This is the Cascading.Multitool (Multitool) application.

Multitool provides a simple command line interface for building data processing jobs. Think of this as grep, sed, and awk for Hadoop, which also supports joins between multiple data-sets.

For example, with $HADOOP_HOME/bin/ in your PATH, the following command,

$ hadoop jar multitool-<release-date>.jar source=input.txt select=Monday sink=outputDir

will start a Hadoop job to read from the source file input.txt, grep all lines with the word Monday, then output the results into the outputDir directory.

Multitool will inherit the underlying Hadoop configuration, so if the default FileSystem is HDFS, all paths will be relative to the cluster filesystem, not local. Using fully qualified urls will override the defaults (file://some/path or s3n:/bucket/file).

This application is built with Cascading.

Cascading is a feature rich API for defining and executing complex, scale-free, and fault tolerant data processing workflows on a Hadoop cluster. It can be found at http://www.cascading.org/

This step is not necessary if you wish to run Multitool directly from the uncompressed distribution folder or Multitool was pre-installed with your Hadoop distribution.

To see if Multitool has already been added to your PATH, type:

$ which multitool

To install for all users into /usr/local/bin:

$ sudo ./bin/multitool install

or for the current user only into ~/.multitool:

$ ./bin/multitool install

For detailed instructions:

$ ./bin/multitool help install

Choose the method that best suites your environment.

If you are running Multitool on AWS Elastic MapReduce, you need to follow the Elastic MapReduce instructions on the AWS site, which typically expect the multitool-<release-date>.jar to be uploaded to AWS S3.

The environment variable HADOOP_HOME should always be set first before using Multitool.

To run from the command line with the jar, Hadoop should be in the path:

$ hadoop jar multitool-<release-date>.jar <args>

...or if Multitool has been installed based on the instructions above:

$ multitool source=data/artist.100.txt cut=0 sink=output

This will cut the first fields out of the file artists.100.txt and save the results to output file.

If no args are given, a comprehensive list of commands will be printed. That list is also available as COMMANDS.md in this directory.

For more detailed examples of using Multitool, see also: http://cascading.org/multitool/

Copying:

$ ./bin/multitool source=input.txt sink=outputDir

Copying while removing the first header line, and overwriting output:

$ ./bin/multitool source=input.txt source.skipheader=true sink=outputDir sink.replace=true

Filter out data:

$ ./bin/multitool source=input.txt "reject=some words" sink=outputDir

For a more complex example:

$ ./bin/multitool source=data/topic.100.txt cut=0 \
"pgen=(\b[12][09][0-9]{2}\b)" group=0 count=0 group=1 \
sink=output sink.replace=true sink.parts=1

This will find all years in the input file, count them, and sort them by counts.

To build Multitool, you may download the source code from GitHub:

https://github.com/cascading/cascading.multitool

This release will pull all dependencies from the relevant maven repos, including http://conjars.org

To build a jar,

$ ant retrieve jar

To test,

$ ant test

See apl.txt in this directory.