Cascading is compiled against the latest stable release of Apache Hadoop and made available through a Maven repository hosted on Conjars as a convenience to developers.
Many developers of open-source project and internal proprietary projects use Maven to pull down Cascading, its dependencies, and any project dependencies. Tools like Ant, Ivy, Maven, Leiningen, SBT, and Gradle use Maven repositories for builds.
Because Cascading is built against the latest stable Apache Hadoop distribution, it will not run reliably on a Hadoop distribution (other than Apache) that may have changed public Hadoop API calls in an incompatible way.
They verify that a pre-built version Cascading, built against the latest stable Apache Hadoop distribution, is 'binary compatible' with another custom Hadoop build.
If a Java API changes, it may be source compatible with a consumer of that API, if the consumer (Cascading) is recompiled against the new API signature.
But often a compatible change handled by a recompile isn't binary compatible. That is, compiling against one version of the API may succeed, but executing the compiled signature against a slightly modified API may fail.
This is most commonly seen if the return type of a method is changed.
See the Compatibility page for a current listing.
After forking this project on GitHub, review the section titles "Testing a different distribution" in the README file.
Send an email to support@cascading.org.
No.
This is the simplest way to have Gradle find your Hadoop build. But you can update the test CLASSPATH to include jars from a local directory, or download a tar/zip file and unpack it into the CLASSPATH.
See the Gradle documentation for more details.
None.
The point of this test is to run Cascading tests on top of the new Hadoop build in question. Not to run the new Hadoop build in question tests.
Cascading is the foundation for a number of open-source and proprietary projects. If the Cascading tests pass, it is assumed any projects built on Cascading will work, within limits of course.
The Cascading SDK is a downloadable archive of many pre-built Cascading applications.
The compatibility test does not need the SDK to be installed, you can clone the compatibility test directly from GitHub and run the tests.
If the Cascading compatibility tests do pass on a custom Hadoop build, then its is generally safe to assume the tools distributed in the SDK will also run without errors.
The compatibility test is included in the SDK as a convenience.
The Apache foundation is the home and legal copyright holder for the Hadoop project. Both the source code and binary releases are hosted on the Apache Hadoop project page.
Per the Apache License, any organization can compile and redistribute Hadoop for nearly any purpose, without any significant limitations (see a lawyer for details).
The Hadoop release downloaded from the Apache Hadoop project page is the Apache Hadoop distribution. No other organization providing a release is entitled to call their Hadoop download Apache, per the rules of the Apache foundation.
Cascading is compiled, tested, and distributed for the latest stable version of Apache Hadoop. Pre-built binaries of Cascading will not run on any other distribution unless the Hadoop distribution in question is source and binary compatible with the Apache Hadoop release.