In this tutorial, we will learn to store data files using Ambari HDFS Files View. We will implement pig latin scripts to process, analyze and manipulate data files of truck drivers statistics. Let’s build our own Pig Latin Scripts now.
Pig is a high level scripting language that is used with Apache Hadoop. Pig excels at describing data analysis problems as data flows. Pig is complete in that you can do all the required data manipulations in Apache Hadoop with Pig. In addition through the User Defined Functions(UDF) facility in Pig you can have Pig invoke code in many languages like JRuby, Jython and Java. Conversely you can execute Pig scripts in other languages. The result is that you can use Pig as a component to build larger and more complex applications that tackle real business problems.
A good example of a
Pig application is the
ETL transaction model that describes how a process will extract data from a source, transform it according to a rule set and then load it into a datastore. Pig can ingest data from files, streams or other sources using the User Defined Functions(UDF). Once it has the data it can perform select, iteration, and other transforms over the data. Again the UDF feature allows passing the data to more complex algorithms for the transform. Finally Pig can store the results into the Hadoop Data File System.
Pig scripts are translated into a series of
MapReduce jobs that are run on the
Apache Hadoop cluster. As part of the translation the Pig interpreter does perform optimizations to speed execution on Apache Hadoop. We are going to write a Pig script that will do our data analysis task.
Tez – Hindi for “speed” provides a general-purpose, highly customizable framework that creates simplifies data-processing tasks across both small scale (low-latency) and large-scale (high throughput) workloads in Hadoop. It generalizes the MapReduce paradigm to a more powerful framework by providing the ability to execute a complex DAG (directed acyclic graph) of tasks for a single job so that projects in the Apache Hadoop ecosystem such as Apache Hive, Apache Pig and Cascading can meet requirements for human-interactive response times and extreme throughput at petabyte scale (clearly MapReduce has been a key driver in achieving this).
We are going to read in a truck driver statistics files. We are going to compute the sum of hours and miles logged driven by a truck driver for an year. Once we have the sum of hours and miles logged, we will extend the script to translate a driver id field into the name of the drivers by joining two different files.
Download the driver data file from here.
Once you have the file you will need to
unzip the file into a directory. We will be uploading two csv files –
We start by selecting the
HDFS Files view from the Off-canvas menu at the top. The
HDFS Files view allows us to view the Hortonworks Data Platform(HDP) file store. This is separate from the local file system. For the Hortonworks Sandbox, it will be part of the file system in the Hortonworks Sandbox VM.
/user/maria_dev and click on the Upload button to select the files we want to upload into the Hortonworks Sandbox environment.
Click on the browse button to open a dialog box. Navigate to where you stored the
drivers.csv file on your local disk and select
drivers.csv and click again
upload. Do the same thing for
timesheet.csv. When you are done you will see there are two new files in your directory.
Now that we have our data files, we can start writing our
Pig script. Click on the
Pig View from the Off-canvas menu.
You can also explore Grunt shell from the terminal which is used to write Pig Latin scripts. There are 3 execute modes of accessing Grunt shell:
pig -x localto enter the shell
pig -x mapreduceto enter the shell
pig -x tezto enter the shell
Default is mapreduce, so if you just type
pig, it will use mapreduce as the execution mode.
Explore this link to explore more about the grunt shell.
We see the
Pig user interface in our browser window. On the left we can choose between our
saved Pig Scripts,
UDFs and the
Pig Jobs executed in the past. To the right of this menu bar we see our
saved Pig Scripts.
To get started push the button
"New Script" at the top right and fill in a name for your script. If you leave the gap “Script HDFS Location” empty, it will be filled automatically.
After clicking on “create”, a new page opens.
At the center is the composition area where we will be writing our script. At top right of the composition area are buttons to
Explain and perform a Syntax check of the current script.
At the left are buttons to save, copy or delete the script and at the very bottom we can add a argument.
The first thing we need to do is load the data. We use the load statement for this. The
PigStorage function is what does the loading and we pass it a
comma as the data
delimiter. Our code is:
drivers = LOAD 'drivers.csv' USING PigStorage(',');
To filter out the first row of the data we have to add this line:
raw_drivers = FILTER drivers BY $0>1;
The next thing we want to do is name the fields. We will use a
FOREACH statement to iterate through the raw_drivers data object. We can use
Pig Helper that is at the bottom of the composition area to provide us with a template. We will click on
Pig Helper, select Data processing functions and then click on the
FOREACH template. We can then replace each element by hitting the tab key.
