Reviving MAHOUT-202 (HBase DataModel for taste)

[David Arthur <mu...@gmail.com>]

I've submitted a patch that adds support for HBase DataModel for the taste framework. It's pretty simple (naïve) in that it just stores a preference as a single cell in HBase (well, really two cells since the rating is duplicated). It uses a FastIDSet as a cache for the user/item ids, but does no other caching. I wrote a simple test script (not included in the patch) that generates some random ratings and hits various DataModel. It seems to perform reasonably well (considering it's all running on a single box). I'd like to write some unit tests, but this requires starting up HBase so I'm not exactly sure where that belongs.

If this is something the community thinks would be useful, I'd love to contribute it to the project. Feedback is, of course, welcome.

Cheers
-David