Releases: lancedb/lance
Releases · lancedb/lance
v0.5.5: Fix ARM wheel build, allow custom AWS credentials when writing dataset
What's Changed
- allow aws cred injection in dataset write as well by @chebbyChefNEQ in #1059
- Arm quick fix by @unkn-wn in #1057
Full Changelog: v0.5.4...v0.5.5
v0.5.4: Contribution guide, ARM Linux build, and other improvments
What's Changed
- feat: upgrade arrow, make write Send by @wjones127 in #1033
- Bring commented out clippy rule back after GHA rust upgrade by @eddyxu in #1039
- contributing guide by @changhiskhan in #1036
- use a large runner for linux tests by @changhiskhan in #1038
- bugfix: remove extra character from jq filter in GHA by @gsilvestrin in #941
- upgrade arrow version in duckdb extension by @changhiskhan in #1040
- Docs: Dataset write takes reader by owner now by @dnsco in #1042
- allow custom AWS credentials when constructing object store by @chebbyChefNEQ in #1049
- fix: don't compare metadata when checking schema in append by @wjones127 in #1051
- Add support for ARM architecture by @unkn-wn in #1043
New Contributors
Full Changelog: v0.5.3...v0.5.4
v0.5.3 Create empty table, Support writing iterator of RecordBatch
What's Changed
- [rust] Added support for creating empty dataset by @trueutkarsh in #1016
- [Rust] Use unaligned load in cosine on Intel CPUs by @eddyxu in #1035
- cache aws credentials until 30s before expiry by @chebbyChefNEQ in #1034
- fix record batch writing to dataset by @TevinWang in #1030
- docs: update format docs with feature flags and row indices by @wjones127 in #1037
Full Changelog: v0.5.2...v0.5.3
v0.5.2: Azure blob storage, dot product metrics type, SQL parsing bug fix, and various other improvements
What's Changed
- Support Azure blob storage as backend by @alikefia in #1017
- wire in DefaultCredentialsChain from aws_config as the credential provider for object_store by @chebbyChefNEQ in #917
- split clippy to a separate workflow, disable incremental build for CI by @chebbyChefNEQ in #1018
- [rust][python] allow create fragment from existed datafile by @LiWeiJie in #981
- Support Dot product as metric type by @eddyxu in #1021
- use default value for null boolean by @wjones127 in #1028
- Throw error instead of panic if kmeans training does not have sufficient data. by @eddyxu in #1029
- Improve support for timestamp cmp and remove custom sql parser by @wjones127 in #1026
New Contributors
Full Changelog: v0.5.1...v0.5.2
v0.5.1: Add object store construction params
What's Changed
- allow object store wrapper as part of dataset construction param by @chebbyChefNEQ in #1014
Full Changelog: v0.5.0...v0.5.1
v0.5.0: Deletion and performance improvements
What's Changed
New features
- Delete rows in a dataset with
Dataset.delete()
- Add delete() method to Python API by @wjones127 in #953
- feat: handle deletes in count_rows, updater, merge by @wjones127 in #995
- feat: add feature flags to the dataset manifest by @wjones127 in #979
- You can now customize the log level to reduce verbose messages
- Negative numbers are now fully supported in SQL expressions
- [rust] Added support for minus operator for numeric literals by @trueutkarsh in #983
- openblas is now an optional dependency
- Add feature flag for opq by @TevinWang in #998
- Remove a directory (dataset) recursively. by @eddyxu in #1011
Bugfixes
- fix s3 get_range issues: 416, range not satisfiable by @LiWeiJie in #975
- fix: make sure we are always running with a single partition by @wjones127 in #977
- fix: handle NaN values in argmin by @wjones127 in #1000
Performance Improvments
- [Rust] Reduce arrow overhead during kmean training by @eddyxu in #990
- Zero copy during kmean membership computation by @eddyxu in #992
- [Rust] Improve kmean training performance by removing dynamic dispatch by @eddyxu in #996
Other
- upgrade to arrow 40 and datafusion 26 by @chebbyChefNEQ in #1005
- bump arrow versions on duckdb intergration as well by @chebbyChefNEQ in #1007
- Remove a directory (dataset) recursively. by @eddyxu in #1011
New Contributors
- @trueutkarsh made their first contribution in #983
- @TevinWang made their first contribution in #998
Full Changelog: v0.4.21...v0.5.0
v0.4.21 Allow users to provide IVF centroids during indexing
What's Changed
- [Rust] Use lance local object store by @eddyxu in #959
- [Rust] Allow to provide pre-existing IVF_PQ centroids during index creation by @eddyxu in #963
- remove deletion_vector method and handle Result from reading deletion file properly by @chebbyChefNEQ in #965
- [Python] Pass IVF centroids via create_index API by @eddyxu in #970
- implement deletion vector handling in index scan by @chebbyChefNEQ in #958
Full Changelog: v0.4.20...v0.4.21
v0.4.20 Cast and timestamp literals in filters
What's Changed
- Add support for cast in expressions by @wjones127 in #957
Full Changelog: v0.4.19...v0.4.20
v0.4.19: new merge() function
What's Changed
- Merge a pre-computed column / array. by @eddyxu in #824
- Use backtick to escape column names by @wjones127 in #935
- seed l2 benchmark data generation and unroll ARM Neon simd by @chebbyChefNEQ in #937
- add a benchmark for cosine distance and unroll ARM Neon simd by @chebbyChefNEQ in #942
Full Changelog: v0.4.18...v0.4.19
v0.4.18 Release
What's Changed
- [Python] relax minimum pandas version to 1.4 by @changhiskhan in #921
- [Doc] Update README.md by @jaichopra in #924
- [Rust] argmin benchmark by @chebbyChefNEQ in #927
- [Docs]: add branding to docs by @wjones127 in #928
- [Rust] Add deletion files to Lance format by @wjones127 in #920
- [Rust] Allow recreate index with the same name on the same column by @eddyxu in #929
- [Doc] make sure we can publish on docs.rs by @wjones127 in #925
Full Changelog: v0.4.17...v0.4.18