Apache Hadoop

A course for undergraduates

Homework Labs
with
Professors Notes

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

Table of Contents

General Notes on Homework Labs ................................................................................................... 3

Lecture 1 Lab: Using HDFS .................................................................................................................. 5
Lecture 2 Lab: Running a MapReduce Job ................................................................................... 11
Lecture 3 Lab: Writing a MapReduce Java Program ................................................................ 16
Lecture 3 Lab: More Practice with MapReduce Java Programs ........................................... 24
Lecture 3 Lab: Writing a MapReduce Streaming Program .................................................... 26
Lecture 3 Lab: Writing Unit Tests with the MRUnit Framework ......................................... 29
Lecture 4 Lab: Using ToolRunner and Passing Parameters .................................................. 30
Lecture 4 Lab: Using a Combiner ................................................................................................... 33
Lecture 5 Lab: Testing with LocalJobRunner ............................................................................. 34
Lecture 5 Lab: Logging ....................................................................................................................... 37
Lecture 5 Lab: Using Counters and a Map-Only Job ................................................................. 40
Lecture 6 Lab: Writing a Partitioner ............................................................................................. 42
Lecture 6 Lab: Implementing a Custom WritableComparable ............................................. 45
Lecture 6 Lab: Using SequenceFiles and File Compression .................................................. 47
Lecture 7 Lab: Creating an Inverted Index ................................................................................. 51
Lecture 7 Lab: Calculating Word Co-Occurrence ...................................................................... 55
Lecture 8 Lab: Importing Data with Sqoop ................................................................................. 57
Lecture 8 Lab: Running an Oozie Workflow ............................................................................... 60
Lecture 8 Bonus Lab: Exploring a Secondary Sort Example ................................................. 62
Notes for Upcoming Labs .................................................................................................................. 66
Lecture 9 Lab: Data Ingest With Hadoop Tools ......................................................................... 67
Lecture 9 Lab: Using Pig for ETL Processing .............................................................................. 71
Lecture 9 Lab: Analyzing Ad Campaign Data with Pig ............................................................ 79
Lecture 10 Lab: Analyzing Disparate Data Sets with Pig ....................................................... 87
Lecture 10 Lab: Extending Pig with Streaming and UDFs ...................................................... 93
Lecture 11 Lab: Running Hive Queries from the Shell, Scripts, and Hue .......................... 98
Lecture 11 Lab: Data Management with Hive .......................................................................... 103
Lecture 12 Lab: Gaining Insight with Sentiment Analysis ................................................... 110
Lecture 12 Lab: Data Transformation with Hive .................................................................... 114
Lecture 13 Lab: Interactive Analysis with Impala .................................................................. 123
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

General Notes on Homework Labs

Students complete homework for this course using the student version of the training
Virtual Machine (VM). Cloudera supplies a second VM, the professors VM, in addition to the
student VM. The professors VM comes complete with solutions to the homework labs.
The professors VM contains additional project subdirectories with hints and solutions.
Subdirectories named src/hints and src/solution provide hints (partial solutions) and full
solutions, respectively. The student VM will have src/stubs directories only, no hints or
solutions directories. Full solutions can be distributed to students after homework has been
submitted. In some cases, a lab may require that the previous lab(s) ran successfully,
ensuring that the VM is in the required state. Providing students with the solution to the
previous lab, and having them run the solution, will bring the VM to the required state. This
should be completed prior to running code for the new lab.
Except for the presence of solutions in the professor VM, the student and professor versions
of the training VM are the same. Both VMs run the CentOS 6.3 Linux distribution and come
configured with CDH (Clouderas Distribution, including Apache Hadoop) installed in
pseudo-distributed mode. In addition to core Hadoop, the Hadoop ecosystem tools
necessary to complete the homework labs are also installed (e.g. Pig, Hive, Flume, etc.). Perl,
Python, PHP, and Ruby are installed as well.
Hadoop pseudo-distributed mode is a method of running Hadoop whereby all Hadoop
daemons run on the same machine. It is, essentially, a cluster consisting of a single machine.
It works just like a larger Hadoop cluster, the key difference (apart from speed, of course!) is
that the block replication factor is set to one, since there is only a single DataNode available.
Note: Homework labs are grouped into individual files by lecture number for easy posting of
assignments. The same labs appear in this document, but with references to hints and
solutions where applicable. The students homework labs will reference a stubs
subdirectory, not hints or solutions. Students will typically complete their coding in the
stubs subdirectories.

Getting Started
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

1. The VM is set to automatically log in as the user training. Should you log out at any
time, you can log back in as the user training with the password training.

Working with the Virtual Machine

1. Should you need it, the root password is training. You may be prompted for this if,
for example, you want to change the keyboard layout. In general, you should not need
this password since the training user has unlimited sudo privileges.
2. In some command-line steps in the labs, you will see lines like this:
$ hadoop fs -put shakespeare

/user/training/shakespeare
The dollar sign ($) at the beginning of each line indicates the Linux shell prompt. The
actual prompt will include additional information (e.g., [training@localhost
workspace]$ ) but this is omitted from these instructions for brevity.
The backslash (\) at the end of the first line signifies that the command is not completed,
and continues on the next line. You can enter the code exactly as shown (on two lines),
or you can enter it on a single line. If you do the latter, you should not type in the
backslash.
3. Although many students are comfortable using UNIX text editors like vi or emacs, some
might prefer a graphical text editor. To invoke the graphical editor from the command
line, type gedit followed by the path of the file you wish to edit. Appending & to the
command allows you to type additional commands while the editor is still open. Here is
an example of how to edit a file named myfile.txt:
$ gedit myfile.txt &

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

Lecture 1 Lab: Using HDFS

Files Used in This Exercise:
Data files (local)
~/training_materials/developer/data/shakespeare.tar.gz
~/training_materials/developer/data/access_log.gz

In this lab you will begin to get acquainted with the Hadoop tools. You will manipulate
files in HDFS, the Hadoop Distributed File System.

Set Up Your Environment

1. Before starting the labs, run the course setup script in a terminal window:
$ ~/scripts/developer/training_setup_dev.sh

Hadoop
Hadoop is already installed, configured, and running on your virtual machine.
Most of your interaction with the system will be through a command-line wrapper called
hadoop. If you run this program with no arguments, it prints a help message. To try this,
run the following command in a terminal window:
$ hadoop

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.


The hadoop command is subdivided into several subsystems. For example, there is a
subsystem for working with files in HDFS and another for launching and managing
MapReduce processing jobs.

Step 1: Exploring HDFS

The subsystem associated with HDFS in the Hadoop wrapper program is called FsShell.
This subsystem can be invoked with the command hadoop fs.
1. Open a terminal window (if one is not already open) by double-clicking the Terminal
icon on the desktop.
2. In the terminal window, enter:
$ hadoop fs
You see a help message describing all the commands associated with the FsShell
subsystem.
3. Enter:
$ hadoop fs -ls /
This shows you the contents of the root directory in HDFS. There will be multiple
entries, one of which is /user. Individual users have a home directory under this
directory, named after their username; your username in this course is training,
therefore your home directory is /user/training.
4. Try viewing the contents of the /user directory by running:
$ hadoop fs -ls /user
You will see your home directory in the directory listing.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

5. List the contents of your home directory by running:

$ hadoop fs -ls /user/training
There are no files yet, so the command silently exits. This is different from running
hadoop fs -ls /foo, which refers to a directory that doesnt exist. In this case, an
error message would be displayed.
Note that the directory structure in HDFS has nothing to do with the directory structure
of the local filesystem; they are completely separate namespaces.

Step 2: Uploading Files

Besides browsing the existing filesystem, another important thing you can do with
FsShell is to upload new data into HDFS.
1. Change directories to the local filesystem directory containing the sample data we will
be using in the homework labs.
$ cd ~/training_materials/developer/data
If you perform a regular Linux ls command in this directory, you will see a few files,
including two named shakespeare.tar.gz and
shakespeare-stream.tar.gz. Both of these contain the complete works of
Shakespeare in text format, but with different formats and organizations. For now we
will work with shakespeare.tar.gz.
2. Unzip shakespeare.tar.gz by running:
$ tar zxvf shakespeare.tar.gz
This creates a directory named shakespeare/ containing several files on your local
filesystem.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

3. Insert this directory into HDFS:

$ hadoop fs -put shakespeare /user/training/shakespeare
This copies the local shakespeare directory and its contents into a remote, HDFS
directory named /user/training/shakespeare.
4. List the contents of your HDFS home directory now:
$ hadoop fs -ls /user/training
You should see an entry for the shakespeare directory.
5. Now try the same fs -ls command but without a path argument:
$ hadoop fs -ls
You should see the same results. If you dont pass a directory name to the -ls
command, it assumes you mean your home directory, i.e. /user/training.

Relative paths
If you pass any relative (non-absolute) paths to FsShell commands (or use relative
paths in MapReduce programs), they are considered relative to your home directory.

6. We will also need a sample web server log file, which we will put into HDFS for use in
future labs. This file is currently compressed using GZip. Rather than extract the file to
the local disk and then upload it, we will extract and upload in one step. First, create a
directory in HDFS in which to store it:
$ hadoop fs -mkdir weblog
7. Now, extract and upload the file in one step. The -c option to gunzip uncompresses to
standard output, and the dash (-) in the hadoop fs -put command takes whatever
is being sent to its standard input and places that data in HDFS.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

$ gunzip -c access_log.gz \
| hadoop fs -put - weblog/access_log
8. Run the hadoop fs -ls command to verify that the log file is in your HDFS home
directory.
9. The access log file is quite large around 500 MB. Create a smaller version of this file,
consisting only of its first 5000 lines, and store the smaller version in HDFS. You can use
the smaller version for testing in subsequent labs.
$ hadoop fs -mkdir testlog
$ gunzip -c access_log.gz | head -n 5000 \
| hadoop fs -put - testlog/test_access_log

Step 3: Viewing and Manipulating Files

Now lets view some of the data you just copied into HDFS.
1. Enter:
$ hadoop fs -ls shakespeare
This lists the contents of the /user/training/shakespeare HDFS directory,
which consists of the files comedies, glossary, histories, poems, and
tragedies.
2. The glossary file included in the compressed file you began with is not strictly a work
of Shakespeare, so lets remove it:
$ hadoop fs -rm shakespeare/glossary
Note that you could leave this file in place if you so wished. If you did, then it would be
included in subsequent computations across the works of Shakespeare, and would skew
your results slightly. As with many real-world big data problems, you make trade-offs
between the labor to purify your input data and the precision of your results.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

3. Enter:
$ hadoop fs -cat shakespeare/histories | tail -n 50
This prints the last 50 lines of Henry IV, Part 1 to your terminal. This command is handy
for viewing the output of MapReduce programs. Very often, an individual output file of a
MapReduce program is very large, making it inconvenient to view the entire file in the
terminal. For this reason, its often a good idea to pipe the output of the fs -cat
command into head, tail, more, or less.
4. To download a file to work with on the local filesystem use the fs -get command.
This command takes two arguments: an HDFS path and a local path. It copies the HDFS
contents into the local filesystem:
$ hadoop fs -get shakespeare/poems ~/shakepoems.txt
$ less ~/shakepoems.txt

Other Commands
There are several other operations available with the hadoop fs command to perform
most common filesystem manipulations: mv, cp, mkdir, etc.
1. Enter:
$ hadoop fs
This displays a brief usage report of the commands available within FsShell. Try
playing around with a few of these commands if you like.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

10

Lecture 2 Lab: Running a MapReduce

Job
Files and Directories Used in this Exercise
Source directory: ~/workspace/wordcount/src/solution
Files:
WordCount.java: A simple MapReduce driver class.
WordMapper.java: A mapper class for the job.
SumReducer.java: A reducer class for the job.
wc.jar: The compiled, assembled WordCount program

In this lab you will compile Java files, create a JAR, and run MapReduce jobs.
In addition to manipulating files in HDFS, the wrapper program hadoop is used to launch
MapReduce jobs. The code for a job is contained in a compiled JAR file. Hadoop loads the JAR
into HDFS and distributes it to the worker nodes, where the individual tasks of the
MapReduce job are executed.
One simple example of a MapReduce job is to count the number of occurrences of each word
in a file or set of files. In this lab you will compile and submit a MapReduce job to count the
number of occurrences of every word in the works of Shakespeare.

Compiling and Submitting a MapReduce Job

1. In a terminal window, change to the lab source directory, and list the contents:
$ cd ~/workspace/wordcount/src
$ ls
List the files in the solution package directory:
$ ls solution
The package contains the following Java files:
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

11

WordCount.java: A simple MapReduce driver class.

