Details
-
Improvement
-
Status: Open
-
Minor
-
Resolution: Unresolved
-
None
-
None
-
None
-
None
Description
Motivation: Kotlin has recently gained a lot of popularity and people are really starting to talk about it and see its merits; and language-wise it should be perfect for writing data-transformation code as it has better support for functional programming and useful type stuff than Java, without the performance overhead and implicit mysteries of Scala.
Attached is a patch for a proof-of-concept Kotlin API for Crunch. It makes heavy use of inlined functions so that the compiled bytecode ends up looking a lot like standard Crunch code (I ran it through a decompiler), and should be just as performant.
It also uses reified type parameters to construct PTypes automatically from TypeReferences, so in most cases it's not necessary to provide a PType for transformations.
See the example job and unit tests to get a quick idea of what kind of thing is possible.
I've decided to share an unfinished patch with you all so that I can get some feedback if we think this is a useful addition to the Crunch project at all, and whether the general approach seems sensible.
Still left TODO (if we decide to finish this and add it):
- PGroupedTable API
- Extensible PType inference (I'm thinking something like register<MyType>(pTypeForMyType))
- Support for data classes via kotlin-reflect
- Fill in any other gaps (mapWithContext, counters, ... ?)