04-struct-union.ppt

Structures and Unions in C
Alan L. Cox
alc@rice.edu

Objectives
Be able to use compound data structures in
programs
Be able to pass compound data structures as
function arguments, either by value or by
reference
Be able to do simple bit-vector manipulations
Cox Structures and Unions 2

Structures
Compound data:
A date is
 an int month and
 an int day and
 an int year
Unlike Java, C doesn’t
automatically define functions
for initializing and printing …
struct ADate {
int month;
int day;
int year;
};
struct ADate date;
date.month = 2;
date.day = 4;
date.year = 2021;

Structure Representation & Size
sizeof(struct …) =
sum of sizeof(field)
+ alignment padding
Processor- and compiler-specific
62
61 EF BE AD DE
c1 c2 i
padding
struct CharCharInt {
char c1;
char c2;
int i;
} foo;
foo.c1 = ’a’;
foo.c2 = ’b’;
foo.i = 0xDEADBEEF;
x86 uses “little-endian” representation

Typedef
Mechanism for creating new type names
 New names are an alias for some other type
 May improve the portability and/or clarity of the
program
typedef long int64_t;
typedef struct ADate {
int month;
int day;
int year;
} Date;
int64_t i = 100000000000;
Date d = { 2, 4, 2021 };
Overload existing type
names for portability
Simplify complex type names

Constants
Allow consistent use of the same constant
throughout the program
 Improves clarity of the program
 Reduces likelihood of simple errors
 Easier to update constants in the program
int array[10];
for (i=0; i<10; i++) {
…
}
#define SIZE 10
int array[SIZE];
for (i=0; i<SIZE; i++) {
…
}
Preprocessor directive
Constant names are
capitalized by convention
Define once,
use throughout
the program

Arrays of Structures
Date birthdays[NFRIENDS];
bool
check_birthday(Date today)
{
int i;
for (i = 0; i < NFRIENDS; i++) {
if ((today.month == birthdays[i].month) &&
(today.day == birthdays[i].day))
return (true);
return (false);
}
Constant
Array declaration
Array index, then
structure field

Pointers to Structures
Date
create_date1(int month,
int day,
int year)
{
Date d;
d.month = month;
d.day = day;
d.year = year;
return (d);
}
void
create_date2(Date *d,
int month,
int day,
int year)
{
d->month = month;
d->day = day;
d->year = year;
}
Copies date
Pass-by-reference
Date today;
today = create_date1(2, 4, 2021);
create_date2(&today, 2, 4, 2021);

Pointers to Structures (cont.)
void
create_date2(Date *d,
int month,
int day,
int year)
{
d->month = month;
d->day = day;
d->year = year;
}
void
fun_with_dates(void)
{
Date today;
create_date2(&today, 2, 4, 2021);
}
today.month:
today.day:
today.year:
0x1000
0x1004
0x1008
month: 2
day: 4
year: 2021
0x30A0
0x30A4
0x30A8
d: 0x1000
0x3098
2
4
2021

Pointers to Structures (cont.)
Date *
create_date3(int month,
int day,
int year)
{
Date *d;
d->month = month;
d->day = day;
d->year = year;
return (d);
}
What is d pointing to?!?!
(more on this later)

Abstraction in C
struct widget;
struct widget *widget_create(void);
int widget_op(struct widget *widget, int operand);
void widget_destroy(struct widget *widget);
From the #include file widget.h:
From the file widget.c:
#include “widget.h”
struct widget {
int x;
…
};
Definition is hidden!

Collections of Bools (Bit Vectors)
Byte, word, ... can represent many Booleans
One per bit, e.g., 00100101 = false, false, true, ..., true
Bit-wise operations:
Bit-wise AND: 00100101 & 10111100 == 00100100
Bit-wise OR: 00100101 | 10111100 == 10111101
Bit-wise NOT: ~ 00100101 == 11011010
Bit-wise XOR: 00100101 ^ 10111100 == 10011001

Operations on Bit Vectors
const unsigned int low_three_bits_mask = 0x7;
unsigned int bit_vec = 0x15;
0…00 0111
0…01 0101
Always use C’s unsigned types for bit vectors
A mask indicates which bit positions we are interested in
0…00 0101 == 0…01 0101 & 0…00 0111
important_bits = bit_vec & low_three_bits_mask;
Selecting bits:
Result = ?

0…00 0111
0…01 0101
bit_vec |= low_three_bits_mask;
Setting bits:
Result = ?
0…01 0111 == 0…01 0101 | 0…00 0111

0…00 0111
0…01 0101
bit_vec &= ~low_three_bits_mask;
Clearing bits:
Result = ?
0…01 0000 == 0…01 0101 & ~0…00 0111

Bit-field Structures
Special syntax packs
structure values more
tightly
Similar to bit vectors, but
arguably easier to read
 Nonetheless, bit vectors
are more commonly
used.
Padded to be an integral
number of words
 Placement is compiler-
specific.
1 1 0 1 1 0 … …
f1 f2 f3
struct Flags {
int f1:3;
unsigned int f2:1;
unsigned int f3:2;
} my_flags;
my_flags.f1 = -2;
my_flags.f2 = 1;
my_flags.f3 = 2;

Unions
Choices:
An element is
 an int i or
 a char c
sizeof(union …) =
maximum of sizeof(field)
EF BE AD DE
c
i
padding
union AnElt {
int i;
char c;
} elt1, elt2;
elt1.i = 4;
elt2.c = ’a’;
elt2.i = 0xDEADBEEF;

Unions
A union value doesn’t “know” which case it
contains
union AnElt {
int i;
char c;
} elt1, elt2;
elt1.i = 4;
elt2.c = ’a’;
elt2.i = 0xDEADBEEF;
if (elt1 currently has a char) …
How should your program keep track
whether elt1, elt2 hold an int or
a char?
?
?
Basic answer: Another variable holds
that info

Tagged Unions
Tag every value with its case
I.e., pair the type info together with the union
Implicit in Java, Scheme, ML, …
Enum must be external to struct,
so constants are globally visible.
Struct field must be named.
enum Union_Tag { IS_INT, IS_CHAR };
struct TaggedUnion {
enum Union_Tag tag;
union {
int i;
char c;
} data;
};

Next Time
Memory Allocation

04-struct-union.ppt

More Related Content

Similar to 04-struct-union.ppt (20)

More from Md Polash (6)

Recently uploaded (20)

04-struct-union.ppt