godfreyhe commented on issue #9141: [FLINK-12249][table] Fix type equivalence check problems for Window Aggregates URL: https://github.com/apache/flink/pull/9141#issuecomment-513070691 > @dawidwys Thanks for your comments. I think you are right. > > We had a discussion just now. (with @godfreyhe @wuchong @sunjincheng121 ) We think that, to solve the problem fundamentally, the current window aggregate should not extend from `org.apache.calcite.rel.core.Aggregate`, instead it should extend from `SingleRel`. This makes sure the right semantics of window aggregate. > > To correct this, the changes may be somehow very large as there are many existing window related rules. Also, this is a problem a long time ago and the probability of occurrence is relatively small. > > Considering the reasons above, I think it would be good if we can fix this issue later after release-1.9? > > What do you think? @dawidwys @godfreyhe @wuchong @sunjincheng121 > > Best, Hequn yes, it's a big issue to refactor WindowAggregate completely using solution3 mentioned in [FLINK-12249](https://issues.apache.org/jira/browse/FLINK-12249). "Creating special window aggregate call function" is not a proper approach, how about users defined aggregate function? currently, i think we should make WindowAggregate does not extend from Aggregate in this pr, which had been validated in a blink minor branch. and do the clean refactor after 1.9 ?
---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services
