Dandandan commented on pull request #9007: URL: https://github.com/apache/arrow/pull/9007#issuecomment-750923525
I think we need better estimates for how many rows are filtered in a `WHERE` clause (for which we need the column level statistics). Currently, the reordering on TPC-H q12 is not applied, because the (very selective) `FILTER` which returns the unknown estimate. Did you run it with the order reversed / the optimization applied? I think it might actually be slower that way, because of the selective filter that can be pushed down (not 100% sure though)? ---------------------------------------------------------------- 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: [email protected]
