siddharthteotia commented on PR #9117: URL: https://github.com/apache/pinot/pull/9117#issuecomment-1198343680
> I read the original issue but am not convinced about the motivation to emit a metric when the order by has a MV column. > My guess is you want to identify the queries that have multi valued columns in the order by fragment and ask them to remove them? Yes, if you read the original PR, there are 2 categories of MV expression in ORDER BY queries where 1 is already failing in the engine (with not a user friendly exception). Then there is another category where MV expression in ORDER BY is ignored. See this comment thread - https://github.com/apache/pinot/pull/9078#issuecomment-1192039642 We agreed on the original PR that users should write more sensible queries and throwing proper generic exception for any presence of MV expression in ORDER BY indicating it is not supported is best. However, we wanted to first find out if anyone is running 2nd category of query through this metric, work with them to change (if any) and then come back to throw the exception properly. > IMO, a better solution would be to simply remove the multi-value columns in order by fragment since it was getting ignored anyway. I don't think that's a good solution. Users should simply not write such queries. It's engine's responsibility to throw meaningful exceptions when users are writing bad / unsupported queries as opposed to silently ignoring things (one of the cases today) or throwing very implementation specific exception (another case) > We can probably throw exception few versions later The problem will still remain on how to do this smoothly as opposed to a "potentially" breaking change where we suddenly start throwing exception for the 2nd case I mentioned above. -- 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]
