Hadoop Pig

Clogeny Technologies http://www.clogeny.com
(US) 408-556-9645
(India) +91 20 661 43 482
What is Pig
Provides abstraction for processing large datasets
A scripting language for exploring large datasets
Made up of two pieces:
• Pig Latin, a language used to express data flows
• The execution enviornment to run Pig Latin programs
 Local execution in a single JVM
 Distributed execution on Hadoop cluster
Provides much richer datasets, typically multivalued and
nested
Made up of series of operations or transformations, that are
applied to the input data to produce output
Under the covers, it turns the transformations into a series
of MapReduce jobs

(US) 408-556-9645
(India) +91 20 661 43 482
What Pig is NOT
Not suitable for small data sets
In some cases, Pig doesn’t perform as well as
programs written in MapReduce

(US) 408-556-9645
(India) +91 20 661 43 482

(US) 408-556-9645
(India) +91 20 661 43 482
Installing Pig
Step 1: make yourself root user
Step 2: Install pig
• $ yum –y install pig
Step 3: Install Hbase
• $ yum –y install hbase

(US) 408-556-9645
(India) +91 20 661 43 482
Installing Pig
Step 4: Set environment variables for pig
• $ vi ~/.bashrc
• Add the following lines to the end of the file
 export PIG_CONF_DIR=/usr/lib/pig/conf
 export PIG_CLASSPATH=/usr/lib/hbase/hbase-0.94.6-
cdh4.3.0-security.jar:/usr/lib/zookeeper/zookeeper-3.4.5-
cdh4.3.0.jar
Step 5: Configure Hadoop Properties
• Add the following in /usr/lib/pig/conf/pig.properties
 fs.default.name=hdfs://localhost/
 mapred.job.tracker=localhost:8021

(US) 408-556-9645
(India) +91 20 661 43 482
Pig Execution Types
Local Mode
• Runs in a single JVM and accesses the local filesystem
• Suitable only for small datasets and when trying out Pig
• The execution type is set using the -x or -exectype
option
• To run in local mode:
 $ pig -x local
 This starts Grunt, the Pig interactive shell

(US) 408-556-9645
(India) +91 20 661 43 482
Pig Execution Types
MapReduce Mode
• In MapReduce mode, Pig translates queries into
MapReduce jobs and runs them on a Hadoop cluster
• Used when you want to run Pig on large datasets
• Uses the fs.default.name and mapred.job.tracker
properties to connect to Hadoop
• To run in MapReduce Mode:
 $ pig

(US) 408-556-9645
(India) +91 20 661 43 482
Running Pig Programs
Script
• Run a script file that contains Pig commands
• $ pig script.pig
Grunt
• An interactive shell for running Pig commands
• You can run pig scripts from Grunt using run and exec
Embedded
• Run Pig programs from Java using PigServer class
• For programmatic access to Grunt, use PigRunner

(US) 408-556-9645
(India) +91 20 661 43 482
Running a sample Pig Job
Step 1: Go to user Joe
• $ su - joe
Step 2: Start the pig grunt command prompt
• $ pig
Step 3: Check if the directories from the previous
MR job exists
• grunt> ls
• It should display:
 hdfs://localhost:8020/user/joe/input <dir>
 hdfs://localhost:8020/user/joe/output <dir>

(US) 408-556-9645
(India) +91 20 661 43 482
Running a sample Pig Job
Step 4: Run the same MR job as a pig job
• grunt> A = LOAD 'input';
• grunt> B = FILTER A BY $0 MATCHES '.*dfs[a-z.]+.*';
• grunt> DUMP B;
Step 5: Check the status of the job
• You can check the status of a job while it is running at
http://<virtual machine ip>:50030

(US) 408-556-9645
(India) +91 20 661 43 482
An Example
Copy a sample file in HDFS
• $ su
• $ cat input/ncdc/micro-tab/sample.txt
• $ hadoop fs -copyFromLocal input/ncdc/micro-
tab/sample.txt .
Start the Grunt shell
• $ pig
Load the data from sample.txt into a datastore
• grunt> records = LOAD 'sample.txt' AS (year:chararray,
temperature:int, quality:int);

