jimczi opened a new pull request, #16905: URL: https://github.com/apache/lucene/pull/16905
Since #16883, merges of vectors and stored fields that open the data file again use the reader's own context with sequential access, the same as sequential reads by searches. A directory can no longer tell a merge apart, so for example `DirectIODirectory` doesn't use direct I/O for these merges anymore. - `Lucene99FlatVectorsReader` and `Lucene90CompressingStoredFieldsReader` open the data file again for merges with `IOContext.merge()`, as before #16883. - Sequential vector values open it with a search context, even on a reader that a merge opened. An input is still shared whenever it is efficient for the context, so merges and sequential values read through the same input unless the directory serves them differently. - Add `LeafReader#getFloatVectorValues(String, DataAccessHint)`, implemented by `CodecReader` through its vectors reader. The default implementation ignores the access pattern; `ExitableDirectoryReader` and `SlowCodecReaderWrapper` forward it. -- 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]
