Lucene: Status=Open

[Save this search]

Author relation

Updated ago

Comment count

Reaction count

Review Requested

User

Last comment user

Type

Filters: Status (Open), Issue type, Author relation, Created, Updated, Updated ago, Comment count, Reaction count, Review Requested, Mentioned, Reviewed, Commented, User, Last comment user, Draft, Component, Type, Labels, Commits?, Reporter, Assignee

#17018 PR: Handle cases where the span query field doesn't have any indexed docs, but the underlying terms do.

2.9 minutes ago 6 comments 0 votes 0 watches github-actions[bot], john-wagster, thecoop

In that case, construct some stats from terms that are usable by scorers that need them.

john-wagster 3 minutes ago:	I was trying to think through what this is actually doing and it will probably miscount sometimes I ... So like: doc0: james doc1: james, sally doc2: greta, sally, james doc3: lisa, bob doc4: sally, ... I think that would be more accurate?
john-wagster 2.1 hours ago:	nit: extra line

#17013 PR: Off-heap bulk scoring for Lucene104 scalar quantized vectors

20.9 minutes ago 32 comments 0 votes 0 watches github-actions[bot], kaivalnp, msokolov

Description Off-heap bulk scoring for Lucene104 scalar quantized vectors. Partially addresses 15155. _**Disclaimer**: much of the code and benchmarks are AI-assisted, with me in the loop

kaivalnp 20.9 minutes ago:	Thanks for the review @msokolov! I've tried to incorporate your feedback and named things more appropriately.
kaivalnp 37.1 minutes ago:	These are still documents, we get queries as a QuantizedByteVectorValues in the asymmetric indexing ... During indexing, queries are still copied to heap (using queryVectors.vectorValue), but this is ... I think the QueryScorer name was confusing, I'll rename to ScorerBase (and consequently, ...

#17025 PR: Add FuzzyAutomatonCache to share Levenshtein automata across fuzzy queries

2.5 hours ago 0 comments 0 votes 0 watches anil884, github-actions[bot]

Description FuzzyQuery rebuilds its Levenshtein automata every time it is rewritten. They are shared across the segments of one rewrite, but the same token searched on another field, ... For fuzzy autocomplete over several fields this is a large part of the rewrite cost.

#17024 PR: Allocate postings enum buffers lazily

5.1 hours ago 1 comments 0 votes 0 watches github-actions[bot], jpountz, msokolov, uchiha-asha

Every BlockPostingsEnum allocates a FixedBitSet even for PACKED encoding, and a freqBuffer array ... This change allocates the bitset on first use and shares a read-only all-ones frequency buffer ... The change shows considerable timing and allocation gains in high-cardinality keyword sort and the ...

uchiha-asha 5.5 hours ago:

@jpountz @msokolov Could you please review this when available. Thanks!

#16709 PR: Add SegmentIVF vector codec to Lucene sandbox

8.2 hours ago 13 comments 0 votes 0 watches RKSPD, github-actions[bot], gsmiller, mikemccand

SegmentIVF: a fast(er) Lucene-native inverted-file vector codec SegmentIVF is a Lucene KNN codec ... It is a further development of IVFaster 16567 with some cool new features. Before going into the implementation, it is important to describe the architectural philosophy ...

mikemccand 8.2 hours ago:	@RKSPD It'd be great to push this forwards! This would be a breath of fresh air compared to HNSW heavy options Lucene has today (actually, ... It'd give options to Lucene users, and also enhance our testing since we could pick vector search ...
github-actions[bot] 25 days ago:	This PR has not had activity in the past 2 weeks, labeling it as stale. If the PR is waiting for review, notify the dev@lucene.apache.org list. Thank you for your contribution!

#17020 PR: Reduce CachingTokenFilter memory by not copying stale bytes

11.8 hours ago 3 comments 0 votes 0 watches github-actions[bot], javanna, reugn, uschindler

CharTermAttributeImplclone copies builder, which holds the bytes of the last term passed to ... That term can be much longer than the current one, and can come from an earlier analysis that ... CachingTokenFilter calls captureState() for every token, so each cached token keeps its own copy of ...

javanna 11.8 hours ago:	hey @uschindler just checking you are ok with this change. It looks like you added this copy bytes line several years back.
javanna 11.9 hours ago:	This looks sounds to me, I left a small comment on testing.

#17022 PR: Rename length to numTerms in FieldInvertState.

13.6 hours ago 0 comments 0 votes 0 watches github-actions[bot], vsop-479

### Description

«
1
2
3
4
5
»

2,739 issues

[22.9 msec search, 23.9 msec total]