Github user JamesRTaylor commented on the pull request:
https://github.com/apache/phoenix/pull/156#issuecomment-205637233
Thanks for the patch, @chrajeshbabu. This is a big improvement over your
initial approach. It's still complex, though. There's an on going cost to
complicating the code to this degree that would be good to minimize.
One big thing missing is an overview of all the moving parts. Perhaps that
could live in BaseResultIterators? For example - this exception is thrown under
these conditions and this state is added to the scan to know how to adjust the
scan ranges, etc. etc. Especially that hasReferences bit in the coprocessor -
what's that all about?
I'd like to understand how ChunkedResultIterator complicates things too. I
think it'd be ok to document that splits during aggregate queries aren't
handled if your HBase version is less that XXX if it'd significantly simplify
this patch (as that code is essentially deprecated).
Another potential, different approach would be for Phoenix to universally
handle the split during scan case (rather than letting the HBase client scanner
handle it for non aggregate case and Phoenix handle it for the aggregate case).
Would that simplify things?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
