Rich-T-kid commented on issue #26976: URL: https://github.com/apache/datafusion/issues/26976#issuecomment-6724988692
Closing this DataFusion-side workaround in favor of the Arrow-native fix in apache/arrow-rs#12126 (for apache/arrow-rs#11748). The direct reader path converts only distinct dictionary values and preserves the encoded keys instead of casting/rebuilding every row. With RLE-to-dictionary enabled in both DataFusion builds and the candidate requesting Dictionary(Int32, Utf8), the balanced ClickBench A/B improved Q23 from 8.70 s to 5.00 s (1.74x), Q10 by 3.06x, Q11 by 2.59x, and Q22 by 1.44x. TPC-H and TPC-DS SF1 were directionally positive at -4.3% and -3.1% for the sum of per-query medians. The remaining DataFusion schema/request integration stays tracked by #26530. -- 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]
