clintropolis commented on PR #12418: URL: https://github.com/apache/druid/pull/12418#issuecomment-1097140531
>hmm. I believe the part of the PR that fixes the bug, is using the row signature of the subquery. If you remove that and just explain the subQuery, the overall query will still fail with the same exception. Sorry yes, I confused myself, it is the outer scan query that is invalid, not the subquery here because the outer cannot order by the aliased time of the subquery, i mixed them up. But, it still is reporting that the query is valid because of the way this explain works, while calling `toDruidQuery` on the same fails with the `CannotBuildQueryException` we see when we try to transform the query after the valid check, so to me that seems it is not in fact a valid query. The worry I think is that changing explain to also explain the subquery might result in extra work being done which could result in increased planning time for nested queries, which is reasonable to be concerned about. Though, i also wonder if by not actually validating here if we are allowing other work to proceed which should not have gotten so far, like in this case if explain failed then it would never proceed to the rule transformation to fail again. I'm not really sure what is best though... or how to know -- 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]
