This chapter describes each of the LINQ query operators. As well as serving as a reference, two of the sections, “Projecting” on page 449 and “Joining” on page 449, cover a number of conceptual areas: • Projecting object hierarchies • • Joining with Select, SelectMany, Join, and GroupJoin • • Query expressions with multiple range variables • All of the examples in this chapter assume that a names array is defined as follows: string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; Examples that query a database assume that a variable called dbContext is instanti‐ ated as var dbContext = new NutshellContext(); where NutshellContext is defined as follows: public class NutshellContext : DbContext { public DbSet<Customer> Customers { get; set; } public DbSet<Purchase> Purchases { get; set; } protected override void OnModelCreating(ModelBuilder modelBuilder) { modelBuilder.Entity<Customer>(entity => { entity.ToTable("Customer"); entity.Property(e => e.Name).IsRequired(); // Column is not nullable }); modelBuilder.Entity<Purchase>(entity => { entity.ToTable("Purchase"); entity.Property(e => e.Date).IsRequired(); 447
- 466. entity.Property(e => e.Description).IsRequired(); }); } } public class Customer { public int ID { get; set; } public string Name { get; set; } public virtual List<Purchase> Purchases { get; set; } = new List<Purchase>(); } public class Purchase { public int ID { get; set; } public int? CustomerID { get; set; } public DateTime Date { get; set; } public string Description { get; set; } public decimal Price { get; set; } public virtual Customer Customer { get; set; } } All of the examples in this chapter are preloaded into LINQ‐ Pad, along with a sample database with a matching schema. You can download LINQPad from http://www.linqpad.net. Here are corresponding SQL Server table definitions: CREATE TABLE Customer ( ID int NOT NULL IDENTITY PRIMARY KEY, Name nvarchar(30) NOT NULL ) CREATE TABLE Purchase ( ID int NOT NULL IDENTITY PRIMARY KEY, CustomerID int NOT NULL REFERENCES Customer(ID), Date datetime NOT NULL, Description nvarchar(30) NOT NULL, Price decimal NOT NULL ) Overview In this section, we provide an overview of the standard query operators. They fall into three categories: • Sequence in, sequence out (sequence→sequence) • • Sequence in, single element or scalar value out • • Nothing in, sequence out (generation methods) • 448 | Chapter 9: LINQ Operators
- 467. We first present each of the three categories and the query operators they include, and then we take up each individual query operator in detail. Sequence→Sequence Most query operators fall into this category—accepting one or more sequences as input and emitting a single output sequence. Figure 9-1 illustrates those operators that restructure the shape of the sequences. Figure 9-1. Shape-changing operators Filtering IEnumerable<TSource> →IEnumerable<TSource> Returns a subset of the original elements. Where, Take, TakeLast, TakeWhile, Skip, SkipLast, SkipWhile, Distinct, DistinctBy Projecting IEnumerable<TSource>→IEnumerable<TResult> Transforms each element with a lambda function. SelectMany flattens nested sequences; Select and SelectMany perform inner joins, left outer joins, cross joins, and non-equi joins with EF Core. Select, SelectMany Joining IEnumerable<TOuter>, IEnumerable<TInner>→IEnumerable<TResult> Meshes elements of one sequence with another. Join and GroupJoin operators are designed to be efficient with local queries and support inner and left outer joins. The Zip operator enumerates two sequences in step, applying a function over each Overview | 449 LINQ Operators
- 468. element pair. Rather than naming the type arguments TOuter and TInner, the Zip operator names them TFirst and TSecond: IEnumerable<TFirst>, IEnumerable<TSecond>→IEnumerable<TResult> Join, GroupJoin, Zip Ordering IEnumerable<TSource>→IOrderedEnumerable<TSource> Returns a reordering of a sequence. OrderBy, OrderByDescending, ThenBy, ThenByDescending, Reverse Grouping IEnumerable<TSource>→IEnumerable<IGrouping<TKey,TElement>> IEnumerable<TSource>→IEnumerable<TElement[]> Groups a sequence into subsequences. GroupBy, Chunk Set operators IEnumerable<TSource>, IEnumerable<TSource>→IEnumerable<TSource> Takes two same-typed sequences and returns their commonality, sum, or difference. Concat, Union, UnionBy, Intersect, IntersectBy, Except, ExceptBy Conversion methods: Import IEnumerable→IEnumerable<TResult> OfType, Cast Conversion methods: Export IEnumerable<TSource>→An array, list, dictionary, lookup, or sequence ToArray, ToList, ToDictionary, ToLookup, AsEnumerable, AsQueryable Sequence→Element or Value The following query operators accept an input sequence and emit a single element or value. Element operators IEnumerable<TSource>→TSource Picks a single element from a sequence. 450 | Chapter 9: LINQ Operators
- 469. First, FirstOrDefault, Last, LastOrDefault, Single, SingleOrDefault, ElementAt, ElementAtOrDefault, MinBy, MaxBy, DefaultIfEmpty Aggregation methods IEnumerable<TSource>→scalar Performs a computation across a sequence, returning a scalar value (typically a number). Aggregate, Average, Count, LongCount, Sum, Max, Min Quantifiers IEnumerable<TSource>→bool An aggregation returning true or false. All, Any, Contains, SequenceEqual Void→Sequence In the third and final category are query operators that produce an output sequence from scratch. Generation methods void→IEnumerable<TResult> Manufactures a simple sequence. Empty, Range, Repeat Filtering IEnumerable<TSource>→IEnumerable<TSource> Method Description SQL equivalents Where Returns a subset of elements that satisfy a given condition WHERE Take Returns the first count elements and discards the rest WHERE ROW_NUMBER()... or TOP n subquery Skip Ignores the first count elements and returns the rest WHERE ROW_NUMBER()... or NOT IN (SELECT TOP n...) TakeLast Takes only the last count elements Exception thrown SkipLast Takes only the last count elements Exception thrown TakeWhile Emits elements from the input sequence until the predicate is false Exception thrown Filtering | 451 LINQ Operators
- 470. Method Description SQL equivalents SkipWhile Ignores elements from the input sequence until the predicate is false, and then emits the rest Exception thrown Distinct, DistinctBy Returns a sequence that excludes duplicates SELECT DISTINCT... The “SQL equivalents” column in the reference tables in this chapter do not necessarily correspond to what an IQueryable implementation such as EF Core will produce. Rather, it indi‐ cates what you’d typically use to do the same job if you were writing the SQL query yourself. Where there is no simple translation, the column is left blank. Where there is no trans‐ lation at all, the column reads “Exception thrown.” Enumerable implementation code, when shown, excludes checking for null arguments and indexing predicates. With each of the filtering methods, you always end up with either the same number or fewer elements than you started with. You can never get more! The elements are also identical when they come out; they are not transformed in any way. Where Argument Type Source sequence IEnumerable<TSource> Predicate TSource => bool or (TSource,int) => boola a Prohibited with LINQ to SQL and Entity Framework Query syntax where bool-expression Enumerable.Where implementation The internal implementation of Enumerable.Where, null checking aside, is function‐ ally equivalent to the following: public static IEnumerable<TSource> Where<TSource> (this IEnumerable<TSource> source, Func <TSource, bool> predicate) { foreach (TSource element in source) if (predicate (element)) yield return element; } 452 | Chapter 9: LINQ Operators
- 471. Overview Where returns the elements from the input sequence that satisfy the given predicate. For instance: string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; IEnumerable<string> query = names.Where (name => name.EndsWith ("y")); // Harry // Mary // Jay In query syntax: IEnumerable<string> query = from n in names where n.EndsWith ("y") select n; A where clause can appear more than once in a query and be interspersed with let, orderby, and join clauses: from n in names where n.Length > 3 let u = n.ToUpper() where u.EndsWith ("Y") select u; // HARRY // MARY Standard C# scoping rules apply to such queries. In other words, you cannot refer to a variable prior to declaring it with a range variable or a let clause. Indexed filtering Where’s predicate optionally accepts a second argument, of type int. This is fed with the position of each element within the input sequence, allowing the predicate to use this information in its filtering decision. For example, the following skips every second element: IEnumerable<string> query = names.Where ((n, i) => i % 2 == 0); // Tom // Harry // Jay An exception is thrown if you use indexed filtering in EF Core. SQL LIKE comparisons in EF Core The following methods on string translate to SQL’s LIKE operator: Contains, StartsWith, EndsWith Filtering | 453 LINQ Operators
- 472. For instance, c.Name.Contains ("abc") translates to customer.Name LIKE '%abc%' (or more accurately, a parameterized version of this). Contains lets you compare only against a locally evaluated expression; to compare against another column, you must use the EF.Functions.Like method: ... where EF.Functions.Like (c.Description, "%" + c.Name + "%") EF.Functions.Like also lets you perform more complex comparisons (e.g., LIKE 'abc%def%'). < and > string comparisons in EF Core You can perform order comparison on strings with string’s CompareTo method; this maps to SQL’s < and > operators: dbContext.Purchases.Where (p => p.Description.CompareTo ("C") < 0) WHERE x IN (…, …, …) in EF Core With EF Core, you can apply the Contains operator to a local collection within a filter predicate. For instance: string[] chosenOnes = { "Tom", "Jay" }; from c in dbContext.Customers where chosenOnes.Contains (c.Name) ... This maps to SQL’s IN operator. In other words: WHERE customer.Name IN ("Tom", "Jay") If the local collection is an array of entities or nonscalar types, EF Core might instead emit an EXISTS clause. Take, TakeLast, Skip, and SkipLast Argument Type Source sequence IEnumerable<TSource> Number of elements to take or skip int Take emits the first n elements and discards the rest; Skip discards the first n ele‐ ments and emits the rest. The two methods are useful together when implementing a web page allowing a user to navigate through a large set of matching records. For instance, suppose that a user searches a book database for the term “mercury”, and there are 100 matches. The following returns the first 20: IQueryable<Book> query = dbContext.Books .Where (b => b.Title.Contains ("mercury")) .OrderBy (b => b.Title) .Take (20); 454 | Chapter 9: LINQ Operators
- 473. The next query returns books 21 to 40: IQueryable<Book> query = dbContext.Books .Where (b => b.Title.Contains ("mercury")) .OrderBy (b => b.Title) .Skip (20).Take (20); EF Core translates Take and Skip to the ROW_NUMBER function in SQL Server 2005, or a TOP n subquery in earlier versions of SQL Server. The TakeLast and SkipLast methods take or skip the last n elements. From .NET 6, the Take method is overloaded to accept a Range variable. This overload can subsume the functionality of all four methods; for instance, Take(5..) is equivalent to Skip(5), and Take(..^5) is equivalent to SkipLast(5). TakeWhile and SkipWhile Argument Type Source sequence IEnumerable<TSource> Predicate TSource => bool or (TSource,int) => bool TakeWhile enumerates the input sequence, emitting each item until the given predi‐ cate is false. It then ignores the remaining elements: int[] numbers = { 3, 5, 2, 234, 4, 1 }; var takeWhileSmall = numbers.TakeWhile (n => n < 100); // { 3, 5, 2 } SkipWhile enumerates the input sequence, ignoring each item until the given predi‐ cate is false. It then emits the remaining elements: int[] numbers = { 3, 5, 2, 234, 4, 1 }; var skipWhileSmall = numbers.SkipWhile (n => n < 100); // { 234, 4, 1 } TakeWhile and SkipWhile have no translation to SQL and throw an exception if used in an EF Core query. Distinct and DistinctBy Distinct returns the input sequence, stripped of duplicates. You can optionally pass in a custom equality comparer. The following returns distinct letters in a string: char[] distinctLetters = "HelloWorld".Distinct().ToArray(); string s = new string (distinctLetters); // HeloWrd We can call LINQ methods directly on a string because string implements IEnumerable<char>. The DistinctBy method was introduced in .NET 6 and lets you specify a key selector to be applied before performing equality comparison. The result of the following expression is {1,2,3}: Filtering | 455 LINQ Operators
- 474. new[] { 1.0, 1.1, 2.0, 2.1, 3.0, 3.1 }.DistinctBy (n => Math.Round (n, 0)) Projecting IEnumerable<TSource>→ IEnumerable<TResult> Method Description SQL equivalents Select Transforms each input element with the given lambda expression SELECT SelectMany Transforms each input element, and then flattens and concatenates the resultant subsequences INNER JOIN, LEFT OUTER JOIN, CROSS JOIN When querying a database, Select and SelectMany are the most versatile joining constructs; for local queries, Join and GroupJoin are the most efficient joining constructs. Select Argument Type Source sequence IEnumerable<TSource> Result selector TSource => TResult or (TSource,int) => TResulta a Prohibited with EF Core Query syntax select projection-expression Enumerable implementation public static IEnumerable<TResult> Select<TSource,TResult> (this IEnumerable<TSource> source, Func<TSource,TResult> selector) { foreach (TSource element in source) yield return selector (element); } Overview With Select, you always get the same number of elements that you started with. Each element, however, can be transformed in any manner by the lambda function. The following selects the names of all fonts installed on the computer (from System.Drawing): IEnumerable<string> query = from f in FontFamily.Families select f.Name; 456 | Chapter 9: LINQ Operators
- 475. foreach (string name in query) Console.WriteLine (name); In this example, the select clause converts a FontFamily object to its name. Here’s the lambda equivalent: IEnumerable<string> query = FontFamily.Families.Select (f => f.Name); Select statements are often used to project into anonymous types: var query = from f in FontFamily.Families select new { f.Name, LineSpacing = f.GetLineSpacing (FontStyle.Bold) }; A projection with no transformation is sometimes used with query syntax to satisfy the requirement that the query end in a select or group clause. The following selects fonts supporting strikeout: IEnumerable<FontFamily> query = from f in FontFamily.Families where f.IsStyleAvailable (FontStyle.Strikeout) select f; foreach (FontFamily ff in query) Console.WriteLine (ff.Name); In such cases, the compiler omits the projection when translating to fluent syntax. Indexed projection The selector expression can optionally accept an integer argument, which acts as an indexer, providing the expression with the position of each input in the input sequence. This works only with local queries: string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; IEnumerable<string> query = names .Select ((s,i) => i + "=" + s); // { "0=Tom", "1=Dick", ... } Select subqueries and object hierarchies You can nest a subquery in a select clause to build an object hierarchy. The following example returns a collection describing each directory under Path.Get TempPath(), with a subcollection of files under each directory: string tempPath = Path.GetTempPath(); DirectoryInfo[] dirs = new DirectoryInfo (tempPath).GetDirectories(); var query = from d in dirs where (d.Attributes & FileAttributes.System) == 0 select new { DirectoryName = d.FullName, Created = d.CreationTime, Projecting | 457 LINQ Operators
- 476. Files = from f in d.GetFiles() where (f.Attributes & FileAttributes.Hidden) == 0 select new { FileName = f.Name, f.Length, } }; foreach (var dirFiles in query) { Console.WriteLine ("Directory: " + dirFiles.DirectoryName); foreach (var file in dirFiles.Files) Console.WriteLine (" " + file.FileName + " Len: " + file.Length); } The inner portion of this query can be called a correlated subquery. A subquery is correlated if it references an object in the outer query—in this case, it references d, the directory being enumerated. A subquery inside a Select allows you to map one object hierarchy to another, or map a relational object model to a hierarchical object model. With local queries, a subquery within a Select causes double-deferred execution. In our example, the files aren’t filtered or projected until the inner foreach statement enumerates. Subqueries and joins in EF Core Subquery projections work well in EF Core, and you can use them to do the work of SQL-style joins. Here’s how we retrieve each customer’s name along with their high-value purchases: var query = from c in dbContext.Customers select new { c.Name, Purchases = (from p in dbContext.Purchases where p.CustomerID == c.ID && p.Price > 1000 select new { p.Description, p.Price }) .ToList() }; foreach (var namePurchases in query) { Console.WriteLine ("Customer: " + namePurchases.Name); foreach (var purchaseDetail in namePurchases.Purchases) Console.WriteLine (" - $$$: " + purchaseDetail.Price); } Note the use of ToList in the subquery. EF Core 3 cannot cre‐ ate queryables from the subquery result when that subquery references the DbContext. This issue is being tracked by the EF Core team and might be resolved in a future release. 458 | Chapter 9: LINQ Operators
- 477. This style of query is ideally suited to interpreted queries. The outer query and subquery are processed as a unit, preventing unnecessary round-tripping. With local queries, however, it’s inefficient because every combination of outer and inner ele‐ ments must be enumerated to get the few matching combina‐ tions. A better choice for local queries is Join or GroupJoin, described in the following sections. This query matches up objects from two disparate collections, and it can be thought of as a “join.” The difference between this and a conventional database join (or subquery) is that we’re not flattening the output into a single two-dimensional result set. We’re mapping the relational data to hierarchical data, rather than to flat data. Here’s the same query simplified by using the Purchases collection navigation property on the Customer entity: from c in dbContext.Customers select new { c.Name, Purchases = from p in c.Purchases // Purchases is List<Purchase> where p.Price > 1000 select new { p.Description, p.Price } }; (EF Core 3 does not require ToList when performing the subquery on a navigation property.) Both queries are analogous to a left outer join in SQL in the sense that we get all customers in the outer enumeration, regardless of whether they have any purchases. To emulate an inner join—whereby customers without high-value purchases are excluded—we would need to add a filter condition on the purchases collection: from c in dbContext.Customers where c.Purchases.Any (p => p.Price > 1000) select new { c.Name, Purchases = from p in c.Purchases where p.Price > 1000 select new { p.Description, p.Price } }; This is slightly untidy, however, in that we’ve written the same predicate (Price > 1000) twice. We can avoid this duplication with a let clause: from c in dbContext.Customers let highValueP = from p in c.Purchases where p.Price > 1000 select new { p.Description, p.Price } where highValueP.Any() select new { c.Name, Purchases = highValueP }; Projecting | 459 LINQ Operators
- 478. This style of query is flexible. By changing Any to Count, for instance, we can modify the query to retrieve only customers with at least two high-value purchases: ... where highValueP.Count() >= 2 select new { c.Name, Purchases = highValueP }; Projecting into concrete types In the examples so far, we’ve instantiated anonymous types in the output. It can also be useful to instantiate (ordinary) named classes, which you populate with object initializers. Such classes can include custom logic and can be passed between methods and assemblies without using type information. A typical example is a custom business entity. A custom business entity is simply a class that you write with some properties but is designed to hide lower-level (database-related) details. You might exclude foreign key fields from business-entity classes, for instance. Assuming that we wrote custom entity classes called Customer Entity and PurchaseEntity, here’s how we could project into them: IQueryable<CustomerEntity> query = from c in dbContext.Customers select new CustomerEntity { Name = c.Name, Purchases = (from p in c.Purchases where p.Price > 1000 select new PurchaseEntity { Description = p.Description, Value = p.Price } ).ToList() }; // Force query execution, converting output to a more convenient List: List<CustomerEntity> result = query.ToList(); When created to transfer data between tiers in a program or between separate systems, custom business entity classes are often called data transfer objects (DTO). DTOs contain no business logic. Notice that so far, we’ve not had to use a Join or SelectMany statement. This is because we’re maintaining the hierarchical shape of the data, as illustrated in Fig‐ ure 9-2. With LINQ, you can often avoid the traditional SQL approach of flattening tables into a two-dimensional result set. 460 | Chapter 9: LINQ Operators
- 479. Figure 9-2. Projecting an object hierarchy SelectMany Argument Type Source sequence IEnumerable<TSource> Result selector TSource => IEnumerable<TResult> or (TSource,int) => IEnumerable<TResult>a a Prohibited with EF Core Query syntax from identifier1 in enumerable-expression1 from identifier2 in enumerable-expression2 ... Enumerable implementation public static IEnumerable<TResult> SelectMany<TSource,TResult> (IEnumerable<TSource> source, Func <TSource,IEnumerable<TResult>> selector) { foreach (TSource element in source) foreach (TResult subElement in selector (element)) yield return subElement; } Overview SelectMany concatenates subsequences into a single flat output sequence. Recall that for each input element, Select yields exactly one output element. In contrast, SelectMany yields 0..n output elements. The 0..n elements come from a subsequence or child sequence that the lambda expression must emit. Projecting | 461 LINQ Operators
- 480. You can use SelectMany to expand child sequences, flatten nested collections, and join two collections into a flat output sequence. Using the conveyor belt analogy, SelectMany funnels fresh material onto a conveyor belt. With SelectMany, each input element is the trigger for the introduction of fresh material. The fresh material is emitted by the selector lambda expression and must be a sequence. In other words, the lambda expression must emit a child sequence per input element. The final result is a concatenation of the child sequences emitted for each input element. Starting with a simple example, suppose that we have the following array of names, string[] fullNames = { "Anne Williams", "John Fred Smith", "Sue Green" }; that we want to convert to a single flat collection of words—in other words: "Anne", "Williams", "John", "Fred", "Smith", "Sue", Green" SelectMany is ideal for this task, because we’re mapping each input element to a variable number of output elements. All we must do is come up with a selector expression that converts each input element to a child sequence. string.Split does the job nicely: it takes a string and splits it into words, emitting the result as an array: string testInputElement = "Anne Williams"; string[] childSequence = testInputElement.Split(); // childSequence is { "Anne", "Williams" }; So, here’s our SelectMany query and the result: IEnumerable<string> query = fullNames.SelectMany (name => name.Split()); foreach (string name in query) Console.Write (name + "|"); // Anne|Williams|John|Fred|Smith|Sue|Green| If you replace SelectMany with Select, you get the same results in hierarchical form. The following emits a sequence of string arrays, requiring nested foreach statements to enumerate: IEnumerable<string[]> query = fullNames.Select (name => name.Split()); foreach (string[] stringArray in query) foreach (string name in stringArray) Console.Write (name + "|"); The benefit of SelectMany is that it yields a single flat result sequence. SelectMany is supported in query syntax and is invoked by having an additional generator—in other words, an extra from clause in the query. The from keyword has two meanings in query syntax. At the start of a query, it introduces the original range variable and input sequence. Anywhere else in the query, it translates to SelectMany. Here’s our query in query syntax: 462 | Chapter 9: LINQ Operators
- 481. IEnumerable<string> query = from fullName in fullNames from name in fullName.Split() // Translates to SelectMany select name; Note that the additional generator introduces a new range variable—in this case, name. The old range variable stays in scope, however, and we can subsequently access both. Multiple range variables In the preceding example, both name and fullName remain in scope until the query either ends or reaches an into clause. The extended scope of these variables is the killer scenario for query syntax over fluent syntax. To illustrate, we can take the preceding query and include fullName in the final projection: IEnumerable<string> query = from fullName in fullNames from name in fullName.Split() select name + " came from " + fullName; Anne came from Anne Williams Williams came from Anne Williams John came from John Fred Smith ... Behind the scenes, the compiler must pull some tricks to let you access both variables. A good way to appreciate this is to try writing the same query in fluent syntax. It’s tricky! It becomes yet more difficult if you insert a where or orderby clause before projecting: from fullName in fullNames from name in fullName.Split() orderby fullName, name select name + " came from " + fullName; The problem is that SelectMany emits a flat sequence of child elements—in our case, a flat collection of words. The original “outer” element from which it came (fullName) is lost. The solution is to “carry” the outer element with each child, in a temporary anonymous type: from fullName in fullNames from x in fullName.Split().Select (name => new { name, fullName } ) orderby x.fullName, x.name select x.name + " came from " + x.fullName; The only change here is that we’re wrapping each child element (name) in an anonymous type that also contains its fullName. This is similar to how a let clause is resolved. Here’s the final conversion to fluent syntax: Projecting | 463 LINQ Operators