WordMapper.java: A mapper class for the job.
SumReducer.java: A reducer class for the job.
Examine these files if you wish, but do not change them. Remain in this directory while
you execute the following commands.
2. Before compiling, examine the classpath Hadoop is configured to use:
$ hadoop classpath

This shows lists the locations where the Hadoop core API classes are installed.
3. Compile the three Java classes:
$ javac -classpath `hadoop classpath` solution/*.java
Note: in the command above, the quotes around hadoop classpath are
backquotes. This runs the hadoop classpath command and uses its output as
part of the javac command.
The compiled (.class) files are placed in the solution directory.
4. Collect your compiled Java files into a JAR file:
$ jar cvf wc.jar solution/*.class
5. Submit a MapReduce job to Hadoop using your JAR file to count the occurrences of each
word in Shakespeare:
$ hadoop jar wc.jar solution.WordCount \
shakespeare wordcounts
This hadoop jar command names the JAR file to use (wc.jar), the class whose main
method should be invoked (solution.WordCount), and the HDFS input and output
directories to use for the MapReduce job.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

12

Your job reads all the files in your HDFS shakespeare directory, and places its output
in a new HDFS directory called wordcounts.
6. Try running this same command again without any change:
$ hadoop jar wc.jar solution.WordCount \
shakespeare wordcounts
Your job halts right away with an exception, because Hadoop automatically fails if your
job tries to write its output into an existing directory. This is by design; since the result
of a MapReduce job may be expensive to reproduce, Hadoop prevents you from
accidentally overwriting previously existing files.
7. Review the result of your MapReduce job:
$ hadoop fs -ls wordcounts
This lists the output files for your job. (Your job ran with only one Reducer, so there
should be one file, named part-r-00000, along with a _SUCCESS file and a _logs
directory.)
8. View the contents of the output for your job:
$ hadoop fs -cat wordcounts/part-r-00000 | less
You can page through a few screens to see words and their frequencies in the works of
Shakespeare. (The spacebar will scroll the output by one screen; the letter 'q' will quit
the less utility.) Note that you could have specified wordcounts/* just as well in this
command.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

13

Wildcards in HDFS file paths

Take care when using wildcards (e.g. *) when specifying HFDS filenames; because of
how Linux works, the shell will attempt to expand the wildcard before invoking hadoop,
and then pass incorrect references to local files instead of HDFS files. You can prevent
this by enclosing the wildcarded HDFS filenames in single quotes, e.g. hadoop fs
cat 'wordcounts/*'

9. Try running the WordCount job against a single file:

$ hadoop jar wc.jar solution.WordCount \
shakespeare/poems pwords
When the job completes, inspect the contents of the pwords HDFS directory.
10. Clean up the output files produced by your job runs:
$ hadoop fs -rm -r wordcounts pwords

Stopping MapReduce Jobs

It is important to be able to stop jobs that are already running. This is useful if, for example,
you accidentally introduced an infinite loop into your Mapper. An important point to
remember is that pressing ^C to kill the current process (which is displaying the
MapReduce job's progress) does not actually stop the job itself.
A MapReduce job, once submitted to Hadoop, runs independently of the initiating process,
so losing the connection to the initiating process does not kill the job. Instead, you need to
tell the Hadoop JobTracker to stop the job.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

14

1. Start another word count job like you did in the previous section:
$ hadoop jar wc.jar solution.WordCount shakespeare \
count2
2. While this job is running, open another terminal window and enter:
$ mapred job -list
This lists the job ids of all running jobs. A job id looks something like:
job_200902131742_0002
3. Copy the job id, and then kill the running job by entering:
$ mapred job -kill jobid
The JobTracker kills the job, and the program running in the original terminal
completes.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

15

Lecture 3 Lab: Writing a MapReduce

Java Program
Projects and Directories Used in this Exercise
Eclipse project: averagewordlength
Java files:
AverageReducer.java (Reducer)
LetterMapper.java (Mapper)
AvgWordLength.java (driver)
Test data (HDFS):
shakespeare
Exercise directory: ~/workspace/averagewordlength

In this lab, you will write a MapReduce job that reads any text input and computes the
average length of all words that start with each character.
For any text input, the job should report the average length of words that begin with a, b,
and so forth. For example, for input:
No now is definitely not the time
The output would be:
N

2.0

3.0

10.0

2.0

3.5

(For the initial solution, your program should be case-sensitive as shown in this example.)

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

16

The Algorithm
The algorithm for this program is a simple one-pass MapReduce program:
The Mapper
The Mapper receives a line of text for each input value. (Ignore the input key.) For each
word in the line, emit the first letter of the word as a key, and the length of the word as a
value. For example, for input value:
No now is definitely not the time
Your Mapper should emit:
N

10

The Reducer
Thanks to the shuffle and sort phase built in to MapReduce, the Reducer receives the keys in
sorted order, and all the values for one key are grouped together. So, for the Mapper output
above, the Reducer receives this:

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

17

(2)

(10)

(2)

(3,3)

(3,4)

The Reducer output should be:

N

2.0

10.0

2.0

3.0

3.5

Step 1: Start Eclipse

There is one Eclipse project for each of the labs that use Java. Using Eclipse will speed up
your development time.
1. Be sure you have run the course setup script as instructed earlier in the General Notes
section. That script sets up the lab workspace and copies in the Eclipse projects you will
use for the remainder of the course.
2. Start Eclipse using the icon on your VM desktop. The projects for this course will appear
in the Project Explorer on the left.

Step 2: Write the Program in Java

There are stub files for each of the Java classes for this lab: LetterMapper.java (the
Mapper), AverageReducer.java (the Reducer), and AvgWordLength.java (the
driver).
If you are using Eclipse, open the stub files (located in the src/stubs package) in the
averagewordlength project. If you prefer to work in the shell, the files are in
~/workspace/averagewordlength/src/stubs.
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

18

You may wish to refer back to the wordcount example (in the wordcount project in
Eclipse or in ~/workspace/wordcount) as a starting point for your Java code. Here are a
few details to help you begin your Java programming:
3. Define the driver
This class should configure and submit your basic job. Among the basic steps here,
configure the job with the Mapper class and the Reducer class you will write, and the
data types of the intermediate and final keys.
4. Define the Mapper
Note these simple string operations in Java:
str.substring(0, 1)

// String : first letter of str

str.length()

// int : length of str

5. Define the Reducer

In a single invocation the reduce() method receives a string containing one letter (the
key) along with an iterable collection of integers (the values), and should emit a single
key-value pair: the letter and the average of the integers.
6. Compile your classes and assemble the jar file
To compile and jar, you may either use the command line javac command as you did
earlier in the Running a MapReduce Job lab, or follow the steps below (Using Eclipse
to Compile Your Solution) to use Eclipse.

Step 3: Use Eclipse to Compile Your Solution

Follow these steps to use Eclipse to complete this lab.
Note: These same steps will be used for all subsequent labs. The instructions will not
be repeated each time, so take note of the steps.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

19

1. Verify that your Java code does not have any compiler errors or warnings.
The Eclipse software in your VM is pre-configured to compile code automatically
without performing any explicit steps. Compile errors and warnings appear as red and
yellow icons to the left of the code.

A red X indicates a compiler error

2. In the Package Explorer, open the Eclipse project for the current lab (i.e.
averagewordlength). Right-click the default package under the src entry and select
Export.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

20


3. Select Java > JAR file from the Export dialog box, then click Next.

4. Specify a location for the JAR file. You can place your JAR files wherever you like, e.g.:

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

21

Note: For more information about using Eclipse, see the Eclipse Reference in
Homework_EclipseRef.docx.

Step 3: Test your program

1. In a terminal window, change to the directory where you placed your JAR file. Run the
hadoop jar command as you did previously in the Running a MapReduce Job lab.
$ hadoop jar avgwordlength.jar stubs.AvgWordLength \
shakespeare wordlengths
2. List the results:
$ hadoop fs -ls wordlengths
A single reducer output file should be listed.
3. Review the results:
$ hadoop fs -cat wordlengths/*
The file should list all the numbers and letters in the data set, and the average length of
the words starting with them, e.g.:
1
2
3
4
5
6
7
8
9
A
B
C

1.02
1.0588235294117647
1.0
1.5
1.5
1.5
1.0
1.5
1.0
3.891394576646375
5.139302507836991
6.629694233531706

This example uses the entire Shakespeare dataset for your input; you can also try it with
just one of the files in the dataset, or with your own test data.
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

22

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

23

Lecture 3 Lab: More Practice with

MapReduce Java Programs
Files and Directories Used in this Exercise
Eclipse project: log_file_analysis
Java files:
SumReducer.java the Reducer
LogFileMapper.java the Mapper
ProcessLogs.java the driver class
Test data (HDFS):
weblog (full version)
testlog (test sample set)
Exercise directory: ~/workspace/log_file_analysis

In this lab, you will analyze a log file from a web server to count the number of hits
made from each unique IP address.
Your task is to count the number of hits made from each IP address in the sample
(anonymized) web server log file that you uploaded to the /user/training/weblog
directory in HDFS when you completed the Using HDFS lab.
In the log_file_analysis directory, you will find stubs for the Mapper and Driver.
1. Using the stub files in the log_file_analysis project directory, write Mapper and
Driver code to count the number of hits made from each IP address in the access log file.
Your final result should be a file in HDFS containing each IP address, and the count of
log hits from that address. Note: The Reducer for this lab performs the exact same
function as the one in the WordCount program you ran earlier. You can reuse that
code or you can write your own if you prefer.
2. Build your application jar file following the steps in the previous lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

24

3. Test your code using the sample log data in the /user/training/weblog directory.
Note: You may wish to test your code against the smaller version of the access log you
created in a prior lab (located in the /user/training/testlog HDFS directory)
before you run your code against the full log which can be quite time consuming.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

25

Lecture 3 Lab: Writing a MapReduce

Streaming Program
Files and Directories Used in this Exercise
Project directory: ~/workspace/averagewordlength
Test data (HDFS):
shakespeare

In this lab you will repeat the same task as in the previous lab: writing a program to
calculate average word lengths for letters. However, you will write this as a streaming
program using a scripting language of your choice rather than using Java.
Your virtual machine has Perl, Python, PHP, and Ruby installed, so you can choose any of
theseor even shell scriptingto develop a Streaming solution.
For your Hadoop Streaming program you will not use Eclipse. Launch a text editor to write
your Mapper script and your Reducer script. Here are some notes about solving the problem
in Hadoop Streaming:
1. The Mapper Script
The Mapper will receive lines of text on stdin. Find the words in the lines to produce
the intermediate output, and emit intermediate (key, value) pairs by writing strings of
the form:
key <tab> value <newline>
These strings should be written to stdout.
2. The Reducer Script
For the reducer, multiple values with the same key are sent to your script on stdin as
successive lines of input. Each line contains a key, a tab, a value, and a newline. All lines
with the same key are sent one after another, possibly followed by lines with a different

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

26

key, until the reducing input is complete. For example, the reduce script may receive the
following:
t

For this input, emit the following to stdout:

t

3.5

5.0

Observe that the reducer receives a key with each input line, and must notice when
the key changes on a subsequent line (or when the input is finished) to know when the
values for a given key have been exhausted. This is different than the Java version you
worked on in the previous lab.
3. Run the streaming program:
$ hadoop jar /usr/lib/hadoop-0.20-mapreduce/\
contrib/streaming/hadoop-streaming*.jar \
-input inputDir -output outputDir \
-file pathToMapScript -file pathToReduceScript \
-mapper mapBasename -reducer reduceBasename
(Remember, you may need to delete any previous output before running your program
by issuing: hadoop fs -rm -r dataToDelete.)
4. Review the output in the HDFS directory you specified (outputDir).

Professors Note~
The Perl example is in: ~/workspace/wordcount/perl_solution


Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

27

Professors Note~

Solution in Python
You can find a working solution to this lab written in Python in the directory
~/workspace/averagewordlength/python_sample_solution.
To run the solution, change directory to ~/workspace/averagewordlength and run
this command:
$ hadoop jar /usr/lib/hadoop-0.20-mapreduce\
/contrib/streaming/hadoop-streaming*.jar \
-input shakespeare -output avgwordstreaming \
-file python_sample_solution/mapper.py \
-file python_sample_solution/reducer.py \
-mapper mapper.py -reducer reducer.py

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

28

Lecture 3 Lab: Writing Unit Tests with

the MRUnit Framework
Projects Used in this Exercise
Eclipse project: mrunit
Java files:
SumReducer.java (Reducer from WordCount)
WordMapper.java (Mapper from WordCount)
TestWordCount.java (Test Driver)

In this Exercise, you will write Unit Tests for the WordCount code.
1. Launch Eclipse (if necessary) and expand the mrunit folder.
2. Examine the TestWordCount.java file in the mrunit project stubs package.
Notice that three tests have been created, one each for the Mapper, Reducer, and the
entire MapReduce flow. Currently, all three tests simply fail.
3. Run the tests by right-clicking on TestWordCount.java in the Package Explorer
panel and choosing Run As > JUnit Test.
4. Observe the failure. Results in the JUnit tab (next to the Package Explorer tab) should
indicate that three tests ran with three failures.
5. Now implement the three tests.
6. Run the tests again. Results in the JUnit tab should indicate that three tests ran with no
failures.
7. When you are done, close the JUnit tab.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

29

Lecture 4 Lab: Using ToolRunner and

Passing Parameters
Files and Directories Used in this Exercise
Eclipse project: toolrunner
Java files:
AverageReducer.java (Reducer from AverageWordLength)
LetterMapper.java (Mapper from AverageWordLength)
AvgWordLength.java (driver from AverageWordLength)
Exercise directory: ~/workspace/toolrunner

In this Exercise, you will implement a driver using ToolRunner.

Follow the steps below to start with the Average Word Length program you wrote in an
earlier lab, and modify the driver to use ToolRunner. Then modify the Mapper to reference a
Boolean parameter called caseSensitive; if true, the mapper should treat upper and
lower case letters as different; if false or unset, all letters should be converted to lower case.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

30

Modify the Average Word Length Driver to use Toolrunner

1. Copy the Reducer, Mapper and driver code you completed in the Writing Java
MapReduce Programs lab earlier, in the averagewordlength project.

Copying Source Files

You can use Eclipse to copy a Java source file from one project or package to another by
right-clicking on the file and selecting Copy, then right-clicking the new package and
selecting Paste. If the packages have different names (e.g. if you copy from
averagewordlength.solution to toolrunner.stubs), Eclipse will automatically
change the package directive at the top of the file. If you copy the file using a file
browser or the shell, you will have to do that manually.

2. Modify the AvgWordLength driver to use ToolRunner. Refer to the slides for details.
a. Implement the run method
b. Modify main to call run
3. Jar your solution and test it before continuing; it should continue to function exactly as
it did before. Refer to the Writing a Java MapReduce Program lab for how to assemble
and test if you need a reminder.

Modify the Mapper to use a configuration parameter

4. Modify the LetterMapper class to
a. Override the setup method to get the value of a configuration parameter
called caseSensitive, and use it to set a member variable indicating
whether to do case sensitive or case insensitive processing.
b. In the map method, choose whether to do case sensitive processing (leave the
letters as-is), or insensitive processing (convert all letters to lower-case)
based on that variable.

Pass a parameter programmatically

Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

31

5. Modify the drivers run method to set a Boolean configuration parameter called
caseSensitive. (Hint: Use the Configuration.setBoolean method.)
6. Test your code twice, once passing false and once passing true. When set to true,
your final output should have both upper and lower case letters; when false, it should
have only lower case letters.
Hint: Remember to rebuild your Jar file to test changes to your code.

Pass a parameter as a runtime parameter

7. Comment out the code that sets the parameter programmatically. (Eclipse hint: Select
the code to comment and then select Source > Toggle Comment). Test again, this time
passing the parameter value using D on the Hadoop command line, e.g.:
$ hadoop jar toolrunner.jar stubs.AvgWordLength \
-DcaseSensitive=true shakespeare toolrunnerout
8. Test passing both true and false to confirm the parameter works correctly.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

32

Lecture 4 Lab: Using a Combiner

Files and Directories Used in this Exercise
Eclipse project: combiner
Java files:
WordCountDriver.java (Driver from WordCount)
WordMapper.java (Mapper from WordCount)
SumReducer.java (Reducer from WordCount)
Exercise directory: ~/workspace/combiner

In this lab, you will add a Combiner to the WordCount program to reduce the amount
of intermediate data sent from the Mapper to the Reducer.
Because summing is associative and commutative, the same class can be used for both the
Reducer and the Combiner.

Implement a Combiner
1. Copy WordMapper.java and SumReducer.java from the wordcount project to
the combiner project.
2. Modify the WordCountDriver.java code to add a Combiner for the WordCount
program.
3. Assemble and test your solution. (The output should remain identical to the WordCount
application without a combiner.)

