Direct Hadoop adapter for Couchbase
- InputFormat that splits by Couchbase nodes, splits are vbuckets that existed on each node at job start.
<String,byte>K/V pairs when using mapreduce. API, or
BytesWritablepairs when using mapred. API
- Cascading Tap impl, returns K/V as Strings (expects UTF-8!)
- Reads K/Vs over Couchbase TAP protocol
- Output to Couchbase
- Clean up the build stuff (don't force deps on cascading, probably other things.)
- Tests (and run them somewhere other than my laptop!)
- An InputFormat that reads from a view could also be nice.