If you have a function that returns a streamed dataset it will only be executed on the 1st node. We need a syntax that will execute it on all the nodes in parallel.
- ALLNODES(<call>) - would have a confusing meaning in roxie, we don't want it run on all roxie slaves (or do we?!)
- DATASET(<call>, LOCAL|DISTRIBUTED) ?
- Something else? Any suggestions?
A requirement for allowing 3rd party distributed file readers to be written as plugins (e.g., to read amazon s3).