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

33

Lecture 5 Lab: Testing with

LocalJobRunner
Files and Directories Used in this Exercise
Eclipse project: toolrunner
Test data (local):
~/training_materials/developer/data/shakespeare
Exercise directory: ~/workspace/toolrunner

In this lab, you will practice running a job locally for debugging and testing purposes.
In the Using ToolRunner and Passing Parameters lab, you modified the Average Word
Length program to use ToolRunner. This makes it simple to set job configuration properties
on the command line.

Run the Average Word Length program using

LocalJobRunner on the command line
1. Run the Average Word Length program again. Specify jt=local to run the job locally
instead of submitting to the cluster, and fs=file:/// to use the local file system
instead of HDFS. Your input and output files should refer to local files rather than HDFS
files.
Note: Use the program you completed in the ToolRunner lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

34

$ hadoop jar toolrunner.jar stubs.AvgWordLength \

-fs=file:/// -jt=local \
~/training_materials/developer/data/shakespeare \
localout
2. Review the job output in the local output folder you specified.

Optional: Run the Average Word Length program using

LocalJobRunner in Eclipse
1. In Eclipse, locate the toolrunner project in the Package Explorer. Open the stubs
package.
2. Right click on the driver class (AvgWordLength) and select Run As > Run
Configurations
3. Ensure that Java Application is selected in the run types listed in the left pane.
4. In the Run Configuration dialog, click the New launch configuration button:

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

35

5. On the Main tab, confirm that the Project and Main class are set correctly for your
project, e.g. Project: toolrunner and Main class: stubs.AvgWordLength
6. Select the Arguments tab and enter the input and output folders. (These are local, not
HDFS folders, and are relative to the run configurations working folder, which by default
is the project folder in the Eclipse workspace: e.g. ~/workspace/toolrunner.)

7. Click the Run button. The program will run locally with the output displayed in the
Eclipse console window.

8. Review the job output in the local output folder you specified.

Note: You can re-run any previous configurations using the Run or Debug history buttons on
the Eclipse tool bar.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

36

Lecture 5 Lab: Logging

Files and Directories Used in this Exercise
Eclipse project: logging
Java files:
AverageReducer.java (Reducer from ToolRunner)
LetterMapper.java (Mapper from ToolRunner)
AvgWordLength.java (driver from ToolRunner)
Test data (HDFS):
shakespeare
Exercise directory: ~/workspace/logging

In this lab, you will practice using log4j with MapReduce.

Modify the Average Word Length program you built in the Using ToolRunner and Passing
Parameters lab so that the Mapper logs a debug message indicating whether it is comparing
with or without case sensitivity.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

37

Enable Mapper Logging for the Job

1. Before adding additional logging messages, try re-running the toolrunner lab solution
with Mapper debug logging enabled by adding
-Dmapred.map.child.log.level=DEBUG
to the command line. E.g.
$ hadoop jar toolrunner.jar stubs.AvgWordLength \
-Dmapred.map.child.log.level=DEBUG shakespeare outdir
2. Take note of the Job ID in the terminal window or by using the maprep job command.
3. When the job is complete, view the logs. In a browser on your VM, visit the Job Tracker
UI: http://localhost:50030/jobtracker.jsp. Find the job you just ran in the
Completed Jobs list and click its Job ID. E.g.:

4. In the task summary, click map to view the map tasks.

5. In the list of tasks, click on the map task to view the details of that task.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

38

6. Under Task Logs, click All. The logs should include both INFO and DEBUG messages.
E.g.:

Add Debug Logging Output to the Mapper

7. Copy the code from the toolrunner project to the logging project stubs package.
(Use your solution from the ToolRunner lab.)
8. Use log4j to output a debug log message indicating whether the Mapper is doing case
sensitive or insensitive mapping.

Build and Test Your Code

9. Following the earlier steps, test your code with Mapper debug logging enabled. View the
map task logs in the Job Tracker UI to confirm that your message is included in the log.
(Hint: Search for LetterMapper in the page to find your message.)
10. Optional: Try running map logging set to INFO (the default) or WARN instead of DEBUG
and compare the log output.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

39

Lecture 5 Lab: Using Counters and a

Map-Only Job
Files and Directories Used in this Exercise
Eclipse project: counters
Java files:
ImageCounter.java (driver)
ImageCounterMapper.java (Mapper)
Test data (HDFS):
weblog (full web server access log)
testlog (partial data set for testing)
Exercise directory: ~/workspace/counters

In this lab you will create a Map-only MapReduce job.

Your application will process a web servers access log to count the number of times gifs,
jpegs, and other resources have been retrieved. Your job will report three figures: number
of gif requests, number of jpeg requests, and number of other requests.

Hints
1. You should use a Map-only MapReduce job, by setting the number of Reducers to 0 in
the driver code.
2. For input data, use the Web access log file that you uploaded to the HDFS
/user/training/weblog directory in the Using HDFS lab.
Note: Test your code against the smaller version of the access log in the
/user/training/testlog directory before you run your code against the full log in
the /user/training/weblog directory.
3. Use a counter group such as ImageCounter, with names gif, jpeg and other.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

40

4. In your driver code, retrieve the values of the counters after the job has completed and
report them using System.out.println.
5. The output folder on HDFS will contain Mapper output files which are empty, because
the Mappers did not write any data.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

41

Lecture 6 Lab: Writing a Partitioner

Files and Directories Used in this Exercise
Eclipse project: partitioner
Java files:
MonthPartitioner.java (Partitioner)
ProcessLogs.java (driver)
CountReducer.java (Reducer)
LogMonthMapper.java (Mapper)
Test data (HDFS):
weblog (full web server access log)
testlog (partial data set for testing)
Exercise directory: ~/workspace/partitioner

In this Exercise, you will write a MapReduce job with multiple Reducers, and create a
Partitioner to determine which Reducer each piece of Mapper output is sent to.

The Problem
In the More Practice with Writing MapReduce Java Programs lab you did previously, you
built the code in log_file_analysis project. That program counted the number of hits
for each different IP address in a web log file. The final output was a file containing a list of
IP addresses, and the number of hits from that address.
This time, you will perform a similar task, but the final output should consist of 12 files, one
each for each month of the year: January, February, and so on. Each file will contain a list of
IP addresses, and the number of hits from that address in that month.
We will accomplish this by having 12 Reducers, each of which is responsible for processing
the data for a particular month. Reducer 0 processes January hits, Reducer 1 processes
February hits, and so on.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

42

Note: We are actually breaking the standard MapReduce paradigm here, which says that all
the values from a particular key will go to the same Reducer. In this example, which is a very
common pattern when analyzing log files, values from the same key (the IP address) will go
to multiple Reducers, based on the month portion of the line.

Write the Mapper

1. Starting with the LogMonthMapper.java stub file, write a Mapper that maps a log
file output line to an IP/month pair. The map method will be similar to that in the
LogFileMapper class in the log_file_analysis project, so you may wish to start
by copying that code.
2. The Mapper should emit a Text key (the IP address) and Text value (the month). E.g.:
Input: 96.7.4.14 - - [24/Apr/2011:04:20:11 -0400] "GET
/cat.jpg HTTP/1.1" 200 12433
Output key: 96.7.4.14
Output value: Apr
Hint: In the Mapper, you may use a regular expression to parse to log file data if you are
familiar with regex processing (see file Homework_RegexRef.docx for reference).
Remember that the log file may contain unexpected data that is, lines that do not
conform to the expected format. Be sure that your code copes with such lines.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

43

Write the Partitioner

3. Modify the MonthPartitioner.java stub file to create a Partitioner that sends the
(key, value) pair to the correct Reducer based on the month. Remember that the
Partitioner receives both the key and value, so you can inspect the value to determine
which Reducer to choose.

Modify the Driver

4. Modify your driver code to specify that you want 12 Reducers.
5. Configure your job to use your custom Partitioner.

Test your Solution

6. Build and test your code. Your output directory should contain 12 files named part-r000xx. Each file should contain IP address and number of hits for month xx.
Hints:

Write unit tests for your Partitioner!

You may wish to test your code against the smaller version of the access log in the
/user/training/testlog directory before you run your code against the full
log in the /user/training/weblog directory. However, note that the test data
may not include all months, so some result files will be empty.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

44

Lecture 6 Lab: Implementing a Custom

WritableComparable
Files and Directories Used in this Exercise
Eclipse project: writables
Java files:
StringPairWritable implements a WritableComparable type
StringPairMapper Mapper for test job
StringPairTestDriver Driver for test job
Data file:
~/training_materials/developer/data/nameyeartestdata (small set of data
for the test job)
Exercise directory: ~/workspace/writables

In this lab, you will create a custom WritableComparable type that holds two strings.
Test the new type by creating a simple program that reads a list of names (first and last) and
counts the number of occurrences of each name.
The mapper should accepts lines in the form:
lastname firstname other data
The goal is to count the number of times a lastname/firstname pair occur within the dataset.
For example, for input:
Smith Joe 1963-08-12 Poughkeepsie, NY
Smith Joe 1832-01-20 Sacramento, CA
Murphy Alice 2004-06-02 Berlin, MA
We want to output:
(Smith,Joe)
2
(Murphy,Alice) 1
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

45

Note: You will use your custom WritableComparable type in a future lab, so make sure it
is working with the test job now.

StringPairWritable
You need to implement a WritableComparable object that holds the two strings. The stub
provides an empty constructor for serialization, a standard constructor that will be given
two strings, a toString method, and the generated hashCode and equals methods. You
will need to implement the readFields, write, and compareTo methods required by
WritableComparables.
Note that Eclipse automatically generated the hashCode and equals methods in the stub
file. You can generate these two methods in Eclipse by right-clicking in the source code and
choosing Source > Generate hashCode() and equals().

Name Count Test Job

The test job requires a Reducer that sums the number of occurrences of each key. This is the
same function that the SumReducer used previously in wordcount, except that SumReducer
expects Text keys, whereas the reducer for this job will get StringPairWritable keys. You
may either re-write SumReducer to accommodate other types of keys, or you can use the
LongSumReducer Hadoop library class, which does exactly the same thing.
You can use the simple test data in
~/training_materials/developer/data/nameyeartestdata to make sure
your new type works as expected.
You may test your code using local job runner or by submitting a Hadoop job to the (pseudo-
)cluster as usual. If you submit the job to the cluster, note that you will need to copy your
test data to HDFS first.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

46

Lecture 6 Lab: Using SequenceFiles and

File Compression
Files and Directories Used in this Exercise
Eclipse project: createsequencefile
Java files:
CreateSequenceFile.java (a driver that converts a text file to a sequence file)
ReadCompressedSequenceFile.java (a driver that converts a compressed sequence
file to text)
Test data (HDFS):
weblog (full web server access log)
Exercise directory: ~/workspace/createsequencefile

In this lab you will practice reading and writing uncompressed and compressed
SequenceFiles.
First, you will develop a MapReduce application to convert text data to a SequenceFile. Then
you will modify the application to compress the SequenceFile using Snappy file compression.
When creating the SequenceFile, use the full access log file for input data. (You uploaded the
access log file to the HDFS /user/training/weblog directory when you performed the
Using HDFS lab.)
After you have created the compressed SequenceFile, you will write a second MapReduce
application to read the compressed SequenceFile and write a text file that contains the
original log file text.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

47

Write a MapReduce program to create sequence files from

text files
1. Determine the number of HDFS blocks occupied by the access log file:
a. In a browser window, start the Name Node Web UI. The URL is
http://localhost:50070
b. Click Browse the filesystem.
c. Navigate to the /user/training/weblog/access_log file.
d. Scroll down to the bottom of the page. The total number of blocks occupied by
the access log file appears in the browser window.
2. Complete the stub file in the createsequencefile project to read the access log file
and create a SequenceFile. Records emitted to the SequenceFile can have any key you
like, but the values should match the text in the access log file. (Hint: You can use Map-
only job using the default Mapper, which simply emits the data passed to it.)
Note: If you specify an output key type other than LongWritable, you must call
job.setOutputKeyClass not job.setMapOutputKeyClass. If you specify an
output value type other than Text, you must call job.setOutputValueClass not
job.setMapOutputValueClass.
3. Build and test your solution so far. Use the access log as input data, and specify the
uncompressedsf directory for output.
4. Examine the initial portion of the output SequenceFile using the following command:
$ hadoop fs -cat uncompressedsf/part-m-00000 | less
Some of the data in the SequenceFile is unreadable, but parts of the SequenceFile
should be recognizable:

The string SEQ, which appears at the beginning of a SequenceFile

The Java classes for the keys and values

Text from the access log file

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

48

5. Verify that the number of files created by the job is equivalent to the number of blocks
required to store the uncompressed SequenceFile.

Compress the Output

6. Modify your MapReduce job to compress the output SequenceFile. Add statements to
your driver to configure the output as follows:

Compress the output file.

Use block compression.

Use the Snappy compression codec.

7. Compile the code and run your modified MapReduce job. For the MapReduce output,
specify the compressedsf directory.
8. Examine the first portion of the output SequenceFile. Notice the differences between the
uncompressed and compressed SequenceFiles:

The compressed SequenceFile specifies the

org.apache.hadoop.io.compress.SnappyCodec compression codec in
its header.

You cannot read the log file text in the compressed file.

9. Compare the file sizes of the uncompressed and compressed SequenceFiles in the
uncompressedsf and compressedsf directories. The compressed SequenceFiles
should be smaller.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

49

Write another MapReduce program to uncompress the files

10. Starting with the provided stub file, write a second MapReduce program to read the
compressed log file and write a text file. This text file should have the same text data as
the log file, plus keys. The keys can contain any values you like.
11. Compile the code and run your MapReduce job.
For the MapReduce input, specify the compressedsf directory in which you created
the compressed SequenceFile in the previous section.
For the MapReduce output, specify the compressedsftotext directory.
12. Examine the first portion of the output in the compressedsftotext directory.
You should be able to read the textual log file entries.

Optional: Use command line options to control compression

13. If you used ToolRunner for your driver, you can control compression using command
line arguments. Try commenting out the code in your driver where you call. Then test
setting the mapred.output.compressed option on the command line, e.g.:
$ hadoop jar sequence.jar \
stubs.CreateUncompressedSequenceFile \
-Dmapred.output.compressed=true \
weblog outdir
14. Review the output to confirm the files are compressed.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

50

Lecture 7 Lab: Creating an Inverted

Index
Files and Directories Used in this Exercise
Eclipse project: inverted_index
Java files:
IndexMapper.java (Mapper)
IndexReducer.java (Reducer)
InvertedIndex.java (Driver)
Data files:
~/training_materials/developer/data/invertedIndexInput.tgz
Exercise directory: ~/workspace/inverted_index

In this lab, you will write a MapReduce job that produces an inverted index.
For this lab you will use an alternate input, provided in the file
invertedIndexInput.tgz. When decompressed, this archive contains a directory of
files; each is a Shakespeare play formatted as follows:
0

HAMLET

1
2
3

DRAMATIS PERSONAE

4
5
6

CLAUDIUS

king of Denmark. (KING CLAUDIUS:)

7
8

HAMLET

son to the late, and nephew to the present king.

