Rich-T-kid commented on PR #26789: URL: https://github.com/apache/datafusion/pull/26789#issuecomment-6847278990
I investigated the TPC-DS Q9 flag before rerunning it. At exact PR head `678d97de`, with the bot settings (`32G`, RLE dictionary reads off), I preconditioned independent flag-off/flag-on contexts with Q1-Q8 and then ran 160 interleaved Q9 pairs: - total: 37.446 ms off vs 37.476 ms on (+0.030 ms, +0.08%) - physical planning: +0.014 ms - execution: +0.016 ms The physical plans are byte-identical, contain no dictionary fields, and all 30 aggregates are global, so this optimizer rejects each one without rewriting it. Every scan is filtered and numeric, so Q9 neither requests nor attaches physical string statistics. The enabled-side cache's seven physical-stat entries were unchanged before and after Q9. This does not reproduce the bot's 53.25 -> 57.48 ms result or identify a flag-dependent Q9 code path. I am requesting one focused exact-head TPC-DS repeat next. If Q9 repeats there, I will keep treating it as a real regression; if it does not, the existing result is consistent with the five-sample variance seen locally. -- 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]
