Release v15.0.0-beta.10 · lance-format/lance

What's Changed

feat(namespace): preserve Retry-After metadata on throttling errors by @hello-peter-tang in #10400
fix(clustering): restore brings back the target's definitions, keeping only the allocation mark by @LuQQiu in #10629

feat(cache): new cache metrics by @a-erofeev in #9781
feat(index): add composable segment statistics APIs by @ivscheianu in #9783
feat: add FileFragment::update_columns_from_stream by @LuciferYang in #9826
feat(python): export merge-insert inserted row filters by @XuQianJin-Stars in #10066
feat(python): expose ZoneMap zone statistics by @XuQianJin-Stars in #10068
feat(python): expose plan, execute, merge and commit of optimize indices by @zhangyue19921010 in #10328
feat(java): expose distributed optimize indices as plan, execute, merge and commit by @zhangyue19921010 in #10332
feat(index): plan tuple predicates over ordered key columns by @Ali2Arslan in #10347
feat(index): preserve native RaBitQ partition results by @BubbleCal in #10683
feat(index): stream stored minhash signatures by @BubbleCal in #10685

fix(merge_insert): stop indexed merges failing or hanging under a bounded pool by @wombatu-kun in #10581
fix(build): merge macOS arm64 startup code into text by @app/lance-gatefixer in #10594
fix(scan): keep materialized row masks replayable by @BubbleCal in #10694
fix(compaction): avoid materializing blob payload batches by @ChilePiquin in #10696
fix(deps): update zstd-safe and remove expired advisory ignore by @Xuanwo in #10703

docs(fts): clarify exclusive UTF-8 token length limit by @karthiksenv in #10320
docs: revert three-layer stack split and restore engine-first quickstart order by @Xuanwo in #10692

perf(fts): prune candidate scoring in the combined_fields scan with M… by @sbrunk in #9630
perf: speed up sparse MinHash bucket scans by @kamronis in #9733
perf(frag-reuse): index long row-address remap chains by fragment by @brendanclement in #9813
perf(python): coalesce stratified sampling reads by @WanSuiMm in #9882
perf(index): stream shard partitions in bounded batches when merging vector segments by @LuciferYang in #10324
perf(encoding): avoid zero-filling FSST decode buffers and trim full-zip text batches by @zhangyue19921010 in #10369
perf(encoding): align mini-block dictionaries for 64-bit offsets by @wombatu-kun in #10394
perf(encoding): avoid per-value buffer slicing in fixed-width packed structs by @zhangyue19921010 in #10398
perf(index): share translated views for whole RaBitQ scoring by @BubbleCal in #10677

refactor(index): decode the HNSW graph batch once for both remap paths by @LuciferYang in #10372
feat(index): expose RaBitQ runtime policy and diagnostic snapshots by @BubbleCal in #10689

Full Changelog: v15.0.0-beta.9...v15.0.0-beta.10