9
10

POLONIUS

lord chamberlain. (LORD POLONIUS:)

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

51

...
Each line contains:

Line number
separator: a tab character
value: the line of text

This format can be read directly using the KeyValueTextInputFormat class provided in
the Hadoop API. This input format presents each line as one record to your Mapper, with the
part before the tab character as the key, and the part after the tab as the value.
Given a body of text in this form, your indexer should produce an index of all the words in
the text. For each word, the index should have a list of all the locations where the word
appears. For example, for the word honeysuckle your output should look like this:
honeysuckle

2kinghenryiv@1038,midsummernightsdream@2175,...

The index should contain such an entry for every word in the text.

Prepare the Input Data

1. Extract the invertedIndexInput directory and upload to HDFS:
$ cd ~/training_materials/developer/data
$ tar zxvf invertedIndexInput.tgz
$ hadoop fs -put invertedIndexInput invertedIndexInput

Define the MapReduce Solution

Remember that for this program you use a special input format to suit the form of your data,
so your driver class will include a line like:
job.setInputFormatClass(KeyValueTextInputFormat.class);
Dont forget to import this class for your use.
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

52

Retrieving the File Name

Note that the lab requires you to retrieve the file name - since that is the name of the play.
The Context object can be used to retrieve the name of the file like this:
FileSplit fileSplit = (FileSplit) context.getInputSplit();
Path path = fileSplit.getPath();
String fileName = path.getName();

Build and Test Your Solution

Test against the invertedIndexInput data you loaded above.

Hints
You may like to complete this lab without reading any further, or you may find the following
hints about the algorithm helpful.

The Mapper
Your Mapper should take as input a key and a line of words, and emit as intermediate values
each word as key, and the key as value.
For example, the line of input from the file hamlet:
282 Have heaven and earth together
produces intermediate output:

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

53

Have

hamlet@282

heaven

hamlet@282

and

hamlet@282

earth

hamlet@282

together

hamlet@282

The Reducer
Your Reducer simply aggregates the values presented to it for the same key, into one value.
Use a separator like , between the values listed.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

54

Lecture 7 Lab: Calculating Word CoOccurrence

Files and Directories Used in this Exercise
Eclipse project: word_co-occurrence
Java files:
WordCoMapper.java (Mapper)
SumReducer.java (Reducer from WordCount)
WordCo.java (Driver)
Test directory (HDFS):
shakespeare
Exercise directory: ~/workspace/word_co-occurence

In this lab, you will write an application that counts the number of times words
appear next to each other.
Test your application using the files in the shakespeare folder you previously copied into
HDFS in the Using HDFS lab.
Note that this implementation is a specialization of Word Co-Occurrence as we describe it in
the notes; in this case we are only interested in pairs of words which appear directly
next to each other.
1. Change directories to the word_co-occurrence directory within the labs directory.
2. Complete the Driver and Mapper stub files; you can use the standard SumReducer from
the WordCount project as your Reducer. Your Mappers intermediate output should be
in the form of a Text object as the key, and an IntWritable as the value; the key will be
word1,word2, and the value will be 1.

Extra Credit
If you have extra time, please complete these additional challenges:
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

55

Challenge 1: Use the StringPairWritable key type from the Implementing a

Custom WritableComparable lab. Copy your completed solution (from the writables
project) into the current project.
Challenge 2: Write a second MapReduce job to sort the output from the first job so that
the list of pairs of words appears in ascending frequency.
Challenge 3: Sort by descending frequency instead (sort that the most frequently
occurring word pairs are first in the output.) Hint: You will need to extend
org.apache.hadoop.io.LongWritable.Comparator.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

56

Lecture 8 Lab: Importing Data with

Sqoop
In this lab you will import data from a relational database using Sqoop. The data you
load here will be used subsequent labs.
Consider the MySQL database movielens, derived from the MovieLens project from
University of Minnesota. (See note at the end of this lab.) The database consists of several
related tables, but we will import only two of these: movie, which contains about 3,900
movies; and movierating, which has about 1,000,000 ratings of those movies.

Review the Database Tables

First, review the database tables to be loaded into Hadoop.
1. Log on to MySQL:
$ mysql --user=training --password=training movielens
2. Review the structure and contents of the movie table:
mysql> DESCRIBE movie;
. . .
mysql> SELECT * FROM movie LIMIT 5;
3. Note the column names for the table:
____________________________________________________________________________________________

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

57

4. Review the structure and contents of the movierating table:

mysql> DESCRIBE movierating;

mysql> SELECT * FROM movierating LIMIT 5;

5. Note these column names:
____________________________________________________________________________________________
6. Exit mysql:
mysql> quit

Import with Sqoop

You invoke Sqoop on the command line to perform several commands. With it you can
connect to your database server to list the databases (schemas) to which you have access,
and list the tables available for loading. For database access, you provide a connect string to
identify the server, and - if required - your username and password.
1. Show the commands available in Sqoop:
$ sqoop help
2. List the databases (schemas) in your database server:
$ sqoop list-databases \
--connect jdbc:mysql://localhost \
--username training --password training
(Note: Instead of entering --password training on your command line, you may
prefer to enter -P, and let Sqoop prompt you for the password, which is then not visible
when you type it.)

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

58

3. List the tables in the movielens database:

$ sqoop list-tables \
--connect jdbc:mysql://localhost/movielens \
--username training --password training
4. Import the movie table into Hadoop:
$ sqoop import \
--connect jdbc:mysql://localhost/movielens \
--username training --password training \
--fields-terminated-by '\t' --table movie
5. Verify that the command has worked.
$ hadoop fs -ls movie
$ hadoop fs -tail movie/part-m-00000
6. Import the movierating table into Hadoop.
Repeat the last two steps, but for the movierating table.

This is the end of the lab.

Note:
This lab uses the MovieLens data set, or subsets thereof. This data is freely available for
academic purposes, and is used and distributed by Cloudera with the express permission
of the UMN GroupLens Research Group. If you would like to use this data for your own
research purposes, you are free to do so, as long as you cite the GroupLens Research
Group in any resulting publications. If you would like to use this data for commercial
purposes, you must obtain explicit permission. You may find the full dataset, as well as
detailed license terms, at http://www.grouplens.org/node/73

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

59

Lecture 8 Lab: Running an Oozie

Workflow
Files and Directories Used in this Exercise
Exercise directory: ~/workspace/oozie_labs
Oozie job folders:
lab1-java-mapreduce
lab2-sort-wordcount

In this lab, you will inspect and run Oozie workflows.

1. Start the Oozie server
$ sudo /etc/init.d/oozie start
2. Change directories to the lab directory:
$ cd ~/workspace/oozie-labs
3. Inspect the contents of the job.properties and workflow.xml files in the lab1java-mapreduce/job folder. You will see that this is the standard WordCount job.
In the job.properties file, take note of the jobs base directory (lab1-javamapreduce), and the input and output directories relative to that. (These are HDFS
directories.)
4. We have provided a simple shell script to submit the Oozie workflow. Inspect the
run.sh script and then run:

$ ./run.sh lab1-java-mapreduce
Notice that Oozie returns a job identification number.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

60

5. Inspect the progress of the job:

$ oozie job -oozie http://localhost:11000/oozie \
-info job_id
6. When the job has completed, review the job output directory in HDFS to confirm that
the output has been produced as expected.
7. Repeat the above procedure for lab2-sort-wordcount. Notice when you inspect
workflow.xml that this workflow includes two MapReduce jobs which run one after
the other, in which the output of the first is the input for the second. When you inspect
the output in HDFS you will see that the second job sorts the output of the first job into
descending numerical order.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

61

Lecture 8 Bonus Lab: Exploring a

Secondary Sort Example
Files and Directories Used in this Exercise
Eclipse project: secondarysort
Data files:
~/training_materials/developer/data/nameyeartestdata
Exercise directory: ~/workspace/secondarysort

In this lab, you will run a MapReduce job in different ways to see the effects of various
components in a secondary sort program.
The program accepts lines in the form
lastname firstname birthdate
The goal is to identify the youngest person with each last name. For example, for input:
Murphy Joanne 1963-08-12
Murphy Douglas 1832-01-20
Murphy Alice 2004-06-02
We want to write out:
Murphy Alice 2004-06-02
All the code is provided to do this. Following the steps below you are going to progressively
add each component to the job to accomplish the final goal.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

62

Build the Program

1. In Eclipse, review but do not modify the code in the secondarysort project example
package.
2. In particular, note the NameYearDriver class, in which the code to set the partitioner,
sort comparator, and group comparator for the job is commented out. This allows us to
set those values on the command line instead.
3. Export the jar file for the program as secsort.jar.
4. A small test datafile called nameyeartestdata has been provided for you, located in
the secondary sort project folder. Copy the datafile to HDFS, if you did not already do so
in the Writables lab.

Run as a Map-only Job

5. The Mapper for this job constructs a composite key using the StringPairWritable
type. See the output of just the mapper by running this program as a Map-only job:
$ hadoop jar secsort.jar example.NameYearDriver \
-Dmapred.reduce.tasks=0 nameyeartestdata secsortout
6. Review the output. Note the key is a string pair of last name and birth year.

Run using the default Partitioner and Comparators

7. Re-run the job, setting the number of reduce tasks to 2 instead of 0.
8. Note that the output now consists of two files; one each for the two reduce tasks. Within
each file, the output is sorted by last name (ascending) and year (ascending). But it isnt
sorted between files, and records with the same last name may be in different files
(meaning they went to different reducers).

Run using the custom partitioner

9. Review the code of the custom partitioner class: NameYearPartitioner.
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

63

10. Re-run the job, adding a second parameter to set the partitioner class to use:
-Dmapreduce.partitioner.class=example.NameYearPartitioner
11. Review the output again, this time noting that all records with the same last name have
been partitioned to the same reducer.
However, they are still being sorted into the default sort order (name, year ascending).
We want it sorted by name ascending/year descending.

Run using the custom sort comparator

12. The NameYearComparator class compares Name/Year pairs, first comparing the
names and, if equal, compares the year (in descending order; i.e. later years are
considered less than earlier years, and thus earlier in the sort order.) Re-run the job
using NameYearComparator as the sort comparator by adding a third parameter:
-D mapred.output.key.comparator.class=
example.NameYearComparator
13. Review the output and note that each reducers output is now correctly partitioned and
sorted.

Run with the NameYearReducer

14. So far weve been running with the default reducer, which is the Identity Reducer, which
simply writes each key/value pair it receives. The actual goal of this job is to emit the
record for the youngest person with each last name. We can do this easily if all records
for a given last name are passed to a single reduce call, sorted in descending order,
which can then simply emit the first value passed in each call.
15. Review the NameYearReducer code and note that it emits
16. Re-run the job, using the reducer by adding a fourth parameter:
-Dmapreduce.reduce.class=example.NameYearReducer
Alas, the job still isnt correct, because the data being passed to the reduce method is
being grouped according to the full key (name and year), so multiple records with the
same last name (but different years) are being output. We want it to be grouped by
name only.
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

64

Run with the custom group comparator

17. The NameComparator class compares two string pairs by comparing only the name
field and disregarding the year field. Pairs with the same name will be grouped into the
same reduce call, regardless of the year. Add the group comparator to the job by adding
a final parameter:
-Dmapred.output.value.groupfn.class=
example.NameComparator
18. Note the final output now correctly includes only a single record for each different last
name, and that that record is the youngest person with that last name.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

65

Notes for Upcoming Labs

VM Services Customization
For the remainder of the labs, there are services that must be running in your VM, and
others that are optional. It is strongly recommended that you run the following command
whenever you start the VM:

$ ~/scripts/analyst/toggle_services.sh
This will conserve memory and increase performance of the virtual machine. After running
this command, you may safely ignore any messages about services that have already been
started or shut down.

Data Model Reference

For your convenience, you will find a reference document depicting the structure for the
tables you will use in the following labs. See file: Homework_DataModelRef.docx

Regular Expression (Regex) Reference

For your convenience, you will find a reference document describing regular expressions
syntax. See file: Homework_RegexRef.docx

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

66

Lecture 9 Lab: Data Ingest With Hadoop

Tools
In this lab you will practice using the Hadoop command line utility to interact with
Hadoops Distributed Filesystem (HDFS) and use Sqoop to import tables from a
relational database to HDFS.

Prepare your Virtual Machine

Launch the VM if you havent already done so, and then run the following command to boost
performance by disabling services that are not needed for this class:
$ ~/scripts/analyst/toggle_services.sh

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

67

Step 1: Setup HDFS

1. Open a terminal window (if one is not already open) by double-clicking the Terminal
icon on the desktop. Next, change to the directory for this lab by running the following
command:
$ cd $ADIR/exercises/data_ingest
2. To see the contents of your home directory, run the following command:
$ hadoop fs -ls /user/training
3. If you do not specify a path, hadoop fs assumes you are referring to your home
directory. Therefore, the following command is equivalent to the one above:
$ hadoop fs -ls
4. Most of your work will be in the /dualcore directory, so create that now:
$ hadoop fs -mkdir /dualcore

Step 2: Importing Database Tables into HDFS with Sqoop

Dualcore stores information about its employees, customers, products, and orders in a
MySQL database. In the next few steps, you will examine this database before using Sqoop to
import its tables into HDFS.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

68

1. Log in to MySQL and select the dualcore database:

$ mysql --user=training --password=training dualcore
2. Next, list the available tables in the dualcore database (mysql> represents the
MySQL client prompt and is not part of the command):
mysql> SHOW TABLES;
3. Review the structure of the employees table and examine a few of its records:
mysql> DESCRIBE employees;
mysql> SELECT emp_id, fname, lname, state, salary FROM
employees LIMIT 10;
4. Exit MySQL by typing quit, and then hit the enter key:
mysql> quit
5. Next, run the following command, which imports the employees table into the
/dualcore directory created earlier using tab characters to separate each field:
$ sqoop import \
--connect jdbc:mysql://localhost/dualcore \
--username training --password training \
--fields-terminated-by '\t' \
--warehouse-dir /dualcore \
--table employees

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

69

6. Revise the previous command and import the customers table into HDFS.
7. Revise the previous command and import the products table into HDFS.
8. Revise the previous command and import the orders table into HDFS.
9. Next, you will import the order_details table into HDFS. The command is slightly
different because this table only holds references to records in the orders and
products table, and lacks a primary key of its own. Consequently, you will need to
specify the --split-by option and instruct Sqoop to divide the import work among
map tasks based on values in the order_id field. An alternative is to use the -m 1
option to force Sqoop to import all the data with a single task, but this would
significantly reduce performance.
$ sqoop import \
--connect jdbc:mysql://localhost/dualcore \
--username training --password training \
--fields-terminated-by '\t' \
--warehouse-dir /dualcore \
--table order_details \
--split-by=order_id

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

70

Lecture 9 Lab: Using Pig for ETL

Processing
In this lab you will practice using Pig to explore, correct, and reorder data in files
from two different ad networks. You will first experiment with small samples of this
data using Pig in local mode, and once you are confident that your ETL scripts work as
you expect, you will use them to process the complete data sets in HDFS by using Pig
in MapReduce mode.
IMPORTANT: Since this lab builds on the previous one, it is important that you successfully
complete the previous lab before starting this lab.

