alamb opened a new pull request, #26543: URL: https://github.com/apache/datafusion/pull/26543
## Which issue does this PR close? - N/A: pre-release testing of arrow/parquet `60.1.0` - Related to apache/arrow-rs#11750 (release prep PR for `60.1.0`) **Note: this is a testing PR and is NOT intended to be merged.** It pins DataFusion to the `60.1.0` release candidate commit to find any correctness or performance problems before the release is published. ## Rationale for this change The `60.1.0` release contains ~300 commits since `60.0.0` (see the [draft CHANGELOG](https://github.com/alamb/arrow-rs/blob/alamb/prepare_60.1.0/CHANGELOG.md)), including a lot of parquet reader work (`ParquetPushDecoder::scan_plan` (apache/arrow-rs#10555), fused same-projection predicates (apache/arrow-rs#10859), deferred dictionary page decode on skip (apache/arrow-rs#11168), 64-bit word `RowSelection` trimming (apache/arrow-rs#11639), RLE dictionary read improvements (apache/arrow-rs#11474), INT96 statistics conversion (apache/arrow-rs#11348)) and many cast / kernel fixes. Running the DataFusion CI and benchmarks against the release candidate lets us find regressions while they can still be fixed before the release. This is the follow on to #26384, which tested arrow-rs `main` as of 2026-10-15 (dd240116c7). This PR pins to the release candidate commit, which is 58 commits later. ## What changes are included in this PR? - Add a `[patch.crates-io]` section pinning all arrow/parquet crates to apache/arrow-rs#11750 (commit `ad3ac04ae3457499f1607aee79fe71c31f4002e3`, fetched from my fork so the pin stays stable) - `Cargo.lock` update (only the arrow/parquet entries change, now `60.1.0`) Plus the same code and test changes that were needed for #26384, which will also be needed in the real upgrade PR: 1. Replace the one use of `FixedSizeListArray::value_offset` (in `unnest.rs`) with `value_offset_at`: the former was deprecated in apache/arrow-rs#11070 and fails `cargo clippy -D warnings` 2. **`interleave` reuses shared buffers for byte view arrays** (apache/arrow-rs#11281): the `spill::tests::test_gc_*` tests now build the repeated-buffer view array by hand 3. **Casting a naive timestamp to a named time zone resolves DST-ambiguous and nonexistent local times** (apache/arrow-rs#11038): `timestamps_timezone.slt` now expects the PostgreSQL answers instead of `Cannot cast timezone to different timezone` errors ## What is the testing strategy for this PR? - CI runs against the pinned `60.1.0` release candidate - `clickbench_partitioned` benchmark via the benchmark bot (see comments below) to compare against `main` (`60.0.0`) ## Are there any user-facing changes? No -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