- 482. IEnumerable<string> query = fullNames .SelectMany (fName => fName.Split() .Select (name => new { name, fName } )) .OrderBy (x => x.fName) .ThenBy (x => x.name) .Select (x => x.name + " came from " + x.fName); Thinking in query syntax As we just demonstrated, there are good reasons to use query syntax if you need multiple range variables. In such cases, it helps to not only use query syntax but also to think directly in its terms. There are two basic patterns when writing additional generators. The first is expanding and flattening subsequences. To do this, you call a property or method on an existing range variable in your additional generator. We did this in the previous example: from fullName in fullNames from name in fullName.Split() Here, we’ve expanded from enumerating full names to enumerating words. An analogous EF Core query is when you expand collection navigation properties. The following query lists all customers along with their purchases: IEnumerable<string> query = from c in dbContext.Customers from p in c.Purchases select c.Name + " bought a " + p.Description; Tom bought a Bike Tom bought a Holiday Dick bought a Phone Harry bought a Car ... Here, we’ve expanded each customer into a subsequence of purchases. The second pattern is performing a cartesian product, or cross join, in which every element of one sequence is matched with every element of another. To do this, introduce a generator whose selector expression returns a sequence unrelated to a range variable: int[] numbers = { 1, 2, 3 }; string[] letters = { "a", "b" }; IEnumerable<string> query = from n in numbers from l in letters select n.ToString() + l; // RESULT: { "1a", "1b", "2a", "2b", "3a", "3b" } This style of query is the basis of SelectMany-style joins. 464 | Chapter 9: LINQ Operators
- 483. Joining with SelectMany You can use SelectMany to join two sequences simply by filtering the results of a cross product. For instance, suppose that we want to match players for a game. We could start as follows: string[] players = { "Tom", "Jay", "Mary" }; IEnumerable<string> query = from name1 in players from name2 in players select name1 + " vs " + name2; //RESULT: { "Tom vs Tom", "Tom vs Jay", "Tom vs Mary", // "Jay vs Tom", "Jay vs Jay", "Jay vs Mary", // "Mary vs Tom", "Mary vs "Jay", "Mary vs Mary" } The query reads “For every player, reiterate every player, selecting player 1 versus player 2.” Although we got what we asked for (a cross join), the results are not useful until we add a filter: IEnumerable<string> query = from name1 in players from name2 in players where name1.CompareTo (name2) < 0 orderby name1, name2 select name1 + " vs " + name2; //RESULT: { "Jay vs Mary", "Jay vs Tom", "Mary vs Tom" } The filter predicate constitutes the join condition. Our query can be called a non- equi join because the join condition doesn’t use an equality operator. SelectMany in EF Core SelectMany in EF Core can perform cross joins, non-equi joins, inner joins, and left outer joins. You can use SelectMany with both predefined associations and ad hoc relationships—just as with Select. The difference is that SelectMany returns a flat rather than a hierarchical result set. An EF Core cross join is written just as in the preceding section. The following query matches every customer to every purchase (a cross join): var query = from c in dbContext.Customers from p in dbContext.Purchases select c.Name + " might have bought a " + p.Description; More typically, though, you’d want to match customers to only their own purchases. You achieve this by adding a where clause with a joining predicate. This results in a standard SQL-style equi-join: var query = from c in dbContext.Customers from p in dbContext.Purchases where c.ID == p.CustomerID select c.Name + " bought a " + p.Description; Projecting | 465 LINQ Operators
- 484. This translates well to SQL. In the next section, we see how it extends to support outer joins. Reformulating such queries with LINQ’s Join operator actually makes them less extensi‐ ble—LINQ is opposite to SQL in this sense. If you have collection navigation properties in your entities, you can express the same query by expanding the subcollection instead of filtering the cross product: from c in dbContext.Customers from p in c.Purchases select new { c.Name, p.Description }; The advantage is that we’ve eliminated the joining predicate. We’ve gone from filtering a cross product to expanding and flattening. You can add where clauses to such a query for additional filtering. For instance, if we want only customers whose names started with “T”, we could filter as follows: from c in dbContext.Customers where c.Name.StartsWith ("T") from p in c.Purchases select new { c.Name, p.Description }; This EF Core query would work equally well if the where clause were moved one line down because the same SQL is generated in both cases. If it is a local query, however, moving the where clause down would make it less efficient. With local queries, you should filter before joining. You can introduce new tables into the mix with additional from clauses. For instance, if each purchase had purchase item child rows, you could produce a flat result set of customers with their purchases, each with their purchase detail lines as follows: from c in dbContext.Customers from p in c.Purchases from pi in p.PurchaseItems select new { c.Name, p.Description, pi.Detail }; Each from clause introduces a new child table. To include data from a parent table (via a navigation property), you don’t add a from clause—you simply navigate to the property. For example, if each customer has a salesperson whose name you want to query, just do this: from c in dbContext.Customers select new { Name = c.Name, SalesPerson = c.SalesPerson.Name }; You don’t use SelectMany in this case because there’s no subcollection to flatten. Parent navigation properties return a single item. 466 | Chapter 9: LINQ Operators
- 485. Outer joins with SelectMany We saw previously that a Select subquery yields a result analogous to a left outer join: from c in dbContext.Customers select new { c.Name, Purchases = from p in c.Purchases where p.Price > 1000 select new { p.Description, p.Price } }; In this example, every outer element (customer) is included, regardless of whether the customer has any purchases. But suppose that we rewrite this query with SelectMany so that we can obtain a single flat collection rather than a hierarchical result set: from c in dbContext.Customers from p in c.Purchases where p.Price > 1000 select new { c.Name, p.Description, p.Price }; In the process of flattening the query, we’ve switched to an inner join: customers are now included only for whom one or more high-value purchases exist. To get a left outer join with a flat result set, we must apply the DefaultIfEmpty query operator on the inner sequence. This method returns a sequence with a single null element if its input sequence has no elements. Here’s such a query, price predicate aside: from c in dbContext.Customers from p in c.Purchases.DefaultIfEmpty() select new { c.Name, p.Description, Price = (decimal?) p.Price }; This works perfectly with EF Core, returning all customers—even if they have no purchases. But if we were to run this as a local query, it would crash because when p is null, p.Description and p.Price throw a NullReferenceException. We can make our query robust in either scenario, as follows: from c in dbContext.Customers from p in c.Purchases.DefaultIfEmpty() select new { c.Name, Descript = p == null ? null : p.Description, Price = p == null ? (decimal?) null : p.Price }; Let’s now reintroduce the price filter. We cannot use a where clause as we did before, because it would execute after DefaultIfEmpty: from c in dbContext.Customers from p in c.Purchases.DefaultIfEmpty() where p.Price > 1000... Projecting | 467 LINQ Operators
- 486. The correct solution is to splice the Where clause before DefaultIfEmpty with a subquery: from c in dbContext.Customers from p in c.Purchases.Where (p => p.Price > 1000).DefaultIfEmpty() select new { c.Name, Descript = p == null ? null : p.Description, Price = p == null ? (decimal?) null : p.Price }; EF Core translates this to a left outer join. This is an effective pattern for writing such queries. If you’re used to writing outer joins in SQL, you might be tempted to overlook the simpler option of a Select subquery for this style of query in favor of the awkward but familiar SQL-centric flat approach. The hierarchical result set from a Select subquery is often better suited to outer join–style queries because there are no additional nulls to deal with. Joining Method Description SQL equivalents Join Applies a lookup strategy to match elements from two collections, emitting a flat result set INNER JOIN GroupJoin Similar to Join, but emits a hierarchical result set INNER JOIN, LEFT OUTER JOIN Zip Enumerates two sequences in step (like a zipper), applying a function over each element pair Exception thrown Join and GroupJoin IEnumerable<TOuter>, IEnumerable<TInner>→IEnumerable<TResult> Join arguments Argument Type Outer sequence IEnumerable<TOuter> Inner sequence IEnumerable<TInner> Outer key selector TOuter => TKey Inner key selector TInner => TKey Result selector (TOuter,TInner) => TResult 468 | Chapter 9: LINQ Operators
- 487. GroupJoin arguments Argument Type Outer sequence IEnumerable<TOuter> Inner sequence IEnumerable<TInner> Outer key selector TOuter => TKey Inner key selector TInner => TKey Result selector (TOuter,IEnumerable<TInner>) => TResult Query syntax from outer-var in outer-enumerable join inner-var in inner-enumerable on outer-key-expr equals inner-key-expr [ into identifier ] Overview Join and GroupJoin mesh two input sequences into a single output sequence. Join emits flat output; GroupJoin emits hierarchical output. Join and GroupJoin provide an alternative strategy to Select and SelectMany. The advantage of Join and GroupJoin is that they execute efficiently over local in-memory collections because they first load the inner sequence into a keyed lookup, avoiding the need to repeatedly enumerate over every inner element. The disadvantage is that they offer the equivalent of inner and left outer joins only; cross joins and non-equi joins must still be done using Select/SelectMany. With EF Core queries, Join and GroupJoin offer no real benefits over Select and SelectMany. Table 9-1 summarizes the differences between each of the joining strategies. Table 9-1. Joining strategies Strategy Result shape Local query efficiency Inner joins Left outer joins Cross joins Non- equi joins Select + SelectMany Flat Bad Yes Yes Yes Yes Select + Select Nested Bad Yes Yes Yes Yes Join Flat Good Yes — — — GroupJoin Nested Good Yes Yes — — GroupJoin + SelectMany Flat Good Yes Yes — — Joining | 469 LINQ Operators
- 488. Join The Join operator performs an inner join, emitting a flat output sequence. The following query lists all customers alongside their purchases without using a navigation property: IQueryable<string> query = from c in dbContext.Customers join p in dbContext.Purchases on c.ID equals p.CustomerID select c.Name + " bought a " + p.Description; The results match what we would get from a SelectMany-style query: Tom bought a Bike Tom bought a Holiday Dick bought a Phone Harry bought a Car To see the benefit of Join over SelectMany, we must convert this to a local query. We can demonstrate this by first copying all customers and purchases to arrays and then querying the arrays: Customer[] customers = dbContext.Customers.ToArray(); Purchase[] purchases = dbContext.Purchases.ToArray(); var slowQuery = from c in customers from p in purchases where c.ID == p.CustomerID select c.Name + " bought a " + p.Description; var fastQuery = from c in customers join p in purchases on c.ID equals p.CustomerID select c.Name + " bought a " + p.Description; Although both queries yield the same results, the Join query is considerably faster because its implementation in Enumerable preloads the inner collection (purchases) into a keyed lookup. The query syntax for join can be written in general terms, as follows: join inner-var in inner-sequence on outer-key-expr equals inner-key-expr Join operators in LINQ differentiate between the outer sequence and inner sequence. Syntactically: • The outer sequence is the input sequence (customers, in this case). • • The inner sequence is the new collection you introduce (purchases, in this • case). Join performs inner joins, meaning customers without purchases are excluded from the output. With inner joins, you can swap the inner and outer sequences in the query and still get the same results: from p in purchases // p is now outer join c in customers on p.CustomerID equals c.ID // c is now inner ... 470 | Chapter 9: LINQ Operators
- 489. You can add further join clauses to the same query. If each purchase, for instance, has one or more purchase items, you could join the purchase items, as follows: from c in customers join p in purchases on c.ID equals p.CustomerID // first join join pi in purchaseItems on p.ID equals pi.PurchaseID // second join ... purchases acts as the inner sequence in the first join and as the outer sequence in the second join. You could obtain the same results (inefficiently) using nested foreach statements, as follows: foreach (Customer c in customers) foreach (Purchase p in purchases) if (c.ID == p.CustomerID) foreach (PurchaseItem pi in purchaseItems) if (p.ID == pi.PurchaseID) Console.WriteLine (c.Name + "," + p.Price + "," + pi.Detail); In query syntax, variables from earlier joins remain in scope—just as they do with SelectMany-style queries. You’re also permitted to insert where and let clauses in between join clauses. Joining on multiple keys You can join on multiple keys with anonymous types, as follows: from x in sequenceX join y in sequenceY on new { K1 = x.Prop1, K2 = x.Prop2 } equals new { K1 = y.Prop3, K2 = y.Prop4 } ... For this to work, the two anonymous types must be structured identically. The compiler then implements each with the same internal type, making the joining keys compatible. Joining in fluent syntax The following query syntax join from c in customers join p in purchases on c.ID equals p.CustomerID select new { c.Name, p.Description, p.Price }; in fluent syntax is as follows: customers.Join ( // outer collection purchases, // inner collection c => c.ID, // outer key selector p => p.CustomerID, // inner key selector (c, p) => new { c.Name, p.Description, p.Price } // result selector ); Joining | 471 LINQ Operators
- 490. The result selector expression at the end creates each element in the output sequence. If you have additional clauses prior to projecting, such as orderby in this example: from c in customers join p in purchases on c.ID equals p.CustomerID orderby p.Price select c.Name + " bought a " + p.Description; you must manufacture a temporary anonymous type in the result selector in fluent syntax. This keeps both c and p in scope following the join: customers.Join ( // outer collection purchases, // inner collection c => c.ID, // outer key selector p => p.CustomerID, // inner key selector (c, p) => new { c, p } ) // result selector .OrderBy (x => x.p.Price) .Select (x => x.c.Name + " bought a " + x.p.Description); Query syntax is usually preferable when joining; it’s less fiddly. GroupJoin GroupJoin does the same work as Join, but instead of yielding a flat result, it yields a hierarchical result, grouped by each outer element. It also allows left outer joins. GroupJoin is not currently supported in EF Core. The query syntax for GroupJoin is the same as for Join, but is followed by the into keyword. Here’s the most basic example, using a local query: Customer[] customers = dbContext.Customers.ToArray(); Purchase[] purchases = dbContext.Purchases.ToArray(); IEnumerable<IEnumerable<Purchase>> query = from c in customers join p in purchases on c.ID equals p.CustomerID into custPurchases select custPurchases; // custPurchases is a sequence An into clause translates to GroupJoin only when it appears directly after a join clause. After a select or group clause, it means query continuation. The two uses of the into key‐ word are quite different, although they have one feature in common: they both introduce a new range variable. The result is a sequence of sequences, which we could enumerate as follows: foreach (IEnumerable<Purchase> purchaseSequence in query) foreach (Purchase p in purchaseSequence) Console.WriteLine (p.Description); 472 | Chapter 9: LINQ Operators
- 491. This isn’t very useful, however, because purchaseSequence has no reference to the customer. More commonly, you’d do this: from c in customers join p in purchases on c.ID equals p.CustomerID into custPurchases select new { CustName = c.Name, custPurchases }; This gives the same results as the following (inefficient) Select subquery: from c in customers select new { CustName = c.Name, custPurchases = purchases.Where (p => c.ID == p.CustomerID) }; By default, GroupJoin does the equivalent of a left outer join. To get an inner join—whereby customers without purchases are excluded—you need to filter on custPurchases: from c in customers join p in purchases on c.ID equals p.CustomerID into custPurchases where custPurchases.Any() select ... Clauses after a group-join into operate on subsequences of inner child elements, not individual child elements. This means that to filter individual purchases, you’d need to call Where before joining: from c in customers join p in purchases.Where (p2 => p2.Price > 1000) on c.ID equals p.CustomerID into custPurchases ... You can construct lambda queries with GroupJoin as you would with Join. Flat outer joins You run into a dilemma if you want both an outer join and a flat result set. GroupJoin gives you the outer join; Join gives you the flat result set. The solution is to first call GroupJoin, then DefaultIfEmpty on each child sequence, and then finally SelectMany on the result: from c in customers join p in purchases on c.ID equals p.CustomerID into custPurchases from cp in custPurchases.DefaultIfEmpty() select new { CustName = c.Name, Price = cp == null ? (decimal?) null : cp.Price }; DefaultIfEmpty emits a sequence with a single null value if a subsequence of purchases is empty. The second from clause translates to SelectMany. In this role, it Joining | 473 LINQ Operators
- 492. expands and flattens all the purchase subsequences, concatenating them into a single sequence of purchase elements. Joining with lookups The Join and GroupJoin methods in Enumerable work in two steps. First, they load the inner sequence into a lookup. Second, they query the outer sequence in combination with the lookup. A lookup is a sequence of groupings that can be accessed directly by key. Another way to think of it is as a dictionary of sequences—a dictionary that can accept many elements under each key (sometimes called a multidictionary). Lookups are read-only and defined by the following interface: public interface ILookup<TKey,TElement> : IEnumerable<IGrouping<TKey,TElement>>, IEnumerable { int Count { get; } bool Contains (TKey key); IEnumerable<TElement> this [TKey key] { get; } } The joining operators—like other sequence-emitting opera‐ tors—honor deferred or lazy execution semantics. This means the lookup is not built until you begin enumerating the output sequence (and then the entire lookup is built right then). You can create and query lookups manually as an alternative strategy to using the joining operators when dealing with local collections. There are a couple of benefits to doing so: • You can reuse the same lookup over multiple queries—as well as in ordinary • imperative code. • Querying a lookup is an excellent way of understanding how Join and Group • Join work. The ToLookup extension method creates a lookup. The following loads all purchases into a lookup—keyed by their CustomerID: ILookup<int?,Purchase> purchLookup = purchases.ToLookup (p => p.CustomerID, p => p); The first argument selects the key; the second argument selects the objects that are to be loaded as values into the lookup. Reading a lookup is rather like reading a dictionary except that the indexer returns a sequence of matching items rather than a single matching item. The following enumerates all purchases made by the customer whose ID is 1: foreach (Purchase p in purchLookup [1]) Console.WriteLine (p.Description); 474 | Chapter 9: LINQ Operators
- 493. With a lookup in place, you can write SelectMany/Select queries that execute as efficiently as Join/GroupJoin queries. Join is equivalent to using SelectMany on a lookup: from c in customers from p in purchLookup [c.ID] select new { c.Name, p.Description, p.Price }; Tom Bike 500 Tom Holiday 2000 Dick Bike 600 Dick Phone 300 ... Adding a call to DefaultIfEmpty makes this into an outer join: from c in customers from p in purchLookup [c.ID].DefaultIfEmpty() select new { c.Name, Descript = p == null ? null : p.Description, Price = p == null ? (decimal?) null : p.Price }; GroupJoin is equivalent to reading the lookup inside a projection: from c in customers select new { CustName = c.Name, CustPurchases = purchLookup [c.ID] }; Enumerable implementations Here’s the simplest valid implementation of Enumerable.Join, null checking aside: public static IEnumerable <TResult> Join <TOuter,TInner,TKey,TResult> ( this IEnumerable <TOuter> outer, IEnumerable <TInner> inner, Func <TOuter,TKey> outerKeySelector, Func <TInner,TKey> innerKeySelector, Func <TOuter,TInner,TResult> resultSelector) { ILookup <TKey, TInner> lookup = inner.ToLookup (innerKeySelector); return from outerItem in outer from innerItem in lookup [outerKeySelector (outerItem)] select resultSelector (outerItem, innerItem); } Joining | 475 LINQ Operators
- 494. GroupJoin’s implementation is like that of Join but simpler: public static IEnumerable <TResult> GroupJoin <TOuter,TInner,TKey,TResult> ( this IEnumerable <TOuter> outer, IEnumerable <TInner> inner, Func <TOuter,TKey> outerKeySelector, Func <TInner,TKey> innerKeySelector, Func <TOuter,IEnumerable<TInner>,TResult> resultSelector) { ILookup <TKey, TInner> lookup = inner.ToLookup (innerKeySelector); return from outerItem in outer select resultSelector (outerItem, lookup [outerKeySelector (outerItem)]); } The Zip Operator IEnumerable<TFirst>, IEnumerable<TSecond>→IEnumerable<TResult> The Zip operator enumerates two sequences in step (like a zipper), returning a sequence based on applying a function over each element pair. For instance, the following: int[] numbers = { 3, 5, 7 }; string[] words = { "three", "five", "seven", "ignored" }; IEnumerable<string> zip = numbers.Zip (words, (n, w) => n + "=" + w); produces a sequence with the following elements: 3=three 5=five 7=seven Extra elements in either input sequence are ignored. Zip is not supported by EF Core. Ordering IEnumerable<TSource>→IOrderedEnumerable<TSource> Method Description SQL equivalents OrderBy, ThenBy Sorts a sequence in ascending order ORDER BY ... OrderByDescending, ThenByDescending Sorts a sequence in descending order ORDER BY ... DESC Reverse Returns a sequence in reverse order Exception thrown Ordering operators return the same elements in a different order. 476 | Chapter 9: LINQ Operators
- 495. OrderBy, OrderByDescending, ThenBy, and ThenByDescending OrderBy and OrderByDescending arguments Argument Type Input sequence IEnumerable<TSource> Key selector TSource => TKey Return type = IOrderedEnumerable<TSource> ThenBy and ThenByDescending arguments Argument Type Input sequence IOrderedEnumerable<TSource> Key selector TSource => TKey Query syntax orderby expression1 [descending] [, expression2 [descending] ... ] Overview OrderBy returns a sorted version of the input sequence, using the keySelector expression to make comparisons. The following query emits a sequence of names in alphabetical order: IEnumerable<string> query = names.OrderBy (s => s); The following sorts names by length: IEnumerable<string> query = names.OrderBy (s => s.Length); // Result: { "Jay", "Tom", "Mary", "Dick", "Harry" }; The relative order of elements with the same sorting key (in this case, Jay/Tom and Mary/Dick) is indeterminate—unless you append a ThenBy operator: IEnumerable<string> query = names.OrderBy (s => s.Length).ThenBy (s => s); // Result: { "Jay", "Tom", "Dick", "Mary", "Harry" }; ThenBy reorders only elements that had the same sorting key in the preceding sort. You can chain any number of ThenBy operators. The following sorts first by length, then by the second character, and finally by the first character: names.OrderBy (s => s.Length).ThenBy (s => s[1]).ThenBy (s => s[0]); Ordering | 477 LINQ Operators
- 496. Here’s the equivalent in query syntax: from s in names orderby s.Length, s[1], s[0] select s; The following variation is incorrect—it will actually order first by s[1] and then by s.Length (or in the case of a database query, it will order only by s[1] and discard the former ordering): from s in names orderby s.Length orderby s[1] ... LINQ also provides OrderByDescending and ThenByDescending operators, which do the same things, emitting the results in reverse order. The following EF Core query retrieves purchases in descending order of price, with those of the same price listed alphabetically: dbContext.Purchases.OrderByDescending (p => p.Price) .ThenBy (p => p.Description); In query syntax: from p in dbContext.Purchases orderby p.Price descending, p.Description select p; Comparers and collations In a local query, the key selector objects themselves determine the ordering algo‐ rithm via their default IComparable implementation (see Chapter 7). You can override the sorting algorithm by passing in an IComparer object. The following performs a case-insensitive sort: names.OrderBy (n => n, StringComparer.CurrentCultureIgnoreCase); Passing in a comparer is not supported in query syntax or in any way by EF Core. When querying a database, the comparison algorithm is determined by the participating column’s collation. If the collation is case sensitive, you can request a case-insensitive sort by calling ToUpper in the key selector: from p in dbContext.Purchases orderby p.Description.ToUpper() select p; IOrderedEnumerable and IOrderedQueryable The ordering operators return special subtypes of IEnumerable<T>. Those in Enumerable return IOrderedEnumerable<TSource>; those in Queryable return IOrderedQueryable<TSource>. These subtypes allow a subsequent ThenBy operator to refine rather than replace the existing ordering. 478 | Chapter 9: LINQ Operators
- 497. The additional members that these subtypes define are not publicly exposed, so they present like ordinary sequences. The fact that they are different types comes into play when building queries progressively: IOrderedEnumerable<string> query1 = names.OrderBy (s => s.Length); IOrderedEnumerable<string> query2 = query1.ThenBy (s => s); If we instead declare query1 of type IEnumerable<string>, the second line would not compile—ThenBy requires an input of type IOrderedEnumerable<string>. You can avoid worrying about this by implicitly typing range variables: var query1 = names.OrderBy (s => s.Length); var query2 = query1.ThenBy (s => s); Implicit typing can create problems of its own, though. The following will not compile: var query = names.OrderBy (s => s.Length); query = query.Where (n => n.Length > 3); // Compile-time error The compiler infers query to be of type IOrderedEnumerable<string>, based on OrderBy’s output sequence type. However, the Where on the next line returns an ordinary IEnumerable<string>, which cannot be assigned back to query. You can work around this either with explicit typing or by calling AsEnumerable() after OrderBy: var query = names.OrderBy (s => s.Length).AsEnumerable(); query = query.Where (n => n.Length > 3); // OK The equivalent in interpreted queries is to call AsQueryable. Grouping Method Description SQL equivalents GroupBy Groups a sequence into subsequences GROUP BY Chunk Groups a sequence into arrays of a fixed size GroupBy IEnumerable<TSource>→IEnumerable<IGrouping<TKey,TElement>> Argument Type Input sequence IEnumerable<TSource> Key selector TSource => TKey Element selector (optional) TSource => TElement Comparer (optional) IEqualityComparer<TKey> Grouping | 479 LINQ Operators
