zacharymorn commented on pull request #101: URL: https://github.com/apache/lucene/pull/101#issuecomment-840934854
Hi @jpountz, upon further reviewing the result above, I found one issue for this particular scorer implementation (but not for the other BulkScorer based implementations). Currently this BMM scorer would also be used for nested boolean queries such as `+body:interview +(body:at body:united)` (AndMedOrHighHigh), but in the jira ticket you had suggested to use BMM for top-level (flat?) boolean query only. Do you think this will need to be fixed? I took a quick look at the code and it seems the query and weight interfaces do not provide an API to tell whether they are top-level or not currently, so this may involve some changes in `BooleanWeight` code I think. On the other hand, from the results above it looks like 2 out of 3 instances, having BMM scorer to run for nested queries doesn’t hurt its performance (the `AndMedOrHighHigh ` one). So I’m wondering maybe it’s ok to leave it as is, as BMM may still give a boost for the pure disjunction part of the query? The one result that does show negative impact to `AndMedOrHighHigh` also shows impact to `OrHighMed`, so it’s a bit strange and may need further looking into to see the cause. -- 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. 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
