benwtrent commented on code in PR #12314:
URL: https://github.com/apache/lucene/pull/12314#discussion_r1243611229
##########
lucene/core/src/java/org/apache/lucene/search/VectorScorer.java:
##########
@@ -85,6 +96,26 @@ public boolean advanceExact(int doc) throws IOException {
return vectorDoc == doc;
}
+ @Override
+ public Map<Integer, Float> scoreMultiValued(BitSet acceptedDocs) throws
IOException {
+ Map<Integer, Float> docToScore = new HashMap<>();
+ for (int vectorId = values.nextDoc(); vectorId != NO_MORE_DOCS; vectorId
= values.nextDoc()) {
+ int docID = values.ordToDoc(vectorId);
+ if (acceptedDocs.get(docID)) {
+ float currentScore = similarity.compare(query, values.vectorValue());
+ docToScore.putIfAbsent(docID, currentScore);
+ docToScore.computeIfPresent(docID,
+ (key, previousScore) -> Math.max(previousScore,
currentScore));
+ }
+ }
+ return docToScore;
+ }
Review Comment:
I don't see how we can do this here. Instead of randomly iterating the
BitSet and going to the exact document, we are iterating every vector in the
scorer and checking its membership in `acceptedDocs`.
Even multi-valued `*VectorValues` classes need to be able to iterate to the
exact document. Otherwise searching with a filter like this is untenable.
--
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]