- 498. Query syntax group element-expression by key-expression Overview GroupBy organizes a flat input sequence into sequences of groups. For example, the following organizes all of the files in Path.GetTempPath() by extension: string[] files = Directory.GetFiles (Path.GetTempPath()); IEnumerable<IGrouping<string,string>> query = files.GroupBy (file => Path.GetExtension (file)); Or, with implicit typing: var query = files.GroupBy (file => Path.GetExtension (file)); Here’s how to enumerate the result: foreach (IGrouping<string,string> grouping in query) { Console.WriteLine ("Extension: " + grouping.Key); foreach (string filename in grouping) Console.WriteLine (" - " + filename); } Extension: .pdf -- chapter03.pdf -- chapter04.pdf Extension: .doc -- todo.doc -- menu.doc -- Copy of menu.doc ... Enumerable.GroupBy works by reading the input elements into a temporary dictio‐ nary of lists so that all elements with the same key end up in the same sublist. It then emits a sequence of groupings. A grouping is a sequence with a Key property: public interface IGrouping <TKey,TElement> : IEnumerable<TElement>, IEnumerable { TKey Key { get; } // Key applies to the subsequence as a whole } By default, the elements in each grouping are untransformed input elements unless you specify an elementSelector argument. The following projects each input ele‐ ment to uppercase: files.GroupBy (file => Path.GetExtension (file), file => file.ToUpper()); An elementSelector is independent of the keySelector. In our case, this means that the Key on each grouping is still in its original case: 480 | Chapter 9: LINQ Operators
- 499. Extension: .pdf -- CHAPTER03.PDF -- CHAPTER04.PDF Extension: .doc -- TODO.DOC Note that the subcollections are not emitted in alphabetical order of key. GroupBy merely groups; it does not sort. In fact, it preserves the original ordering. To sort, you must add an OrderBy operator: files.GroupBy (file => Path.GetExtension (file), file => file.ToUpper()) .OrderBy (grouping => grouping.Key); GroupBy has a simple and direct translation in query syntax: group element-expr by key-expr Here’s our example in query syntax: from file in files group file.ToUpper() by Path.GetExtension (file); As with select, group “ends” a query—unless you add a query continuation clause: from file in files group file.ToUpper() by Path.GetExtension (file) into grouping orderby grouping.Key select grouping; Query continuations are often useful in a group by query. The next query filters out groups that have fewer than five files in them: from file in files group file.ToUpper() by Path.GetExtension (file) into grouping where grouping.Count() >= 5 select grouping; A where after a group by is equivalent to HAVING in SQL. It applies to each subsequence or grouping as a whole rather than the individual elements. Sometimes, you’re interested purely in the result of an aggregation on a grouping and so can abandon the subsequences: string[] votes = { "Dogs", "Cats", "Cats", "Dogs", "Dogs" }; IEnumerable<string> query = from vote in votes group vote by vote into g orderby g.Count() descending select g.Key; string winner = query.First(); // Dogs Grouping | 481 LINQ Operators
- 500. GroupBy in EF Core Grouping works in the same way when querying a database. If you have navigation properties set up, you’ll find, however, that the need to group arises less frequently than with standard SQL. For instance, to select customers with at least two purcha‐ ses, you don’t need to group; the following query does the job nicely: from c in dbContext.Customers where c.Purchases.Count >= 2 select c.Name + " has made " + c.Purchases.Count + " purchases"; An example of when you might use grouping is to list total sales by year: from p in dbContext.Purchases group p.Price by p.Date.Year into salesByYear select new { Year = salesByYear.Key, TotalValue = salesByYear.Sum() }; LINQ’s grouping is more powerful than SQL’s GROUP BY in that you can fetch all detail rows without any aggregation: from p in dbContext.Purchases group p by p.Date.Year Date.Year However, this doesn’t work in EF Core. An easy workaround is to call .AsEnumera ble() just before grouping so that the grouping happens on the client. This is no less efficient as long as you perform any filtering before grouping so that you only fetch the data you need from the server. Another departure from traditional SQL comes in there being no obligation to project the variables or expressions used in grouping or sorting. Grouping by multiple keys You can group by a composite key, using an anonymous type: from n in names group n by new { FirstLetter = n[0], Length = n.Length }; Custom equality comparers You can pass a custom equality comparer into GroupBy, in a local query, to change the algorithm for key comparison. Rarely is this required, though, because changing the key selector expression is usually sufficient. For instance, the following creates a case-insensitive grouping: group n by n.ToUpper() 482 | Chapter 9: LINQ Operators
- 501. Chunk IEnumerable<TSource>→IEnumerable<TElement[]> Argument Type Input sequence IEnumerable<TSource> size int Introduced in .NET 6, Chunk groups a sequence into chunks of a given size (or fewer, if there aren’t enough elements): foreach (int[] chunk in new[] { 1, 2, 3, 4, 5, 6, 7, 8 }.Chunk (3)) Console.WriteLine (string.Join (", ", chunk)); Output: 1, 2, 3 4, 5, 6 7, 8 Set Operators IEnumerable<TSource>, IEnumerable<TSource>→IEnumerable<TSource> Method Description SQL equivalents Concat Returns a concatenation of elements in each of the two sequences UNION ALL Union, UnionBy Returns a concatenation of elements in each of the two sequences, excluding duplicates UNION Intersect, IntersectBy Returns elements present in both sequences WHERE ... IN (...) Except, ExceptBy Returns elements present in the first but not the second sequence EXCEPT or WHERE ... NOT IN (...) Concat, Union, and UnionBy Concat returns all the elements of the first sequence, followed by all the elements of the second. Union does the same but removes any duplicates: int[] seq1 = { 1, 2, 3 }, seq2 = { 3, 4, 5 }; IEnumerable<int> concat = seq1.Concat (seq2), // { 1, 2, 3, 3, 4, 5 } union = seq1.Union (seq2); // { 1, 2, 3, 4, 5 } Specifying the type argument explicitly is useful when the sequences are differently typed but the elements have a common base type. For instance, with the reflection API (Chapter 18), methods and properties are represented with MethodInfo and Set Operators | 483 LINQ Operators
- 502. PropertyInfo classes, which have a common base class called MemberInfo. We can concatenate methods and properties by stating that base class explicitly when calling Concat: MethodInfo[] methods = typeof (string).GetMethods(); PropertyInfo[] props = typeof (string).GetProperties(); IEnumerable<MemberInfo> both = methods.Concat<MemberInfo> (props); In the next example, we filter the methods before concatenating: var methods = typeof (string).GetMethods().Where (m => !m.IsSpecialName); var props = typeof (string).GetProperties(); var both = methods.Concat<MemberInfo> (props); This example relies on interface type parameter variance: methods is of type IEnumerable<MethodInfo>, which requires a covariant conversion to IEnumerable <MemberInfo>. It’s a good illustration of how variance makes things work more like you’d expect. UnionBy (introduced in .NET 6) takes a keySelector, which is used in determin‐ ing whether an element is a duplicate. In the following example, we perform a case-insensitive union: string[] seq1 = { "A", "b", "C" }; string[] seq2 = { "a", "B", "c" }; var union = seq1.UnionBy (seq2, x => x.ToUpperInvariant()); // union is { "A", "b", "C" } In this case, the same thing can be accomplished with Union, if we supply an equality comparer: var union = seq1.Union (seq2, StringComparer.InvariantCultureIgnoreCase); Intersect, Intersect By, Except, and ExceptBy Intersect returns the elements that two sequences have in common. Except returns the elements in the first input sequence that are not present in the second: int[] seq1 = { 1, 2, 3 }, seq2 = { 3, 4, 5 }; IEnumerable<int> commonality = seq1.Intersect (seq2), // { 3 } difference1 = seq1.Except (seq2), // { 1, 2 } difference2 = seq2.Except (seq1); // { 4, 5 } Enumerable.Except works internally by loading all of the elements in the first collection into a dictionary and then removing from the dictionary all elements present in the second sequence. The equivalent in SQL is a NOT EXISTS or NOT IN subquery: SELECT number FROM numbers1Table WHERE number NOT IN (SELECT number FROM numbers2Table) 484 | Chapter 9: LINQ Operators
- 503. The IntersectBy and ExceptBy methods (from .NET 6) let you specify a key selector that’s applied before performing equality comparison (see the discussion on UnionBy in the preceding section). Conversion Methods LINQ deals primarily in sequences; in other words, collections of type IEnumera ble<T>. The conversion methods convert to and from other types of collections: Method Description OfType Converts IEnumerable to IEnumerable<T>, discarding wrongly typed elements Cast Converts IEnumerable to IEnumerable<T>, throwing an exception if there are any wrongly typed elements ToArray Converts IEnumerable<T> to T[] ToList Converts IEnumerable<T> to List<T> ToDictionary Converts IEnumerable<T> to Dictionary<TKey,TValue> ToLookup Converts IEnumerable<T> to ILookup<TKey,TElement> AsEnumerable Upcasts to IEnumerable<T> AsQueryable Casts or converts to IQueryable<T> OfType and Cast OfType and Cast accept a nongeneric IEnumerable collection and emit a generic IEnumerable<T> sequence that you can subsequently query: ArrayList classicList = new ArrayList(); // in System.Collections classicList.AddRange ( new int[] { 3, 4, 5 } ); IEnumerable<int> sequence1 = classicList.Cast<int>(); Cast and OfType differ in their behavior when encountering an input element that’s of an incompatible type. Cast throws an exception; OfType ignores the incompatible element. Continuing the preceding example: DateTime offender = DateTime.Now; classicList.Add (offender); IEnumerable<int> sequence2 = classicList.OfType<int>(), // OK - ignores offending DateTime sequence3 = classicList.Cast<int>(); // Throws exception The rules for element compatibility exactly follow those of C#’s is operator, and therefore consider only reference conversions and unboxing conversions. We can see this by examining the internal implementation of OfType: public static IEnumerable<TSource> OfType <TSource> (IEnumerable source) { foreach (object element in source) if (element is TSource) Conversion Methods | 485 LINQ Operators
- 504. yield return (TSource)element; } Cast has an identical implementation, except that it omits the type compatibility test: public static IEnumerable<TSource> Cast <TSource> (IEnumerable source) { foreach (object element in source) yield return (TSource)element; } A consequence of these implementations is that you cannot use Cast to perform numeric or custom conversions (for these, you must perform a Select operation instead). In other words, Cast is not as flexible as C#’s cast operator: int i = 3; long l = i; // Implicit numeric conversion int->long int i2 = (int) l; // Explicit numeric conversion long->int We can demonstrate this by attempting to use OfType or Cast to convert a sequence of ints to a sequence of longs: int[] integers = { 1, 2, 3 }; IEnumerable<long> test1 = integers.OfType<long>(); IEnumerable<long> test2 = integers.Cast<long>(); When enumerated, test1 emits zero elements and test2 throws an exception. Examining OfType’s implementation, it’s fairly clear why. After substituting TSource, we get the following expression: (element is long) This returns false for an int element, due to the lack of an inheritance relationship. The reason that test2 throws an exception when enumerated is more subtle. Notice in Cast’s implementation that element is of type object. When TSource is a value type, the CLR assumes this is an unboxing conversion and synthesizes a method that reproduces the scenario described in the section “Boxing and Unboxing” on page 126: int value = 123; object element = value; long result = (long) element; // exception Because the element variable is declared of type object, an object-to-long cast is performed (an unboxing) rather than an int-to-long numeric conversion. Unboxing operations require an exact type match, so the object-to-long unbox fails when given an int. 486 | Chapter 9: LINQ Operators
- 505. As we suggested previously, the solution is to use an ordinary Select: IEnumerable<long> castLong = integers.Select (s => (long) s); OfType and Cast are also useful in downcasting elements in a generic input sequence. For instance, if you have an input sequence of type IEnumerable<Fruit>, OfType<Apple> would return just the apples. This is particularly useful in LINQ to XML (see Chapter 10). Cast has query syntax support: simply precede the range variable with a type: from TreeNode node in myTreeView.Nodes ... ToArray, ToList, ToDictionary, ToHashSet, and ToLookup ToArray, ToList, and ToHashSet emit the results into an array, List<T> or Hash Set<T>. When they execute, these operators force the immediate enumeration of the input sequence. For examples, refer to “Deferred Execution” on page 408 in Chapter 8. ToDictionary and ToLookup accept the following arguments: Argument Type Input sequence IEnumerable<TSource> Key selector TSource => TKey Element selector (optional) TSource => TElement Comparer (optional) IEqualityComparer<TKey> ToDictionary also forces immediate execution of a sequence, writing the results to a generic Dictionary. The keySelector expression you provide must evaluate to a unique value for each element in the input sequence; otherwise, an exception is thrown. In contrast, ToLookup allows many elements of the same key. We described lookups in “Joining with lookups” on page 474. AsEnumerable and AsQueryable AsEnumerable upcasts a sequence to IEnumerable<T>, forcing the compiler to bind subsequent query operators to methods in Enumerable instead of Queryable. For an example, see “Combining Interpreted and Local Queries” on page 428. AsQueryable downcasts a sequence to IQueryable<T> if it implements that inter‐ face. Otherwise, it instantiates an IQueryable<T> wrapper over the local query. Conversion Methods | 487 LINQ Operators
- 506. Element Operators IEnumerable<TSource>→ TSource Method Description SQL equivalents First, FirstOrDefault Returns the first element in the sequence, optionally satisfying a predicate SELECT TOP 1 ... ORDER BY ... Last, LastOrDefault Returns the last element in the sequence, optionally satisfying a predicate SELECT TOP 1 ... ORDER BY ... DESC Single, SingleOrDefault Equivalent to First/FirstOrDefault, but throws an exception if there is more than one match ElementAt, ElementAtOrDefault Returns the element at the specified position Exception thrown MinBy, MaxBy Returns the element with the smallest or largest value Exception thrown DefaultIfEmpty Returns a single-element sequence whose value is default(TSource) if the sequence has no elements OUTER JOIN Methods ending in “OrDefault” return default(TSource) rather than throwing an exception if the input sequence is empty or if no elements match the supplied predicate. default(TSource) is null for reference type elements, false for the bool type, and zero for numeric types. First, Last, and Single Argument Type Source sequence IEnumerable<TSource> Predicate (optional) TSource => bool The following example demonstrates First and Last: int[] numbers = { 1, 2, 3, 4, 5 }; int first = numbers.First(); // 1 int last = numbers.Last(); // 5 int firstEven = numbers.First (n => n % 2 == 0); // 2 int lastEven = numbers.Last (n => n % 2 == 0); // 4 The following demonstrates First versus FirstOrDefault: int firstBigError = numbers.First (n => n > 10); // Exception int firstBigNumber = numbers.FirstOrDefault (n => n > 10); // 0 488 | Chapter 9: LINQ Operators
- 507. To prevent an exception, Single requires exactly one matching element; SingleOr Default requires one or zero matching elements: int onlyDivBy3 = numbers.Single (n => n % 3 == 0); // 3 int divBy2Err = numbers.Single (n => n % 2 == 0); // Error: 2 & 4 match int singleError = numbers.Single (n => n > 10); // Error int noMatches = numbers.SingleOrDefault (n => n > 10); // 0 int divBy2Error = numbers.SingleOrDefault (n => n % 2 == 0); // Error Single is the “fussiest” in this family of element operators. FirstOrDefault and LastOrDefault are the most tolerant. In EF Core, Single is often used to retrieve a row from a table by primary key: Customer cust = dataContext.Customers.Single (c => c.ID == 3); ElementAt Argument Type Source sequence IEnumerable<TSource> Index of element to return int ElementAt picks the nth element from the sequence: int[] numbers = { 1, 2, 3, 4, 5 }; int third = numbers.ElementAt (2); // 3 int tenthError = numbers.ElementAt (9); // Exception int tenth = numbers.ElementAtOrDefault (9); // 0 Enumerable.ElementAt is written such that if the input sequence happens to imple‐ ment IList<T>, it calls IList<T>’s indexer. Otherwise, it enumerates n times and then returns the next element. ElementAt is not supported in EF Core. MinBy and MaxBy MinBy and MaxBy (introduced in .NET 6) return the element with the smallest or largest value, as determined by a keySelector: string[] names = { "Tom", "Dick", "Harry", "Mary", "Jay" }; Console.WriteLine (names.MaxBy (n => n.Length)); // Harry In contrast, Min and Max (which we will cover in the following section) return the smallest or largest value itself: Console.WriteLine (names.Max (n => n.Length)); // 5 If two or more elements share a minimum/maximum value, MinBy/MaxBy returns the first: Console.WriteLine (names.MinBy (n => n.Length)); // Tom Element Operators | 489 LINQ Operators
- 508. If the input sequence is empty, MinBy and MaxBy return null if the element type is nullable (or throw an exception if the element type is not nullable). DefaultIfEmpty DefaultIfEmpty returns a sequence containing a single element whose value is default(TSource) if the input sequence has no elements; otherwise, it returns the input sequence unchanged. You use this in writing flat outer joins: see “Outer joins with SelectMany” on page 467 and “Flat outer joins” on page 473. Aggregation Methods IEnumerable<TSource>→scalar Method Description SQL equivalents Count, LongCount Returns the number of elements in the input sequence, optionally satisfying a predicate COUNT (...) Min, Max Returns the smallest or largest element in the sequence MIN (...), MAX (...) Sum, Average Calculates a numeric sum or average over elements in the sequence SUM (...), AVG (...) Aggregate Performs a custom aggregation Exception thrown Count and LongCount Argument Type Source sequence IEnumerable<TSource> Predicate (optional) TSource => bool Count simply enumerates over a sequence, returning the number of items: int fullCount = new int[] { 5, 6, 7 }.Count(); // 3 The internal implementation of Enumerable.Count tests the input sequence to see whether it happens to implement ICollection<T>. If it does, it simply calls ICollec tion<T>.Count; otherwise, it enumerates over every item, incrementing a counter. You can optionally supply a predicate: int digitCount = "pa55w0rd".Count (c => char.IsDigit (c)); // 3 LongCount does the same job as Count but returns a 64-bit integer, allowing for sequences of greater than two billion elements. 490 | Chapter 9: LINQ Operators
- 509. Min and Max Argument Type Source sequence IEnumerable<TSource> Result selector (optional) TSource => TResult Min and Max return the smallest or largest element from a sequence: int[] numbers = { 28, 32, 14 }; int smallest = numbers.Min(); // 14; int largest = numbers.Max(); // 32; If you include a selector expression, each element is first projected: int smallest = numbers.Max (n => n % 10); // 8; A selector expression is mandatory if the items themselves are not intrinsically comparable—in other words, if they do not implement IComparable<T>: Purchase runtimeError = dbContext.Purchases.Min (); // Error decimal? lowestPrice = dbContext.Purchases.Min (p => p.Price); // OK A selector expression determines not only how elements are compared, but also the final result. In the preceding example, the final result is a decimal value, not a purchase object. To get the cheapest purchase, you need a subquery: Purchase cheapest = dbContext.Purchases .Where (p => p.Price == dbContext.Purchases.Min (p2 => p2.Price)) .FirstOrDefault(); In this case, you could also formulate the query without an aggregation by using an OrderBy followed by FirstOrDefault. Sum and Average Argument Type Source sequence IEnumerable<TSource> Result selector (optional) TSource => TResult Sum and Average are aggregation operators that are used in a similar manner to Min and Max: decimal[] numbers = { 3, 4, 8 }; decimal sumTotal = numbers.Sum(); // 15 decimal average = numbers.Average(); // 5 (mean value) The following returns the total length of each of the strings in the names array: int combinedLength = names.Sum (s => s.Length); // 19 Aggregation Methods | 491 LINQ Operators
- 510. Sum and Average are fairly restrictive in their typing. Their definitions are hard‐ wired to each of the numeric types (int, long, float, double, decimal, and their nullable versions). In contrast, Min and Max can operate directly on anything that implements IComparable<T>—such as a string, for instance. Further, Average always returns either decimal, float, or double, according to the following table: Selector type Result type decimal decimal float float int, long, double double This means that the following does not compile (“cannot convert double to int”): int avg = new int[] { 3, 4 }.Average(); But this will compile: double avg = new int[] { 3, 4 }.Average(); // 3.5 Average implicitly upscales the input values to prevent loss of precision. In this example, we averaged integers and got 3.5 without needing to resort to an input element cast: double avg = numbers.Average (n => (double) n); When querying a database, Sum and Average translate to the standard SQL aggrega‐ tions. The following query returns customers whose average purchase was more than $500: from c in dbContext.Customers where c.Purchases.Average (p => p.Price) > 500 select c.Name; Aggregate Aggregate allows you to specify a custom accumulation algorithm for implement‐ ing unusual aggregations. Aggregate is not supported in EF Core and is somewhat specialized in its use cases. The following demonstrates how Aggregate can do the work of Sum: int[] numbers = { 1, 2, 3 }; int sum = numbers.Aggregate (0, (total, n) => total + n); // 6 The first argument to Aggregate is the seed, from which accumulation starts. The second argument is an expression to update the accumulated value, given a fresh element. You can optionally supply a third argument to project the final result value from the accumulated value. 492 | Chapter 9: LINQ Operators
- 511. Most problems for which Aggregate has been designed can be solved as easily with a foreach loop—and with more familiar syntax. The advantage of using Aggregate is that with large or complex aggregations, you can automatically parallelize the operation with PLINQ (see Chapter 22). Unseeded aggregations You can omit the seed value when calling Aggregate, in which case the first element becomes the implicit seed, and aggregation proceeds from the second element. Here’s the preceding example, unseeded: int[] numbers = { 1, 2, 3 }; int sum = numbers.Aggregate ((total, n) => total + n); // 6 This gives the same result as before, but we’re actually doing a different calculation. Before, we were calculating 0 + 1 + 2 + 3; now we’re calculating 1 + 2 + 3. We can better illustrate the difference by multiplying instead of adding: int[] numbers = { 1, 2, 3 }; int x = numbers.Aggregate (0, (prod, n) => prod * n); // 0*1*2*3 = 0 int y = numbers.Aggregate ( (prod, n) => prod * n); // 1*2*3 = 6 As you’ll see in Chapter 22, unseeded aggregations have the advantage of being parallelizable without requiring the use of special overloads. However, there are some traps with unseeded aggregations. Traps with unseeded aggregations The unseeded aggregation methods are intended for use with delegates that are commutative and associative. If used otherwise, the result is either unintuitive (with ordinary queries) or nondeterministic (in the case that you parallelize the query with PLINQ). For example, consider the following function: (total, n) => total + n * n This is neither commutative nor associative. (For example, 1 + 2 * 2 != 2 + 1 * 1.) Let’s see what happens when we use it to sum the square of the numbers 2, 3, and 4: int[] numbers = { 2, 3, 4 }; int sum = numbers.Aggregate ((total, n) => total + n * n); // 27 Instead of calculating 2*2 + 3*3 + 4*4 // 29 it calculates: 2 + 3*3 + 4*4 // 27 We can fix this in a number of ways. First, we could include 0 as the first element: int[] numbers = { 0, 2, 3, 4 }; Aggregation Methods | 493 LINQ Operators
- 512. Not only is this inelegant, but it will still give incorrect results if parallelized— because PLINQ uses the function’s assumed associativity by selecting multiple ele‐ ments as seeds. To illustrate, if we denote our aggregation function as follows: f(total, n) => total + n * n LINQ to Objects would calculate this: f(f(f(0, 2),3),4) whereas PLINQ might do this: f(f(0,2),f(3,4)) with the following result: First partition: a = 0 + 2*2 (= 4) Second partition: b = 3 + 4*4 (= 19) Final result: a + b*b (= 365) OR EVEN: b + a*a (= 35) There are two good solutions. The first is to turn this into a seeded aggregation with 0 as the seed. The only complication is that with PLINQ, we’d need to use a special overload in order for the query not to execute sequentially (see “Optimizing PLINQ” on page 919). The second solution is to restructure the query such that the aggregation function is commutative and associative: int sum = numbers.Select (n => n * n).Aggregate ((total, n) => total + n); Of course, in such simple scenarios you can (and should) use the Sum operator instead of Aggregate: int sum = numbers.Sum (n => n * n); You can actually go quite far just with Sum and Average. For instance, you can use Average to calculate a root-mean- square: Math.Sqrt (numbers.Average (n => n * n)) You can even calculate standard deviation: double mean = numbers.Average(); double sdev = Math.Sqrt (numbers.Average (n => { double dif = n - mean; return dif * dif; })); Both are safe, efficient, and fully parallelizable. In Chapter 22, we give a practical example of a custom aggregation that can’t be reduced to Sum or Average. 494 | Chapter 9: LINQ Operators
