yinli-systems commented on PR #25303: URL: https://github.com/apache/datafusion/pull/25303#issuecomment-5800786472
Rebased onto current `main` after #25185 and reran the key `Utf8View`/dictionary matrix. The new hash-cache path is included, but the profitability boundary remains material: dictionaries are 13–35% faster at NDV 1,000 and 44–73% slower at NDV 10,000 in the tested 16B/64B cases. I updated the PR description with the full method and current numbers; both 153-test configurations, fmt, and default/all-feature package clippy pass. @Rich-T-kid @LiaCastaneda, could you check whether this benchmark surface captures the dimensions needed before the #24117 cost rule is reviewed? -- 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]
