An Introduction to Apache Pig For Absolute Beginners!

Dhanya Thailappan 08 Aug, 2021 • 6 min read

This article was published as a part of the Data Science Blogathon

This article is focused on Apache Pig. It is a high-level platform for processing and analyzing a huge amount of data.

OVERVIEW

If we see the top-level overview of Pig, then Pig is an abstraction over MapReduce. Pig runs on Hadoop. So, it makes use of both the Hadoop Distributed File System (HDFS) and Hadoop’s processing system, MapReduce. Data flows are executed
by an engine. It is used to analyze data sets as data flows. It includes a high-level language called Pig Latin for expressing these data flows.

The input for Pig is Pig Latin which will be converted into MapReduce jobs. Pig uses MapReduce tricks to do all of its data processing. It combines Pig Latin scripts into a series of one or more MapReduce jobs that in turn executes.

Apache Pig was designed by Yahoo as it is easy to learn and work with. So, Pig makes Hadoop quite easy. Apache Pig was developed because MapReduce programming was getting quite difficult and many MapReduce users are not comfortable with declarative languages. Now, Pig is an open-source project under Apache.

1. FEATURES OF PIG

Let’s look at some of the features of Pig.

It has a rich set of operators such as join, sort, etc.
It is easy to program as it is similar to SQL.
The tasks in Apache Pig have been converted into MapReduce jobs automatically. The programmers need to focus only on the semantics of the language and not on MapReduce.
Own functions can be created using Pig.
Functions in other programming languages such as java can be embedded in Pig Latin scripts.
Apache Pig can handle
all kinds of data such as structured, unstructured, and semi-structured data and
stores the result in HDFS.

2. PIG VS MAPREDUCE

Let’s see the difference between Pig and MapReduce.

Pig has several advantages over MapReduce.

Apache Pig is a data flow language. It means that it allows users to describe how data from one or more inputs should be read, processed, and then stored to one or more outputs in parallel. While MapReduce on the other hand is a programming style.

Apache Pig is a high-level language while MapReduce is a compiled java code.

The syntax for Pig for performing join and multiple files is very intuitive and quite simple like SQL. MapReduce code
becomes complex if you want to write joining operations.

The learning curve for Apache Pig is very small. Expertise in Java and MapReduce libraries is a must
to run MapReduce code.

Apache Pig scripts can do the equivalent of multiple lines of MapReduce code and MapReduce code takes more lines of codes to perform the same operations.

Apache Pig is easy to debug and test while MapReduce programs take a lot of time for coding, testing, etc. Pig Latin is less costly than MapReduce.

3. PIG ARCHITECTURE

Now let’s see Pig architecture.

Source

Pig sits on top of Hadoop. Pig scripts can be run either on Grunt shell or on the Pig server. Execution engine of Pig the passes optimizes and compiles the script and finally converts into MapReduce jobs. It uses HDFS to store intermediate data between MapReduce jobs and then writes its output to HDFS.

4. PIG EXECUTION OPTIONS

Apache Pig can run two run modes. Both of them produce the same results.

4.1. Local mode

Command on gateways

pig -x local

4.2. Hadoop mode

Command on gateways

pig -exectype mapreduce

Apache Pig can be run in three ways in the above two modes.

Interactive mode / Grunt shell: enter Pig commands manually by using the grunt shell
Batch mode / Script file: Place Pig commands in a script file and run the script
Embedded program/UDF: embed Pig commands in java and run the scripts

5. PIG GRUNT SHELL COMMANDS

Grunt shells can be used to write Pig Latin scripts. The shell commands can be invoked by using fs and sh commands. Let’s see some basic
Pig commands.

5.1. fs command

fs command lets you run HDFS commands from Pig

5.1.1 To list all directories in HDFS

grunt> fs -ls;

Now, all the files in HDFS will be displayed.

5.1.2. To create a new directory mydir in HDFS

grunt> fs -mkdir mydir/;

The above command will create a new directory called mydir in HDFS.

5.1.3. To remove a directory

grunt> fs -rmdir mydir;

The above command will remove the created directory mydir.