FOREACH statement will iterate through the raw_drivers data object and
GENERATE pulls out selected fields and assigns them names. The new data object we are creating is then named
driver_details. Our code will now be:
drivers_details = FOREACH raw_drivers GENERATE $0 AS driverId, $1 AS name;
timesheet data and then filter out the first row of the data to remove column headings and then use
FOREACH statement to iterate each row and
GENERATE to pull out selected fields and assign them names.
timesheet = LOAD 'timesheet.csv' USING PigStorage(','); raw_timesheet = FILTER timesheet by $0>1; timesheet_logged = FOREACH raw_timesheet GENERATE $0 AS driverId, $2 AS hours_logged, $3 AS miles_logged;
The next line of code is a
GROUP statement that groups the elements in
timesheet_logged by the
driverId field. So the
grp_logged object will then be indexed by
driverId. In the next statement as we iterate through
grp_logged we will go through driverId by driverId. Type in the code:
grp_logged = GROUP timesheet_logged by driverId;
In the next
FOREACH statement, we are going to find the sum of hours and miles logged by each driver. The code for this is:
sum_logged = FOREACH grp_logged GENERATE group as driverId, SUM(timesheet_logged.hours_logged) as sum_hourslogged, SUM(timesheet_logged.miles_logged) as sum_mileslogged;
Now that we have the sum of hours and miles logged, we need to join this with the
driver_details data object so we can pick up the name of the driver. The result will be a dataset with
driverId, name, hours logged and miles logged. At the end we
DUMP the data to the output.
join_sum_logged = JOIN sum_logged by driverId, drivers_details by driverId; join_data = FOREACH join_sum_logged GENERATE $0 as driverId, $4 as name, $1 as hours_logged, $2 as miles_logged; dump join_data;
Let’s take a look at our script. The first thing to notice is we never really address single rows of data to the left of the equals sign and on the right we just describe what we want to do for each row. We just assume things are applied to all the rows. We also have powerful operators like
JOIN to sort rows by a key and to build new data objects.
At this point we can save our script. Let’s execute our code by clicking on the
Execute button at the top right of the composition area, which opens a new page.
As the jobs are run we will get status boxes where we will see logs, error message, the output of our script and our code at the bottom.
If you scroll down to the “Logs…” and click on the link you can see the log file of your jobs. We should always check the Logs to check if your script was executed correctly.
NOTE : Notice, in the Logs, the script typically takes a little around 47 seconds to finish on our single node pseudo cluster.
So we have created a simple
Pig script that reads in some comma separated data.
Once we have that set of records in Pig we pull out the driverId, hours logged and miles logged fields from each row.
We then group them by driverId with one statement, GROUP.
Then we find the sum of hours and miles logged for each driverId.
This is finally mapped to the driver name by joining two datasets and we produce our final dataset.
As mentioned before
Pig operates on data flows. We consider each group of rows together and we specify how we operate on them as a group. As the datasets get larger and/or add fields our
Pig script will remain pretty much the same because it is concentrating on how we want to manipulate the data.
drivers = LOAD 'drivers.csv' USING PigStorage(','); raw_drivers = FILTER drivers BY $0>1; drivers_details = FOREACH raw_drivers GENERATE $0 AS driverId, $1 AS name; timesheet = LOAD 'timesheet.csv' USING PigStorage(','); raw_timesheet = FILTER timesheet by $0>1; timesheet_logged = FOREACH raw_timesheet GENERATE $0 AS driverId, $2 AS hours_logged, $3 AS miles_logged; grp_logged = GROUP timesheet_logged by driverId; sum_logged = FOREACH grp_logged GENERATE group as driverId, SUM(timesheet_logged.hours_logged) as sum_hourslogged, SUM(timesheet_logged.miles_logged) as sum_mileslogged; join_sum_logged = JOIN sum_logged by driverId, drivers_details by driverId; join_data = FOREACH join_sum_logged GENERATE $0 as driverId, $4 as name, $1 as hours_logged, $2 as miles_logged; dump join_data;
Let’s run the same Pig script with Tez by clicking on
Execute on Tez button near the
Notice the time after the execution completes:
On our machine it took around 16 seconds with Pig using the Tez engine. That is more than 2X faster than Pig using MapReduce even without any specific optimization in the script for Tez.
Tez definitely lives up to it’s name.