JingsongLi opened a new pull request, #30: URL: https://github.com/apache/paimon-vector-index/pull/30
## Summary Documents the v1 vector index storage formats and adds golden fixtures for the current v1 writers/readers before release. This keeps the compatibility contract explicit: v1 is the first released disk format, and pre-release layouts are not supported. ## Changes - Add `STORAGE_FORMAT.md` covering v1 headers, sections, list payloads, shared delta-varint ID encoding, shared HNSW graph encoding, and the release compatibility policy. - Add inspectable hex golden fixtures for IVF-FLAT, IVF-PQ, IVF-PQ 4-bit, IVF-HNSW-FLAT, and IVF-HNSW-SQ. - Add an integration test that verifies current writers produce the fixture bytes and current readers can load/search those fixture bytes. - Stabilize the parallel HNSW recall test by using a wider search and a schedule-tolerant recall threshold while keeping the degree-bound assertion. ## Testing - [x] `cargo fmt` - [x] `cargo test -p paimon-vindex-core --test storage_format_fixtures` - [x] `cargo test -p paimon-vindex-core hnsw::tests::test_hnsw_parallel_build_large_partition_recall_tracks_exact_search -- --exact --nocapture` repeated 5 times - [x] `cargo test -p paimon-vindex-core` ## Notes This PR does not add compatibility for old pre-release layouts. The documented policy treats v1 as the first release format and requires future incompatible disk changes to bump the format version. -- 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]
