Skip to content

A Spark-based data comparison tool at scale which facilitates software development engineers to compare a plethora of pair combinations of possible data sources. Multiple execution modes in multiple environments enable the user to generate a diff report as a Java/Scala-friendly DataFrame or as a file for future use. Comes with out of the box Spa…

License

FINRAOS/MegaSparkDiff

Repository files navigation

Black Duck Security Risk

MegaSparkDiff

MegaSparkDiff is an open source tool that helps you compare any pair
combination of data sets that are of the following types: 

(HDFS, JDBC, S3, Hbase, Text Files, Hive, Json, DynamoDb). 

MegaSparkDiff can run on 
(a) Amazon EMR (Elastic Map Reduce),
(b) Amazon EC2 instances and cloud environments with compatible Spark distributions.
(c) DataBricks Interactive Notebooks with Visualizations via displayhtml function.

How to Use from Within a Java or Scala Project

<dependency>
    <groupId>org.finra.megasparkdiff</groupId>
    <artifactId>mega-spark-diff</artifactId>
    <version>0.4.0</version>
</dependency>

SparkFactory

parallelizes source/target data.

The data sources can be in following forms:
    Text File
    HDFS File
    SQL query over a JDBC data source
    Hive Table
    Json File
    DynamoDb Table

SparkCompare

Compares pair combinations of supported sources,
Please note in case of comparing a schema-based source to a non-schema based source, the SparkCompare
class will attempt to flatten the schema based source to delimited values and then do the comparison. The delimiter
can be specified while launching the compare job.

How to use via shell script in EMR

There will exist a shell script named a3a.sh that will wrap around
this Java/Scala project.  This script will accept several parameters
related to source definitions, output destination, and run
configurations, as well as which two data sets to compare.

The parameters are as follows:
    -ds=<data_source_folder>: The folder where the database
        connection parameters and data queries reside
    -od=<output_directory>: The directory where MegaSparkDiff will write
        its output
    -rc=<run_config_file_name>: The file that will be used to load
        any special run and Spark configurations.  This parameter is
        optional
        
To specify a data set to compare, pass in the name of one of the
data queries found in a config file inside <data_source_folder>
prepended by "--".  The program will execute the queries assigned to
the names passed into the command line, store them into tables, and
perform the comparison.

Example call:
    ./msd.sh -ds=./data_sources/ -od=output --shraddha --carlos
    
Additionally, the user will have the option to add JDBC Driver jar
files by including them in the classpath.  This is to enable them to
extract from whichever database they choose.

Run tests on Windows

  1. Download Hadoop winutils
  2. Extract to some path, e.g. C:\Users\MegaSparkDiffFan\bin
  3. Run tests while defining hadoop.home.dir, e.g. mvn test -Dhadoop.home.dir=C:\Users\MegaSparkDiffFan

About

A Spark-based data comparison tool at scale which facilitates software development engineers to compare a plethora of pair combinations of possible data sources. Multiple execution modes in multiple environments enable the user to generate a diff report as a Java/Scala-friendly DataFrame or as a file for future use. Comes with out of the box Spa…

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages