uschindler commented on PR #15089: URL: https://github.com/apache/lucene/pull/15089#issuecomment-3214867277
Actually there are multiple problems: We need to fix this also in 10.x branch. But as the stupid method (which I would never have been put into IndexInput/DataInput) may be implemented by subclasses (I know many who implemented their own MemorySegmentIndexInput and consorts) where update might break, so we should be a bit careful. I'd like to remove the GroupVInt stuff completely from IndexInput and IndexOutput. I was already against the whole lambda and the implementation in general 2 years ago. The funny thing: The improvement by the In short: If you ask me: Let's remeove the "improved" method. It is a bit faster in microbenchmarks but the maintenance cost is too expensive and the risk of failing in benchmarks is too hight. Actually Mike's benchmark sees no difference at all because the "microbenchmark" is so specialized and due to callsite pollution it was never working correctly. I Im still against adding the burden to downstream users to implement a customly optimized readGroupVInt method in their own directory implementation that have never shouwn any use in real benchmarks. This few percent of improvement by calling readGroupVInt in a mincrobenchmark in isolation is a no-go for good code-quality. And that code in those three IndexInputs that implement our specialcase is a mess and I am against it. Let's remove the mmap/nio/bytebuffers directory optimization, commit it and let Mike benchmark it. If we see no difference larger than 5%, lets kill it forever! -- 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
