Requirement

In real time scenario, data files contain many records. Also, there may be many data files available. In that case, it’s good to find a suitable approach to find out the output. Here, we want total number of records available in data files. So the requirement is to how to find the number of records using Map Reduce.

Components Involved

• HDFS: Here it is used to store the input data which will get pass to the Map Reduce as an input and also store the map reduce output.
• Map Reduce: Here it is used to process the data/file available in HDFS and store the output to the HDFS.

Sample Data

Let’s have a look at the sample data, how it will look like:

Here, the sample data contains user’s info which is a comma(,) separated. You can download the sample data file from here.

Solution

The solution has many steps from setting up to the output of the code to the execution and validates the output. Let’s go ahead one by one:

Step 1: Input Data Preparation

The first step includes data preparation. In real time scenario, you will have your data. The data preparation means that it should present at the location from where our map reduce job can get it as an input file.

Let’s keep the sample data file at the local path. In my case, file name is “sampledataForDuplicate” and local path is “/home/NN/HadoopRepo/MapReduce/resources/recordCount”.

We need to move this data file to an HDFS location which will be input path for the map reduce.

It will copy the data file from local to HDFS location.

Step 2: Create Maven Project

In order to write map reduce program, create a maven project. We will add the dependency for the required package.

Follow the below steps in order to create the maven project:

• Open Eclipse
• Create a maven project

Step 3: Resolve Dependency

Add below dependency in pom.XML file and resolve dependency using cmd:

Step 4: Write Mapper

Once you are done with all above steps, write a mapper class which will take an input file. It will read the file and store each word of the file with key-value pair. Here using a java program to write the mapper.

Step 5: Write Reducer

In this step, we will take the mapper data as an input and process it. The actual logic has been written here.

Find the code below:

Step 6: Write Driver

In order to execute the mapper and reducer, let’s create a driver class which will call mapper and reducer. Find below the driver class code:

Step 7: Package Preparation

In this step, we will create a package(.jar) of the project. Follow the below steps:

• Open CMD
• Head to the maven project
• Use the command: mvn package

You can also download the package which I have built for this requirement:

Step 8: Execution

All the setup has been done. Let’s execute the job and validate the output. In order to execute the map reduce, use below command:
Format: hadoop jar <path of jar> <driver class with package name> <input data path of HDFS> <output path at HDFS>

Ex:
hadoop jar /home/NN/HadoopRepo/MapReduce/MapReduceForRecordCount-0.0.1-SNAPSHOT.jar com.bdp.mapreduce.recordcount.driver.RecordCountDriver /user/bdp/mapreduce/recordCount /user/bdp/mapreduce/out/recordCount

Step 9: Validate Output

Check the output at HDFS path.

Wrapping Up

In this post, we have written a map reduce to count the number of records of the data file(s). This is similar to SQL count(*) command. Here, this MapReduce job contains three class map, reduce and driver. Mapper reads the input, Reducer counts the records and Driver running setting all the configuration to run the map reduce.