peter-toth commented on PR #59916: URL: https://github.com/apache/spark/pull/59916#issuecomment-6634032335
@dongjoon-hyun, thanks for round 3! It is addressed in [becfca3](https://github.com/apache/spark/pull/59916/commits/becfca30884990d3f82a700f1fd36d5ef5c1de60): - 1 to 3: scratch runs at a3b03fe confirmed all three. Each failed where master returns rows, with AQE on and off. Instead of following more DPP filters, the rule now adds a bloom filter for a key whose DPP filters are all dropped only if nothing that its subquery evaluates can fail. Then a filter that narrows only the join's side, a DPP filter or another bloom filter, leaves the subquery only more work. This also covers round-2 item 3, so the kept-DPP check is gone. - 2: the up-cast check leaves out a decimal to an integral type. - Also, as you asked in round-2 item 5, the determinism and CTE checks now look only at the plan that the bloom filter's subquery reads. A non-deterministic column that the subquery does not read no longer keeps the bloom filter out. - The rest is answered in the threads. -- 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]
