Rich-T-kid commented on PR #26789: URL: https://github.com/apache/datafusion/pull/26789#issuecomment-6688950159
Closing this WIP because the Arrow row-cache optimization does not close the gap to DataFusion's default Utf8View path. In controlled TPC-H SF1 runs, the cache gave Q4 a 0.84 ms incremental improvement, but the combined Q4 result was not significantly faster than default; Q12, Q20, and Q22 were still 13.3%, 11.9%, and 13.4% slower. The requested Arrow tip also does not currently integrate: its RefCell cache makes RowConverter non-Sync, which conflicts with DataFusion's Accumulator contract. The last buildable Mutex-based cache was used for the performance check. The remaining cost is broader than repeated dictionary row encoding, so this optimizer rule is not worth carrying in its current form. -- 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]
