NVIDIA / NVIDIA/cuvs

[BUG] Lucene: isolate versioned providers and repair quantized format factories

Open
#2,615 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Cuda
Stars
854
Forks
236
Avg merge
3d 3h
Merged PRs (30d)
62

Description

Problem

LuceneProvider currently keeps one process-wide instance regardless of the requested Lucene version. Its quantized HNSW factories also expose return descriptors that cannot represent the corresponding Lucene HNSW formats, and the scalar constructor arguments do not match Lucene's maxConn, beamWidth order.

Proposed change
  • Cache providers by version/capability with safe publication.
  • Add correctly typed KnnVectorsFormat factory methods.
  • Retain existing JVM method descriptors as deprecated compatibility entry points with actionable failures.
  • Retain the misspelled binary factory as a deprecated delegating alias.
  • Preserve reflected IOException, RuntimeException, and Error targets.
  • Add construction, concurrency, descriptor, parameter-order, and CPU-fallback tests.

Do not change the released binary codec's on-disk layout in this work.

Coordinates with #2410, #2512, and #2586.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by locating LuceneProvider and the quantized HNSW format factories, then trace their version selection, JVM descriptors, scalar argument order, and reflected exception targets. Review the construction, concurrency, descriptor, parameter-order, and CPU-fallback tests; done means versioned providers are safely isolated, compatibility aliases fail or delegate as specified, and the released binary codec layout is unchanged.

Written by the indexing model from the issue text.

Assessment

Tech stack
java
Domain
search
Issue type
Bug
Difficulty
5/5
Estimated time
Over a week
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
42/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.