zacharymorn commented on pull request #418: URL: https://github.com/apache/lucene/pull/418#issuecomment-983322022
> Thanks for making progress on this and sorry for the slow review cycles. No worry and thanks for the review and feedback! > I just had a quick look and it seems that some computations can be move up-front such as deciding what field we use as a lead since field weights do not change during query execution? Yup I can move it out to avoid repeated computation there. Just a quick check though, from the perf results so far (https://github.com/apache/lucene/pull/418#discussion_r754091056 & https://github.com/apache/lucene/pull/418#discussion_r759006970), it seems using the most weighted field to decide on the leading impacts may not yield the best result, as matches in most weighted field may also give a high docId upTo bound. If we were to drop that approach and revert the commit https://github.com/apache/lucene/pull/418/commits/db2446f0337e5860bb8c731df1f6c16281efb6bb that introduced it, then it won't be computing most weighted field at all. Are you trying to see if perf tests can be further improved if most weighted field only gets computed once? > I wonder if there's things we can do to make the code simpler, such as sharing the logic to combine multiple terms together within the same field with SynonymQuery. Yeah I have the same thought there. I wasn't sure earlier as `SynonymQuery` and `CombinedFieldsQuery` sit in different modules / packages, so extracting common code out may potentially require a new public specialized API / class (in the `core` module ?). Let me give that a try and see how it looks! -- 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: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