Background Information
Dualcore has recently started using online advertisements to attract new customers to its e-
commerce site. Each of the two ad networks they use provides data about the ads theyve
placed. This includes the site where the ad was placed, the date when it was placed, what
keywords triggered its display, whether the user clicked the ad, and the per-click cost.
Unfortunately, the data from each network is in a different format. Each file also contains
some invalid records. Before we can analyze the data, we must first correct these problems
by using Pig to:

Filter invalid records

Reorder fields

Correct inconsistencies

Write the corrected data to HDFS

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

71

Step #1: Working in the Grunt Shell

In this step, you will practice running Pig commands in the Grunt shell.
1. Change to the directory for this lab:
$ cd $ADIR/exercises/pig_etl
2. Copy a small number of records from the input file to another file on the local file
system. When you start Pig, you will run in local mode. For testing, you can work faster
with small local files than large files in HDFS.
It is not essential to choose a random sample here just a handful of records in the
correct format will suffice. Use the command below to capture the first 25 records so
you have enough to test your script:
$ head -n 25 $ADIR/data/ad_data1.txt > sample1.txt
3. Start the Grunt shell in local mode so that you can work with the local sample1.txt
file.
$ pig -x local
A prompt indicates that you are now in the Grunt shell:
grunt>
4. Load the data in the sample1.txt file into Pig and dump it:
grunt> data = LOAD 'sample1.txt';
grunt> DUMP data;
You should see the 25 records that comprise the sample data file.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

72

5. Load the first two columns data from the sample file as character data, and then dump
that data:
grunt> first_2_columns = LOAD 'sample1.txt' AS
(keyword:chararray, campaign_id:chararray);
grunt> DUMP first_2_columns;
6. Use the DESCRIBE command in Pig to review the schema of first_2_cols:
grunt> DESCRIBE first_2_columns;
The schema appears in the Grunt shell.
Use the DESCRIBE command while performing these labs any time you would like to
review schema definitions.
7. See what happens if you run the DESCRIBE command on data. Recall that when you
loaded data, you did not define a schema.
grunt> DESCRIBE data;
8. End your Grunt shell session:
grunt> QUIT;

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

73

Step #2: Processing Input Data from the First Ad Network

In this step, you will process the input data from the first ad network. First, you will create a
Pig script in a file, and then you will run the script. Many people find working this way
easier than working directly in the Grunt shell.
1. Edit the first_etl.pig file to complete the LOAD statement and read the data from
the sample you just created. The following table shows the format of the data in the file.
For simplicity, you should leave the date and time fields separate, so each will be of
type chararray, rather than converting them to a single field of type datetime.

Index
0
1
2
3
4
5
6
7
8

Field
keyword
campaign_id
date
time
display_site
was_clicked
cpc
country
placement

Data Type
chararray
chararray
chararray
chararray
chararray
int
int
chararray
chararray

Description
Keyword that triggered ad
Uniquely identifies the ad
Date of ad display
Time of ad display
Domain where ad shown
Whether ad was clicked
Cost per click, in cents
Name of country in which ad ran
Where on page was ad displayed

Example
tablet
A3
05/29/2013
15:49:21
www.example.com
1
106
USA
TOP

2. Once you have edited the LOAD statement, try it out by running your script in local
mode:
$ pig -x local first_etl.pig
Make sure the output looks correct (i.e., that you have the fields in the expected order
and the values appear similar in format to that shown in the table above) before you
continue with the next step.
3. Make each of the following changes, running your script in local mode after each one to
verify that your change is correct:
a. Update your script to filter out all records where the country field does not
contain USA.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

74

b. We need to store the fields in a different order than we received them. Use a
FOREACH GENERATE statement to create a new relation containing the
fields in the same order as shown in the following table (the country field is
not included since all records now have the same value):

Index
0
1
2
3
4
5
6
7

Field
campaign_id
date
time
keyword
display_site
placement
was_clicked
cpc

Description
Uniquely identifies the ad
Date of ad display
Time of ad display
Keyword that triggered ad
Domain where ad shown
Where on page was ad displayed
Whether ad was clicked
Cost per click, in cents

c. Update your script to convert the keyword field to uppercase and to remove
any leading or trailing whitespace (Hint: You can nest calls to the two built-in
functions inside the FOREACH GENERATE statement from the last
statement).
4. Add the complete data file to HDFS:
$ hadoop fs -put $ADIR/data/ad_data1.txt /dualcore
5. Edit first_etl.pig and change the path in the LOAD statement to match the path of
the file you just added to HDFS (/dualcore/ad_data1.txt).
6. Next, replace DUMP with a STORE statement that will write the output of your
processing as tab-delimited records to the /dualcore/ad_data1 directory.
7. Run this script in Pigs MapReduce mode to analyze the entire file in HDFS:
$ pig first_etl.pig
If your script fails, check your code carefully, fix the error, and then try running it again.
Dont forget that you must remove output in HDFS from a previous run before you
execute the script again.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

75

8. Check the first 20 output records that your script wrote to HDFS and ensure they look
correct (you can ignore the message cat: Unable to write to output stream; this simply
happens because you are writing more data with the fs -cat command than you are
reading with the head command):
$ hadoop fs -cat /dualcore/ad_data1/part* | head -20
a. Are the fields in the correct order?
b. Are all the keywords now in uppercase?

Step #3: Processing Input Data from the Second Ad Network

Now that you have successfully processed the data from the first ad network, continue by
processing data from the second one.
1. Create a small sample of the data from the second ad network that you can test locally
while you develop your script:
$ head -n 25 $ADIR/data/ad_data2.txt > sample2.txt

2. Edit the second_etl.pig file to complete the LOAD statement and read the data from
the sample you just created (Hint: The fields are comma-delimited). The following table
shows the order of fields in this file:

Index
0
1
2
3
4
5
6
7

Field
campaign_id
date
time
display_site
placement
was_clicked
cpc
keyword

Data Type
chararray
chararray
chararray
chararray
chararray
int
int
chararray

Description
Uniquely identifies the ad
Date of ad display
Time of ad display
Domain where ad shown
Where on page was ad displayed
Whether ad was clicked
Cost per click, in cents
Keyword that triggered ad

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

Example
A3
05/29/2013
15:49:21
www.example.com
TOP
Y
106
tablet

76

3. Once you have edited the LOAD statement, use the DESCRIBE keyword and then run
your script in local mode to check that the schema matches the table above:
$ pig -x local second_etl.pig
4. Replace DESCRIBE with a DUMP statement and then make each of the following
changes to second_etl.pig, running this script in local mode after each change to
verify what youve done before you continue with the next step:
d. This ad network sometimes logs a given record twice. Add a statement to the
second_etl.pig file so that you remove any duplicate records. If you have
done this correctly, you should only see one record where the
display_site field has a value of siliconwire.example.com.

e. As before, you need to store the fields in a different order than you received
them. Use a FOREACH GENERATE statement to create a new relation
containing the fields in the same order you used to write the output from first
ad network (shown again in the table below) and also use the UPPER and
TRIM functions to correct the keyword field as you did earlier:
Index
0
1
2
3
4
5
6
7

Field
campaign_id
date
time
keyword
display_site
placement
was_clicked
cpc

Description
Uniquely identifies the ad
Date of ad display
Time of ad display
Keyword that triggered ad
Domain where ad shown
Where on page was ad displayed
Whether ad was clicked
Cost per click, in cents

f. The date field in this data set is in the format MM-DD-YYYY, while the data
you previously wrote is in the format MM/DD/YYYY. Edit the FOREACH
GENERATE statement to call the REPLACE(date, '-', '/') function
to correct this.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

77

5. Once you are sure the script works locally, add the full data set to HDFS:
$ hadoop fs -put $ADIR/data/ad_data2.txt /dualcore
6. Edit the script to have it LOAD the file you just added to HDFS, and then replace the
DUMP statement with a STORE statement to write your output as tab-delimited records
to the /dualcore/ad_data2 directory.
7. Run your script against the data you added to HDFS:
$ pig second_etl.pig
8. Check the first 15 output records written in HDFS by your script:
$ hadoop fs -cat /dualcore/ad_data2/part* | head -15
a. Do you see any duplicate records?
b. Are the fields in the correct order?
c. Are all the keywords in uppercase?

d. Is the date field in the correct (MM/DD/YYYY) format?

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

78

Lecture 9 Lab: Analyzing Ad Campaign

Data with Pig
During the previous lab, you performed ETL processing on data sets from two online
ad networks. In this lab, you will write Pig scripts that analyze this data to optimize
advertising, helping Dualcore to save money and attract new customers.
IMPORTANT: Since this lab builds on the previous one, it is important that you successfully
complete the previous lab before starting this lab.

Step #1: Find Low Cost Sites

Both ad networks charge a fee only when a user clicks on Dualcores ad. This is ideal for
Dualcore since their goal is to bring new customers to their site. However, some sites and
keywords are more effective than others at attracting people interested in the new tablet
being advertised by Dualcore. With this in mind, you will begin by identifying which sites
have the lowest total cost.
1. Change to the directory for this lab:
$ cd $ADIR/exercises/analyze_ads
2. Obtain a local subset of the input data by running the following command:
$ hadoop fs -cat /dualcore/ad_data1/part* \
| head -n 100 > test_ad_data.txt
You can ignore the message cat: Unable to write to output stream, which appears
because you are writing more data with the fs -cat command than you are reading
with the head command.
Note: As mentioned in the previous lab, it is faster to test Pig scripts by using a local
subset of the input data. Although explicit steps are not provided for creating local data
subsets in upcoming labs, doing so will help you perform the labs more quickly.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

79

3. Open the low_cost_sites.pig file in your editor, and then make the following
changes:
a. Modify the LOAD statement to read the sample data in the
test_ad_data.txt file.
b. Add a line that creates a new relation to include only records where
was_clicked has a value of 1.
c. Group this filtered relation by the display_site field.
d. Create a new relation that includes two fields: the display_site and
the total cost of all clicks on that site.
e. Sort that new relation by cost (in ascending order)
f. Display just the first three records to the screen
4. Once you have made these changes, try running your script against the sample data:
$ pig x local low_cost_sites.pig
5. In the LOAD statement, replace the test_ad_data.txt file with a file glob (pattern)
that will load both the /dualcore/ad_data1 and /dualcore/ad_data2
directories (and does not load any other data, such as the text files from the previous
lab).
6. Once you have made these changes, try running your script against the data in HDFS:
$ pig low_cost_sites.pig
Question: Which three sites have the lowest overall cost?

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

80

Step #2: Find High Cost Keywords

The terms users type when doing searches may prompt the site to display a Dualcore
advertisement. Since online advertisers compete for the same set of keywords, some of
them cost more than others. You will now write some Pig Latin to determine which
keywords have been the most expensive for Dualcore overall.
1. Since this will be a slight variation on the code you have just written, copy that file as
high_cost_keywords.pig:
$ cp low_cost_sites.pig high_cost_keywords.pig
2. Edit the high_cost_keywords.pig file and make the following three changes:
a. Group by the keyword field instead of display_site
b. Sort in descending order of cost
c. Display the top five results to the screen instead of the top three as before
3. Once you have made these changes, try running your script against the data in HDFS:
$ pig high_cost_keywords.pig
Question: Which five keywords have the highest overall cost?

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

81

Bonus Lab #1: Count Ad Clicks

One important statistic we havent yet calculated is the total number of clicks the ads have
received. Doing so will help the marketing director plan the next ad campaign budget.
1. Change to the bonus_01 subdirectory of the current lab:
$ cd bonus_01
2. Edit the total_click_count.pig file and implement the following:
a. Group the records (filtered by was_clicked == 1) so that you can call
the aggregate function in the next step.
b. Invoke the COUNT function to calculate the total of clicked ads (Hint:
Because we shouldnt have any null records, you can use the COUNT
function instead of COUNT_STAR, and the choice of field you supply to the
function is arbitrary).
c. Display the result to the screen
3. Once you have made these changes, try running your script against the data in HDFS:
$ pig total_click_count.pig
Question: How many clicks did we receive?

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

82

Bonus Lab #2: Estimate the Maximum Cost of the Next Ad

Campaign
When you reported the total number of clicks, the Marketing Director said that the goal is to
get about three times that amount during the next campaign. Unfortunately, because the
cost is based on the site and keyword, it isnt clear how much to budget for that campaign.
You can help by estimating the worst case (most expensive) cost based on 50,000 clicks. You
will do this by finding the most expensive ad and then multiplying it by the number of clicks
desired in the next campaign.
1. Because this code will be similar to the code you wrote in the previous step, start by
copying that file as project_next_campaign_cost.pig:
$ cp total_click_count.pig project_next_campaign_cost.pig
2. Edit the project_next_campaign_cost.pig file and make the following
modifications:
a. Since you are trying to determine the highest possible cost, you should
not limit your calculation to the cost for ads actually clicked. Remove the
FILTER statement so that you consider the possibility that any ad might
be clicked.
b. Change the aggregate function to the one that returns the maximum value
in the cpc field (Hint: Dont forget to change the name of the relation this
field belongs to, in order to account for the removal of the FILTER
statement in the previous step).
c. Modify your FOREACH...GENERATE statement to multiply the value
returned by the aggregate function by the total number of clicks we
expect to have in the next campaign
d. Display the resulting value to the screen.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

83

3. Once you have made these changes, try running your script against the data in HDFS:
$ pig project_next_campaign_cost.pig
Question: What is the maximum you expect this campaign might cost?