(US) 408-556-9645
(India) +91 20 661 43 482
An Example
Filter out invalid records
• grunt> filtered_records = FILTER records BY temperature !=
9999 AND (quality == 0 OR quality == 1 OR quality == 4 OR
quality == 5 OR quality == 9);
Group the filtered records on data column “year”
• grunt> grouped_records = GROUP filtered_records BY
year;
Get the max temperature for each group
• grunt> max_temp = FOREACH grouped_records GENERATE
group,MAX(filtered_records.temperature);
Get the output
• grunt> DUMP max_temp;

(US) 408-556-9645
(India) +91 20 661 43 482
Example Explained
Loading the data from sample.txt creates a
datastore with the following schema:
• records: {year: chararray,temperature: int,quality: int}
Actual data is loaded only when some related
datastore is DUMP’ed or STORE’ed
When a filter is applied a new datastore is created
from the records detastore with the same schema
but with filtered data
• filtered_records: {year: chararray,temperature:
int,quality: int}

(US) 408-556-9645
(India) +91 20 661 43 482
Example Explained
When a group is created for a particular data field,
for each unique value of that field a list of tuples is
created
The tuple list contains all data fields grouped
together
The length of that list is equal to the number of
occurrences of the group value in original data
The structure of the grouped_record dataset is
• grouped_records: {group: chararray,filtered_records:
{(year: chararray,temperature: int,quality: int)}}

(US) 408-556-9645
(India) +91 20 661 43 482
Example Explained
To get some aggregated value from a group you
should apply some aggregation on one or more of
the fields which were not a part of the group
The structure of the aggregated data set is
• max_temp: {group: chararray,int}
Now when you say DUMP max_temp all the actual
data loading, filtering, grouping, aggregation takes
place as MR jobs
The mappers are used for loading and filtering and
the reducers for grouping and aggregation

(US) 408-556-9645
(India) +91 20 661 43 482
Example Explained
records:
– (1950,0,1)
– (1950,22,1)
– (1950,-11,1)
– (1949,111,1)
– (1949,78,1)
filtered_records:
– (1950,0,1)
– (1950,22,1)
– (1950,-11,1)
– (1949,111,1)
– (1949,78,1)

(US) 408-556-9645
(India) +91 20 661 43 482
Example Explained
grouped_records:
– (1949,{(1949,111,1),(1949,78,1)})
– (1950,{(1950,0,1),(1950,22,1),(1950,-11,1)})
max_temp:
– (1949,111)
– (1950,22)

(US) 408-556-9645
(India) +91 20 661 43 482
Generating examples
Creating a dataset to test your examples is a pain
Using random datasets doesn’t work as filter and
join operations tend to remove all random data
With ILLUSTRATE, Pig provides a reasonably
complete and concise sample dataset
• grunt> ILLUSTRATE max_temp;

(US) 408-556-9645
(India) +91 20 661 43 482
Generating examples

(US) 408-556-9645
(India) +91 20 661 43 482
Comparison with Databases
Relational Databases Pig
SQL statements are a set of
constraints that, taken together,
define the output
Pig Latin program is a step-by-step
transformation of input relation to
output
RDBMS store data in tables, with
tightly predefined schemas
In Pig you can define the schema
at runtime
SQL operates on flatter data
structures
Pig Latin supports nested data
structures
RDBMS have features to support
online, low latency queries, such
as transactions and indexes
Such features are absent in Pig

(US) 408-556-9645
(India) +91 20 661 43 482
Multiquery Execution
STORE and DUMP are different commands but act
same in interactive mode
DUMP is a diagnostic tool and always triggers
execution
STORE in batch mode triggers execution efficiently
Pig parses whole script in batch mode and checks
for any optimizations