- 513. Quantifiers IEnumerable<TSource>→bool Method Description SQL equivalents Contains Returns true if the input sequence contains the given element WHERE ... IN (...) Any Returns true if any elements satisfy the given predicate WHERE ... IN (...) All Returns true if all elements satisfy the given predicate WHERE (...) SequenceEqual Returns true if the second sequence has identical elements to the input sequence Contains and Any The Contains method accepts an argument of type TSource; Any accepts an optional predicate. Contains returns true if the given element is present: bool hasAThree = new int[] { 2, 3, 4 }.Contains (3); // true; Any returns true if the given expression is true for at least one element. We can rewrite the preceding query with Any as follows: bool hasAThree = new int[] { 2, 3, 4 }.Any (n => n == 3); // true; Any can do everything that Contains can do, and more: bool hasABigNumber = new int[] { 2, 3, 4 }.Any (n => n > 10); // false; Calling Any without a predicate returns true if the sequence has one or more elements. Here’s another way to write the preceding query: bool hasABigNumber = new int[] { 2, 3, 4 }.Where (n => n > 10).Any(); Any is particularly useful in subqueries and is used often when querying databases; for example: from c in dbContext.Customers where c.Purchases.Any (p => p.Price > 1000) select c All and SequenceEqual All returns true if all elements satisfy a predicate. The following returns customers whose purchases are less than $100: dbContext.Customers.Where (c => c.Purchases.All (p => p.Price < 100)); SequenceEqual compares two sequences. To return true, each sequence must have identical elements, in the identical order. You can optionally provide an equality comparer; the default is EqualityComparer<T>.Default. Quantifiers | 495 LINQ Operators
- 514. Generation Methods void→IEnumerable<TResult> Method Description Empty Creates an empty sequence Repeat Creates a sequence of repeating elements Range Creates a sequence of integers Empty, Repeat, and Range are static (nonextension) methods that manufacture simple local sequences. Empty Empty manufactures an empty sequence and requires just a type argument: foreach (string s in Enumerable.Empty<string>()) Console.Write (s); // <nothing> In conjunction with the ?? operator, Empty does the reverse of DefaultIfEmpty. For example, suppose that we have a jagged array of integers and we want to get all the integers into a single flat list. The following SelectMany query fails if any of the inner arrays is null: int[][] numbers = { new int[] { 1, 2, 3 }, new int[] { 4, 5, 6 }, null // this null makes the query below fail. }; IEnumerable<int> flat = numbers.SelectMany (innerArray => innerArray); Empty in conjunction with ?? fixes the problem: IEnumerable<int> flat = numbers .SelectMany (innerArray => innerArray ?? Enumerable.Empty <int>()); foreach (int i in flat) Console.Write (i + " "); // 1 2 3 4 5 6 Range and Repeat Range accepts a starting index and count (both integers): foreach (int i in Enumerable.Range (5, 3)) Console.Write (i + " "); // 5 6 7 Repeat accepts an element to repeat, and the number of repetitions: foreach (bool x in Enumerable.Repeat (true, 3)) Console.Write (x + " "); // True True True 496 | Chapter 9: LINQ Operators
- 515. 10 LINQ to XML .NET provides a number of APIs for working with XML data. The primary choice for general-purpose XML document processing is LINQ to XML. LINQ to XML comprises a lightweight, LINQ-friendly XML document object model (DOM), plus a set of supplementary query operators. In this chapter, we concentrate entirely on LINQ to XML. In Chapter 11, we cover the forward-only XML reader/writer, and in the online supplement, we cover the types for working with schemas and stylesheets. .NET also includes the legacy XmlDocument-based DOM, which we don’t cover. The LINQ to XML DOM is extremely well designed and highly performant. Even without LINQ, the LINQ to XML DOM is valuable as a lightweight façade over the low-level XmlReader and XmlWriter classes. All LINQ to XML types are defined in the System.Xml.Linq namespace. Architectural Overview This section starts with a very brief introduction to the concept of a DOM, and then explains the rationale behind LINQ to XML’s DOM. What Is a DOM? Consider the following XML file: <?xml version="1.0" encoding="utf-8"?> <customer id="123" status="archived"> <firstname>Joe</firstname> <lastname>Bloggs</lastname> </customer> 497
- 516. As with all XML files, we start with a declaration and then a root element, whose name is customer. The customer element has two attributes, each with a name (id and status) and value ("123" and "archived"). Within customer, there are two child elements, firstname and lastname, each having simple text content ("Joe" and "Bloggs"). Each of these constructs—declaration, element, attribute, value, and text content— can be represented with a class. And if such classes have collection properties for storing child content, we can assemble a tree of objects to fully describe a document. This is called a Document Object Model, or DOM. The LINQ to XML DOM LINQ to XML comprises two things: • An XML DOM, which we call the X-DOM • • A set of about 10 supplementary query operators • As you might expect, the X-DOM consists of types such as XDocument, XElement, and XAttribute. Interestingly, the X-DOM types are not tied to LINQ—you can load, instantiate, update, and save an X-DOM without ever writing a LINQ query. Conversely, you could use LINQ to query a DOM created of the older W3C- compliant types. However, this would be frustrating and limiting. The distinguish‐ ing feature of the X-DOM is that it’s LINQ-friendly, meaning: • It has methods that emit useful IEnumerable sequences upon which you can • query. • Its constructors are designed such that you can build an X-DOM tree through a • LINQ projection. X-DOM Overview Figure 10-1 shows the core X-DOM types. The most frequently used of these types is XElement. XObject is the root of the inheritance hierarchy; XElement and XDocument are roots of the containership hierarchy. 498 | Chapter 10: LINQ to XML
- 517. Figure 10-1. Core X-DOM types Figure 10-2 shows the X-DOM tree created from the following code: string xml = @"<customer id='123' status='archived'> <firstname>Joe</firstname> <lastname>Bloggs<!--nice name--></lastname> </customer>"; XElement customer = XElement.Parse (xml); Figure 10-2. A simple X-DOM tree X-DOM Overview | 499 LINQ to XML
- 518. XObject is the abstract base class for all XML content. It defines a link to the Parent element in the containership tree as well as an optional XDocument. XNode is the base class for most XML content excluding attributes. The distinguish‐ ing feature of XNode is that it can sit in an ordered collection of mixed-type XNodes. For instance, consider the following XML: <data> Hello world <subelement1/> <!--comment--> <subelement2/> </data> Within the parent element <data>, there’s first an XText node (Hello world), then an XElement node, then an XComment node, and then a second XElement node. In contrast, an XAttribute will tolerate only other XAttributes as peers. Although an XNode can access its parent XElement, it has no concept of child nodes: this is the job of its subclass XContainer. XContainer defines members for dealing with children and is the abstract base class for XElement and XDocument. XElement introduces members for managing attributes—as well as a Name and Value. In the (fairly common) case of an element having a single XText child node, the Value property on XElement encapsulates this child’s content for both get and set operations, cutting unnecessary navigation. Thanks to Value, you can mostly avoid working directly with XText nodes. XDocument represents the root of an XML tree. More precisely, it wraps the root XElement, adding an XDeclaration, processing instructions, and other root-level “fluff.” Unlike with the W3C DOM, its use is optional: you can load, manipulate, and save an X-DOM without ever creating an XDocument! The nonreliance on XDocument also means that you can efficiently and easily move a node subtree to another X-DOM hierarchy. Loading and Parsing Both XElement and XDocument provide static Load and Parse methods to build an X-DOM tree from an existing source: • Load builds an X-DOM from a file, URI, Stream, TextReader, or XmlReader. • • Parse builds an X-DOM from a string. • For example: XDocument fromWeb = XDocument.Load ("http://albahari.com/sample.xml"); XElement fromFile = XElement.Load (@"e:mediasomefile.xml"); XElement config = XElement.Parse ( @"<configuration> 500 | Chapter 10: LINQ to XML
- 519. <client enabled='true'> <timeout>30</timeout> </client> </configuration>"); In later sections, we describe how to traverse and update an X-DOM. As a quick preview, here’s how to manipulate the config element we just populated: foreach (XElement child in config.Elements()) Console.WriteLine (child.Name); // client XElement client = config.Element ("client"); bool enabled = (bool) client.Attribute ("enabled"); // Read attribute Console.WriteLine (enabled); // True client.Attribute ("enabled").SetValue (!enabled); // Update attribute int timeout = (int) client.Element ("timeout"); // Read element Console.WriteLine (timeout); // 30 client.Element ("timeout").SetValue (timeout * 2); // Update element client.Add (new XElement ("retries", 3)); // Add new element Console.WriteLine (config); // Implicitly call config.ToString() Here’s the result of that last Console.WriteLine: <configuration> <client enabled="false"> <timeout>60</timeout> <retries>3</retries> </client> </configuration> XNode also provides a static ReadFrom method that instantiates and populates any type of node from an XmlReader. Unlike Load, it stops after reading one (complete) node, so you can continue to read manually from the XmlReader afterward. You can also do the reverse and use an XmlReader or XmlWriter to read or write an XNode, via its CreateReader and CreateWriter methods. We describe XML readers and writers and how to use them with the X-DOM in Chapter 11. Saving and Serializing Calling ToString on any node converts its content to an XML string—format‐ ted with line breaks and indentation as we just saw. (You can disable the line breaks and indentation by specifying SaveOptions.DisableFormatting when call‐ ing ToString.) XElement and XDocument also provide a Save method that writes an X-DOM to a file, Stream, TextWriter, or XmlWriter. If you specify a file, an XML declaration is X-DOM Overview | 501 LINQ to XML
- 520. automatically written. There is also a WriteTo method defined in the XNode class, which accepts just an XmlWriter. We describe in more detail the handling of XML declarations when saving in “Documents and Declarations” on page 515. Instantiating an X-DOM Rather than using the Load or Parse methods, you can build an X-DOM tree by manually instantiating objects and adding them to a parent via XContainer’s Add method. To construct an XElement and XAttribute, simply provide a name and value: XElement lastName = new XElement ("lastname", "Bloggs"); lastName.Add (new XComment ("nice name")); XElement customer = new XElement ("customer"); customer.Add (new XAttribute ("id", 123)); customer.Add (new XElement ("firstname", "Joe")); customer.Add (lastName); Console.WriteLine (customer.ToString()); Here’s the result: <customer id="123"> <firstname>Joe</firstname> <lastname>Bloggs<!--nice name--></lastname> </customer> A value is optional when constructing an XElement—you can provide just the ele‐ ment name and add content later. Notice that when we did provide a value, a simple string sufficed—we didn’t need to explicitly create and add an XText child node. The X-DOM does this work automatically, so you can deal simply with “values.” Functional Construction In our preceding example, it’s difficult to glean the XML structure from the code. X-DOM supports another mode of instantiation, called functional construction (from functional programming). With functional construction, you build an entire tree in a single expression: XElement customer = new XElement ("customer", new XAttribute ("id", 123), new XElement ("firstname", "joe"), new XElement ("lastname", "bloggs", new XComment ("nice name") ) ); 502 | Chapter 10: LINQ to XML
- 521. 1 The X-DOM actually optimizes this step internally by storing simple text content in a string. The XTEXT node is not actually created until you call Nodes( ) on the XContainer. This has two benefits. First, the code resembles the shape of the XML. Second, it can be incorporated into the select clause of a LINQ query. For example, the following query projects from an EF Core entity class into an X-DOM: XElement query = new XElement ("customers", from c in dbContext.Customers.AsEnumerable() select new XElement ("customer", new XAttribute ("id", c.ID), new XElement ("firstname", c.FirstName), new XElement ("lastname", c.LastName, new XComment ("nice name") ) ) ); We examine this further later in this chapter in “Projecting into an X-DOM” on page 525. Specifying Content Functional construction is possible because the constructors for XElement (and XDocument) are overloaded to accept a params object array: public XElement (XName name, params object[] content) The same holds true for the Add method in XContainer: public void Add (params object[] content) Hence, you can specify any number of child objects of any type when building or appending an X-DOM. This works because anything counts as legal content. To see how, we need to examine how each content object is processed internally. Here are the decisions made by XContainer, in order: 1. If the object is null, it’s ignored. 1. 2. If the object is based on XNode or XStreamingElement, it’s added as is to the 2. Nodes collection. 3. If the object is an XAttribute, it’s added to the Attributes collection. 3. 4. If the object is a string, it’s wrapped in an XText node and added to Nodes.1 4. 5. If the object implements IEnumerable, it’s enumerated, and the same rules are 5. applied to each element. 6. Otherwise, the object is converted to a string, wrapped in an XText node, and 6. then added to Nodes.2 Instantiating an X-DOM | 503 LINQ to XML
- 522. 2 See footnote 1. Everything ends up in one of two buckets: Nodes or Attributes. Furthermore, any object is valid content because it can always ultimately call ToString on it and treat it as an XText node. Before calling ToString on an arbitrary type, XContainer first tests whether it is one of the following types: float, double, decimal, bool, DateTime, DateTimeOffset, TimeSpan If so, it calls an appropriate typed ToString method on the XmlConvert helper class instead of calling ToString on the object itself. This ensures that the data is round-trippable and compliant with standard XML formatting rules. Automatic Deep Cloning When a node or attribute is added to an element (whether via functional construc‐ tion or an Add method), the node or attribute’s Parent property is set to that element. A node can have only one parent element: if you add an already parented node to a second parent, the node is automatically deep-cloned. In the following example, each customer has a separate copy of address: var address = new XElement ("address", new XElement ("street", "Lawley St"), new XElement ("town", "North Beach") ); var customer1 = new XElement ("customer1", address); var customer2 = new XElement ("customer2", address); customer1.Element ("address").Element ("street").Value = "Another St"; Console.WriteLine ( customer2.Element ("address").Element ("street").Value); // Lawley St This automatic duplication keeps X-DOM object instantiation free of side effects— another hallmark of functional programming. Navigating and Querying As you might expect, the XNode and XContainer classes define methods and proper‐ ties for traversing the X-DOM tree. Unlike a conventional DOM, however, these functions don’t return a collection that implements IList<T>. Instead, they return either a single value or a sequence that implements IEnumerable<T>—upon which you are then expected to execute a LINQ query (or enumerate with a foreach). This allows for advanced queries as well as simple navigation tasks—using familiar LINQ query syntax. 504 | Chapter 10: LINQ to XML
- 523. Element and attribute names are case sensitive in the X-DOM, just as they are in XML. Child Node Navigation Return type Members Works on XNode FirstNode { get; } XContainer LastNode { get; } XContainer IEnumerable<XNode> Nodes() XContainer* DescendantNodes() XContainer* DescendantNodesAndSelf() XElement* XElement Element (XName) XContainer IEnumerable<XElement> Elements() XContainer* Elements (XName) XContainer* Descendants() XContainer* Descendants (XName) XContainer* DescendantsAndSelf() XElement* DescendantsAndSelf (XName) XElement* bool HasElements { get; } XElement Functions marked with an asterisk in the third column of this and other tables also operate on sequences of the same type. For instance, you can call Nodes on either an XContainer or a sequence of XContainer objects. This is possible because of extension methods defined in System.Xml.Linq—the supple‐ mentary query operators we talked about in the overview. FirstNode, LastNode, and Nodes FirstNode and LastNode give you direct access to the first or last child node; Nodes returns all children as a sequence. All three functions consider only direct descendants: var bench = new XElement ("bench", new XElement ("toolbox", new XElement ("handtool", "Hammer"), new XElement ("handtool", "Rasp") ), new XElement ("toolbox", new XElement ("handtool", "Saw"), new XElement ("powertool", "Nailgun") ), new XComment ("Be careful with the nailgun") ); Navigating and Querying | 505 LINQ to XML
- 524. foreach (XNode node in bench.Nodes()) Console.WriteLine (node.ToString (SaveOptions.DisableFormatting) + "."); This is the output: <toolbox><handtool>Hammer</handtool><handtool>Rasp</handtool></toolbox>. <toolbox><handtool>Saw</handtool><powertool>Nailgun</powertool></toolbox>. <!--Be careful with the nailgun-->. Retrieving elements The Elements method returns just the child nodes of type XElement: foreach (XElement e in bench.Elements()) Console.WriteLine (e.Name + "=" + e.Value); // toolbox=HammerRasp // toolbox=SawNailgun The following LINQ query finds the toolbox with the nail gun: IEnumerable<string> query = from toolbox in bench.Elements() where toolbox.Elements().Any (tool => tool.Value == "Nailgun") select toolbox.Value; RESULT: { "SawNailgun" } The next example uses a SelectMany query to retrieve the hand tools in all toolboxes: IEnumerable<string> query = from toolbox in bench.Elements() from tool in toolbox.Elements() where tool.Name == "handtool" select tool.Value; RESULT: { "Hammer", "Rasp", "Saw" } Elements itself is equivalent to a LINQ query on Nodes. Our preceding query could be started as follows: from toolbox in bench.Nodes().OfType<XElement>() where ... Elements can also return just the elements of a given name: int x = bench.Elements ("toolbox").Count(); // 2 This is equivalent to the following: int x = bench.Elements().Where (e => e.Name == "toolbox").Count(); // 2 Elements is also defined as an extension method accepting IEnumerable <XContainer> or, more precisely, it accepts an argument of this type: IEnumerable<T> where T : XContainer This allows it to work with sequences of elements, too. Using this method, we can rewrite the query that finds the hand tools in all toolboxes as follows: 506 | Chapter 10: LINQ to XML
- 525. from tool in bench.Elements ("toolbox").Elements ("handtool") select tool.Value; The first call to Elements binds to XContainer’s instance method; the second call to Elements binds to the extension method. Retrieving a single element The method Element (singular) returns the first matching element of the given name. Element is useful for simple navigation, as follows: XElement settings = XElement.Load ("databaseSettings.xml"); string cx = settings.Element ("database").Element ("connectString").Value; Element is equivalent to calling Elements() and then applying LINQ’s FirstOr Default query operator with a name-matching predicate. Element returns null if the requested element doesn’t exist. Element("xyz").Value will throw a NullReferenceExcep tion if element xyz does not exist. If you’d prefer a null to an exception, either use the null-conditional operator— Element("xyz")?.Value—or cast the XElement to a string instead of querying its Value property. In other words: string xyz = (string) settings.Element ("xyz"); This works because XElement defines an explicit string con‐ version—just for this purpose! Retrieving descendants XContainer also provides Descendants and DescendantNodes methods that return child elements or nodes plus all of their children, and so on (the entire tree). Descendants accepts an optional element name. Returning to our earlier example, we can use Descendants to find all of the hand tools: Console.WriteLine (bench.Descendants ("handtool").Count()); // 3 Both parent and leaf nodes are included, as the following example demonstrates: foreach (XNode node in bench.DescendantNodes()) Console.WriteLine (node.ToString (SaveOptions.DisableFormatting)); Here’s the output: <toolbox><handtool>Hammer</handtool><handtool>Rasp</handtool></toolbox> <handtool>Hammer</handtool> Hammer <handtool>Rasp</handtool> Rasp <toolbox><handtool>Saw</handtool><powertool>Nailgun</powertool></toolbox> <handtool>Saw</handtool> Saw <powertool>Nailgun</powertool> Nailgun <!--Be careful with the nailgun--> Navigating and Querying | 507 LINQ to XML
- 526. The next query extracts all comments anywhere within the X-DOM that contain the word “careful”: IEnumerable<string> query = from c in bench.DescendantNodes().OfType<XComment>() where c.Value.Contains ("careful") orderby c.Value select c.Value; Parent Navigation All XNodes have a Parent property and AncestorXXX methods for parent navigation. A parent is always an XElement: Return type Members Works on XElement Parent { get; } XNode Enumerable<XElement> Ancestors() XNode Ancestors (XName) XNode AncestorsAndSelf() XElement AncestorsAndSelf (XName) XElement If x is an XElement, the following always prints true: foreach (XNode child in x.Nodes()) Console.WriteLine (child.Parent == x); However, the same is not the case if x is an XDocument. XDocument is peculiar: it can have children but can never be anyone’s parent! To access the XDocument, you instead use the Document property; this works on any object in the X-DOM tree. Ancestors returns a sequence whose first element is Parent and whose next ele‐ ment is Parent.Parent, and so on, until the root element. You can navigate to the root element with the LINQ query AncestorsAndSelf().Last(). Another way to achieve the same thing is to call Document .Root, although this works only if an XDocument is present. 508 | Chapter 10: LINQ to XML
- 527. Peer Node Navigation Return type Members Defined in bool IsBefore (XNode node) XNode IsAfter (XNode node) XNode XNode PreviousNode { get; } XNode NextNode { get; } XNode IEnumerable<XNode> NodesBeforeSelf() XNode NodesAfterSelf() XNode IEnumerable<XElement> ElementsBeforeSelf() XNode ElementsBeforeSelf (XName name) XNode ElementsAfterSelf() XNode ElementsAfterSelf (XName name) XNode With PreviousNode and NextNode (and FirstNode/LastNode), you can traverse nodes with the feel of a linked list. This is noncoincidental: internally, nodes are stored in a linked list. XNode internally uses a singly linked list, so PreviousNode is not performant. Attribute Navigation Return type Members Defined in bool HasAttributes { get; } XElement XAttribute Attribute (XName name) XElement FirstAttribute { get; } XElement LastAttribute { get; } XElement IEnumerable<XAttribute> Attributes() XElement Attributes (XName name) XElement In addition, XAttribute defines PreviousAttribute and NextAttribute properties as well as Parent. The Attributes method that accepts a name returns a sequence with either zero or one element; an element cannot have duplicate attribute names in XML. Navigating and Querying | 509 LINQ to XML
- 528. Updating an X-DOM You can update elements and attributes in the following ways: • Call SetValue or reassign the Value property. • • Call SetElementValue or SetAttributeValue. • • Call one of the RemoveXXX methods. • • Call one of the AddXXX or ReplaceXXX methods, specifying fresh content. • You can also reassign the Name property on XElement objects. Simple Value Updates Members Works on SetValue (object value) XElement, XAttribute Value { get; set } XElement, XAttribute The SetValue method replaces an element or attribute’s content with a simple value. Setting the Value property does the same but accepts string data only. We describe both of these functions in detail later in “Working with Values” on page 513. An effect of calling SetValue (or reassigning Value) is that it replaces all child nodes: XElement settings = new XElement ("settings", new XElement ("timeout", 30) ); settings.SetValue ("blah"); Console.WriteLine (settings.ToString()); // <settings>blah</settings> Updating Child Nodes and Attributes Category Members Works on Add Add (params object[] content) XContainer AddFirst (params object[] content) XContainer Remove RemoveNodes() XContainer RemoveAttributes() XElement RemoveAll() XElement Update ReplaceNodes (params object[] content) XContainer ReplaceAttributes (params object[] content) XElement ReplaceAll (params object[] content XElement SetElementValue (XName name, object value) XElement SetAttributeValue (XName name, object value) XElement 510 | Chapter 10: LINQ to XML
- 529. The most convenient methods in this group are the last two: SetElementValue and SetAttributeValue. They serve as shortcuts for instantiating an XElement or XAttribute and then Adding it to a parent, replacing any existing element or attribute of that name: XElement settings = new XElement ("settings"); settings.SetElementValue ("timeout", 30); // Adds child node settings.SetElementValue ("timeout", 60); // Update it to 60 Add appends a child node to an element or document. AddFirst does the same thing but inserts at the beginning of the collection rather than the end. You can remove all child nodes or attributes in one hit with RemoveNodes or RemoveAttributes. RemoveAll is equivalent to calling both methods. The ReplaceXXX methods are equivalent to Removing and then Adding. They take a snapshot of the input, so e.ReplaceNodes(e.Nodes()) works as expected. Updating Through the Parent Members Works on AddBeforeSelf (params object[] content) XNode AddAfterSelf (params object[] content) XNode Remove() XNode, XAttribute ReplaceWith (params object[] content) XNode The methods AddBeforeSelf, AddAfterSelf, Remove, and ReplaceWith don’t oper‐ ate on the node’s children. Instead, they operate on the collection in which the node itself is in. This requires that the node have a parent element—otherwise, an exception is thrown. AddBeforeSelf and AddAfterSelf are useful for inserting a node into an arbitrary position: XElement items = new XElement ("items", new XElement ("one"), new XElement ("three") ); items.FirstNode.AddAfterSelf (new XElement ("two")); Here’s the result: <items><one /><two /><three /></items> Inserting into an arbitrary position within a long sequence of elements is efficient because nodes are stored internally in a linked list. The Remove method removes the current node from its parent. ReplaceWith does the same and then inserts some other content at the same position: XElement items = XElement.Parse ("<items><one/><two/><three/></items>"); items.FirstNode.ReplaceWith (new XComment ("One was here")); Updating an X-DOM | 511 LINQ to XML
