benwtrent commented on issue #15427: URL: https://github.com/apache/lucene/issues/15427#issuecomment-3656625575
@mikemccand @msokolov I was thinking about this, and talking with somebody working in JVector land, and they mentioned some of their merging plans. During merge, we know ALL the vectors. Incremental updates should really only ever be done during initial flush. However, during merge, we know all the vectors and have all the previous graphs. So, during initial candidate gathering, why not search the old graphs and translate their ordinals to the new vector ordinals? Then during candidate retrieval, we see everything. There is no longer a need for checking for backlinks for the majority (maybe all?) vectors. In this way, this "2nd pass" thing sort of goes away. -- 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]