(US) 408-556-9645
(India) +91 20 661 43 482
• $ hadoop fs –copyFromLocal input/pig/multiquey/A .
• $ pig
• grunt> A = LOAD 'A';
• grunt> B = FILTER A BY $1 == 'banana';
• grunt> C = FILTER A BY $1 != 'banana';
• grunt> STORE B INTO 'output/b';
• grunt> STORE C INTO 'output/c';
Relations B and C are both derived from A, so to save
reading A twice, Pig can run this script as a single
MapReduce job by reading A once and writing two
output files from the job, one for each of B and C.
This feature is called multiquery execution

(US) 408-556-9645
(India) +91 20 661 43 482
$ hadoop fs –copyFromLocal
input/pig/multiquery/A .
$ hadoop fs –mkdir output
$ vi multiquery.pig
Copy the script from previous slide in the file
$ pig multiquery.pig
The output of the MR job will show only 1 mapper
and 0 reducers

(US) 408-556-9645
(India) +91 20 661 43 482
Pig Latin Relational Operators
Category Operator Description
Loading and Sorting LOAD Load data from filesystem or other
storage into a relation
STORE Saves a relation to the filesystem or other
storage
DUMP Prints a relation to the console
Filtering FILTER Removes unwanted rows from a relation
DISTINCT Removes duplicate rows from a relation
FOREACH…GENERATE Adds or removes fields from a relation
MAPREDUCE Runs a MapReduce job using a relation as
input
STREAM Transforms a relation using an external
program
SAMPLE Selects a random sample of a relation

(US) 408-556-9645
(India) +91 20 661 43 482
Pig Latin Relational Operators
Category Operator Description
Grouping and joining JOIN Joins two or more relations
COGROUP Groups the data in two or more relations
GROUP Groups the data in a single relation
CROSS Creates the cross-product of two or more
relations
Sorting ORDER Sorts a relation by one or more fields
LIMIT Limits the size of a relation to a maximum
number of tuples
Combining and Splitting UNION Combines two or more relations into one
SPLIT Splits a relation into two or more relations

(US) 408-556-9645
(India) +91 20 661 43 482
Pig Latin Diagnostic Operators
Operator Description
DESCRIBE Prints a relations schema
EXPLAIN Prints the logical and physical plans
ILLUSTRATE Shows a sample execution of the logical
plan, using a generated subset of the
input

(US) 408-556-9645
(India) +91 20 661 43 482
Pig Latin macro and UDF statements
Operator Description
REGISTER Registers a JAR file with the Pig runtime
DEFINE Creates an alias for a macro, UDF,
streaming script, or command
specification
IMPORT Import macros defined in a separate file
into a script

(US) 408-556-9645
(India) +91 20 661 43 482
Pig Latin Commands
Category Command Description
Hadoop Filesystem cat Prints the contents of one or more files
cd Changes the current directory
copyFromLocal Copies a local file or directory to a Hadoop filesystem
copyToLocal Copies a file or directory on a Hadoop filesystem to
the local filesystem
cp Copies a file or directory to another directory
fs Accesses Hadoop’s filesystem shell
ls Lists files
mkdir Creates a new directory
mv Moves a file or directory to another directory
pwd Prints the path of the current working directory
rm Deletes a file or directory
rmf Forcibly deletes a file or directory

(US) 408-556-9645
(India) +91 20 661 43 482
Pig Latin Commands
Category Command Description
Hadoop
MapReduce Utility
kill Kills a MapReduce job
exec Runs a script in a new Grunt shell in batch mode
help Shows the available commands and options
quit Exits the interpreter
run Runs a script within the existing Grunt shell
set Sets Pig options and MapReduce job properties
sh Run a shell command from within Grunt
debug Runs the job in debug mode. You can also set the
debug level
job.name Gives a pig job a meaningful name

(US) 408-556-9645
(India) +91 20 661 43 482
Expressions
Category Expressions Description Examples
Constant Literal Constant value 1.0, 'a'
Field (by position) $n Field in position n (zero-based) $0
Field (by name) f Field named f year
Field (disambiguate) r::f Field named f from relation r
after grouping
or joining
A::year
Projection c.$n, c.f Field in container c (relation, bag,
or tuple)
by position, by name
records.$0,
records.year
Map lookup m#k Value associated with key k in
map m
items#'Coat'
Cast (t) f Cast of field f to type t (int) year
Conditional x ? y : z Bincond/ternary; y if x evaluates
to true, z
Otherwise
quality == 0 ? 0 : 1