- 530. Here’s the result: <items><!--one was here--><two /><three /></items> Removing a sequence of nodes or attributes Thanks to extension methods in System.Xml.Linq, you can also call Remove on a sequence of nodes or attributes. Consider this X-DOM: XElement contacts = XElement.Parse ( @"<contacts> <customer name='Mary'/> <customer name='Chris' archived='true'/> <supplier name='Susan'> <phone archived='true'>012345678<!--confidential--></phone> </supplier> </contacts>"); The following removes all customers: contacts.Elements ("customer").Remove(); The following removes all archived contacts (so Chris disappears): contacts.Elements().Where (e => (bool?) e.Attribute ("archived") == true) .Remove(); If we replaced Elements() with Descendants(), all archived elements throughout the DOM would disappear, yielding this result: <contacts> <customer name="Mary" /> <supplier name="Susan" /> </contacts> The next example removes all contacts that feature the comment “confidential” anywhere in their tree: contacts.Elements().Where (e => e.DescendantNodes() .OfType<XComment>() .Any (c => c.Value == "confidential") ).Remove(); This is the result: <contacts> <customer name="Mary" /> <customer name="Chris" archived="true" /> </contacts> Contrast this with the following simpler query, which strips all comment nodes from the tree: contacts.DescendantNodes().OfType<XComment>().Remove(); 512 | Chapter 10: LINQ to XML
- 531. Internally, the Remove method first reads all matching ele‐ ments into a temporary list and then enumerates over the temporary list to perform the deletions. This prevents errors that could otherwise result from deleting and querying at the same time. Working with Values XElement and XAttribute both have a Value property of type string. If an element has a single XText child node, XElement’s Value property acts as a convenient shortcut to the content of that node. With XAttribute, the Value property is simply the attribute’s value. Despite the storage differences, the X-DOM provides a consistent set of operations for working with element and attribute values. Setting Values There are two ways to assign a value: call SetValue or assign the Value property. SetValue is more flexible because it accepts not just strings but other simple data types, too: var e = new XElement ("date", DateTime.Now); e.SetValue (DateTime.Now.AddDays(1)); Console.Write (e.Value); // 2019-10-02T16:39:10.734375+09:00 We could have instead just set the element’s Value property, but this would mean manually converting the DateTime to a string. This is more complicated than calling ToString—it requires the use of XmlConvert for an XML-compliant result. When you pass a value into XElement or XAttribute’s constructor, the same auto‐ matic conversion takes place for nonstring types. This ensures that DateTimes are correctly formatted; true is written in lowercase, and double.NegativeInfinity is written as “-INF”. Getting Values To go the other way around and parse a Value back to a base type, you simply cast the XElement or XAttribute to the desired type. It sounds like it shouldn’t work—but it does! For instance: XElement e = new XElement ("now", DateTime.Now); DateTime dt = (DateTime) e; XAttribute a = new XAttribute ("resolution", 1.234); double res = (double) a; An element or attribute doesn’t store DateTimes or numbers natively—they’re always stored as text and then parsed as needed. It also doesn’t “remember” the original type, so you must cast it correctly to prevent a runtime error. To make Working with Values | 513 LINQ to XML
- 532. your code robust, you can put the cast in a try/catch block, catching a Format Exception. Explicit casts on XElement and XAttribute can parse to the following types: • All standard numeric types • • string, bool, DateTime, DateTimeOffset, TimeSpan, and Guid • • Nullable<> versions of the aforementioned value types • Casting to a nullable type is useful in conjunction with the Element and Attribute methods, because if the requested name doesn’t exist, the cast still works. For instance, if x has no timeout element, the first line generates a runtime error and the second line does not: int timeout = (int) x.Element ("timeout"); // Error int? timeout = (int?) x.Element ("timeout"); // OK; timeout is null. You can factor away the nullable type in the final result with the ?? operator. The following evaluates to 1.0 if the resolution attribute doesn’t exist: double resolution = (double?) x.Attribute ("resolution") ?? 1.0; Casting to a nullable type won’t get you out of trouble, though, if the element or attribute exists and has an empty (or improperly formatted) value. For this, you must catch a FormatException. You can also use casts in LINQ queries. The following returns “John”: var data = XElement.Parse ( @"<data> <customer id='1' name='Mary' credit='100' /> <customer id='2' name='John' credit='150' /> <customer id='3' name='Anne' /> </data>"); IEnumerable<string> query = from cust in data.Elements() where (int?) cust.Attribute ("credit") > 100 select cust.Attribute ("name").Value; Casting to a nullable int prevents a NullReferenceException in the case of Anne, who has no credit attribute. Another solution would be to add a predicate to the where clause: where cust.Attributes ("credit").Any() && (int) cust.Attribute... The same principles apply in querying element values. Values and Mixed Content Nodes Given the value of Value, you might wonder when you’d ever need to deal directly with XText nodes. The answer is when you have mixed content. For example: <summary>An XAttribute is <bold>not</bold> an XNode</summary> 514 | Chapter 10: LINQ to XML
- 533. A simple Value property is not enough to capture summary’s content. The summary element contains three children: an XText node, followed by an XElement, followed by another XText node. Here’s how to construct it: XElement summary = new XElement ("summary", new XText ("An XAttribute is "), new XElement ("bold", "not"), new XText (" an XNode") ); Interestingly, we can still query summary’s Value—without getting an exception. Instead, we get a concatenation of each child’s value: An XAttribute is not an XNode It’s also legal to reassign summary’s Value, at the cost of replacing all previous children with a single new XText node. Automatic XText Concatenation When you add simple content to an XElement, the X-DOM appends to the existing XText child rather than creating a new one. In the following examples, e1 and e2 end up with just one child XText element whose value is HelloWorld: var e1 = new XElement ("test", "Hello"); e1.Add ("World"); var e2 = new XElement ("test", "Hello", "World"); If you specifically create XText nodes, however, you end up with multiple children: var e = new XElement ("test", new XText ("Hello"), new XText ("World")); Console.WriteLine (e.Value); // HelloWorld Console.WriteLine (e.Nodes().Count()); // 2 XElement doesn’t concatenate the two XText nodes, so the nodes’ object identities are preserved. Documents and Declarations XDocument As we said previously, an XDocument wraps a root XElement and allows you to add an XDeclaration, processing instructions, a document type, and root-level comments. An XDocument is optional and can be ignored or omitted: unlike with the W3C DOM, it does not serve as glue to keep everything together. An XDocument provides the same functional constructors as XElement. And because it’s based on XContainer, it also supports the AddXXX, RemoveXXX, and ReplaceXXX methods. Unlike XElement, however, an XDocument can accept only limited content: • A single XElement object (the “root”) • • A single XDeclaration object • Documents and Declarations | 515 LINQ to XML
- 534. • A single XDocumentType object (to reference a document type definition • [DTD]) • Any number of XProcessingInstruction objects • • Any number of XComment objects • Of these, only the root XElement is mandatory in order to have a valid XDocument. The XDeclaration is optional—if omitted, default settings are applied during serialization. The simplest valid XDocument has just a root element: var doc = new XDocument ( new XElement ("test", "data") ); Notice that we didn’t include an XDeclaration object. The file generated by calling doc.Save would still contain an XML declaration, however, because one is gener‐ ated by default. The next example produces a simple but correct XHTML file, illustrating all the constructs that an XDocument can accept: var styleInstruction = new XProcessingInstruction ( "xml-stylesheet", "href='styles.css' type='text/css'"); var docType = new XDocumentType ("html", "-//W3C//DTD XHTML 1.0 Strict//EN", "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd", null); XNamespace ns = "http://www.w3.org/1999/xhtml"; var root = new XElement (ns + "html", new XElement (ns + "head", new XElement (ns + "title", "An XHTML page")), new XElement (ns + "body", new XElement (ns + "p", "This is the content")) ); var doc = new XDocument ( new XDeclaration ("1.0", "utf-8", "no"), new XComment ("Reference a stylesheet"), styleInstruction, docType, root); doc.Save ("test.html"); The resultant test.html reads as follows: <?xml version="1.0" encoding="utf-8" standalone="no"?> <!--Reference a stylesheet--> <?xml-stylesheet href='styles.css' type='text/css'?> 516 | Chapter 10: LINQ to XML
- 535. <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"> <html xmlns="http://www.w3.org/1999/xhtml"> <head> <title>An XHTML page</title> </head> <body> <p>This is the content</p> </body> </html> XDocument has a Root property that serves as a shortcut for accessing a document’s single XElement. The reverse link is provided by XObject’s Document property, which works for all objects in the tree: Console.WriteLine (doc.Root.Name.LocalName); // html XElement bodyNode = doc.Root.Element (ns + "body"); Console.WriteLine (bodyNode.Document == doc); // True Recall that a document’s children have no Parent: Console.WriteLine (doc.Root.Parent == null); // True foreach (XNode node in doc.Nodes()) Console.Write (node.Parent == null); // TrueTrueTrueTrue An XDeclaration is not an XNode and does not appear in the document’s Nodes collection—unlike comments, processing instructions, and the root element. Instead, it’s assigned to a dedicated property called Declaration. This is why “True” is repeated four and not five times in the last example. XML Declarations A standard XML file starts with a declaration such as the following: <?xml version="1.0" encoding="utf-8" standalone="yes"?> An XML declaration ensures that the file will be correctly parsed and understood by a reader. XElement and XDocument follow these rules in emitting XML declarations: • Calling Save with a filename always writes a declaration. • • Calling Save with an XmlWriter writes a declaration unless the XmlWriter is • instructed otherwise. • The ToString method never emits an XML declaration. • You can instruct an XmlWriter not to produce a declaration by setting the OmitXmlDeclaration and ConformanceLevel properties of an XmlWriterSettings object when constructing the XmlWriter. We describe this in Chapter 11. Documents and Declarations | 517 LINQ to XML
- 536. The presence or absence of an XDeclaration object has no effect on whether an XML declaration is written. The purpose of an XDeclaration is instead to hint the XML serialization, in two ways: • What text encoding to use • • What to put in the XML declaration’s encoding and standalone attributes • (should a declaration be written) XDeclaration’s constructor accepts three arguments, which correspond to the attributes version, encoding, and standalone. In the following example, test.xml is encoded in UTF-16: var doc = new XDocument ( new XDeclaration ("1.0", "utf-16", "yes"), new XElement ("test", "data") ); doc.Save ("test.xml"); Whatever you specify for the XML version is ignored by the XML writer: it always writes "1.0". The encoding must use an IETF code such as "utf-16", just as it would appear in the XML declaration. Writing a declaration to a string Suppose that we want to serialize an XDocument to a string, including the XML declaration. Because ToString doesn’t write a declaration, we’d need to use an XmlWriter instead: var doc = new XDocument ( new XDeclaration ("1.0", "utf-8", "yes"), new XElement ("test", "data") ); var output = new StringBuilder(); var settings = new XmlWriterSettings { Indent = true }; using (XmlWriter xw = XmlWriter.Create (output, settings)) doc.Save (xw); Console.WriteLine (output.ToString()); This is the result: <?xml version="1.0" encoding="utf-16" standalone="yes"?> <test>data</test> Notice that we have UTF-16 in the output, even though we explicitly requested UTF-8 in an XDeclaration! This might look like a bug, but in fact, XmlWriter is being remarkably smart. Because we’re writing to a string and not a file or stream, it’s impossible to apply any encoding other than UTF-16—the format in which strings are internally stored. Hence, XmlWriter writes "utf-16" so as not to lie. 518 | Chapter 10: LINQ to XML
- 537. This also explains why the ToString method doesn’t emit an XML declaration. Imagine that instead of calling Save, you did the following to write an XDocument to a file: File.WriteAllText ("data.xml", doc.ToString()); As it stands, data.xml would lack an XML declaration, making it incomplete but still parsable (you can infer the text encoding). But if ToString() emitted an XML declaration, data.xml would actually contain an incorrect declaration (encod ing="utf-16"), which might prevent it from being read at all because WriteAllText encodes using UTF-8. Names and Namespaces Just as .NET types can have namespaces, so too can XML elements and attributes. XML namespaces achieve two things. First, rather like namespaces in C#, they help prevent naming collisions. This can become an issue when you merge data from one XML file into another. Second, namespaces assign absolute meaning to a name. The name “nil,” for instance, could mean anything. Within the http://www.w3.org/ 2001/xmlschema-instance namespace, however, “nil” means something equivalent to null in C# and comes with specific rules on how it can be applied. Because XML namespaces are a significant source of confusion, we first cover namespaces in general, and then move on to how they’re used in LINQ to XML. Namespaces in XML Suppose that we want to define a customer element in the namespace OReilly.Nut shell.CSharp. There are two ways to proceed. The first is to use the xmlns attribute: <customer xmlns="OReilly.Nutshell.CSharp"/> xmlns is a special reserved attribute. When used in this manner, it performs two functions: • It specifies a namespace for the element in question. • • It specifies a default namespace for all descendant elements. • This means that in the following example, address and postcode implicitly reside in the OReilly.Nutshell.CSharp namespace: <customer xmlns="OReilly.Nutshell.CSharp"> <address> <postcode>02138</postcode> </address> </customer> Names and Namespaces | 519 LINQ to XML
- 538. If we want address and postcode to have no namespace, we’d need to do this: <customer xmlns="OReilly.Nutshell.CSharp"> <address xmlns=""> <postcode>02138</postcode> <!-- postcode now inherits empty ns --> </address> </customer> Prefixes The other way to specify a namespace is with a prefix. A prefix is an alias that you assign to a namespace to save typing. There are two steps in using a prefix—defining the prefix and using it. You can do both together: <nut:customer xmlns:nut="OReilly.Nutshell.CSharp"/> Two distinct things are happening here. On the right, xmlns:nut="..." defines a prefix called nut and makes it available to this element and all its descendants. On the left, nut:customer assigns the newly allocated prefix to the customer element. A prefixed element does not define a default namespace for descendants. In the following XML, firstname has an empty namespace: <nut:customer xmlns:nut="OReilly.Nutshell.CSharp"> <firstname>Joe</firstname> </customer> To give firstname the OReilly.Nutshell.CSharp prefix, you must do this: <nut:customer xmlns:nut="OReilly.Nutshell.CSharp"> <nut:firstname>Joe</firstname> </customer> You can also define a prefix—or prefixes—for the convenience of your descendants, without assigning any of them to the parent element itself. The following defines two prefixes, i and z, while leaving the customer element itself with an empty namespace: <customer xmlns:i="http://www.w3.org/2001/XMLSchema-instance" xmlns:z="http://schemas.microsoft.com/2003/10/Serialization/"> ... </customer> If this were the root node, the whole document would have i and z at its fingertips. Prefixes are convenient when elements need to draw from multiple namespaces. Notice that both namespaces in this example are URIs. Using URIs (that you own) is standard practice: it ensures namespace uniqueness. So, in real life, our customer element would more likely be: <customer xmlns="http://oreilly.com/schemas/nutshell/csharp"/> or: <nut:customer xmlns:nut="http://oreilly.com/schemas/nutshell/csharp"/> 520 | Chapter 10: LINQ to XML
- 539. Attributes You can assign namespaces to attributes, too. The main difference is that an attribute always requires a prefix. For instance: <customer xmlns:nut="OReilly.Nutshell.CSharp" nut:id="123" /> Another difference is that an unqualified attribute always has an empty namespace: it never inherits a default namespace from a parent element. Attributes tend not to need namespaces because their meaning is usually local to the element. An exception is with general-purpose or metadata attributes such as the nil attribute defined by W3C: <customer xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"> <firstname>Joe</firstname> <lastname xsi:nil="true"/> </customer> This indicates unambiguously that lastname is nil (null in C#) and not an empty string. Because we’ve used the standard namespace, a general-purpose parsing util‐ ity could know with certainty our intention. Specifying Namespaces in the X-DOM So far in this chapter, we’ve used just simple strings for XElement and XAttribute names. A simple string corresponds to an XML name with an empty namespace— rather like a .NET type defined in the global namespace. There are a couple of ways to specify an XML namespace. The first is to enclose it in braces, before the local name: var e = new XElement ("{http://domain.com/xmlspace}customer", "Bloggs"); Console.WriteLine (e.ToString()); This yields the resulting XML: <customer xmlns="http://domain.com/xmlspace">Bloggs</customer> The second (and more performant) approach is to use the XNamespace and XName types. Here are their definitions: public sealed class XNamespace { public string NamespaceName { get; } } public sealed class XName // A local name with optional namespace { public string LocalName { get; } public XNamespace Namespace { get; } // Optional } Both types define implicit casts from string, so the following is legal: Names and Namespaces | 521 LINQ to XML
- 540. XNamespace ns = "http://domain.com/xmlspace"; XName localName = "customer"; XName fullName = "{http://domain.com/xmlspace}customer"; XNamespace also overloads the + operator, allowing you to combine a namespace and name into an XName without using braces: XNamespace ns = "http://domain.com/xmlspace"; XName fullName = ns + "customer"; Console.WriteLine (fullName); // {http://domain.com/xmlspace}customer All constructors and methods in the X-DOM that accept an element or attribute name actually accept an XName object rather than a string. The reason you can substitute a string—as in all our examples to date—is because of the implicit cast. Specifying a namespace is the same whether for an element or an attribute: XNamespace ns = "http://domain.com/xmlspace"; var data = new XElement (ns + "data", new XAttribute (ns + "id", 123) ); The X-DOM and Default Namespaces The X-DOM ignores the concept of default namespaces until it comes time to actually output XML. This means that when you construct a child XElement, you must give it a namespace explicitly if needed; it will not inherit from the parent: XNamespace ns = "http://domain.com/xmlspace"; var data = new XElement (ns + "data", new XElement (ns + "customer", "Bloggs"), new XElement (ns + "purchase", "Bicycle") ); The X-DOM does, however, apply default namespaces when reading and outputting XML: Console.WriteLine (data.ToString()); OUTPUT: <data xmlns="http://domain.com/xmlspace"> <customer>Bloggs</customer> <purchase>Bicycle</purchase> </data> Console.WriteLine (data.Element (ns + "customer").ToString()); OUTPUT: <customer xmlns="http://domain.com/xmlspace">Bloggs</customer> If you construct XElement children without specifying namespaces, in other words: XNamespace ns = "http://domain.com/xmlspace"; var data = new XElement (ns + "data", new XElement ("customer", "Bloggs"), new XElement ("purchase", "Bicycle") 522 | Chapter 10: LINQ to XML
- 541. ); Console.WriteLine (data.ToString()); you get this result instead: <data xmlns="http://domain.com/xmlspace"> <customer xmlns="">Bloggs</customer> <purchase xmlns="">Bicycle</purchase> </data> Another trap is failing to include a namespace when navigating an X-DOM: XNamespace ns = "http://domain.com/xmlspace"; var data = new XElement (ns + "data", new XElement (ns + "customer", "Bloggs"), new XElement (ns + "purchase", "Bicycle") ); XElement x = data.Element (ns + "customer"); // ok XElement y = data.Element ("customer"); // null If you build an X-DOM tree without specifying namespaces, you can subsequently assign every element to a single namespace, as follows: foreach (XElement e in data.DescendantsAndSelf()) if (e.Name.Namespace == "") e.Name = ns + e.Name.LocalName; Prefixes The X-DOM treats prefixes just as it treats namespaces: purely as a serialization function. This means that you can choose to completely ignore the issue of pre‐ fixes—and get by! The only reason you might want to do otherwise is for efficiency when outputting to an XML file. For example, consider this: XNamespace ns1 = "http://domain.com/space1"; XNamespace ns2 = "http://domain.com/space2"; var mix = new XElement (ns1 + "data", new XElement (ns2 + "element", "value"), new XElement (ns2 + "element", "value"), new XElement (ns2 + "element", "value") ); By default, XElement will serialize this as follows: <data xmlns="http://domain.com/space1"> <element xmlns="http://domain.com/space2">value</element> <element xmlns="http://domain.com/space2">value</element> <element xmlns="http://domain.com/space2">value</element> </data> As you can see, there’s a bit of unnecessary duplication. The solution is not to change the way you construct the X-DOM, but instead to hint the serializer prior to writing the XML. Do this by adding attributes defining prefixes that you want to see applied. This is typically done on the root element: Names and Namespaces | 523 LINQ to XML
- 542. mix.SetAttributeValue (XNamespace.Xmlns + "ns1", ns1); mix.SetAttributeValue (XNamespace.Xmlns + "ns2", ns2); This assigns the prefix “ns1” to our XNamespace variable ns1, and “ns2” to ns2. The X-DOM automatically picks up these attributes when serializing and uses them to condense the resulting XML. Here’s the result now of calling ToString on mix: <ns1:data xmlns:ns1="http://domain.com/space1" xmlns:ns2="http://domain.com/space2"> <ns2:element>value</ns2:element> <ns2:element>value</ns2:element> <ns2:element>value</ns2:element> </ns1:data> Prefixes don’t change the way you construct, query, or update the X-DOM—for these activities, you ignore the presence of prefixes and continue to use full names. Prefixes come into play only when converting to and from XML files or streams. Prefixes are also honored in serializing attributes. In the following example, we record a customer’s date of birth and credit as "nil" using the W3C-standard attribute. The highlighted line ensures that the prefix is serialized without unneces‐ sary namespace repetition: XNamespace xsi = "http://www.w3.org/2001/XMLSchema-instance"; var nil = new XAttribute (xsi + "nil", true); var cust = new XElement ("customers", new XAttribute (XNamespace.Xmlns + "xsi", xsi), new XElement ("customer", new XElement ("lastname", "Bloggs"), new XElement ("dob", nil), new XElement ("credit", nil) ) ); This is its XML: <customers xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"> <customer> <lastname>Bloggs</lastname> <dob xsi:nil="true" /> <credit xsi:nil="true" /> </customer> </customers> For brevity, we predeclared the nil XAttribute so that we could use it twice in building the DOM. You’re allowed to reference the same attribute twice because it’s automatically duplicated as required. Annotations You can attach custom data to any XObject with an annotation. Annotations are intended for your own private use and are treated as black boxes by X-DOM. If you’ve ever used the Tag property on a Windows Forms or Windows Presentation 524 | Chapter 10: LINQ to XML
- 543. Foundation (WPF) control, you’ll be familiar with the concept—the difference is that you have multiple annotations, and your annotations can be privately scoped. You can create an annotation that other types cannot even see—let alone overwrite. The following methods on XObject add and remove annotations: public void AddAnnotation (object annotation) public void RemoveAnnotations<T>() where T : class The following methods retrieve annotations: public T Annotation<T>() where T : class public IEnumerable<T> Annotations<T>() where T : class Each annotation is keyed by its type, which must be a reference type. The following adds and then retrieves a string annotation: XElement e = new XElement ("test"); e.AddAnnotation ("Hello"); Console.WriteLine (e.Annotation<string>()); // Hello You can add multiple annotations of the same type and then use the Annotations method to retrieve a sequence of matches. A public type such as string doesn’t make a great key, however, because code in other types can interfere with your annotations. A better approach is to use an internal or (nested) private class: class X { class CustomData { internal string Message; } // Private nested type static void Test() { XElement e = new XElement ("test"); e.AddAnnotation (new CustomData { Message = "Hello" } ); Console.Write (e.Annotations<CustomData>().First().Message); // Hello } } To remove annotations, you must also have access to the key’s type: e.RemoveAnnotations<CustomData>(); Projecting into an X-DOM So far, we’ve shown how to use LINQ to get data out of an X-DOM. You can also use LINQ queries to project into an X-DOM. The source can be anything over which LINQ can query, such as the following: • EF Core entity classes • • A local collection • • Another X-DOM • Projecting into an X-DOM | 525 LINQ to XML
- 544. Regardless of the source, the strategy is the same in using LINQ to emit an X-DOM: first write a functional construction expression that produces the desired X-DOM shape and then build a LINQ query around the expression. For instance, suppose that we want to retrieve customers from a database into the following XML: <customers> <customer id="1"> <name>Sue</name> <buys>3</buys> </customer> ... </customers> We start by writing a functional construction expression for the X-DOM using simple literals: var customers = new XElement ("customers", new XElement ("customer", new XAttribute ("id", 1), new XElement ("name", "Sue"), new XElement ("buys", 3) ) ); We then turn this into a projection and build a LINQ query around it: var customers = new XElement ("customers", // We must call AsEnumerable() due to a bug in EF Core. from c in dbContext.Customers.AsEnumerable() select new XElement ("customer", new XAttribute ("id", c.ID), new XElement ("name", c.Name), new XElement ("buys", c.Purchases.Count) ) ); The call to AsEnumerable is required due to a bug in EF Core (a fix is scheduled for a later release). After the bug is fixed, removing the call to AsEnumerable will improve efficiency by preventing a round-trip with each call to c.Purchases.Count. Here’s the result: <customers> <customer id="1"> <name>Tom</name> <buys>3</buys> </customer> <customer id="2"> <name>Harry</name> <buys>2</buys> </customer> 526 | Chapter 10: LINQ to XML
