mikemccand commented on PR #15633: URL: https://github.com/apache/lucene/pull/15633#issuecomment-3823646765
Whoa, yeah, `seekCeil` is scary for bloom filter -- it just punts and falls back to the wrapped `TermsEnum` right? It cannot know the next term anyways (it only stores the hashed bit positions). Do you have any benchy results to test the impact? I don't think luceneutil has DV updates benchy? We (Amazon) have been befuddled by poor bloom filter performance (relative to what we expect, which might simply be too much!) ... but we are not using `updateDocValues`, more just for `PKLookup`. I think we make this opto in other places ... I wonder if those places should also just trust the `TermsEnum` impl to handle reuse. Though, I do expect trie / block tree terms dict to still see gains if you visit terms in order... -- 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]
