jpountz opened a new pull request, #14701: URL: https://github.com/apache/lucene/pull/14701
`MaxScoreBulkScorer` and `BlockMaxConjunctionBulkScorer` currently evaluate hits in a doc-at-a-time (DAAT) fashion, meaning that they they look at all their clauses to find the next doc and so forth until all docs from the window are evaluated. This changes evaluation to run in a more term-at-a-time fashion (TAAT) within scoring windows, meaning that each clause is fully evaluated within the window before moving on to the next clause. Note that this isn't completely new, `BooleanScorer` has been doing this to exhaustively evaluate disjunctive queries, by loading their matches into a bit set, one clause at a time. Also note that this is a bit different from traditional TAAT as this is scoped to small-ish windows of doc IDs, not the entire doc ID space. This in-turn allows these scorers to take advantage of the new `Scorer#nextDocsAndScores` API, and provides a good speedup. A downside is that we may need to perform more memory copying in some cases, and evaluate a bit more documents, but the change still looks like a win in general. -- 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
