costin commented on PR #16623: URL: https://github.com/apache/lucene/pull/16623#issuecomment-6132495155
@msokolov, thanks for looking at this. The PR improves the cost of applying deletes and opening soft-deletes readers on segments whose deleted docs were loaded from disk by re-enabling a fast path that existed before #15413. Salvatore first ran into this on a write-heavy workload, where the first delete after a reader snapshot paid the full per-bit copy. Re: Afaict, luceneutil tasks don't cover this code path since the index has no deletions and queries only call Bits.get on live docs. SoftDeletesReaderBenchmark (in the PR) aims to measure the end-to-end reader cost in a soft-deletes setup (such as Elasticsearch) by wrapping a DirectoryReader across different hard and soft delete rates. -- 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]
