hhhizzz commented on PR #23274: URL: https://github.com/apache/datafusion/pull/23274#issuecomment-4853651513
> I think this design should be implemented as the first PR for #7065 🤔 . It does not seem possible to introduce an intermediate step like this PR that only splits out the `EmitTo` part, since it is tightly coupled with the execution logic and the accumulator physical layout. > > I agree that the initial blocked-state PR may become large, so we will need to figure out how to arrange and split the work better. For coordination, the initial blocked-state PR would depend on #22710 being finished first. Thanks for the guidance. That makes sense to me. I was trying to split the work into the smallest possible first step, but I agree that if `EmitTo::FirstBlock` is too tightly coupled with the execution logic and accumulator layout, introducing it as an API-only PR may be premature. I will pause this PR for now and follow the blocked-state / #7065 direction instead. I can keep the current API experiment and benchmark results as reference material, and revisit the API shape when the initial blocked-state implementation is ready after the #22710 stream-splitting work. Thanks again for helping clarify the right sequencing. -- 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]
