Release v0.27.0 · lance-format/lance

What's Changed

feat!: allow take_blobs API to take on row address, indices or row ids by @eddyxu in #3755

feat: add public accessors for lance scan fields by @wkalt in #3739
feat: store data file size in manifest by @wjones127 in #3750
feat: expose public accessor for LanceScanExec config by @wkalt in #3757
feat: fix various spans by @wjones127 in #3738
feat(build): add make clean target for Python temporary files by @Jay-ju in #3756
feat: add tracing spans to ScanExec and TakeExec by @wjones127 in #3766
feat: make it possible to filter on large_binary by @westonpace in #3797
feat: split large reads into smaller parallel reads by @westonpace in #3801
feat: introduce new FilteredReadExec which will replace the initial scan/take by @westonpace in #3782
feat: support LANCE_FILE_WRITER_MAX_PAGE_BYTES env by @yanghua in #3815

fix: prevent infinite loop of writing manifest by @wjones127 in #3760
fix: revert IO optimizations for now until we can test more by @wjones127 in #3763
fix: don't lose rows during bitmap remap if rows not being remapped by @westonpace in #3765
fix: flat scan frags not covered by scalar index when doing prefilter search by @westonpace in #3768
fix: partition not used yet panic by @BubbleCal in #3769
perf: reduce retry thrashing in merge_insert by @wjones127 in #3780
fix: fix mapdataset batch getitems by @Jay-ju in #3783
fix: revert tracing change from #3766 by @wjones127 in #3800
fix: limit the batch size to 64 for indexing multivectors by @BubbleCal in #3810
fix: logic bug in RowIdTreeMap union by @westonpace in #3811
fix: handle bitmap index overflow by implementing multi-batch read&write #3515 by @HaochengLIU in #3789
fix: fix python tracing by @westonpace in #3807

docs: how to set blob metadata by @eddyxu in #3759
docs:clarify instructions on where to run maturin develop by @HaochengLIU in #3767
docs: fix typos by @omahs in #3803

Full Changelog: v0.26.1...v0.27.0