Professors Note~
You can compare your solution to the one in the bonus_02/sample_solution/
subdirectory.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

84

Bonus Lab #3: Calculating Click-Through Rate (CTR)

The calculations you did at the start of this lab provided a rough idea about the success of
the ad campaign, but didnt account for the fact that some sites display Dualcores ads more
than others. This makes it difficult to determine how effective their ads were by simply
counting the number of clicks on one site and comparing it to the number of clicks on
another site. One metric that would allow Dualcore to better make such comparisons is the
Click-Through Rate (http://tiny.cloudera.com/ade03a), commonly abbreviated as
CTR. This value is simply the percentage of ads shown that users actually clicked, and can be
calculated by dividing the number of clicks by the total number of ads shown.
1. Change to the bonus_03 subdirectory of the current lab:
$ cd ../bonus_03
2. Edit the lowest_ctr_by_site.pig file and implement the following:
a. Within the nested FOREACH, filter the records to include only records
where the ad was clicked.
b. Create a new relation on the line that follows the FILTER statement
which counts the number of records within the current group
c. Add another line below that to calculate the click-through rate in a new
field named ctr
d. After the nested FOREACH, sort the records in ascending order of
clickthrough rate and display the first three to the screen.
3. Once you have made these changes, try running your script against the data in HDFS:
$ pig lowest_ctr_by_site.pig
Question: Which three sites have the lowest click through rate?
If you still have time remaining, modify your script to display the three keywords with the
highest click-through rate.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

85

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

86

Lecture 10 Lab: Analyzing Disparate Data

Sets with Pig
In this lab, you will practice combining, joining, and analyzing the product sales data
previously exported from Dualcores MySQL database so you can observe the effects
that the recent advertising campaign has had on sales.
IMPORTANT: Since this lab builds on the previous one, it is important that you successfully
complete the previous lab before starting this lab.

Step #1: Show Per-Month Sales Before and After Campaign

Before we proceed with more sophisticated analysis, you should first calculate the number
of orders Dualcore received each month for the three months before their ad campaign
began (February April, 2013), as well as for the month during which their campaign ran
(May, 2013).
1. Change to the directory for this lab:
$ cd $ADIR/exercises/disparate_datasets
2. Open the count_orders_by_period.pig file in your editor. We have provided the
LOAD statement as well as a FILTER statement that uses a regular expression to match
the records in the data range youll analyze. Make the following additional changes:
a. Following the FILTER statement, create a new relation with just one
field: the orders year and month (Hint: Use the SUBSTRING built-in
function to extract the first part of the order_dtm field, which contains
the month and year).
b. Count the number of orders in each of the months you extracted in the
previous step.
c. Display the count by month to the screen

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

87

3. Once you have made these changes, try running your script against the data in HDFS:
$ pig count_orders_by_period.pig
Question: Does the data suggest that the advertising campaign we started in May led to
a substantial increase in orders?

Step #2: Count Advertised Product Sales by Month

Our analysis from the previous step suggests that sales increased dramatically the same
month Dualcore began advertising. Next, youll compare the sales of the specific product
Dualcore advertised (product ID #1274348) during the same period to see whether the
increase in sales was actually related to their campaign.
You will be joining two data sets during this portion of the lab. Since this is the first join you
have done with Pig, now is a good time to mention a tip that can have a profound effect on
the performance of your script. Filtering out unwanted data from each relation before you
join them, as weve done in our example, means that your script will need to process less
data and will finish more quickly. We will discuss several more Pig performance tips later in
class, but this one is worth learning now.
4. Edit the count_tablet_orders_by_period.pig file and implement the
following:
a. Join the two relations on the order_id field they have in common
b. Create a new relation from the joined data that contains a single field: the
orders year and month, similar to what you did previously in the
count_orders_by_period.pig file.
c. Group the records by month and then count the records in each group
d. Display the results to your screen

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

88

5. Once you have made these changes, try running your script against the data in HDFS:
$ pig count_tablet_orders_by_period.pig
Question: Does the data show an increase in sales of the advertised product
corresponding to the month in which Dualcores campaign was active?

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

89

Bonus Lab #1: Calculate Average Order Size

It appears that Dualcores advertising campaign was successful in generating new orders.
Since they sell this tablet at a slight loss to attract new customers, lets see if customers who
buy this tablet also buy other things. You will write code to calculate the average number of
items for all orders that contain the advertised tablet during the campaign period.
1. Change to the bonus_01 subdirectory of the current lab:
$ cd bonus_01
2. Edit the average_order_size.pig file to calculate the average as described above.
While there are multiple ways to achieve this, it is recommended that you implement
the following:
a. Filter the orders by date (using a regular expression) to include only those
placed during the campaign period (May 1, 2013 through May 31, 2013)
b. Exclude any orders which do not contain the advertised product (product ID
#1274348)
c. Create a new relation containing the order_id and product_id fields for
these orders.
d. Count the total number of products per order
e. Calculate the average number of products for all orders
3. Once you have made these changes, try running your script against the data in HDFS:
$ pig average_order_size.pig
Question: Does the data show that the average order contained at least two items in
addition to the tablet Dualcore advertised?

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

90

Bonus Lab #2: Segment Customers for Loyalty Program

Dualcore is considering starting a loyalty rewards program. This will provide exclusive
benefits to their best customers, which will help to retain them. Another advantage is that it
will also allow Dualcore to capture even more data about the shopping habits of their
customers; for example, Dualcore can easily track their customers in-store purchases when
these customers provide their rewards program number at checkout.
To be considered for the program, a customer must have made at least five purchases from
Dualcore during 2012. These customers will be segmented into groups based on the total
retail price of all purchases each made during that year:

Platinum: Purchases totaled at least $10,000

Gold: Purchases totaled at least $5,000 but less than $10,000

Silver: Purchases totaled at least $2,500 but less than $5,000

Since we are considering the total sales price of orders in addition to the number of orders a
customer has placed, not every customer with at least five orders during 2012 will qualify.
In fact, only about one percent of the customers will be eligible for membership in one of
these three groups.
During this lab, you will write the code needed to filter the list of orders based on date,
group them by customer ID, count the number of orders per customer, and then filter this to
exclude any customer who did not have at least five orders. You will then join this
information with the order details and products data sets in order to calculate the total sales
of those orders for each customer, split them into the groups based on the criteria described
above, and then write the data for each group (customer ID and total sales) into a separate
directory in HDFS.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

91

1. Change to the bonus_02 subdirectory of the current lab:

$ cd ../bonus_02
2. Edit the loyalty_program.pig file and implement the steps described above. The
code to load the three data sets you will need is already provided for you.
3. After you have written the code, run it against the data in HDFS:
$ pig loyalty_program.pig
4. If your script completed successfully, use the hadoop fs -getmerge command to
create a local text file for each group so you can check your work (note that the name of
the directory shown here may not be the same as the one you chose):
$ hadoop fs -getmerge /dualcore/loyalty/platinum platinum.txt
$ hadoop fs -getmerge /dualcore/loyalty/gold gold.txt
$ hadoop fs -getmerge /dualcore/loyalty/silver silver.txt
5. Use the UNIX head and/or tail commands to check a few records and ensure that the
total sales prices fall into the correct ranges:
$ head platinum.txt
$ tail gold.txt
$ head silver.txt
6. Finally, count the number of customers in each group:
$ wc -l platinum.txt
$ wc -l gold.txt
$ wc -l silver.txt

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

92

Lecture 10 Lab: Extending Pig with

Streaming and UDFs
In this lab you will use the STREAM keyword in Pig to analyze metadata from
Dualcores customer service call recordings to identify the cause of a sudden increase
in complaints. You will then use this data in conjunction with a user-defined function
to propose a solution for resolving the problem.
IMPORTANT: Since this lab builds on the previous one, it is important that you successfully
complete the previous lab before starting this lab.

Background Information
Dualcore outsources its call center operations and costs have recently risen due to an
increase in the volume of calls handled by these agents. Unfortunately, Dualcore does not
have access to the call centers database, but they are provided with recordings of these calls
stored in MP3 format. By using Pigs STREAM keyword to invoke a provided Python script,
you can extract the category and timestamp from the files, and then analyze that data to
learn what is causing the recent increase in calls.

Step #1: Extract Call Metadata

Note: Since the Python library we are using for extracting the tags doesn't support HDFS, we
run this script in local mode on a small sample of the call recordings. Because you will use
Pigs local mode, there will be no need to ship the script to the nodes in the cluster.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

93

1. Change to the directory for this lab:

$ cd $ADIR/exercises/extending_pig
2. A Python script (readtags.py) is provided for extracting the metadata from the MP3
files. This script takes the path of a file on the command line and returns a record
containing five tab-delimited fields: the file path, call category, agent ID, customer ID,
and the timestamp of when the agent answered the call.
Your first step is to create a text file containing the paths of the files to analyze, with one
line for each file. You can easily create the data in the required format by capturing the
output of the UNIX find command:
$ find $ADIR/data/cscalls/ -name '*.mp3' > call_list.txt
3. Edit the extract_metadata.pig file and make the following changes:
a. Replace the hardcoded parameter in the SUBSTRING function used to filter
by month with a parameter named MONTH whose value you can assign on the
command line. This will make it easy to check the leading call categories for
different months without having to edit the script.
b. Add the code necessary to count calls by category
c. Display the top three categories (based on number of calls) to the screen.
4. Once you have made these changes, run your script to check the top three categories in
the month before Dualcore started the online advertising campaign:
$ pig -x local -param MONTH=2013-04 extract_metadata.pig
5. Now run the script again, this time specifying the parameter for May:
$ pig -x local -param MONTH=2013-05 extract_metadata.pig
The output should confirm that not only is call volume substantially higher in May, the
SHIPPING_DELAY category has more than twice the amount of calls as the other two.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

94

Step #2: Choose Best Location for Distribution Center

The analysis you just completed uncovered a problem. Dualcores Vice President of
Operations launched an investigation based on your findings and has now confirmed the
cause: their online advertising campaign is indeed attracting many new customers, but
many of them live far from Dualcores only distribution center in Palo Alto, California. All
shipments are transported by truck, so an order can take up to five days to deliver
depending on the customers location.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

95

To solve this problem, Dualcore will open a new distribution center to improve
shipping times.
The ZIP codes for the three proposed sites are 02118, 63139, and 78237. You will
look up the latitude and longitude of these ZIP codes, as well as the ZIP codes of
customers who have recently ordered, using a supplied data set. Once you have the
coordinates, you will invoke the use the HaversineDistInMiles UDF
distributed with DataFu to determine how far each customer is from the three data
centers. You will then calculate the average distance for all customers to each of
these data centers in order to propose the one that will benefit the most customers.
1. Add the tab-delimited file mapping ZIP codes to latitude/longitude points to
HDFS:
$ hadoop fs -mkdir /dualcore/distribution
$ hadoop fs -put $ADIR/data/latlon.tsv \
/dualcore/distribution
2. A script (create_cust_location_data.pig) has been provided to find
the ZIP codes for customers who placed orders during the period of the ad
campaign. It also excludes the ones who are already close to the current facility,
as well as customers in the remote states of Alaska and Hawaii (where orders
are shipped by airplane). The Pig Latin code joins these customers ZIP codes
with the latitude/longitude data set uploaded in the previous step, then writes
those three columns (ZIP code, latitude, and longitude) as the result. Examine
the script to see how it works, and then run it to create the customer location
data in HDFS:
$ pig create_cust_location_data.pig
3. You will use the HaversineDistInMiles function to calculate the distance
from each customer to each of the three proposed warehouse locations. This
function requires us to supply the latitude and longitude of both the customer
and the warehouse. While the script you just executed created the latitude and
longitude for each customer, you must create a data set containing the ZIP code,

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

96

latitude, and longitude for these warehouses. Do this by running the following
UNIX command:
$ egrep '^02118|^63139|^78237' \
$ADIR/data/latlon.tsv > warehouses.tsv
4. Next, add this file to HDFS:
$ hadoop fs -put warehouses.tsv /dualcore/distribution

5. Edit the calc_average_distances.pig file. The UDF is already registered
and an alias for this function named DIST is defined at the top of the script, just
before the two data sets you will use are loaded. You need to complete the rest
of this script:
a. Create a record for every combination of customer and proposed
distribution center location
b. Use the function to calculate the distance from the customer to
the warehouse
c. Calculate the average distance for all customers to each
warehouse
d. Display the result to the screen
6. After you have finished implementing the Pig Latin code described above, run
the script:
$ pig calc_average_distances.pig
Question: Which of these three proposed ZIP codes has the lowest average
mileage to Dualcores customers?

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

97

Lecture 11 Lab: Running Hive Queries

from the Shell, Scripts, and Hue
In this lab you will write HiveQL queries to analyze data in Hive tables that
have been populated with data you placed in HDFS during earlier labs.
IMPORTANT: Since this lab builds on the previous one, it is important that you
successfully complete the previous lab before starting this lab.

Step #1: Running a Query from the Hive Shell

Dualcore ran a contest in which customers posted videos of interesting ways to use
their new tablets. A $5,000 prize will be awarded to the customer whose video
received the highest rating.
However, the registration data was lost due to an RDBMS crash, and the only
information they have is from the videos. The winning customer introduced herself
only as Bridget from Kansas City in her video.
You will need to run a Hive query that identifies the winners record in the customer
database so that Dualcore can send her the $5,000 prize.
1. Change to the directory for this lab:
$ cd $ADIR/exercises/analyzing_sales
2. Start Hive:
$ hive

Hive Prompt
To make it easier to copy queries and paste them into your terminal window, we
do not show the hive> prompt in subsequent steps. Steps prefixed with
$ should be executed on the UNIX command line; the rest should be run in Hive
unless otherwise noted.
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

98

3. Make the query results easier to read by setting the property that will make
Hive show column headers:
set hive.cli.print.header=true;
4. All you know about the winner is that her name is Bridget and she lives in
Kansas City. Use Hive's LIKE operator to do a wildcard search for names such as
"Bridget", "Bridgette" or "Bridgitte". Remember to filter on the customer's city.
Question: Which customer did your query identify as the winner of the $5,000
prize?

Step #2: Running a Query Directly from the Command

Line
You will now run a top-N query to identify the three most expensive products that
Dualcore currently offers.
5. Exit the Hive shell and return to the command line:
quit;
6. Although HiveQL statements are terminated by semicolons in the Hive shell, it is
not necessary to do this when running a single query from the command line
using the -e option. Run the following command to execute the quoted HiveQL
statement:
$ hive -e 'SELECT price, brand, name FROM PRODUCTS
ORDER BY price DESC LIMIT 3'
Question: Which three products are the most expensive?

