asolimando commented on issue #26696: URL: https://github.com/apache/datafusion/issues/26696#issuecomment-6452736063
I think it's still good to have operator-side bounds as safety net (either of #26703 and #26816), as @jayzhan211 suggests, but I guess @2010YOUY01 you are saying that we should try and avoid those cases via plan-time CBO decisions (which @Rich-T-kid is working on as a combination of #26315 and #26789) as much as possible, and I agree. To see this, the safety net PRs bring `clickbench_partitioned` (RLE -> dictionary config on) query runtime total from ~73 s to ~52 s, and the CBO planning rule gets it down to ~35 s on its own, and it works; the operator side bounds are needed to protect cases where the rule mispredicts, and there will always be such cases. What I'd like to see, is a run of #26789 against the default config (dictionary reads off), so we can compare against a baseline run with plain strings instead of dictionaries. -- 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]
