Release v0.34.0 · lance-format/lance

What's Changed

feat(java): java transaction supports DataReplacement operation by @majin1102 in #4454
feat: add contains_tokens udf by @wojiaodoubao in #4420
feat: add retry-based conflict resolution for delete operations (#4159) by @wjones127 in #4292
feat: add fuzz tests for encoding by @Xuanwo in #4480
feat: implement large read io split in format 2.1 by @Xuanwo in #4462
feat(python): expose diff_meta for pylance by @yanghua in #4502
feat: add automatic conflict resolution support for delete operations by @wjones127 in #4407
feat: add a cycle_bool data generator by @westonpace in #4507
feat: allow reading the _rowoffset by @westonpace in #4478
feat: add zonemap filter support to Lance at table level by @HaochengLIU in #4244
feat(rust): support update stable row id for merge insert without index by @yanghua in #4450
fix: handle parsing sql_cast_expr to datafusion expr properly by @imededin in #4541
feat: provide feature flag for shallow cloning by @majin1102 in #4552
feat: add lance file version 2.2 by @Xuanwo in #4568
feat: add more metrics for FTS queries by @BubbleCal in #4560

fix: lance failed to encode 1025 rep/def values by @Xuanwo in #4467
fix: ensure transaction rebase happens in ascending order by @jackye1995 in #4486
fix: posting lists cache conflicts across partitions by @BubbleCal in #4490
fix: the elapsed_time is not correct for IvfSubIndex Node by @BubbleCal in #4491
fix: skip rebasing for strict overwrite by @HubertY in #4377
fix: use LanceBuffer::borrow_to_typed_slice for safe slice cast by @Xuanwo in #4493
fix: avoid column name collision in merge_insert by using __action by @wjones127 in #4499
fix: remove data files with only tombstoned fields by @wjones127 in #4497
fix: address fuzz test #4492 by @Xuanwo in #4503
fix: shrink_to_fit after copy_from_bytes to ensure memory size is correct by @Xuanwo in #4515
fix: fix FTS index cache underutilization due to DeepSizeOf overestimation by @LuQQiu in #4513
fix: full text search BM25 score is not accurate by @BubbleCal in #4525
fix: use java/lang/String to get DeletionFileType value by @fangbo in #4526
fix: incorrect logic for logical or of RowIdMask by @westonpace in #4531
fix: fix spelling in comment in dataset.rs by @cmccabe in #4554
fix: restore behavior of projection plans from schema by @westonpace in #4562
fix: get rust-benchmark.yml working again by @cmccabe in #4567
fix(doc): fix take_blobs python doc wrong parameters format by @yanghua in #4569

perf: use heap to maintain topk results by @BubbleCal in #4505
perf: remove tracing instrumentation from FTS hot paths by @LuQQiu in #4536
perf: fallback to brute force FTS if filters matching fewer rows by @BubbleCal in #4551
perf: skip more documents in FTS flat search by @BubbleCal in #4570

refactor: remove owned LanceBuffer to eliminate runtime try_clone checks by @Xuanwo in #4460
refactor: reduce lance-encoding compile times by @westonpace in #4474
refactor: remove special case in can_cast_types by @cmccabe in #4549
refactor: remove Ray integration from Lance core by @jackye1995 in #4563

Full Changelog: v0.33.0...v0.34.0