Step #3: Running a HiveQL Script

The rules for the contest described earlier require that the winner bought the
advertised tablet from Dualcore between May 1, 2013 and May 31, 2013. Before

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

99

Dualcore can authorize the accounting department to pay the $5,000 prize, you
must ensure that Bridget is eligible. Since this query involves joining data from
several tables, its a perfect case for running it as a Hive script.
1. Study the HiveQL code for the query to learn how it works:
$ cat verify_tablet_order.hql
2. Execute the HiveQL script using the hive commands -f option:
$ hive -f verify_tablet_order.hql
Question: Did Bridget order the advertised tablet in May?

Step #4: Running a Query Through Hue and Beeswax

Another way to run Hive queries is through your Web browser using Hues Beeswax
application. This is especially convenient if you use more than one computer or if
you use a device (such as a tablet) that isnt capable of running Hive itself because
it does not require any software other than a browser.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

100

1. Start the Firefox Web browser by clicking the orange and blue icon near the top
of the VM window, just to the right of the System menu. Once Firefox starts, type
http://localhost:8888/ into the address bar, and then hit the enter key.
2. After a few seconds, you should see Hues login screen. Enter training in
both the username and password fields, and then click the Sign In button. If
prompted to remember the password, decline by hitting the ESC key so you can
practice this step again later if you choose.
Although several Hue applications are available through the icons at the top of
the page, the Beeswax query editor is shown by default.
3. Select default from the database list on the left side of the page.
4. Write a query in the text area that will count the number of records in the
customers table, and then click the Execute button.
Question: How many customers does Dualcore serve?
5. Click the Query Editor link in the upper left corner, and then write and run a
query to find the ten states with the most customers.
Question: Which state has the most customers?

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

101

Bonus Lab #1: Calculating Revenue and Profit

Several more questions are described below and you will need to write the HiveQL
code to answer them. You can use whichever method you like best, including Hive
shell, Hive Script, or Hue, to run your queries.

Which top three products has Dualcore sold more of than any other?
Hint: Remember that if you use a GROUP BY clause in Hive, you must group by
all fields listed in the SELECT clause that are not part of an aggregate function.

What was Dualcores total revenue in May, 2013?

What was Dualcores gross profit (sales price minus cost) in May, 2013?

The results of the above queries are shown in cents. Rewrite the gross profit
query to format the value in dollars and cents (e.g., $2000000.00). To do this,
you can divide the profit by 100 and format the result using the PRINTF
function and the format string "$%.2f".

Professors Note~
There are several ways you could write each query, and you can find one solution
for each problem in the bonus_01/sample_solution/ directory.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

102

Lecture 11 Lab: Data Management

with Hive
In this lab you will practice using several common techniques for creating and
populating Hive tables. You will also create and query a table containing each
of the complex field types we studied: array, map, and struct.
IMPORTANT: Since this lab builds on the previous one, it is important that you
successfully complete the previous lab before starting this lab.
Additionally, many of the commands you will run use environmental variables and
relative file paths. It is important that you use the Hive shell, rather than Hue or
another interface, as you work through the steps that follow.

Step #1: Use Sqoops Hive Import Option to Create a

Table
You used Sqoop in an earlier lab to import data from MySQL into HDFS. Sqoop can
also create a Hive table with the same fields as the source table in addition to
importing the records, which saves you from having to write a CREATE TABLE
statement.
1. Change to the directory for this lab:
$ cd $ADIR/exercises/data_mgmt
2. Execute the following command to import the suppliers table from MySQL as
a new Hive-managed table:
$ sqoop import \
--connect jdbc:mysql://localhost/dualcore \
--username training --password training \
--fields-terminated-by '\t' \
--table suppliers \
--hive-import
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

103

3. Start Hive:
$ hive
4. It is always a good idea to validate data after adding it. Execute the Hive query
shown below to count the number of suppliers in Texas:
SELECT COUNT(*) FROM suppliers WHERE state='TX';
The query should show that nine records match.

Step #2: Create an External Table in Hive

You imported data from the employees table in MySQL in an earlier lab, but it
would be convenient to be able to query this from Hive. Since the data already exists
in HDFS, this is a good opportunity to use an external table.

1. Write and execute a HiveQL statement to create an external table for the tab-
delimited records in HDFS at /dualcore/employees. The data format is
shown below:

Field Name Field Type
STRING
emp_id
STRING
fname
STRING
lname
STRING
address
STRING
city
STRING
state
STRING
zipcode
STRING
job_title
STRING
email
STRING
active
INT
salary

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

104

2. Run the following Hive query to verify that you have created the table correctly.
SELECT job_title, COUNT(*) AS num
FROM employees
GROUP BY job_title
ORDER BY num DESC
LIMIT 3;
It should show that Sales Associate, Cashier, and Assistant Manager are the
three most common job titles at Dualcore.

Step #3: Create and Load a Hive-Managed Table

Next, you will create and then load a Hive-managed table with product ratings data.
1. Create a table named ratings for storing tab-delimited records using this
structure:

Field Name Field Type
TIMESTAMP
posted
INT
cust_id
INT
prod_id
TINYINT
rating
STRING
message

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

105

2. Show the table description and verify that its fields have the correct order,
names, and types:
DESCRIBE ratings;
3. Next, open a separate terminal window (File -> Open Terminal) so you can run
the following shell command. This will populate the table directly by using the
hadoop fs command to copy product ratings data from 2012 to that directory
in HDFS:
$ hadoop fs -put $ADIR/data/ratings_2012.txt \
/user/hive/warehouse/ratings
Leave the window open afterwards so that you can easily switch between Hive
and the command prompt.
4. Next, verify that Hive can read the data we just added. Run the following query
in Hive to count the number of records in this table (the result should be 464):
SELECT COUNT(*) FROM ratings;
5. Another way to load data into a Hive table is through the LOAD DATA command.
The next few commands will lead you through the process of copying a local file
to HDFS and loading it into Hive. First, copy the 2013 ratings data to HDFS:
$ hadoop fs -put $ADIR/data/ratings_2013.txt /dualcore
6. Verify that the file is there:
$ hadoop fs -ls /dualcore/ratings_2013.txt
7. Use the LOAD DATA statement in Hive to load that file into the ratings table:
LOAD DATA INPATH '/dualcore/ratings_2013.txt' INTO
TABLE ratings;

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

106

8. The LOAD DATA INPATH command moves the file to the tables directory.
Verify that the file is no longer present in the original directory:
$ hadoop fs -ls /dualcore/ratings_2013.txt
9. Verify that the file is shown alongside the 2012 ratings data in the tables
directory:
$ hadoop fs -ls /user/hive/warehouse/ratings
10. Finally, count the records in the ratings table to ensure that all 21,997 are
available:
SELECT COUNT(*) FROM ratings;

Step #4: Create, Load, and Query a Table with Complex

Fields
Dualcore recently started a loyalty program to reward their best customers.
Dualcore has a sample of the data that contains information about customers who
have signed up for the program, including their phone numbers (as a map), a list of
past order IDs (as an array), and a struct that summarizes the minimum, maximum,
average, and total value of past orders. You will create the table, populate it with the
provided data, and then run a few queries to practice referencing these types of
fields.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

107

1. Run the following statement in Hive to create the table:

CREATE TABLE loyalty_program
(cust_id INT,
fname STRING,
lname STRING,
email STRING,
level STRING,
phone MAP<STRING, STRING>,
order_ids ARRAY<INT>,
order_value STRUCT<min:INT,
max:INT,
avg:INT,
total:INT>)
ROW FORMAT DELIMITED
FIELDS TERMINATED BY '|'
COLLECTION ITEMS TERMINATED BY ','
MAP KEYS TERMINATED BY ':';
2. Examine the data in loyalty_data.txt to see how it corresponds to the
fields in the table and then load it into Hive:
LOAD DATA LOCAL INPATH 'loyalty_data.txt' INTO TABLE
loyalty_program;
3. Run a query to select the HOME phone number (Hint: Map keys are case-
sensitive) for customer ID 1200866. You should see 408-555-4914 as the result.
4. Select the third element from the order_ids array for customer ID 1200866
(Hint: Elements are indexed from zero). The query should return 5278505.
5. Select the total attribute from the order_value struct for customer ID
1200866. The query should return 401874.


Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

108

Bonus Lab #1: Alter and Drop a Table

1. Use ALTER TABLE to rename the level column to status.
2. Use the DESCRIBE command on the loyalty_program table to verify the
change.
3. Use ALTER TABLE to rename the entire table to reward_program.
4. Although the ALTER TABLE command often requires that we make a
corresponding change to the data in HDFS, renaming a table or column does not.
You can verify this by running a query on the table using the new names (the
result should be SILVER):
SELECT status FROM reward_program WHERE cust_id =
1200866;
5. As sometimes happens in the corporate world, priorities have shifted and the
program is now canceled. Drop the reward_program table.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

109

Lecture 12 Lab: Gaining Insight with

Sentiment Analysis
In this optional lab, you will use Hive's text processing features to analyze
customers comments and product ratings. You will uncover problems and
propose potential solutions.

IMPORTANT: Since this lab builds on the previous one, it is important that you
successfully complete the previous lab before starting this lab.

Background Information
Customer ratings and feedback are great sources of information for both customers
and retailers like Dualcore. However, customer comments are typically free-form
text and must be handled differently. Fortunately, Hive provides extensive support
for text processing.

Step #1: Analyze Numeric Product Ratings

Before delving into text processing, you will begin by analyzing the numeric ratings
customers have assigned to various products.
1. Change to the directory for this lab:
$ cd $ADIR/exercises/sentiment

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

110

2. Start Hive and use the DESCRIBE command to remind yourself of the tables
structure.
3. We want to find the product that customers like most, but must guard against
being misled by products that have few ratings assigned. Run the following
query to find the product with the highest average among all those with at least
50 ratings:
SELECT prod_id, FORMAT_NUMBER(avg_rating, 2) AS
avg_rating
FROM (SELECT prod_id, AVG(rating) AS avg_rating,
COUNT(*) AS num
FROM ratings
GROUP BY prod_id) rated
WHERE num >= 50
ORDER BY avg_rating DESC
LIMIT 1;
4. Rewrite, and then execute, the query above to find the product with the lowest
average among products with at least 50 ratings. You should see that the result
is product ID 1274673 with an average rating of 1.10.

Step #2: Analyze Rating Comments

We observed earlier that customers are very dissatisfied with one of the products
that Dualcore sells. Although numeric ratings can help identify which product that is,
they dont tell Dualcore why customers dont like the product. We could simply read
through all the comments associated with that product to learn this information, but
that approach doesnt scale. Next, you will use Hives text processing support to
analyze the comments.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

111

1. The following query normalizes all comments on that product to lowercase,

breaks them into individual words using the SENTENCES function, and passes
those to the NGRAMS function to find the five most common bigrams (two-word
combinations). Run the query in Hive:
SELECT EXPLODE(NGRAMS(SENTENCES(LOWER(message)), 2, 5))
AS bigrams
FROM ratings
WHERE prod_id = 1274673;
2. Most of these words are too common to provide much insight, though the word
expensive does stand out in the list. Modify the previous query to find the five
most common trigrams (three-word combinations), and then run that query in
Hive.
3. Among the patterns you see in the result is the phrase ten times more. This
might be related to the complaints that the product is too expensive. Now that
youve identified a specific phrase, look at a few comments that contain it by
running this query:
SELECT message
FROM ratings
WHERE prod_id = 1274673
AND message LIKE '%ten times more%'
LIMIT 3;
You should see three comments that say, Why does the red one cost ten times
more than the others?

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

112

4. We can infer that customers are complaining about the price of this item, but
the comment alone doesnt provide enough detail. One of the words (red) in
that comment was also found in the list of trigrams from the earlier query.
Write and execute a query that will find all distinct comments containing the
word red that are associated with product ID 1274673.
5. The previous step should have displayed two comments:

What is so special about red?

Why does the red one cost ten times more than the others?

The second comment implies that this product is overpriced relative to similar
products. Write and run a query that will display the record for product ID
1274673 in the products table.
6. Your query should have shown that the product was a 16GB USB Flash Drive
(Red) from the Orion brand. Next, run this query to identify similar products:
SELECT *
FROM products
WHERE name LIKE '%16 GB USB Flash Drive%'
AND brand='Orion';
The query results show that there are three almost identical products, but the
product with the negative reviews (the red one) costs about ten times as much
as the others, just as some of the comments said.
Based on the cost and price columns, it appears that doing text processing on
the product ratings has helped Dualcore uncover a pricing error.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

113

Lecture 12 Lab: Data Transformation

with Hive
In this lab you will create and populate a table with log data from Dualcores
Web server. Queries on that data will reveal that many customers abandon
their shopping carts before completing the checkout process. You will create
several additional tables, using data from a TRANSFORM script and a supplied
UDF, which you will use later to analyze how Dualcore could turn this problem
into an opportunity.

IMPORTANT: Since this lab builds on the previous one, it is important that you
successfully complete the previous lab before starting this lab.

Step #1: Create and Populate the Web Logs Table

Typical log file formats are not delimited, so you will need to use the RegexSerDe
and specify a pattern Hive can use to parse lines into individual fields you can then
query.
1. Change to the directory for this lab:
$ cd $ADIR/exercises/transform
2. Examine the create_web_logs.hql script to get an idea of how it uses a
RegexSerDe to parse lines in the log file (an example log line is shown in the
comment at the top of the file). When you have examined the script, run it to
create the table in Hive:
$ hive -f create_web_logs.hql

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

114

3. Populate the table by adding the log file to the tables directory in HDFS:
$ hadoop fs -put $ADIR/data/access.log
/dualcore/web_logs
4. Start the Hive shell in another terminal window
5. Verify that the data is loaded correctly by running this query to show the top
three items users searched for on Dualcores Web site:
SELECT term, COUNT(term) AS num FROM
(SELECT LOWER(REGEXP_EXTRACT(request,
'/search\\?phrase=(\\S+)', 1)) AS term
FROM web_logs
WHERE request REGEXP '/search\\?phrase=') terms
GROUP BY term
ORDER BY num DESC
LIMIT 3;
You should see that it returns tablet (303), ram (153) and wifi (148).
Note: The REGEXP operator, which is available in some SQL dialects, is similar
to LIKE, but uses regular expressions for more powerful pattern matching. The
REGEXP operator is synonymous with the RLIKE operator.

