clintropolis commented on PR #20469: URL: https://github.com/apache/druid/pull/20469#issuecomment-6241582930
>A little birdie told me that the following queries would involve regressions or behavior changes on this branch. Please look into them. Thanks for this list, it was a good one: * 1/2: a regression from this PR (an allow-list matching nothing broke the index path), fixed * 3: fixed by #20538 * 4/5: arguments inside join conditions are no longer specialized. The nested-join case (4) also exposed an existing bug that I'll fix separately. * 6: RegexFilteredVirtualColumn and the expression it plans to aren't consistent, so MV_FILTER_REGEX over expressions isn't specialized for now. I don't think this is correct/intended so I'll look into fixing this to make them consistent in a follow-up. * 7/8: changed stuff so an argument is only specialized when it returns null for a null input, since empty multi-value rows behave differently otherwise * 9/10: these now match how the same function already behaves on a plain multi-value column (COALESCE applies per element, and topN doesn't count empty rows), so maybe is ok? 9 specifically feels a bit ambiguous anyway, the coalesce is outside the mv_ function so technically i think should operate on the output as an array instead of to each element, but .. coalesce directly on an mvd applies to each element, this expression in particular is in a weird area. -- 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]