- 545. ... </customers> We can see how this works more clearly by constructing the same query in two steps. First: IEnumerable<XElement> sqlQuery = from c in dbContext.Customers.AsEnumerable() select new XElement ("customer", new XAttribute ("id", c.ID), new XElement ("name", c.Name), new XElement ("buys", c.Purchases.Count) ); This inner portion is a normal LINQ query that projects into XElements. Here’s the second step: var customers = new XElement ("customers", sqlQuery); This constructs the root XElement. The only thing unusual is that the content, sqlQuery, is not a single XElement but an IQueryable<XElement>, which imple‐ ments IEnumerable<XElement>. Remember that in the processing of XML content, collections are automatically enumerated. So, each XElement is added as a child node. Eliminating Empty Elements Suppose in the preceding example that we also wanted to include details of the customer’s most recent high-value purchase. We could do this as follows: var customers = new XElement ("customers", // The AsEnumerable call can be removed when the EF Core bug is fixed. from c in dbContext.Customers.AsEnumerable() let lastBigBuy = (from p in c.Purchases where p.Price > 1000 orderby p.Date descending select p).FirstOrDefault() select new XElement ("customer", new XAttribute ("id", c.ID), new XElement ("name", c.Name), new XElement ("buys", c.Purchases.Count), new XElement ("lastBigBuy", new XElement ("description", lastBigBuy?.Description), new XElement ("price", lastBigBuy?.Price ?? 0m) ) ) ); This emits empty elements, though, for customers with no high-value purchases. (If it were a local query rather than a database query, it would throw a NullRefer enceException.) In such cases, it would be better to omit the lastBigBuy node entirely. We can achieve this by wrapping the constructor for the lastBigBuy element in a conditional operator: Projecting into an X-DOM | 527 LINQ to XML
- 546. select new XElement ("customer", new XAttribute ("id", c.ID), new XElement ("name", c.Name), new XElement ("buys", c.Purchases.Count), lastBigBuy == null ? null : new XElement ("lastBigBuy", new XElement ("description", lastBigBuy.Description), new XElement ("price", lastBigBuy.Price) For customers with no lastBigBuy, a null is emitted instead of an empty XElement. This is what we want, because null content is simply ignored. Streaming a Projection If you’re projecting into an X-DOM only to Save it (or call ToString on it), you can improve memory efficiency through an XStreamingElement. An XStreaming Element is a cut-down version of XElement that applies deferred loading semantics to its child content. To use it, you simply replace the outer XElements with XStrea mingElements: var customers = new XStreamingElement ("customers", from c in dbContext.Customers select new XStreamingElement ("customer", new XAttribute ("id", c.ID), new XElement ("name", c.Name), new XElement ("buys", c.Purchases.Count) ) ); customers.Save ("data.xml"); The queries passed into an XStreamingElement’s constructor are not enumerated until you call Save, ToString, or WriteTo on the element; this prevents loading the whole X-DOM into memory at once. The flipside is that the queries are reevalu‐ ated, should you re-Save. Also, you cannot traverse an XStreamingElement’s child content—it does not expose methods such as Elements or Attributes. XStreamingElement is not based on XObject—or any other class—because it has such a limited set of members. The only members it has, besides Save, ToString, and WriteTo, are the following: • An Add method, which accepts content like the constructor • • A Name property • XStreamingElement does not allow you to read content in a streamed fashion—for this, you must use an XmlReader in conjunction with the X-DOM. We describe how to do this in “Patterns for Using XmlReader/XmlWriter” on page 539. 528 | Chapter 10: LINQ to XML
- 547. 11 Other XML and JSON Technologies In Chapter 10, we covered the LINQ-to-XML API—and XML in general. In this chapter, we explore the low-level XmlReader/XmlWriter classes and the types for working with JavaScript Object Notation (JSON), which has become a popular alternative to XML. In the online supplement, we describe the tools for working with XML schema and stylesheets. XmlReader XmlReader is a high-performance class for reading an XML stream in a low-level, forward-only manner. Consider the following XML file, customer.xml: <?xml version="1.0" encoding="utf-8" standalone="yes"?> <customer id="123" status="archived"> <firstname>Jim</firstname> <lastname>Bo</lastname> </customer> To instantiate an XmlReader, you call the static XmlReader.Create method, passing in a Stream, a TextReader, or a URI string: using XmlReader reader = XmlReader.Create ("customer.xml"); ... Because XmlReader lets you read from potentially slow sources (Streams and URIs), it offers asynchronous versions of most of its methods so that you can easily write nonblocking code. We cover asynchrony in detail in Chapter 14. 529
- 548. To construct an XmlReader that reads from a string: using XmlReader reader = XmlReader.Create ( new System.IO.StringReader (myString)); You can also pass in an XmlReaderSettings object to control parsing and validation options. The following three properties on XmlReaderSettings are particularly useful for skipping over superfluous content: bool IgnoreComments // Skip over comment nodes? bool IgnoreProcessingInstructions // Skip over processing instructions? bool IgnoreWhitespace // Skip over whitespace? In the following example, we instruct the reader not to emit whitespace nodes, which are a distraction in typical scenarios: XmlReaderSettings settings = new XmlReaderSettings(); settings.IgnoreWhitespace = true; using XmlReader reader = XmlReader.Create ("customer.xml", settings); ... Another useful property on XmlReaderSettings is ConformanceLevel. Its default value of Document instructs the reader to assume a valid XML document with a single root node. This is a problem if you want to read just an inner portion of XML, containing multiple nodes: <firstname>Jim</firstname> <lastname>Bo</lastname> To read this without throwing an exception, you must set ConformanceLevel to Fragment. XmlReaderSettings also has a property called CloseInput, which indicates whether to close the underlying stream when the reader is closed (there’s an analogous prop‐ erty on XmlWriterSettings called CloseOutput). The default value for CloseInput and CloseOutput is false. Reading Nodes The units of an XML stream are XML nodes. The reader traverses the stream in textual (depth-first) order. The Depth property of the reader returns the current depth of the cursor. The most primitive way to read from an XmlReader is to call Read. It advances to the next node in the XML stream, rather like MoveNext in IEnumerator. The first call to Read positions the cursor at the first node. When Read returns false, it means the cursor has advanced past the last node, at which point the XmlReader should be closed and abandoned. Two string properties on XmlReader provide access to a node’s content: Name and Value. Depending on the node type, either Name or Value (or both) are populated. 530 | Chapter 11: Other XML and JSON Technologies
- 549. In this example, we read every node in the XML stream, outputting each node type as we go: XmlReaderSettings settings = new XmlReaderSettings(); settings.IgnoreWhitespace = true; using XmlReader reader = XmlReader.Create ("customer.xml", settings); while (reader.Read()) { Console.Write (new string (' ', reader.Depth * 2)); // Write indentation Console.Write (reader.NodeType.ToString()); if (reader.NodeType == XmlNodeType.Element || reader.NodeType == XmlNodeType.EndElement) { Console.Write (" Name=" + reader.Name); } else if (reader.NodeType == XmlNodeType.Text) { Console.Write (" Value=" + reader.Value); } Console.WriteLine (); } The output is as follows: XmlDeclaration Element Name=customer Element Name=firstname Text Value=Jim EndElement Name=firstname Element Name=lastname Text Value=Bo EndElement Name=lastname EndElement Name=customer Attributes are not included in Read-based traversal (see “Reading Attributes” on page 535). NodeType is of type XmlNodeType, which is an enum with these members: None XmlDeclaration Element EndElement Text Attribute Comment Entity EndEntity EntityReference ProcessingInstruction CDATA Document DocumentType DocumentFragment Notation Whitespace SignificantWhitespace Reading Elements Often, you already know the structure of the XML document that you’re reading. To help with this, XmlReader provides a range of methods that read while presuming a XmlReader | 531 Other XML and JSON Technologies
- 550. particular structure. This simplifies your code as well as performing some validation at the same time. XmlReader throws an XmlException if any validation fails. XmlException has LineNumber and LinePosition properties indicating where the error occurred—logging this information is essential if the XML file is large! ReadStartElement verifies that the current NodeType is Element and then calls Read. If you specify a name, it verifies that it matches that of the current element. ReadEndElement verifies that the current NodeType is EndElement and then calls Read. For instance, we could read this: <firstname>Jim</firstname> as follows: reader.ReadStartElement ("firstname"); Console.WriteLine (reader.Value); reader.Read(); reader.ReadEndElement(); The ReadElementContentAsString method does all of this in one hit. It reads a start element, a text node, and an end element, returning the content as a string: string firstName = reader.ReadElementContentAsString ("firstname", ""); The second argument refers to the namespace, which is blank in this example. There are also typed versions of this method, such as ReadElementContentAsInt, which parse the result. Returning to our original XML document: <?xml version="1.0" encoding="utf-8" standalone="yes"?> <customer id="123" status="archived"> <firstname>Jim</firstname> <lastname>Bo</lastname> <creditlimit>500.00</creditlimit> <!-- OK, we sneaked this in! --> </customer> We could read it in as follows: XmlReaderSettings settings = new XmlReaderSettings(); settings.IgnoreWhitespace = true; using XmlReader r = XmlReader.Create ("customer.xml", settings); r.MoveToContent(); // Skip over the XML declaration r.ReadStartElement ("customer"); string firstName = r.ReadElementContentAsString ("firstname", ""); string lastName = r.ReadElementContentAsString ("lastname", ""); decimal creditLimit = r.ReadElementContentAsDecimal ("creditlimit", ""); r.MoveToContent(); // Skip over that pesky comment r.ReadEndElement(); // Read the closing customer tag 532 | Chapter 11: Other XML and JSON Technologies
- 551. The MoveToContent method is really useful. It skips over all the fluff: XML declarations, whitespace, comments, and processing instructions. You can also instruct the reader to do most of this automatically through the properties on XmlReaderSettings. Optional elements In the previous example, suppose that <lastname> was optional. The solution to this is straightforward: r.ReadStartElement ("customer"); string firstName = r. ReadElementContentAsString ("firstname", ""); string lastName = r.Name == "lastname" ? r.ReadElementContentAsString() : null; decimal creditLimit = r.ReadElementContentAsDecimal ("creditlimit", ""); Random element order The examples in this section rely on elements appearing in the XML file in a set order. If you need to cope with elements appearing in any order, the easiest solution is to read that section of the XML into an X-DOM. We describe how to do this later in “Patterns for Using XmlReader/XmlWriter” on page 539. Empty elements The way that XmlReader handles empty elements presents a horrible trap. Consider the following element: <customerList></customerList> In XML, this is equivalent to the following: <customerList/> And yet, XmlReader treats the two differently. In the first case, the following code works as expected: reader.ReadStartElement ("customerList"); reader.ReadEndElement(); In the second case, ReadEndElement throws an exception because there is no sepa‐ rate “end element” as far as XmlReader is concerned. The workaround is to check for an empty element: bool isEmpty = reader.IsEmptyElement; reader.ReadStartElement ("customerList"); if (!isEmpty) reader.ReadEndElement(); In reality, this is a nuisance only when the element in question might contain child elements (such as a customer list). With elements that wrap simple text (such as firstname), you can avoid the entire issue by calling a method such as ReadElementContentAsString. The ReadElementXXX methods handle both kinds of empty elements correctly. XmlReader | 533 Other XML and JSON Technologies
- 552. Other ReadXXX methods Table 11-1 summarizes all ReadXXX methods in XmlReader. Most of these are designed to work with elements. The sample XML fragment shown in bold is the section read by the method described. Table 11-1. Read methods Members Works on NodeType Sample XML fragment Input parameters Data returned ReadContentAsXXX Text <a>x</a> x ReadElement ContentAsXXX Element <a>x</a> x ReadInnerXml Element <a>x</a> x ReadOuterXml Element <a>x</a> <a>x</a> ReadStartElement Element <a>x</a> ReadEndElement Element <a>x</a> ReadSubtree Element <a>x</a> <a>x</a> ReadToDescendant Element <a>x<b></b></a> "b" ReadToFollowing Element <a>x<b></b></a> "b" ReadToNextSibling Element <a>x</a><b></b> "b" ReadAttribute Value Attribute See “Reading Attributes” on page 535 The ReadContentAsXXX methods parse a text node into type XXX. Internally, the XmlConvert class performs the string-to-type conversion. The text node can be within an element or an attribute. The ReadElementContentAsXXX methods are wrappers around corresponding Read ContentAsXXX methods. They apply to the element node rather than the text node enclosed by the element. ReadInnerXml is typically applied to an element, and it reads and returns an element and all its descendants. When applied to an attribute, it returns the value of the attribute. ReadOuterXml is the same except that it includes rather than excludes the element at the cursor position. ReadSubtree returns a proxy reader that provides a view over just the current element (and its descendants). The proxy reader must be closed before the original reader can be safely read again. When the proxy reader is closed, the cursor position of the original reader moves to the end of the subtree. ReadToDescendant moves the cursor to the start of the first descendant node with the specified name/namespace. ReadToFollowing moves the cursor to the start of the first node—regardless of depth—with the specified name/namespace. ReadTo 534 | Chapter 11: Other XML and JSON Technologies
- 553. NextSibling moves the cursor to the start of the first sibling node with the specified name/namespace. There are also two legacy methods: ReadString and ReadElementString behave like ReadContentAsString and ReadElementContentAsString, except that they throw an exception if there’s more than a single text node within the element. You should avoid these methods because they throw an exception if an element contains a comment. Reading Attributes XmlReader provides an indexer giving you direct (random) access to an ele‐ ment’s attributes—by name or position. Using the indexer is equivalent to calling GetAttribute. Given the XML fragment <customer id="123" status="archived"/> we could read its attributes, as follows: Console.WriteLine (reader ["id"]); // 123 Console.WriteLine (reader ["status"]); // archived Console.WriteLine (reader ["bogus"] == null); // True The XmlReader must be positioned on a start element in order to read attributes. After calling ReadStartElement, the attributes are gone forever! Although attribute order is semantically irrelevant, you can access attributes by their ordinal position. We could rewrite the preceding example as follows: Console.WriteLine (reader [0]); // 123 Console.WriteLine (reader [1]); // archived The indexer also lets you specify the attribute’s namespace—if it has one. AttributeCount returns the number of attributes for the current node. Attribute nodes To explicitly traverse attribute nodes, you must make a special diversion from the normal path of just calling Read. A good reason to do so is if you want to parse attribute values into other types, via the ReadContentAsXXX methods. The diversion must begin from a start element. To make the job easier, the forward- only rule is relaxed during attribute traversal: you can jump to any attribute (for‐ ward or backward) by calling MoveToAttribute. MoveToElement returns you to the start element from any‐ place within the attribute node diversion. XmlReader | 535 Other XML and JSON Technologies
- 554. Returning to our previous example: <customer id="123" status="archived"/> we can do this: reader.MoveToAttribute ("status"); string status = reader.ReadContentAsString(); reader.MoveToAttribute ("id"); int id = reader.ReadContentAsInt(); MoveToAttribute returns false if the specified attribute doesn’t exist. You can also traverse each attribute in sequence by calling the MoveToFirstAttri bute and then the MoveToNextAttribute methods: if (reader.MoveToFirstAttribute()) do { Console.WriteLine (reader.Name + "=" + reader.Value); } while (reader.MoveToNextAttribute()); // OUTPUT: id=123 status=archived Namespaces and Prefixes XmlReader provides two parallel systems for referring to element and attribute names: • Name • • NamespaceURI and LocalName • Whenever you read an element’s Name property or call a method that accepts a single name argument, you’re using the first system. This works well if no namespaces or prefixes are present; otherwise, it acts in a crude and literal manner. Namespaces are ignored, and prefixes are included exactly as they were written; for example: Sample fragment Name <customer ...> customer <customer xmlns='blah' ...> customer <x:customer ...> x:customer The following code works with the first two cases: reader.ReadStartElement ("customer"); The following is required to handle the third case: reader.ReadStartElement ("x:customer"); 536 | Chapter 11: Other XML and JSON Technologies
- 555. The second system works through two namespace-aware properties: NamespaceURI and LocalName. These properties take into account prefixes and default namespaces defined by parent elements. Prefixes are automatically expanded. This means that NamespaceURI always reflects the semantically correct namespace for the current element, and LocalName is always free of prefixes. When you pass two name arguments into a method such as ReadStartElement, you’re using this same system. For example, consider the following XML: <customer xmlns="DefaultNamespace" xmlns:other="OtherNamespace"> <address> <other:city> ... We could read this as follows: reader.ReadStartElement ("customer", "DefaultNamespace"); reader.ReadStartElement ("address", "DefaultNamespace"); reader.ReadStartElement ("city", "OtherNamespace"); Abstracting away prefixes is usually exactly what you want. If necessary, you can see what prefix was used through the Prefix property and convert it into a namespace by calling LookupNamespace. XmlWriter XmlWriter is a forward-only writer of an XML stream. The design of XmlWriter is symmetrical to XmlReader. As with XmlTextReader, you construct an XmlWriter by calling Create with an optional settings object. In the following example, we enable indenting to make the output more human-readable and then write a simple XML file: XmlWriterSettings settings = new XmlWriterSettings(); settings.Indent = true; using XmlWriter writer = XmlWriter.Create ("foo.xml", settings); writer.WriteStartElement ("customer"); writer.WriteElementString ("firstname", "Jim"); writer.WriteElementString ("lastname", "Bo"); writer.WriteEndElement(); This produces the following document (the same as the file we read in the first example of XmlReader): <?xml version="1.0" encoding="utf-8"?> <customer> <firstname>Jim</firstname> <lastname>Bo</lastname> </customer> XmlWriter | 537 Other XML and JSON Technologies
- 556. XmlWriter automatically writes the declaration at the top unless you indicate otherwise in XmlWriterSettings by setting OmitXmlDeclaration to true or Con formanceLevel to Fragment. The latter also permits writing multiple root nodes— something that otherwise throws an exception. The WriteValue method writes a single text node. It accepts both string and non‐ string types such as bool and DateTime, internally calling XmlConvert to perform XML-compliant string conversions: writer.WriteStartElement ("birthdate"); writer.WriteValue (DateTime.Now); writer.WriteEndElement(); In contrast, if we call WriteElementString ("birthdate", DateTime.Now.ToString()); the result would be both non-XML-compliant and vulnerable to incorrect parsing. WriteString is equivalent to calling WriteValue with a string. XmlWriter automat‐ ically escapes characters that would otherwise be illegal within an attribute or element, such as &, < >, and extended Unicode characters. Writing Attributes You can write attributes immediately after writing a start element: writer.WriteStartElement ("customer"); writer.WriteAttributeString ("id", "1"); writer.WriteAttributeString ("status", "archived"); To write nonstring values, call WriteStartAttribute, WriteValue, and then Write EndAttribute. Writing Other Node Types XmlWriter also defines the following methods for writing other kinds of nodes: WriteBase64 // for binary data WriteBinHex // for binary data WriteCData WriteComment WriteDocType WriteEntityRef WriteProcessingInstruction WriteRaw WriteWhitespace WriteRaw directly injects a string into the output stream. There is also a WriteNode method that accepts an XmlReader, echoing everything from the given XmlReader. 538 | Chapter 11: Other XML and JSON Technologies
- 557. Namespaces and Prefixes The overloads for the Write* methods allow you to associate an element or attribute with a namespace. Let’s rewrite the contents of the XML file in our previous example. This time we will associate all of the elements with the http://oreilly.com namespace, declaring the prefix o at the customer element: writer.WriteStartElement ("o", "customer", "http://oreilly.com"); writer.WriteElementString ("o", "firstname", "http://oreilly.com", "Jim"); writer.WriteElementString ("o", "lastname", "http://oreilly.com", "Bo"); writer.WriteEndElement(); The output is now as follows: <?xml version="1.0" encoding="utf-8"?> <o:customer xmlns:o='http://oreilly.com'> <o:firstname>Jim</o:firstname> <o:lastname>Bo</o:lastname> </o:customer> Notice how for brevity XmlWriter omits the child element’s namespace declarations when they are already declared by the parent element. Patterns for Using XmlReader/XmlWriter Working with Hierarchical Data Consider the following classes: public class Contacts { public IList<Customer> Customers = new List<Customer>(); public IList<Supplier> Suppliers = new List<Supplier>(); } public class Customer { public string FirstName, LastName; } public class Supplier { public string Name; } Suppose that you want to use XmlReader and XmlWriter to serialize a Contacts object to XML, as in the following: <?xml version="1.0" encoding="utf-8"?> <contacts> <customer id="1"> <firstname>Jay</firstname> <lastname>Dee</lastname> </customer> <customer> <!-- we'll assume id is optional --> <firstname>Kay</firstname> <lastname>Gee</lastname> </customer> <supplier> <name>X Technologies Ltd</name> </supplier> </contacts> Patterns for Using XmlReader/XmlWriter | 539 Other XML and JSON Technologies
- 558. The best approach is not to write one big method, but to encapsulate XML func‐ tionality in the Customer and Supplier types themselves by writing ReadXml and WriteXml methods on these types. The pattern for doing so is straightforward: • ReadXml and WriteXml leave the reader/writer at the same depth when they • exit. • ReadXml reads the outer element, whereas WriteXml writes only its inner • content. Here’s how we would write the Customer type: public class Customer { public const string XmlName = "customer"; public int? ID; public string FirstName, LastName; public Customer () { } public Customer (XmlReader r) { ReadXml (r); } public void ReadXml (XmlReader r) { if (r.MoveToAttribute ("id")) ID = r.ReadContentAsInt(); r.ReadStartElement(); FirstName = r.ReadElementContentAsString ("firstname", ""); LastName = r.ReadElementContentAsString ("lastname", ""); r.ReadEndElement(); } public void WriteXml (XmlWriter w) { if (ID.HasValue) w.WriteAttributeString ("id", "", ID.ToString()); w.WriteElementString ("firstname", FirstName); w.WriteElementString ("lastname", LastName); } } Notice that ReadXml reads the outer start and end element nodes. If its caller did this job instead, Customer couldn’t read its own attributes. The reason for not making WriteXml symmetrical in this regard is twofold: • The caller might need to choose how the outer element is named. • • The caller might need to write extra XML attributes, such as the element’s • subtype (which could then be used to decide which class to instantiate when reading back the element). Another benefit of following this pattern is that it makes your implementation compatible with IXmlSerializable (we cover this in “Serialization” in the online supplement at http://www.albahari.com/nutshell). 540 | Chapter 11: Other XML and JSON Technologies
- 559. The Supplier class is analogous: public class Supplier { public const string XmlName = "supplier"; public string Name; public Supplier () { } public Supplier (XmlReader r) { ReadXml (r); } public void ReadXml (XmlReader r) { r.ReadStartElement(); Name = r.ReadElementContentAsString ("name", ""); r.ReadEndElement(); } public void WriteXml (XmlWriter w) => w.WriteElementString ("name", Name); } With the Contacts class, we must enumerate the customers element in ReadXml, checking whether each subelement is a customer or a supplier. We also need to code around the empty element trap: public void ReadXml (XmlReader r) { bool isEmpty = r.IsEmptyElement; // This ensures we don't get r.ReadStartElement(); // snookered by an empty if (isEmpty) return; // <contacts/> element! while (r.NodeType == XmlNodeType.Element) { if (r.Name == Customer.XmlName) Customers.Add (new Customer (r)); else if (r.Name == Supplier.XmlName) Suppliers.Add (new Supplier (r)); else throw new XmlException ("Unexpected node: " + r.Name); } r.ReadEndElement(); } public void WriteXml (XmlWriter w) { foreach (Customer c in Customers) { w.WriteStartElement (Customer.XmlName); c.WriteXml (w); w.WriteEndElement(); } foreach (Supplier s in Suppliers) { w.WriteStartElement (Supplier.XmlName); s.WriteXml (w); w.WriteEndElement(); } } Patterns for Using XmlReader/XmlWriter | 541 Other XML and JSON Technologies
