jianlirong commented on PR #1178: URL: https://github.com/apache/cloudberry/pull/1178#issuecomment-2992092970
What I want to express is that for Streaming Partial HashAggregate, the data structure NumericAggState for the intermediate state in the first stage aggregation doesn't need to be the same as the state data structure for the final stage aggregation. For the final stage aggregation's state data structure, it should indeed maintain consistency with PG, and the two fields mentioned above could indeed exceed the maximum value of int32. However, for Streaming Partial HashAggregate, we can do some special handling: when the above two fields reach the maximum value of int32, we can stream out the results of the Partial HashAggregate, and then reconstruct a new Partial HashAggregate result starting from zero. Considering that int32 should be very large, even in extreme cases where the number of tuples is very high, the overall execution performance won't actually decline, because in such cases, the cost that NumericAggState brings to motion is almost negligible compared to scanning and pe rforming aggregation operations on more than 2^32 - 1 tuples. -- 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: commits-unsubscr...@cloudberry.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@cloudberry.apache.org For additional commands, e-mail: commits-h...@cloudberry.apache.org