5.1.4. To copy a file to HDFS

grant> fs -put sales.txt sales/;

Here, the file named sales.txt is the source file that will be copied to the destination directory in HDFS i.e. sales.

5.1.5. To quit from grunt shell

grunt> quit;

The above command will exit the grunt shell.

5.2. sh command

sh command lets you run Unix statement from Pig

5.2.1. To display the current date

grunt> sh date;

This command will show the current date.

5.2.2. To lists local files

grunt> sh ls;

This command will display all the files in the local system.

5.2.3. To execute Pig Latin from grunt shell

grunt> run salesreport.pig;

The above command will execute a Pig Latin script file “salesreport.pig” from the grunt shell.

5.2.4. To execute Pig Latin from Unix prompt

$pig salesreport.pig;

The above command will execute a Pig Latin script file “salesreport.pig” from Unix prompt.

6. PIG DATA TYPES

Pig Latin consists of the following datatypes.

6.1. Data Atom

It is a single value. It can be a string or a number. They are of scalar types such as int, float, double, etc.

For example, “john”, 9.0

6.2. Tuple

A tuple is similar to a record with a sequence of fields. It can be of any data type.

For example, (‘john’, ‘james’) is a tuple.

6.3. Data bag

It consists of a collection of tuples which is equivalent to a “table” in SQL. The tuples are non-unique and can have an arbitrary number of fields, each can be of any type.

For example, {(‘john’, ‘James), (‘king’, ‘mark’)} is a data bag which is equivalent to the below table in SQL.

john	James
king	mark

6.4. Data map

This data type
contains a collection of key-value pairs. Here, the key must be a chararray and unique. The values can be of any type.

For example, [name#(‘john’, ‘james’), age#22] is a data map where name, age are keys and (‘john, ‘james’), 22 are values.

7. PIG OPERATORS

Below is the contents of student.txt file.

John,23,Hyderabad
James,45,Hyderabad
Sam,33,Chennai
,56,Delhi
,43,Mumbai

7.1. LOAD

It loads data from the given file system.

A = LOAD 'student.txt' AS (name: chararray, age: int, city: chararray);

The data from the student file with column names as ‘name’, ‘age’, ‘city’ will be loaded into a variable A.

7.2. DUMP

DUMP operator is used to displaying the contents of a relation. Here, the contents of A will be displayed.

DUMP A
//results
(John,23,Hyderabad)
(James,45,Hyderabad)
(Sam,33,Chennai)
(,56,Delhi)
(,43,Mumbai)

7.3. STORE

The store function saves the results to the file system.

STORE A into ‘myoutput’ using PigStorage(‘*’);

Here, the data present in A will be stored into myoutput separated by ‘*’.

DUMP myoutput; 
//results
John*23*Hyderabad
James*45*Hyderabad
Sam*33*Chennai
*56*Delhi
*43*Mumbai

7.4. FILTER

B = FILTER A by name is not null;

The FILTER operator will filter a table with some conditions. Here, the name is the column in A. Non-empty values in the name will be stored in variable B.

DUMP B;
//results
(John,23,Hyderabad)
(James,45,Hyderabad)
(Sam,33,Chennai)

7.5. FOREACH GENERATE

C = FOREACH A GENERATE name, city;

FOREACH operator is used to accessing individual records. Here, the rows present in name and city will be fetched from A and stored into C.

DUMP C
//results
(John,Hyderabad)
(James,Hyderabad)
(Sam,Chennai)
(,Delhi)
(,Mumbai)

8. PIG LATIN SCRIPT EXAMPLE

We have a people file that has employee id, name, and hours as fields.

001,Rajiv,21
002,siddarth,12
003,Rajesh,22

First, load this data into a variable employee. Filter it by hours less than 20 and store in parttime. Order parttime by descending order and store it in another file called part_time. Display the contents.

The script will be

employee = Load ‘people’ as (empid, name, hours);
parttime = FILTER employee BY Hours < 20;
sorted = ORDER parttime by hours DESC;
STORE sorted INTO ‘part_time’;
DUMP sorted;
DESCRIBE sorted;
//results
(003,Rajesh,22)
(001,Rajiv,21)