MRUnit is a unit test driver for MapReduce programs for use with JUnit. See the overview in the Javadoc for more details.
This is a fork of the MRUnit code in Apache Hadoop's contrib directory and has been created to quickly test and develop new features. Notable changes from Apache Hadoop's contrib MRUnit:
- Build system converted to Maven from Ant / Ivy.
- Cleaned up to produce no warnings from javac.
Planned features:
- Port of MRUnit to the "new" API (mapreduce). This will be in addition to the old (mapred) API.
- Support for testing {Input,Output}Formats.
- Support for testing Writable serialization of custom types.
Some members of the Apache Hadoop community feel like the contrib directory is not a good long term home for projects like MRUnit. I agree. I would hope (and intend to file a JIRA at some point) to deprecate and remove MRUnit from the Hadoop contrib tree in favor of a standalone version hosted on Github as a separate project. My belief is that MRUnit will benefit from a more frequent release cycle and an independent schedule. The code will remain under the Apache License.
Should the Apache Hadoop community elect to retain MRUnit in the contrib tree, I intend to attempt to contribute changes upstream to the ASF.
- Aaron Kimball [email protected]
- E. Sammer [email protected]