(US) 408-556-9645
(India) +91 20 661 43 482
Expressions
Category Expressions Description Examples
Arithmetic x + y, x – y
x * y, x / y
x % y
+x, -x
Addition, subtraction
Multiplication, division
Modulo, the remainder of x divided by y
Unary positive, negation
$1 + $2, $1 - $2
$1 * $2, $1 / $2
$1 % $2
+1, –1
Comparison x == y, x != y
x > y, x < y
x >= y, x <= y
x matches y
x is null
x is not null
Equals, does not equal
Greater than, less than
Greater or equal to, less or equal to
Pattern matching with regular exprn
Is null
Is not null
q == 0, t != 9999
q > 0, q < 10
q >= 1, q <=9
q matches '[01459]‘
t is null
t is not null
Boolean x or y
x and y
not x
Logical or
Logical and
Logical negation
q == 0 or q == 1
q == 0 and r == 0
not q matches
'[01459]'
Functional fn(f1,f2,…) Invocation of function fn on fields f1, f2,
etc.
isGood(quality)

(US) 408-556-9645
(India) +91 20 661 43 482
Data Types
Category Type Description Examples
Numeric int
long
float
double
32-bit signed integer
64-bit signed integer
32-bit floating-point number
64-bit floating-point number
1
1L
1.0F
1.0
Text chararray Character array in UTF-16 format 'a’
Binary bytearray Byte array Not supported
Complex tuple
bag
map
Sequence of fields of any type
An unordered collection of tuples,
possibly with duplicates
A set of key-value pairs; keys must be
character arrays, but
values may be any type
(1,‘apple')
{(1,‘apple'),(2)}
['a'#‘apple']

(US) 408-556-9645
(India) +91 20 661 43 482
Schemas
Defining a right schema for a dataset is not mandatory
 grunt> records = LOAD 'sample.txt’ AS (year:int, temperature:int,
quality:int);
 grunt> DESCRIBE records;
– records: {year: int,temperature: int,quality: int}
 grunt> records = LOAD 'sample.txt’ AS (year:chararray,
– records: {year: chararray,temperature: int,quality: int}
 grunt> records = LOAD 'sample.txt’ AS (year, temperature:int,
quality);
– records: {year: bytearray,temperature: int,quality: bytearray}

(US) 408-556-9645
(India) +91 20 661 43 482
Schemas
Not defining the schema
• grunt> records = LOAD 'sample.txt';
• grunt> describe records;
 Schema for records unknown.
• grunt> dump records;
 (1950,0,1)
 (1950,22,1)
 (1950,-11,1)
 (1949,111,1)
 (1949,78,1)

(US) 408-556-9645
(India) +91 20 661 43 482
Schemas
Creating dataset from a dataset whose schema is
undefined
• grunt> projected_records = FOREACH records GENERATE
$0, $1, $2;
• grunt> DUMP projected_records;
 (1950,0,1)
 (1950,22,1)
 (1950,-11,1)
 (1949,111,1)
 (1949,78,1)
• grunt> DESCRIBE projected_records;
 projected_records: {bytearray,bytearray,bytearray}

(US) 408-556-9645
(India) +91 20 661 43 482
Validations and nulls
SQL will fail if we try to load string into a numeric column
In Pig, if the value cannot be cast to the type declared in the
schema, it will substitute a null value
• $ hadoop fs -copyFromLocal input/ncdc/micro-
tab/sample_corrupt.txt .
• grunt> records = LOAD 'sample_corrupt.txt' AS (year:chararray,
• grunt> DUMP records;
 (1950,0,1)
 (1950,22,1)
 (1950,,1)
 (1949,111,1)
 (1949,78,1)

