msokolov commented on PR #12434: URL: https://github.com/apache/lucene/pull/12434#issuecomment-1657264032
@jpountz I think I see what you mean re: the cost of evaluating filters. We have been working on a solution where the most common and restrictive filters are cached and can be applied as pre-filters without needing to recompute for every query, while also joining with other constraints (post-filtering). This has led to increased yield (more high-scoring matches) at acceptable cost increase in many cases but in some cases we see increased costs without much better yield (as compared to applying all the filters as post-filters). So I feel like if the parent doc filter is highly restrictive there could be a similar tradeoff in this case ie even when the cost of the parent doc filter computation is not a consideration you might still see better recall/latency tradeoff using overcollection/post-filtering. I don't think this has any implications for this change really, but I do think it would be interesting to see some evaluation comparing. -- 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
