Rachelint commented on issue #19649: URL: https://github.com/apache/datafusion/issues/19649#issuecomment-3732331211
As I see, there are actually 2 issues `support blocked aggr state` and `improve aggr code readability`. And maybe we should not mix `adding new feature` and `refactoring` into one work, that may add complexity. Code changes of #15591 in `GroupedHashAggregateStream` are not large (176 lines, and many lines are comments). I think we can just add it firstly. And secondly, what we need may be to refactor the `GroupedHashAggregateStream` totally. During working in aggr, I found the complexity of `GroupedHashAggregateStream` is due to we make `only one stream for the whole aggr`, rather than `one stream for partial aggr and another stream for final aggr`. And it lead to that we are forced to mix codes of `partial aggr` and `final aggr` in `GroupedHashAggregateStream`, and finally many many code branches appear. I think if we just fork a new stream rather than totally rewriting to split the `partial and final codes`, I am concerned it will become something like current `GroupedHashAggregateStream` -- 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]