(US) 408-556-9645
(India) +91 20 661 43 482
Validations and nulls
Filtering records on null values
• grunt> corrupt_records = FILTER records BY temperature is null;
• grunt> DUMP corrupt_records;
 (1950,,1)
• grunt> SPLIT records INTO good_records IF temperature is not
null, bad_records IF temperature is null;
• grunt> dump corrupt_records;
 (1950,,1)
• grunt> DUMP good_records;
 (1950,0,1)
 (1950,22,1)
 (1949,111,1)
 (1949,78,1)

(US) 408-556-9645
(India) +91 20 661 43 482
Macros
Provide a way to package reusable pieces of Pig
Latin code
grunt> DEFINE max_by_group(X, group_key, max_field) RETURNS Y
{
A = GROUP $X by $group_key;
$Y = FOREACH A GENERATE group, MAX($X.$max_field);
};

(US) 408-556-9645
(India) +91 20 661 43 482
Macros
Using Macro
• grunt> records = LOAD 'sample.txt' AS (year:chararray, temperature:int,
quality:int);
• grunt> max_temp = max_by_group(records, year, temperature);
 (1949,111)
 (1950,22)
Using Normal Pig Constructs
• grunt> macro_max_by_group_A_0 = GROUP records by (year);
• grunt> max_temp = FOREACH macro_max_by_group_A_0 GENERATE group,
MAX(records.(temperature));
 (1949,111)
 (1950,22)

(US) 408-556-9645
(India) +91 20 661 43 482
UDF
UDF are User Defined Functions
It gives the ability to plug in custom code
We only cover Java UDF’s in this section
UDF’s can also be written in Python and JavaScript
Types of UDF:
• Filter UDF
• Eval UDF
• Load UDF

(US) 408-556-9645
(India) +91 20 661 43 482
Filter UDF
Filter UDF’s are UDF used as filter in Pig
The idea is to change this line
• filtered_records = FILTER records BY temperature != 9999 AND (quality
== 0 OR quality == 1 OR quality == 4 OR quality == 5 OR quality == 9);
To
• filtered_records = FILTER records BY temperature != 9999 AND
isGood(quality);
Filter UDFs are all subclasses of FilterFunc, which itself is a
subclass of EvalFunc.
EvalFunc looks like the following class:
• public abstract class EvalFunc<T> {
public abstract T exec(Tuple input) throws IOException;
}

(US) 408-556-9645
(India) +91 20 661 43 482
Filter UDF
EvalFunc’s only abstract method, exec(), takes a
tuple and returns a single value, the
(parameterized) type T.
For FilterFunc, T is Boolean, so the method should
return true only for those tuples that should not be
filtered out.

package org.vmware.training;
import java.io.IOException;
import java.util.ArrayList;
import java.util.List;
import org.apache.pig.FilterFunc;
import org.apache.pig.FuncSpec;
import
org.apache.pig.backend.executionengine.ExecException;
import org.apache.pig.data.DataType;
import org.apache.pig.data.Tuple;
import org.apache.pig.impl.logicalLayer.FrontendException;
import org.apache.pig.impl.logicalLayer.schema.Schema;

public class IsGoodQuality extends FilterFunc {
@Override
public Boolean exec(Tuple tuple) throws IOException {
if (tuple == null || tuple.size() == 0) {
return false;
}
try {
Object object = tuple.get(0);
if (object == null) {
return false;
}
int i = (Integer) object;
return i == 0 || i == 1 || i == 4 || i == 5 || i == 9;
} catch (ExecException e) {
throw new IOException(e);
}
}
}

(US) 408-556-9645
(India) +91 20 661 43 482
Lets write out the code
$ mkdir ~/UDF
$ mkdir ~/UDF/IsGoodQuality
$ mkdir
~/UDF/IsGoodQuality/IsGoodQualityClasses
$ cd ~/UDF/IsGoodQuality
$ vim IsGoodQuality.java
Copy the code from previous slide into it

(US) 408-556-9645
(India) +91 20 661 43 482
Lets compile the code and create a jar
Set the classpath
• $ export
CLASSPATH=/usr/lib/hadoop/*:/usr/lib/hadoop/client-
0.20/*:/usr/lib/pig/*:/usr/lib/pig/lib/*
Compile the java class
• $ javac -d IsGoodQualityClasses IsGoodQuality.java
Create a jar
• $ jar -cvf goodquality.jar -C IsGoodQualityClasses/ .