Step #2: Analyze Customer Checkouts

Youve just queried the logs to see what users search for on Dualcores Web site, but
now youll run some queries to learn whether they buy. As on many Web sites,
customers add products to their shopping carts and then follow a checkout
process to complete their purchase. Since each part of this four-step process can be
identified by its URL in the logs, we can use a regular expression to easily identify
them:

Step Request URL

Description

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

115

1
2
3
4

/cart/checkout/step1-viewcart
/cart/checkout/step2-shippingcost
/cart/checkout/step3-payment
/cart/checkout/step4-receipt

View list of items added to cart

Notify customer of shipping cost
Gather payment information
Show receipt for completed order

1. Run the following query in Hive to show the number of requests for each step of
the checkout process:
SELECT COUNT(*), request
FROM web_logs
WHERE request REGEXP '/cart/checkout/step\\d.+'
GROUP BY request;
The results of this query highlight a major problem. About one out of every
three customers abandons their cart after the second step. This might mean
millions of dollars in lost revenue, so lets see if we can determine the cause.
2. The log files cookie field stores a value that uniquely identifies each user
session. Since not all sessions involve checkouts at all, create a new table
containing the session ID and number of checkout steps completed for just
those sessions that do:
CREATE TABLE checkout_sessions AS
SELECT cookie, ip_address, COUNT(request) AS
steps_completed
FROM web_logs
WHERE request REGEXP '/cart/checkout/step\\d.+'
GROUP BY cookie, ip_address;

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

116

3. Run this query to show the number of people who abandoned their cart after
each step:
SELECT steps_completed, COUNT(cookie) AS num
FROM checkout_sessions
GROUP BY steps_completed;
You should see that most customers who abandoned their order did so after the
second step, which is when they first learn how much it will cost to ship their
order.

Step #3: Use TRANSFORM for IP Geolocation

Based on what you've just seen, it seems likely that customers abandon their carts
due to high shipping costs. The shipping cost is based on the customer's location and
the weight of the items they've ordered. Although this information is not in the
database (since the order wasn't completed), we can gather enough data from the
logs to estimate them.
We don't have the customer's address, but we can use a process known as "IP
geolocation" to map the computer's IP address in the log file to an approximate
physical location. Since this isn't a built-in capability of Hive, you'll use a provided
Python script to TRANSFORM the ip_address field from the
checkout_sessions table to a ZIP code, as part of HiveQL statement that creates
a new table called cart_zipcodes.

Regarding TRANSFORM and UDF Examples in this

Exercise
During this lab, you will use a Python script for IP geolocation and a UDF to
calculate shipping costs. Both are implemented merely as a simulation
compatible with the fictitious data we use in class and intended to work even
when Internet access is unavailable. The focus of these labs is on how to use
external scripts and UDFs, rather than how the code for the examples works
internally.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

117

1. Examine the create_cart_zipcodes.hql script and observe the following:

a. It creates a new table called cart_zipcodes based on select
statement.
b. That select statement transforms the ip_address, cookie, and
steps_completed fields from the checkout_sessions table
using a Python script.
c. The new table contains the ZIP code instead of an IP address, plus the
other two fields from the original table.
2. Examine the ipgeolocator.py script and observe the following:
a. Records are read from Hive on standard input.
b. The script splits them into individual fields using a tab delimiter.
c. The ip_addr field is converted to zipcode, but the cookie and
steps_completed fields are passed through unmodified.
d. The three fields in each output record are delimited with tabs are
printed to standard output.
3. Run the script to create the cart_zipcodes table:
$ hive -f create_cart_zipcodes.hql

Step #4: Extract List of Products Added to Each Cart

As described earlier, estimating the shipping cost also requires a list of items in the
customers cart. You can identify products added to the cart since the request URL
looks like this (only the product ID changes from one record to the next):

/cart/additem?productid=1234567

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

118

1. Write a HiveQL statement to create a table called cart_items with two fields:
cookie and prod_id based on data selected the web_logs table. Keep the
following in mind when writing your statement:
a. The prod_id field should contain only the seven-digit product ID
(Hint: Use the REGEXP_EXTRACT function)
b. Add a WHERE clause with REGEXP using the same regular expression
as above so that you only include records where customers are adding
items to the cart.

Professors Note~
If you need a hint on how to write the statement, look at the file:
sample_solution/create_cart_items.hql
2. Execute the HiveQL statement from you just wrote.
3. Verify the contents of the new table by running this query:
SELECT COUNT(DISTINCT cookie) FROM cart_items WHERE
prod_id=1273905;

Professors Note~
If this doesnt return 47, then compare your statement to the file:
sample_solution/create_cart_items.hql. Make the necessary
corrections, and then re-run your statement (after dropping the cart_items
table).

Step #5: Create Tables to Join Web Logs with Product

Data
You now have tables representing the ZIP codes and products associated with
checkout sessions, but you'll need to join these with the products table to get the
weight of these items before you can estimate shipping costs. In order to do some

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

119

more analysis later, well also include total selling price and total wholesale cost in
addition to the total shipping weight for all items in the cart.
1. Run the following HiveQL to create a table called cart_orders with the
information:
CREATE TABLE cart_orders AS
SELECT z.cookie, steps_completed, zipcode,
SUM(shipping_wt) as total_weight,
SUM(price) AS total_price,
SUM(cost) AS total_cost
FROM cart_zipcodes z
JOIN cart_items i
ON (z.cookie = i.cookie)
JOIN products p
ON (i.prod_id = p.prod_id)
GROUP BY z.cookie, zipcode, steps_completed;

Step #6: Create a Table Using a UDF to Estimate

Shipping Cost
We finally have all the information we need to estimate the shipping cost for each
abandoned order. You will use a Hive UDF to calculate the shipping cost given a ZIP
code and the total weight of all items in the order.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

120

1. Before you can use a UDF, you must add it to Hives classpath. Run the following
command in Hive to do that:
ADD JAR geolocation_udf.jar;
2. Next, you must register the function with Hive and provide the name of the UDF
class as well as the alias you want to use for the function. Run the Hive
command below to associate our UDF with the alias CALC_SHIPPING_COST:
CREATE TEMPORARY FUNCTION CALC_SHIPPING_COST AS
'com.cloudera.hive.udf.UDFCalcShippingCost';

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

121

3. Now create a new table called cart_shipping that will contain the session ID,
number of steps completed, total retail price, total wholesale cost, and the
estimated shipping cost for each order based on data from the cart_orders
table:
CREATE TABLE cart_shipping AS
SELECT cookie, steps_completed, total_price,
total_cost,
CALC_SHIPPING_COST(zipcode, total_weight) AS
shipping_cost
FROM cart_orders;
4. Finally, verify your table by running the following query to check a record:
SELECT * FROM cart_shipping WHERE
cookie='100002920697';
This should show that session as having two completed steps, a total retail price
of $263.77, a total wholesale cost of $236.98, and a shipping cost of $9.09.
Note: The total_price, total_cost, and shipping_cost columns in the
cart_shipping table contain the number of cents as integers. Be sure to
divide results containing monetary amounts by 100 to get dollars and cents.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

122

Lecture 13 Lab: Interactive Analysis

with Impala
In this lab you will examine abandoned cart data using the tables created in
the previous lab. You will use Impala to quickly determine how much lost
revenue these abandoned carts represent and use several what if scenarios
to determine whether Dualcore should offer free shipping to encourage
customers to complete their purchases.
IMPORTANT: Since this lab builds on the previous one, it is important that you
successfully complete the previous lab before starting this lab.

Step #1: Start the Impala Shell and Refresh the Cache
1. Issue the following commands to start Impala, then change to the directory for
this lab:
$ sudo service impala-server start
$ sudo service impala-state-store start
$ cd $ADIR/exercises/interactive
2. First, start the Impala shell:
$ impala-shell
3. Since you created tables and modified data in Hive, Impalas cache of the
metastore is outdated. You must refresh it before continuing by entering the
following command in the Impala shell:
REFRESH;

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

123

Step #2: Calculate Lost Revenue

1. First, youll calculate how much revenue the abandoned carts represent.
Remember, there are four steps in the checkout process, so only records in the
cart_shipping table with a steps_completed value of four represent a
completed purchase:
SELECT SUM(total_price) AS lost_revenue
FROM cart_shipping
WHERE steps_completed < 4;

Lost Revenue From Abandoned Shipping Carts

cart_shipping
cookie

steps_completed

total_price total_cost shipping_cost

100054318085

6899

6292

425

100060397203

19218

17520

552

100062224714

7609

7155

556

100064732105

53137

50685

839

100107017704

44928

44200

720

...

...

...

...

...

Sum of total_price where steps_completed < 4

You should see that abandoned carts mean that Dualcore is potentially losing
out on more than $2 million in revenue! Clearly its worth the effort to do
further analysis.
Note: The total_price, total_cost, and shipping_cost columns in the
cart_shipping table contain the number of cents as integers. Be sure to
divide results containing monetary amounts by 100 to get dollars and cents.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

124

2. The number returned by the previous query is revenue, but what counts is
profit. We calculate gross profit by subtracting the cost from the price. Write
and execute a query similar to the one above, but which reports the total lost
profit from abandoned carts.

Professors Note~

If you need a hint on how to write this query, you can check the file:
sample_solution/abandoned_checkout_profit.sql
After running your query, you should see that Dualcore is potentially losing
$111,058.90 in profit due to customers not completing the checkout process.

3. How does this compare to the amount of profit Dualcore receives from
customers who do complete the checkout process? Modify your previous query
to consider only those records where steps_completed = 4, and then
execute it in the Impala shell.
Professors Note~
Check sample_solution/completed_checkout_profit.sql for a hint.

The result should show that Dualcore earns a total of $177,932.93 on completed
orders, so abandoned carts represent a substantial proportion of additional
profits.
4. The previous two queries show the total profit for abandoned and completed
orders, but these arent directly comparable because there were different
numbers of each. It might be the case that one is much more profitable than the
other on a per-order basis. Write and execute a query that will calculate the
average profit based on the number of steps completed during the checkout
process.

Professors Note~
If you need help writing this query, check the file:
sample_solution/checkout_profit_by_step.sql
You should observe that carts abandoned after step two represent an even
higher average profit per order than completed orders.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

125

Step #3: Calculate Cost/Profit for a Free Shipping Offer

You have observed that most carts and the most profitable carts are abandoned
at the point where the shipping cost is displayed to the customer. You will now run
some queries to determine whether offering free shipping, on at least some orders,
would actually bring in more revenue assuming this offer prompted more
customers to finish the checkout process.
1. Run the following query to compare the average shipping cost for orders
abandoned after the second step versus completed orders:
SELECT steps_completed, AVG(shipping_cost) AS ship_cost
FROM cart_shipping
WHERE steps_completed = 2 OR steps_completed = 4
GROUP BY steps_completed;

Average Shipping Cost for Carts Abandoned After Steps 2 and 4

cart_shipping
cookie

steps_completed

total_price total_cost shipping_cost

100054318085

6899

6292

425

100060397203

19218

17520

552

100062224714

7609

7155

556

100064732105

53137

50685

839

100107017704

44928

44200

720

...

...

...

...

...

Average of shipping_cost where steps_completed = 2 or 4

You will see that the shipping cost of abandoned orders was almost 10%
higher than for completed purchases. Offering free shipping, at least for
some orders, might actually bring in more money than passing on the
cost and risking abandoned orders.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

126

2. Run the following query to determine the average profit per order over the
entire month for the data you are analyzing in the log file. This will help you to
determine whether Dualcore could absorb the cost of offering free shipping:
SELECT AVG(price - cost) AS profit
FROM products p
JOIN order_details d
ON (d.prod_id = p.prod_id)
JOIN orders o
ON (d.order_id = o.order_id)
WHERE YEAR(order_date) = 2013
AND MONTH(order_date) = 05;

Average Profit per Order, May 2013

products

order_details

prod_id

price cost

1273641

1839

1275

1273642

1949

1273643

orders

product_id

order_id

order_date

6547914

1273641

6547914

2013-05-01 00:02:08

721

6547914

1273644

6547915

2013-05-01 00:02:55

6547914

1273645

6547916

2013-05-01 00:06:15
2013-06-12 00:10:41

order_id

2149

845

1273644

2029

763

6547915

1273645

6547917

1273645

1909

1234

6547916

1273641

6547918

2013-06-12 00:11:30

...

...

...

...

...

...

...

Average the profit...

... on orders made in

May, 2013

You should see that the average profit for all orders during May was
$7.80. An earlier query you ran showed that the average shipping cost
was $8.83 for completed orders and $9.66 for abandoned orders, so
clearly Dualcore would lose money by offering free shipping on all
orders. However, it might still be worthwhile to offer free shipping on
orders over a certain amount.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

127

3. Run the following query, which is a slightly revised version of the previous one,
to determine whether offering free shipping only on orders of $10 or more
would be a good idea:
SELECT AVG(price - cost) AS profit
FROM products p
JOIN order_details d
ON (d.prod_id = p.prod_id)
JOIN orders o
ON (d.order_id = o.order_id)
WHERE YEAR(order_date) = 2013
AND MONTH(order_date) = 05
AND PRICE >= 1000;

You should see that the average profit on orders of $10 or more was
$9.09, so absorbing the cost of shipping would leave very little profit.

4. Repeat the previous query, modifying it slightly each time to find the average
profit on orders of at least $50, $100, and $500.

You should see that there is a huge spike in the amount of profit for
orders of $500 or more (Dualcore makes $111.05 on average for these
orders).

5. How much does shipping cost on average for orders totaling $500 or more?
Write and run a query to find out.
Professors Note~
The file sample_solution/avg_shipping_cost_50000.sql contains
the solution.

You should see that the average shipping cost is $12.28, which happens
to be about 11% of the profit brought in on those orders.

6. Since Dualcore wont know in advance who will abandon their cart, they would
have to absorb the $12.28 average cost on all orders of at least $500. Would the
extra money they might bring in from abandoned carts offset the added cost of
Copyright 2010-2014 Cloudera, Inc. All rights reserved.
Not to be reproduced without prior written consent.

128

free shipping for customers who would have completed their purchases
anyway? Run the following query to see the total profit on completed
purchases:
SELECT SUM(total_price - total_cost) AS total_profit
FROM cart_shipping
WHERE total_price >= 50000
AND steps_completed = 4;

After running this query, you should see that the total profit for
completed orders is $107,582.97.

7. Now, run the following query to find the potential profit, after subtracting
shipping costs, if all customers completed the checkout process:
SELECT gross_profit - total_shipping_cost AS
potential_profit
FROM (SELECT
SUM(total_price - total_cost) AS
gross_profit,
SUM(shipping_cost) AS total_shipping_cost
FROM cart_shipping
WHERE total_price >= 50000) large_orders;
Since the result of $120,355.26 is greater than the current profit of $107,582.97
Dualcore currently earns from completed orders, it appears that they could earn
nearly $13,000 more by offering free shipping for all orders of at least $500.
Congratulations! Your hard work analyzing a variety of data with Hadoops
tools has helped make Dualcore more profitable than ever.

This is the end of the lab.

Copyright 2010-2014 Cloudera, Inc. All rights reserved.

Not to be reproduced without prior written consent.

129