- 560. Here’s how to serialize a Contacts object populated with Customers and Suppliers to an XML file: var settings = new XmlWriterSettings(); settings.Indent = true; // To make visual inspection easier using XmlWriter writer = XmlWriter.Create ("contacts.xml", settings); var cts = new Contacts() // Add Customers and Suppliers... writer.WriteStartElement ("contacts"); cts.WriteXml (writer); writer.WriteEndElement(); Here’s how to deserialize from the same file: var settings = new XmlReaderSettings(); settings.IgnoreWhitespace = true; settings.IgnoreComments = true; settings.IgnoreProcessingInstructions = true; using XmlReader reader = XmlReader.Create("contacts.xml", settings); reader.MoveToContent(); var cts = new Contacts(); cts.ReadXml(reader); Mixing XmlReader/XmlWriter with an X-DOM You can fly in an X-DOM at any point in the XML tree where XmlReader or XmlWriter becomes too cumbersome. Using the X-DOM to handle inner elements is an excellent way to combine X-DOM’s ease of use with the low-memory footprint of XmlReader and XmlWriter. Using XmlReader with XElement To read the current element into an X-DOM, you call XNode.ReadFrom, passing in the XmlReader. Unlike XElement.Load, this method is not “greedy” in that it doesn’t expect to see a whole document. Instead, it reads just the end of the current subtree. For instance, suppose that we have an XML logfile structured as follows: <log> <logentry id="1"> <date>...</date> <source>...</source> ... </logentry> ... </log> If there were a million logentry elements, reading the entire thing into an X-DOM would waste memory. A better solution is to traverse each logentry with an XmlReader and then use XElement to process the elements individually: 542 | Chapter 11: Other XML and JSON Technologies
- 561. XmlReaderSettings settings = new XmlReaderSettings(); settings.IgnoreWhitespace = true; using XmlReader r = XmlReader.Create ("logfile.xml", settings); r.ReadStartElement ("log"); while (r.Name == "logentry") { XElement logEntry = (XElement) XNode.ReadFrom (r); int id = (int) logEntry.Attribute ("id"); DateTime date = (DateTime) logEntry.Element ("date"); string source = (string) logEntry.Element ("source"); ... } r.ReadEndElement(); If you follow the pattern described in the previous section, you can slot an XElement into a custom type’s ReadXml or WriteXml method without the caller ever knowing you’ve cheated! For instance, we could rewrite Customer’s ReadXml method, as follows: public void ReadXml (XmlReader r) { XElement x = (XElement) XNode.ReadFrom (r); ID = (int) x.Attribute ("id"); FirstName = (string) x.Element ("firstname"); LastName = (string) x.Element ("lastname"); } XElement collaborates with XmlReader to ensure that namespaces are kept intact, and prefixes are properly expanded—even if defined at an outer level. So, if our XML file reads like this: <log xmlns="http://loggingspace"> <logentry id="1"> ... the XElements we constructed at the logentry level would correctly inherit the outer namespace. Using XmlWriter with XElement You can use an XElement just to write inner elements to an XmlWriter. The fol‐ lowing code writes a million logentry elements to an XML file using XElement— without storing the entire thing in memory: using XmlWriter w = XmlWriter.Create ("logfile.xml"); w.WriteStartElement ("log"); for (int i = 0; i < 1000000; i++) { XElement e = new XElement ("logentry", new XAttribute ("id", i), new XElement ("date", DateTime.Today.AddDays (-1)), new XElement ("source", "test")); Patterns for Using XmlReader/XmlWriter | 543 Other XML and JSON Technologies
- 562. e.WriteTo (w); } w.WriteEndElement (); Using an XElement incurs minimal execution overhead. If we amend this example to use XmlWriter throughout, there’s no measurable difference in execution time. Working with JSON JSON has become a popular alternative to XML. Although it lacks the advanced features of XML (such as namespaces, prefixes, and schemas), it benefits from being simple and uncluttered, with a format similar to what you would get from converting a JavaScript object to a string. Historically, .NET had no built-in support for JSON, and you had to rely on third-party libraries—primarily Json.NET. Although this is no longer the case, the Json.NET library is still popular for a number of reasons: • It’s been around since 2011. • • The same API also runs on older .NET platforms. • • It’s considered to be more functional (as least in the past) than the Microsoft • JSON APIs. The Microsoft JSON APIs have the advantage of having been designed from the ground up to be simple and extremely efficient. Also, from .NET 6, their functional‐ ity has become quite close to that of Json.NET. In this section, we cover the following: • The forward-only reader and writer (Utf8JsonReader and Utf8JsonWriter) • • The JsonDocument read-only DOM reader • • The JsonNode read/write DOM reader/writer • In “Serialization,” in the online supplement at http://www.albahari.com/nutshell, we cover JsonSerializer, which automatically serializes and deserializes JSON to classes. Utf8JsonReader System.Text.Json.Utf8JsonReader is an optimized forward-only reader for UTF-8 encoded JSON text. Conceptually, it’s like the XmlReader introduced earlier in this chapter, and is used in much the same way. 544 | Chapter 11: Other XML and JSON Technologies
- 563. Consider the following JSON file named people.json: { "FirstName":"Sara", "LastName":"Wells", "Age":35, "Friends":["Dylan","Ian"] } The curly braces indicate a JSON object (which contains properties such as "First Name" and "LastName"), whereas the square brackets indicate a JSON array (which contains repeating elements). In this case, the repeating elements are strings, but they could be objects (or other arrays). The following code parses the file by enumerating its JSON tokens. A token is the beginning or end of an object, the beginning or end of an array, the name of a property, or an array or property value (string, number, true, false, or null): byte[] data = File.ReadAllBytes ("people.json"); Utf8JsonReader reader = new Utf8JsonReader (data); while (reader.Read()) { switch (reader.TokenType) { case JsonTokenType.StartObject: Console.WriteLine ($"Start of object"); break; case JsonTokenType.EndObject: Console.WriteLine ($"End of object"); break; case JsonTokenType.StartArray: Console.WriteLine(); Console.WriteLine ($"Start of array"); break; case JsonTokenType.EndArray: Console.WriteLine ($"End of array"); break; case JsonTokenType.PropertyName: Console.Write ($"Property: {reader.GetString()}"); break; case JsonTokenType.String: Console.WriteLine ($" Value: {reader.GetString()}"); break; case JsonTokenType.Number: Console.WriteLine ($" Value: {reader.GetInt32()}"); break; default: Console.WriteLine ($"No support for {reader.TokenType}"); break; } } Working with JSON | 545 Other XML and JSON Technologies
- 564. Here’s the output: Start of object Property: FirstName Value: Sara Property: LastName Value: Wells Property: Age Value: 35 Property: Friends Start of array Value: Dylan Value: Ian End of array End of object Because Utf8JsonReader works directly with UTF-8, it steps through the tokens without first having to convert the input into UTF-16 (the format of .NET strings). Conversion to UTF-16 takes place only when you call a method such as Get String(). Interestingly, Utf8JsonReader’s constructor does not accept a byte array, but rather a ReadOnlySpan<byte> (for this reason, Utf8JsonReader is defined as a ref struct). You can pass in a byte array because there’s an implicit conversion from T[] to ReadOnlySpan<T>. In Chapter 23, we describe how spans work and how you can use them to improve performance by minimizing memory allocations. JsonReaderOptions By default, Utf8JsonReader requires that the JSON conform strictly to the JSON RFC 8259 standard. You can instruct the reader to be more tolerant by passing an instance of JsonReaderOptions to the Utf8JsonReader constructor. The options allow the following: C-Style comments By default, comments in JSON cause a JsonException to be thrown. Setting the CommentHandling property to JsonCommentHandling.Skip causes com‐ ments to be ignored, whereas JsonCommentHandling.Allow causes the reader to recognize them and emit JsonTokenType.Comment tokens when they are encountered. Comments cannot appear in the middle of other tokens. Trailing commas Per the standard, the last property of an object and the last element of an array must not have a trailing comma. Setting the AllowTrailingCommas property to e relaxes this restriction. Control over the maximum nesting depth By default, objects and arrays can nest to 64 levels. Setting the MaxDepth to a different number overrides this setting. 546 | Chapter 11: Other XML and JSON Technologies
- 565. Utf8JsonWriter System.Text.Json.Utf8JsonWriter is a forward-only JSON writer. It supports the following types: • String and DateTime (which is formatted as a JSON string) • • The numeric types Int32, UInt32, Int64, UInt64, Single, Double, and Decimal • (which are formatted as JSON numbers) • bool (formatted as JSON true/false literals) • • JSON null • • Arrays • You can organize these data types into objects in accordance with the JSON stan‐ dard. It also lets you write comments, which are not part of the JSON standard but are often supported by JSON parsers in practice. The following code demonstrates its use: var options = new JsonWriterOptions { Indented = true }; using (var stream = File.Create ("MyFile.json")) using (var writer = new Utf8JsonWriter (stream, options)) { writer.WriteStartObject(); // Property name and value specified in one call writer.WriteString ("FirstName", "Dylan"); writer.WriteString ("LastName", "Lockwood"); // Property name and value specified in separate calls writer.WritePropertyName ("Age"); writer.WriteNumberValue (46); writer.WriteCommentValue ("This is a (non-standard) comment"); writer.WriteEndObject(); } This generates the following output file: { "FirstName": "Dylan", "LastName": "Lockwood", "Age": 46 /*This is a (non-standard) comment*/ } From .NET 6, Utf8JsonWriter has a WriteRawValue method to emit a string or byte array directly into the JSON stream. This is useful in special cases—for instance, if you want a number to be written such that it always includes a decimal point (1.0 rather than 1). In this example, we set the Indented property on JsonWriterOptions to true to improve readability. Had we not done so, the output would be as follows: {"FirstName":"Dylan","LastName":"Lockwood","Age":46...} Working with JSON | 547 Other XML and JSON Technologies
- 566. The JsonWriterOptions also has an Encoder property to control the escaping of strings, and a SkipValidation property to allow structural validation checks to be bypassed (allowing the emission of invalid output JSON). JsonDocument System.Text.Json.JsonDocument parses JSON data into a read-only DOM com‐ posed of JsonElement instances that are generated on demand. Unlike Utf8Json Reader, JsonDocument lets you access elements randomly. JsonDocument is one of two DOM-based APIs for working with JSON, the other being JsonNode (which we will cover in the following section). JsonNode was intro‐ duced in .NET 6, primarily to satisfy the demand for a writable DOM. However, it’s also suitable in read-only scenarios and exposes a somewhat more fluent inter‐ face, backed by a traditional DOM that uses classes for JSON values, arrays, and objects. In contrast, JsonDocument is extremely lightweight, comprising just one class of note (JsonDocument) and two lightweight structs (JsonElement and Json Property) that parse the underlying data on demand. The difference is illustrated in Figure 11-1. In most real-world scenarios, the performance benefits of JsonDocument over JsonNode are negligible, so you can skip to JsonNode if you prefer to learn just one API. Figure 11-1. JSON DOM APIs JsonDocument further improves its efficiency by employing pooled memory to minimize garbage collection. This means that you must dispose the JsonDocument after use; otherwise, its memory will not be returned to the pool. Consequently, when a class stores a JsonDocument in a field, it must also implement IDisposable. Should this be burdensome, con‐ sider using JsonNode instead. 548 | Chapter 11: Other XML and JSON Technologies
- 567. The static Parse method instantiates a JsonDocument from a stream, string, or memory buffer: using JsonDocument document = JsonDocument.Parse (jsonString); ... When calling Parse, you can optionally provide a JsonDocumentOptions object to control the handling of trailing commas, comments, and the maximum nesting depth (for a discussion on how these options work, see “JsonReaderOptions” on page 546). From there, you can access the DOM via the RootElement property: using JsonDocument document = JsonDocument.Parse ("123"); JsonElement root = document.RootElement; Console.WriteLine (root.ValueKind); // Number JsonElement can represent a JSON value (string, number, true/false, null), array, or object; the ValueKind property indicates which. The methods that we describe in the following sections throw an exception if the element isn’t of the kind expected. If you’re not sure of a JSON file’s schema, you can avoid such excep‐ tions by checking ValueKind first (or by using the TryGet* methods). JsonElement also provides two methods that work for any kind of element: GetRawText() returns the inner JSON, and WriteTo writes that element to a Utf8JsonWriter. Reading simple values If the element represents a JSON value, you can obtain its value by calling Get String, GetInt32, GetBoolean, etc.): using JsonDocument document = JsonDocument.Parse ("123"); int number = document.RootElement.GetInt32(); JsonElement also provides methods to parse JSON strings into other commonly used CLR types such as DateTime (and even base-64 binary). There are also TryGet* versions that avoid throwing an exception if the parse fails. Reading JSON arrays If the JsonElement represents an array, you can call the following methods: EnumerateArray() Enumerates all the subitems for a JSON array (as JsonElements). GetArrayLength() Returns the number of elements in the array. You can also use the indexer to return an element at a specific position: Working with JSON | 549 Other XML and JSON Technologies
- 568. using JsonDocument document = JsonDocument.Parse (@"[1, 2, 3, 4, 5]"); int length = document.RootElement.GetArrayLength(); // 5 int value = document.RootElement[3].GetInt32(); // 4 Reading JSON objects If the element represents a JSON object, you can call the following methods: EnumerateObject() Enumerates all of the object’s property names and values. GetProperty (string propertyName) Gets a property by name (returning another JsonElement). Throws an excep‐ tion if the name isn’t present. TryGetProperty (string propertyName, out JsonElement value) Returns an object’s property if present. For example: using JsonDocument document = JsonDocument.Parse (@"{ ""Age"": 32}"); JsonElement root = document.RootElement; int age = root.GetProperty ("Age").GetInt32(); Here’s how we could “discover” the Age property: JsonProperty ageProp = root.EnumerateObject().First(); string name = ageProp.Name; // Age JsonElement value = ageProp.Value; Console.WriteLine (value.ValueKind); // Number Console.WriteLine (value.GetInt32()); // 32 JsonDocument and LINQ JsonDocument lends itself well to LINQ. Given the following JSON file: [ { "FirstName":"Sara", "LastName":"Wells", "Age":35, "Friends":["Ian"] }, { "FirstName":"Ian", "LastName":"Weems", "Age":42, "Friends":["Joe","Eric","Li"] }, { "FirstName":"Dylan", "LastName":"Lockwood", "Age":46, "Friends":["Sara","Ian"] 550 | Chapter 11: Other XML and JSON Technologies
- 569. } ] we can use JsonDocument to query this with LINQ, as follows: using var stream = File.OpenRead (jsonPath); using JsonDocument document = JsonDocument.Parse (json); var query = from person in document.RootElement.EnumerateArray() select new { FirstName = person.GetProperty ("FirstName").GetString(), Age = person.GetProperty ("Age").GetInt32(), Friends = from friend in person.GetProperty ("Friends").EnumerateArray() select friend.GetString() }; Because LINQ queries are lazily evaluated, it’s important to enumerate the query before the document goes out of scope and JsonDocument is implicitly disposed of by virtue of the using statement. Making updates with a JSON writer Although JsonDocument is read-only, you can send the content of a JsonElement to a Utf8JsonWriter with the WriteTo method. This provides a mechanism for emitting a modified version of the JSON. Here’s how we can take the JSON from the preceding example and write it to a new JSON file that includes only people with two or more friends: using var json = File.OpenRead (jsonPath); using JsonDocument document = JsonDocument.Parse (json); var options = new JsonWriterOptions { Indented = true }; using (var outputStream = File.Create ("NewFile.json")) using (var writer = new Utf8JsonWriter (outputStream, options)) { writer.WriteStartArray(); foreach (var person in document.RootElement.EnumerateArray()) { int friendCount = person.GetProperty ("Friends").GetArrayLength(); if (friendCount >= 2) person.WriteTo (writer); } } If you need the ability to update the DOM, however, JsonNode is a better solution. JsonNode JsonNode (in System.Text.Json.Nodes) was introduced in .NET 6, primarily to satisfy the demand for a writable DOM. However, it’s also suitable in read-only Working with JSON | 551 Other XML and JSON Technologies
- 570. scenarios and exposes a somewhat more fluent interface, backed by a traditional DOM that uses classes for JSON values, arrays, and objects (see Figure 11-1). Being classes, they incur a garbage-collection cost, but this is likely to be negligible in most real-world scenarios. JsonNode is still highly optimized and can actually be faster than JsonDocument when the same nodes are read repeatedly (because JsonNode, while lazy, caches the results of parsing). The static Parse method creates a JsonNode from a stream, string, memory buffer, or Utf8JsonReader: JsonNode node = JsonNode.Parse (jsonString); When calling Parse, you can optionally provide a JsonDocumentOptions object to control the handling of trailing commas, comments, and the maximum nesting depth (for a discussion on how these options work, see “JsonReaderOptions” on page 546). Unlike JsonDocument, JsonNode does not require disposal. Calling ToString() on a JsonNode returns a human-readable (indented) JSON string. There is also a ToJsonString() method, which returns a compact JSON string. Parse returns a subtype of JsonNode, which will be JsonValue, JsonObject, or JsonArray. To avoid the clutter of a downcast, JsonNode provides helper methods called AsValue(), AsObject(), and AsArray(): var node = JsonNode.Parse ("123"); // Parses to a JsonValue int number = node.AsValue().GetValue<int>(); // Shortcut for ((JsonValue)node).GetValue<int>(); However, you don’t usually need to call these methods, because the most commonly used members are exposed on the JsonNode class itself: var node = JsonNode.Parse ("123"); int number = node.GetValue<int>(); // Shortcut for node.AsValue().GetValue<int>(); Reading simple values We just saw that you can extract or parse a simple value by calling GetValue with a type parameter. To make this even easier, JsonNode overloads C#’s explicit cast operators, enabling the following shortcut: var node = JsonNode.Parse ("123"); int number = (int) node; The types for which this works comprise the standard numeric types: char, bool, DateTime, DateTimeOffset, and Guid (and their nullable versions), as well as string. If you’re not sure whether parsing will succeed, the following code is required: if (node.AsValue().TryGetValue<int> (out var number)) Console.WriteLine (number); 552 | Chapter 11: Other XML and JSON Technologies
- 571. Nodes that have been parsed from JSON text are internally backed by a JsonElement (part of the JsonDocument read-only JSON API). You can extract the underlying JsonElement as follows: JsonElement je = node.GetValue<JsonElement>(); However, this doesn’t work when the node is instantiated explicitly (as will be the case when we update the DOM). Such nodes are backed not by a JsonElement but by the actual parsed value (see “Making updates with JsonNode” on page 554). Reading JSON arrays A JsonNode that represents a JSON array will be of type JsonArray. JsonArray implements IList<JsonNode>, so you can enumerate over it and access the elements like you would an array or list: var node = JsonNode.Parse (@"[1, 2, 3, 4, 5]"); Console.WriteLine (node.AsArray().Count); // 5 foreach (JsonNode child in node.AsArray()) { ... } As a shortcut, you can access the indexer directly from the JsonNode class: Console.WriteLine ((int)node[0]); // 1 Reading JSON objects A JsonNode that represents a JSON object will be of type JsonObject. JsonObject implements IDictionary<string,JsonNode>, so you can access a member via the indexer, as well as enumerating over the dictionary’s key/value pairs. And as with JsonArray, you can access the indexer directly from the JsonNode class: var node = JsonNode.Parse (@"{ ""Name"":""Alice"", ""Age"": 32}"); string name = (string) node ["Name"]; // Alice int age = (int) node ["Age"]; // 32 Here’s how we could “discover” the Name and Age properties: // Enumerate over the dictionary’s key/value pairs: foreach (KeyValuePair<string,JsonNode> keyValuePair in node.AsObject()) { string propertyName = keyValuePair.Key; // "Name" (then "Age") JsonNode value = keyValuePair.Value; } If you’re not sure whether a property has been defined, the following pattern also works: Working with JSON | 553 Other XML and JSON Technologies
- 572. if (node.AsObject().TryGetPropertyValue ("Name", out JsonNode nameNode)) { ... } Fluent traversal and LINQ You can reach deep into a hierarchy just with indexers. For example, given the following JSON file: [ { "FirstName":"Sara", "LastName":"Wells", "Age":35, "Friends":["Ian"] }, { "FirstName":"Ian", "LastName":"Weems", "Age":42, "Friends":["Joe","Eric","Li"] }, { "FirstName":"Dylan", "LastName":"Lockwood", "Age":46, "Friends":["Sara","Ian"] } ] we can extract the second person’s third friend as follows: string li = (string) node[1]["Friends"][2]; Such a file is also easy to query via LINQ: JsonNode node = JsonNode.Parse (File.ReadAllText (jsonPath)); var query = from person in node.AsArray() select new { FirstName = (string) person ["FirstName"], Age = (int) person ["Age"], Friends = from friend in person ["Friends"].AsArray() select (string) friend }; Unlike JsonDocument, JsonNode is not disposable, so we don’t have to worry about the potential for disposal during lazy enumeration. Making updates with JsonNode JsonObject and JsonArray are mutable, so you can update their content. 554 | Chapter 11: Other XML and JSON Technologies
- 573. The easiest way to replace or add properties to a JsonObject is via the indexer. In the following example, we change the Color property’s value from “Red” to “White” and add a new property called “Valid”: var node = JsonNode.Parse ("{ "Color": "Red" }"); node ["Color"] = "White"; node ["Valid"] = true; Console.WriteLine (node.ToJsonString()); // {"Color":"White","Valid":true} The second line in that example is a shortcut for the following: node ["Color"] = JsonValue.Create ("White"); Rather than assigning the property a simple value, you can assign it a JsonArray or JsonObject. (We will demonstrate how to construct JsonArray and JsonObject instances in the following section.) To remove a property, first cast to JsonObject (or call AsObject) and then call the Remove method: node.AsObject().Remove ("Valid"); (JsonObject also exposes an Add method, which throws an exception if the property already exists.) JsonArray also lets you use the indexer to replace items: var node = JsonNode.Parse ("[1, 2, 3]"); node[0] = 10; Calling AsArray exposes the Add/Insert/Remove/RemoveAt methods. In the follow‐ ing example, we remove the first element in the array and add one to the end: var arrayNode = JsonNode.Parse ("[1, 2, 3]"); arrayNode.AsArray().RemoveAt(0); arrayNode.AsArray().Add (4); Console.WriteLine (arrayNode.ToJsonString()); // [2,3,4] Constructing a JsonNode DOM programmatically JsonArray and JsonObject have constructors that support object initialization syntax, which allows you to build an entire JsonNode DOM in one expression: var node = new JsonArray { new JsonObject { ["Name"] = "Tracy", ["Age"] = 30, ["Friends"] = new JsonArray ("Lisa", "Joe") }, new JsonObject { ["Name"] = "Jordyn", ["Age"] = 25, ["Friends"] = new JsonArray ("Tracy", "Li") } }; Working with JSON | 555 Other XML and JSON Technologies
- 574. This evaluates to the following JSON: [ { "Name": "Tracy", "Age": 30, "Friends": ["Lisa", "Joe"] }, { "Name": "Jordyn", "Age": 25, "Friends": ["Tracy","Li"] } ] 556 | Chapter 11: Other XML and JSON Technologies
- 575. 12 Disposal and Garbage Collection Some objects require explicit tear-down code to release resources such as open files, locks, operating system handles, and unmanaged objects. In .NET parlance, this is called disposal, and it is supported through the IDisposable interface. The managed memory occupied by unused objects must also be reclaimed at some point; this function is known as garbage collection and is performed by the CLR. Disposal differs from garbage collection in that disposal is usually explicitly instiga‐ ted; garbage collection is totally automatic. In other words, the programmer takes care of such things as releasing file handles, locks, and operating system resources, while the CLR takes care of releasing memory. This chapter discusses both disposal and garbage collection, and also describes C# finalizers and the pattern by which they can provide a backup for disposal. Lastly, we discuss the intricacies of the garbage collector and other memory management options. IDisposable, Dispose, and Close .NET defines a special interface for types requiring a tear-down method: public interface IDisposable { void Dispose(); } C#’s using statement provides a syntactic shortcut for calling Dispose on objects that implement IDisposable, using a try/finally block: using (FileStream fs = new FileStream ("myFile.txt", FileMode.Open)) { // ... Write to the file ... } The compiler converts this to the following: 557
- 576. FileStream fs = new FileStream ("myFile.txt", FileMode.Open); try { // ... Write to the file ... } finally { if (fs != null) ((IDisposable)fs).Dispose(); } The finally block ensures that the Dispose method is called even when an excep‐ tion is thrown or the code exits the block early. Similarly, the following syntax ensures disposal as soon as fs goes out of scope: using FileStream fs = new FileStream ("myFile.txt", FileMode.Open); // ... Write to the file ... In simple scenarios, writing your own disposable type is just a matter of implement‐ ing IDisposable and writing the Dispose method: sealed class Demo : IDisposable { public void Dispose() { // Perform cleanup / tear-down. ... } } This pattern works well in simple cases and is appropriate for sealed classes. In “Calling Dispose from a Finalizer” on page 566, we describe a more elaborate pattern that can provide a backup for consumers that forget to call Dispose. With unsealed types, there’s a strong case for following this latter pattern from the outset—otherwise, it becomes very messy if the subtype wants to add such functionality itself. Standard Disposal Semantics .NET follows a de facto set of rules in its disposal logic. These rules are not hardwired to .NET or the C# language in any way; their purpose is to define a consistent protocol to consumers. Here they are: 1. After an object has been disposed, it’s beyond redemption. It cannot be reacti‐ 1. vated, and calling its methods or properties (other than Dispose) throws an ObjectDisposedException. 2. Calling an object’s Dispose method repeatedly causes no error. 2. 3. If disposable object x “owns” disposable object y, x’s Dispose method automati‐ 3. cally calls y’s Dispose method—unless instructed otherwise. 558 | Chapter 12: Disposal and Garbage Collection
