2010YOUY01 commented on issue #19649: URL: https://github.com/apache/datafusion/issues/19649#issuecomment-3727697426
I have a somewhat crazy thought when I was reviewing the previous blocked aggregation PR: should we throw away all the existing aggregation code and rewrite it from scratch for this blocked design? The reason is that the current implementation is built around a single contiguous aggregation state, and dozens of optimizations and features have been developed on top of that assumption. If we try to apply a fundamentally different design (blocked state management) as a patch, it may actually be more challenging than a full rewrite. I’ve done [something similar](https://github.com/apache/datafusion/pull/16996) before, and surprisingly, a complete rewrite turned out to be simpler than patching the existing code. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