(US) 408-556-9645
(India) +91 20 661 43 482
Lets use the UDF in our code
Start the pig console
• $ cd
• $ pig
Register the jar file
• grunt> REGISTER /root/UDF/IsGoodQuality/goodquality.jar;
Use the UDF from the jar
• grunt> records = LOAD 'sample.txt' AS (year:chararray,
• grunt> filtered_records = FILTER records BY temperature !=
9999 AND org.vmware.training.IsGoodQuality(quality);
• grunt> DUMP filtered_records;

(US) 408-556-9645
(India) +91 20 661 43 482
Join
Prerequisite
• $ hadoop fs –copyFromLocal /root/input/pig/join/A .
• $ hadoop fs –copyFromLocal /root/input/pig/join/B .
• $ pig
• grunt> A = LOAD 'A' AS (id:int, item:chararray);
• grunt> B = LOAD 'B' AS (name:chararray, id:int);
• grunt> DUMP A;
 (2,Tie)
 (4,Coat)
 (3,Hat)
 (1,Scarf)
• grunt> DUMP B;
 (Joe,2)
 (Hank,4)
 (Ali,0)
 (Eve,3)
 (Hank,2)

(US) 408-556-9645
(India) +91 20 661 43 482
Join
Inner Join
• grunt> C = JOIN A BY $0, B BY $1;
• grunt> DUMP C;
 (2,Tie,Joe,2)
 (2,Tie,Hank,2)
 (3,Hat,Eve,3)
 (4,Coat,Hank,4)
Outer Join
• grunt> C = JOIN A BY $0 LEFT OUTER, B BY $1;
 (1,Scarf,,)
 (2,Tie,Joe,2)
 (2,Tie,Hank,2)
 (3,Hat,Eve,3)
 (4,Coat,Hank,4)

(US) 408-556-9645
(India) +91 20 661 43 482
COGROUP
Similar to JOIN
Creates nested set of output tuples rather than a
flat structure
• grunt> D = COGROUP A BY $0, B BY $1;
• grunt> DUMP D;
 (0,{},{(Ali,0)})
 (1,{(1,Scarf)},{})
 (2,{(2,Tie)},{(Joe,2),(Hank,2)})
 (3,{(3,Hat)},{(Eve,3)})
 (4,{(4,Coat)},{(Hank,4)})

(US) 408-556-9645
(India) +91 20 661 43 482
CROSS
Joins every tuple in a relation with every tuple in a
second relation
• grunt> I = CROSS A, B;
• grunt> DUMP I;
 (2,Tie,Joe,2)
 (2,Tie,Hank,4)
 (2,Tie,Ali,0)
 (2,Tie,Eve,3)
 (2,Tie,Hank,2)
 (4,Coat,Joe,2)
 (4,Coat,Hank,4) ….

(US) 408-556-9645
(India) +91 20 661 43 482
STREAM
The STREAM operator allows you to transform data in a relation
using an external program or script
It is named by analogy with Hadoop Streaming, which provides a
similar capability for MapReduce
• grunt> DUMP A;
 (2,Tie)
 (4,Coat)
 (3,Hat)
 (1,Scarf)
• grunt> C = STREAM A THROUGH `cut -f 2`;
• grunt> DUMP C;
 (Tie)
 (Coat)
 (Hat)
 (Scarf)

(US) 408-556-9645
(India) +91 20 661 43 482
STREAM
The STREAM operator uses PigStorage to serialize
and deserialize relations to and from the program’s
standard input and output streams
Tuples in A are converted to tab delimited lines
that are passed to the script.
The output of the script is read one line at a time
and split on tabs to create new tuples for the
output relation C.

Hadoop Pig

More Related Content

What's hot (20)

Similar to Hadoop Pig (20)

Recently uploaded (20)

Hadoop Pig