- 577. These rules are also helpful when writing your own types, though they’re not mandatory. Nothing prevents you from writing an “Undispose” method other than, perhaps, the flak you might cop from colleagues! According to rule 3, a container object automatically disposes its child objects. A good example is a Windows Forms container control such as a Form or Panel. The container can host many child controls, yet you don’t dispose every one of them explicitly; closing or disposing the parent control or form takes care of the whole lot. Another example is when you wrap a FileStream in a DeflateStream. Disposing the DeflateStream also disposes the FileStream—unless you instructed otherwise in the constructor. Close and Stop Some types define a method called Close in addition to Dispose. The .NET BCL is not completely consistent on the semantics of a Close method, although in nearly all cases it’s either of the following: • Functionally identical to Dispose • • A functional subset of Dispose • An example of the latter is IDbConnection: a Closed connection can be re-Opened; a Disposed connection cannot. Another example is a Windows Form activated with ShowDialog: Close hides it; Dispose releases its resources. Some classes define a Stop method (e.g., Timer or HttpListener). A Stop method may release unmanaged resources, like Dispose, but unlike Dispose, it allows for re-Starting. When to Dispose A safe rule to follow (in nearly all cases) is “if in doubt, dispose.” Objects wrapping an unmanaged resource handle will nearly always require disposal in order to free the handle. Examples include file or network streams, network sockets, Windows Forms controls, GDI+ pens, brushes, and bitmaps. Conversely, if a type is dispos‐ able, it will often (but not always) reference an unmanaged handle, directly or indirectly. This is because unmanaged handles provide the gateway to the “outside world” of OS resources, network connections, and database locks—the primary means by which objects can create trouble outside of themselves if improperly abandoned. There are, however, three scenarios for not disposing: • When you don’t “own” the object—for example, when obtaining a shared object • via a static field or property • When an object’s Dispose method does something that you don’t want • IDisposable, Dispose, and Close | 559 Disposal & Garbage Collection
- 578. • When an object’s Dispose method is unnecessary by design, and disposing that • object would add complexity to your program The first category is rare. The main cases are in the System.Drawing namespace: the GDI+ objects obtained through static fields or properties (such as Brushes.Blue) must never be disposed because the same instance is used throughout the life of the application. Instances that you obtain through constructors, however (such as new SolidBrush), should be disposed, as should instances obtained through static methods (such as Font.FromHdc). The second category is more common. There are some good examples in the System.IO and System.Data namespaces: Type Disposal function When not to dispose MemoryStream Prevents further I/O When you later need to read/write the stream StreamReader, StreamWriter Flushes the reader/writer and closes the underlying stream When you want to keep the underlying stream open (you must then call Flush on a Stream Writer when you’re done) IDbConnection Releases a database connection and clears the connection string If you need to re-Open it, you should call Close instead of Dispose DbContext (EF Core) Prevents further use When you might have lazily evaluated queries connected to that context MemoryStream’s Dispose method disables only the object; it doesn’t perform any critical cleanup because a MemoryStream holds no unmanaged handles or other such resources. The third category includes the classes such as StringReader and StringWriter. These types are disposable under the duress of their base class rather than through a genuine need to perform essential cleanup. If you happen to instantiate and work with such an object entirely in one method, wrapping it in a using block adds little inconvenience. But if the object is longer lasting, keeping track of when it’s no longer used so that you can dispose of it adds unnecessary complexity. In such cases, you can simply ignore object disposal. Ignoring disposal can sometimes incur a performance cost (see “Calling Dispose from a Finalizer” on page 566). Clearing Fields in Disposal In general, you don’t need to clear an object’s fields in its Dispose method. However, it is good practice to unsubscribe from events that the object has subscribed to internally over its lifetime (for an example, see “Managed Memory Leaks” on page 576). Unsubscribing from such events prevents receiving unwanted event 560 | Chapter 12: Disposal and Garbage Collection
- 579. notifications—and prevents unintentionally keeping the object alive in the eyes of the garbage collector (GC). A Dispose method itself does not cause (managed) memory to be released—this can happen only in garbage collection. It’s also worth setting a field to indicate that the object is disposed so that you can throw an ObjectDisposedException if a consumer later tries to call members on the object. A good pattern is to use a publicly readable automatic property for this: public bool IsDisposed { get; private set; } Although technically unnecessary, it can also be good to clear an object’s own event handlers (by setting them to null) in the Dispose method. This eliminates the possibility of those events firing during or after disposal. Occasionally, an object holds high-value secrets, such as encryption keys. In these cases, it can make sense to clear such data from fields during disposal (to avoid potential discovery by other processes on the machine when the memory is later released to the operating system). The SymmetricAlgorithm class in System .Security.Cryptography does exactly this by calling Array.Clear on the byte array holding the encryption key. Anonymous Disposal Sometimes, it’s useful to implement IDisposable without having to write a class. For instance, suppose that you want to expose methods on a class that suspend and resume event processing: class Foo { int _suspendCount; public void SuspendEvents() => _suspendCount++; public void ResumeEvents() => _suspendCount--; void FireSomeEvent() { if (_suspendCount == 0) ... fire some event ... } ... } Such an API is clumsy to use. Consumers must remember to call ResumeEvents. And to be robust, they must do so in a finally block (in case an exception is thrown): var foo = new Foo(); foo.SuspendEvents(); try { IDisposable, Dispose, and Close | 561 Disposal & Garbage Collection
- 580. ... do stuff ... // Because an exception could be thrown here } finally { foo.ResumeEvents(); // ...we must call this in a finally block } A better pattern is to do away with ResumeEvents and have SuspendEvents return an IDisposable. Consumers can then do this: using (foo.SuspendEvents()) { ... do stuff ... } The problem is that this pushes work onto whoever has to implement the Suspend Events method. Even with a good effort to reduce whitespace, we end up with this extra clutter: public IDisposable SuspendEvents() { _suspendCount++; return new SuspendToken (this); } class SuspendToken : IDisposable { Foo _foo; public SuspendToken (Foo foo) => _foo = foo; public void Dispose() { if (_foo != null) _foo._suspendCount--; _foo = null; // Prevent against consumer disposing twice } } The anonymous disposal pattern solves this problem. With the following reusable class: public class Disposable : IDisposable { public static Disposable Create (Action onDispose) => new Disposable (onDispose); Action _onDispose; Disposable (Action onDispose) => _onDispose = onDispose; public void Dispose() { _onDispose?.Invoke(); // Execute disposal action if non-null. _onDispose = null; // Ensure it can’t execute a second time. } } 562 | Chapter 12: Disposal and Garbage Collection
- 581. we can reduce our SuspendEvents method to the following: public IDisposable SuspendEvents() { _suspendCount++; return Disposable.Create (() => _suspendCount--); } Automatic Garbage Collection Regardless of whether an object requires a Dispose method for custom tear-down logic, at some point the memory it occupies on the heap must be freed. The CLR handles this side of it entirely automatically via an automatic GC. You never deallocate managed memory yourself. For example, consider the following method: public void Test() { byte[] myArray = new byte[1000]; ... } When Test executes, an array to hold 1,000 bytes is allocated on the memory heap. The array is referenced by the variable myArray, stored on the local variable stack. When the method exits, this local variable myArray pops out of scope, meaning that nothing is left to reference the array on the memory heap. The orphaned array then becomes eligible to be reclaimed in garbage collection. In debug mode with optimizations disabled, the lifetime of an object referenced by a local variable extends to the end of the code block to ease debugging. Otherwise, it becomes eligible for collection at the earliest point at which it’s no longer used. Garbage collection does not happen immediately after an object is orphaned. Rather like garbage collection on the street, it happens periodically, although (unlike garbage collection on the street) not to a fixed schedule. The CLR bases its deci‐ sion on when to collect upon a number of factors, such as the available memory, the amount of memory allocation, and the time since the last collection (the GC self-tunes to optimize for an application’s specific memory access patterns). This means that there’s an indeterminate delay between an object being orphaned and being released from memory. This delay can range from nanoseconds to days. The GC doesn’t collect all garbage with every collection. Instead, the memory manager divides objects into genera‐ tions, and the GC collects new generations (recently alloca‐ ted objects) more frequently than old generations (long-lived objects). We discuss this in more detail in “How the GC Works” on page 569. Automatic Garbage Collection | 563 Disposal & Garbage Collection
- 582. Garbage Collection and Memory Consumption The GC tries to strike a balance between the time it spends doing garbage collection and the application’s memory consumption (working set). Consequently, applica‐ tions can consume more memory than they need, particularly if large temporary arrays are constructed. You can monitor a process’s memory consumption via the Windows Task Manager or Resource Monitor—or programmatically by querying a performance counter: // These types are in System.Diagnostics: string procName = Process.GetCurrentProcess().ProcessName; using PerformanceCounter pc = new PerformanceCounter ("Process", "Private Bytes", procName); Console.WriteLine (pc.NextValue()); This queries the private working set, which gives the best overall indication of your program’s memory consumption. Specifically, it excludes memory that the CLR has internally deallocated and is willing to rescind to the OS should another process need it. Roots A root is something that keeps an object alive. If an object is not directly or indirectly referenced by a root, it will be eligible for garbage collection. A root is one of the following: • A local variable or parameter in an executing method (or in any method in its • call stack) • A static variable • • An object on the queue that stores objects ready for finalization (see the next • section) It’s impossible for code to execute in a deleted object, so if there’s any possibility of an (instance) method executing, its object must somehow be referenced in one of these ways. Note that a group of objects that reference one another cyclically are considered dead without a root referee (see Figure 12-1). To put it in another way, objects that cannot be accessed by following the arrows (references) from a root object are unreachable—and therefore subject to collection. 564 | Chapter 12: Disposal and Garbage Collection
- 583. Figure 12-1. Roots Finalizers Prior to an object being released from memory, its finalizer runs, if it has one. A finalizer is declared like a constructor, but it is prefixed by the ˜ symbol: class Test { ˜Test() { // Finalizer logic... } } (Although similar in declaration to a constructor, finalizers cannot be declared as public or static, cannot have parameters, and cannot call the base class.) Finalizers are possible because garbage collection works in distinct phases. First, the GC identifies the unused objects ripe for deletion. Those without finalizers are deleted immediately. Those with pending (unrun) finalizers are kept alive (for now) and are put onto a special queue. At that point, garbage collection is complete, and your program continues execut‐ ing. The finalizer thread then kicks in and starts running in parallel to your pro‐ gram, picking objects off that special queue and running their finalization methods. Prior to each object’s finalizer running, it’s still very much alive—that queue acts as a root object. After it’s been dequeued and the finalizer executed, the object becomes orphaned and will be deleted in the next collection (for that object’s generation). Finalizers | 565 Disposal & Garbage Collection
- 584. Finalizers can be useful, but they come with some provisos: • Finalizers slow the allocation and collection of memory (the GC needs to keep • track of which finalizers have run). • Finalizers prolong the life of the object and any referred objects (they must all • await the next garbage truck for actual deletion). • It’s impossible to predict in what order the finalizers for a set of objects will be • called. • You have limited control over when the finalizer for an object will be called. • • If code in a finalizer blocks, other objects cannot be finalized. • • Finalizers can be circumvented altogether if an application fails to unload • cleanly. In summary, finalizers are somewhat like lawyers—although there are cases in which you really need them, in general you don’t want to use them unless absolutely necessary. If you do use them, you need to be 100% sure you understand what they are doing for you. Here are some guidelines for implementing finalizers: • Ensure that your finalizer executes quickly. • • Never block in your finalizer (see “Blocking” on page 612). • • Don’t reference other finalizable objects. • • Don’t throw exceptions. • The CLR can call an object’s finalizer even if an exception is thrown during construction. For this reason, it pays not to assume that fields are correctly initialized when writing a finalizer. Calling Dispose from a Finalizer A popular pattern is to have the finalizer call Dispose. This makes sense when cleanup is not urgent and hastening it by calling Dispose is more of an optimization than a necessity. Keep in mind that with this pattern you couple memory deal‐ location to resource deallocation—two things with potentially divergent interests (unless the resource is itself memory). You also increase the burden on the finalization thread. This pattern also serves as a backup for cases when a con‐ sumer simply forgets to call Dispose. However, it’s then a good idea to log the failure so that you can fix the bug. There’s a standard pattern for implementing this, as follows: 566 | Chapter 12: Disposal and Garbage Collection
- 585. class Test : IDisposable { public void Dispose() // NOT virtual { Dispose (true); GC.SuppressFinalize (this); // Prevent finalizer from running. } protected virtual void Dispose (bool disposing) { if (disposing) { // Call Dispose() on other objects owned by this instance. // You can reference other finalizable objects here. // ... } // Release unmanaged resources owned by (just) this object. // ... } ~Test() => Dispose (false); } Dispose is overloaded to accept a bool disposing flag. The parameterless version is not declared as virtual and simply calls the enhanced version with true. The enhanced version contains the actual disposal logic and is protected and virtual; this provides a safe point for subclasses to add their own disposal logic. The disposing flag means it’s being called “properly” from the Dispose method rather than in “last-resort mode” from the finalizer. The idea is that when called with disposing set to false, this method should not, in general, reference other objects with finalizers (because such objects might themselves have been finalized and so be in an unpredictable state). This rules out quite a lot! Here are a couple of tasks that the Dispose method can still perform in last-resort mode, when disposing is false: • Releasing any direct references to OS resources (obtained, perhaps, via a • P/Invoke call to the Win32 API) • Deleting a temporary file created on construction • To make this robust, any code capable of throwing an exception should be wrapped in a try/catch block, and the exception, ideally, logged. Any logging should be as simple and robust as possible. Notice that we call GC.SuppressFinalize in the parameterless Dispose method— this prevents the finalizer from running when the GC later catches up with it. Technically, this is unnecessary given that Dispose methods must tolerate repeated calls. However, doing so improves performance because it allows the object (and its referenced objects) to be garbage collected in a single cycle. Finalizers | 567 Disposal & Garbage Collection
- 586. Resurrection Suppose a finalizer modifies a living object such that it refers back to the dying object. When the next garbage collection happens (for the object’s generation), the CLR will see the previously dying object as no longer orphaned—and so it will evade garbage collection. This is an advanced scenario and is called resurrection. To illustrate, suppose that we want to write a class that manages a temporary file. When an instance of that class is garbage collected, we’d like the finalizer to delete the temporary file. It sounds easy: public class TempFileRef { public readonly string FilePath; public TempFileRef (string filePath) { FilePath = filePath; } ~TempFileRef() { File.Delete (FilePath); } } Unfortunately, this has a bug: File.Delete might throw an exception (due to a lack of permissions, perhaps, or the file being in use, or having already been deleted). Such an exception would take down the entire application (as well as preventing other finalizers from running). We could simply “swallow” the exception with an empty catch block, but then we’d never know that anything went wrong. Calling some elaborate error reporting API would also be undesirable because it would burden the finalizer thread, hindering garbage collection for other objects. We want to restrict finalization actions to those that are simple, reliable, and quick. A better option is to record the failure to a static collection, as follows: public class TempFileRef { static internal readonly ConcurrentQueue<TempFileRef> FailedDeletions = new ConcurrentQueue<TempFileRef>(); public readonly string FilePath; public Exception DeletionError { get; private set; } public TempFileRef (string filePath) { FilePath = filePath; } ~TempFileRef() { try { File.Delete (FilePath); } catch (Exception ex) { DeletionError = ex; FailedDeletions.Enqueue (this); // Resurrection } } } Enqueuing the object to the static FailedDeletions collection gives the object another referee, ensuring that it remains alive until the object is eventually dequeued. 568 | Chapter 12: Disposal and Garbage Collection
- 587. ConcurrentQueue<T> is a thread-safe version of Queue<T> and is defined in System.Collections.Concurrent (see Chap‐ ter 22). There are a couple of reasons for using a thread-safe collection. First, the CLR reserves the right to execute finaliz‐ ers on more than one thread in parallel. This means that when accessing shared state such as a static collection, we must consider the possibility of two objects being finalized at once. Second, at some point we’re going to want to dequeue items from FailedDeletions so that we can do something about them. This also must be done in a thread-safe fashion because it could happen while the finalizer is concurrently enqueuing another object. GC.ReRegisterForFinalize A resurrected object’s finalizer will not run a second time—unless you call GC .ReRegisterForFinalize. In the following example, we try to delete a temporary file in a finalizer (as in the last example). But if the deletion fails, we reregister the object so as to try again in the next garbage collection: public class TempFileRef { public readonly string FilePath; int _deleteAttempt; public TempFileRef (string filePath) { FilePath = filePath; } ~TempFileRef() { try { File.Delete (FilePath); } catch { if (_deleteAttempt++ < 3) GC.ReRegisterForFinalize (this); } } } After the third failed attempt, our finalizer will silently give up trying to delete the file. We could enhance this by combining it with the previous example—in other words, adding it to the FailedDeletions queue after the third failure. Be careful to call ReRegisterForFinalize just once in the finalizer method. If you call it twice, the object will be reregis‐ tered twice and will have to undergo two more finalizations! How the GC Works The standard CLR uses a generational mark-and-compact GC that performs auto‐ matic memory management for objects stored on the managed heap. The GC is considered to be a tracing GC in that it doesn’t interfere with every access to an How the GC Works | 569 Disposal & Garbage Collection
- 588. object, but rather wakes up intermittently and traces the graph of objects stored on the managed heap to determine which objects can be considered garbage and therefore collected. The GC initiates a garbage collection upon performing a memory allocation (via the new keyword), either after a certain threshold of memory has been allocated or at other times to reduce the application’s memory footprint. This process can also be initiated manually by calling System.GC.Collect. During a garbage collection, all threads can by frozen (more on this in the next section). The GC begins with its root object references and walks the object graph, marking all the objects it touches as reachable. When this process is complete, all objects that have not been marked are considered unused and are subject to garbage collection. Unused objects without finalizers are immediately discarded; unused objects with finalizers are enqueued for processing on the finalizer thread after the GC is com‐ plete. These objects then become eligible for collection in the next GC for the object’s generation (unless resurrected). The remaining “live” objects are then shifted to the start of the heap (compacted), freeing space for more objects. This compaction serves two purposes: it prevents memory fragmentation, and it allows the GC to employ a very simple strategy when allocating new objects, which is to always allocate memory at the end of the heap. This prevents the potentially time-consuming task of maintaining a list of free memory segments. If there is insufficient space to allocate memory for a new object after garbage collection and the OS is unable to grant further memory, an OutOfMemoryException is thrown. You can obtain information about the current state of the managed heap by calling GC.GetGCMemoryInfo(). From .NET 5, this method has been enhanced to return performance- related data. Optimization Techniques The GC incorporates various optimization techniques to reduce the garbage collec‐ tion time. Generational collection The most important optimization is that the GC is generational. This takes advan‐ tage of the fact that although many objects are allocated and discarded rapidly, cer‐ tain objects are long-lived and thus don’t need to be traced during every collection. Basically, the GC divides the managed heap into three generations. Objects that have just been allocated are in Gen0, and objects that have survived one collection cycle are in Gen1; all other objects are in Gen2. Gen0 and Gen1 are known as ephemeral (short-lived) generations. 570 | Chapter 12: Disposal and Garbage Collection
- 589. The CLR keeps the Gen0 section relatively small (with a typical size of a few hundred KB to a few MB). When the Gen0 section fills up, the GC instigates a Gen0 collection—which happens relatively often. The GC applies a similar memory threshold to Gen1 (which acts as a buffer to Gen2), and so Gen1 collections are relatively quick and frequent, too. Full collections that include Gen2, however, take much longer and so happen infrequently. Figure 12-2 shows the effect of a full collection. Figure 12-2. Heap generations To give some very rough ballpark figures, a Gen0 collection might take less than one millisecond, which is not enough to be noticed in a typical application. A full collection, however, might take as long as 100 ms on a program with large object graphs. These figures depend on numerous factors and so can vary considerably— particularly in the case of Gen2, whose size is unbounded (unlike Gen0 and Gen1). The upshot is that short-lived objects are very efficient in their use of the GC. The StringBuilders created in the following method would almost certainly be collected in a fast Gen0: string Foo() { var sb1 = new StringBuilder ("test"); sb1.Append ("..."); var sb2 = new StringBuilder ("test"); sb2.Append (sb1.ToString()); return sb2.ToString(); } How the GC Works | 571 Disposal & Garbage Collection
- 590. 1 The same thing can occur occasionally in the generational heap due to pinning (see “The fixed Statement” on page 244). The Large Object Heap The GC uses a separate heap called the Large Object Heap (LOH) for objects larger than a certain threshold (currently 85,000 bytes). This prevents the cost of compacting large objects and prevents excessive Gen0 collections—without the LOH, allocating a series of 16 MB objects might trigger a Gen0 collection after every allocation. By default, the LOH is not subject to compaction, because moving large blocks of memory during garbage collection would be prohibitively expensive. This has two consequences: • Allocations can be slower, because the GC can’t always simply allocate objects • at the end of the heap—it must also look in the middle for gaps, and this requires maintaining a linked list of free memory blocks.1 • The LOH is subject to fragmentation. This means that the freeing of an object • can create a hole in the LOH that can be difficult to fill later. For instance, a hole left by an 86,000-byte object can be filled only by an object of between 85,000 bytes and 86,000 bytes (unless adjoined by another hole). Should you anticipate a problem with fragmentation, you can instruct the GC to compact the LOH in the next collection, as follows: GCSettings.LargeObjectHeapCompactionMode = GCLargeObjectHeapCompactionMode.CompactOnce; Another workaround, if your program frequently allocates large arrays, is to use .NET’s array pooling API (see “Array P