GH-38770: [C++][Python] RecordBatch.filter() segfaults if passed a ChunkedArray

[AlenkaF]

Rationale for this change

Filtering a record batch with a boolean mask in the form of a ChunkedArray results in a segmentation fault.

What changes are included in this PR?

In case chunked array is passed as a mask to filter record batch, the code path for pa.Table.filter() is taken resulting in a filtered table.

Are these changes tested?

Yes.

Are there any user-facing changes?

No.

• GitHub Issue: [C++][Python] RecordBatch.filter() segfaults if passed a ChunkedArray #38770

[github-actions]

⚠️ GitHub issue #38770 has been automatically assigned in GitHub to PR creator.

[danepitkin]

LGTM!

[pitrou]

Hmm... if the crash occurs due to a defect in the C++ code, perhaps this can be fixed on the C++ side? (or at least the C++ side could return a proper error)

[AlenkaF]

Hmm... if the crash occurs due to a defect in the C++ code, perhaps this can be fixed on the C++ side? (or at least the C++ side could return a proper error)

Ah yes, you are correct. That is also what Joris ment in the issue thread. Will correct.

[AlenkaF]

@pitrou I have moved the check to C++ but changed the logic to raise an error in case of a chunked mask. This way the bug will be resolved quicker as I would need to figure out some things in order to use FilterTable path =)

[jorisvandenbossche]

Looks good to me!

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 7 benchmarking runs that have been run so far on merge-commit d83af8f.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 24 possible false positives for unstable benchmarks that are known to sometimes produce them.