msokolov commented on issue #14758: URL: https://github.com/apache/lucene/issues/14758#issuecomment-4313214261
The way I was thinking of "use another field to reference underlying vectors" would be like this: 1. create a Knn*VectorField "base" as one does today. 2. create one or more additional FilteredVectorField(String baseVectorField, Query filter) Then it is up to the format to apply the filter during flush/merge, and to mark the field as a "filtered field that depends on 'base' vector field" in the format. Then search operations on this field can be performed by operating on a filtered graph and delegating vector reading to the referenced field. As long as we don't have problems during indexing, I think this pushes all of the complexity down into the codec so that user-level APIs would not have to change? Maybe I'm missing something though! -- 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]
