jtibshirani commented on PR #11843: URL: https://github.com/apache/lucene/pull/11843#issuecomment-1276520789
To give some context, Elasticsearch exposes a query type that performs a kNN exact scan. It iterates through all the `VectorValues` matching a query, and computes the similarity. Since `ExitableDirectoryReader` checked for cancellation on every vector access, it added significant overhead to these queries. This is also relevant to Lucene (even if Elasticsearch didn't have this query type). In `KnnVectorQuery`, we fall back to an exact scan if kNN with filtering is too expensive. I'm guessing we didn't see this in the nightlies because we don't test kNN with filtering plus query cancellation. -- 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]
