GH-48072: [C++][Acero] fix a bug of materialize boolean

[waruto210]

Rationale for this change

Use the correct offset when accessing boolean values from the buffer.

What changes are included in this PR?

In unmaterialized_table_internal.h, use the correct offset when accessing boolean values from the buffer. Also, add a unit test to regress this bug.

Are these changes tested?

Yes

Are there any user-facing changes?

No

This PR contains a "Critical Fix".

Yes, fix a bug that caused incorrect data

• GitHub Issue: [C++][Acero] materializeColumn for boolean type overflow #48072

[github-actions]

⚠️ GitHub issue #48072 has been automatically assigned in GitHub to PR creator.

[waruto210]

@westonpace PTAL

[kou]

+1

Could you rebase on main to run CI?

[waruto210]

+1

Could you rebase on main to run CI?

Rebase done.

[kou]

Sorry for my late check... There are CI failures. Could you rebase on main again?

[thisisnic]

I've rebased this, but it was failing so I had Claude fix up the remaining things, my changes are detailed below.

@kou - any more changes required for this?

---

🤖 :

The new test in sorted_merge_node_test.cc didn't compile on builds using -Werror, because of an implicit int64_t to int32_t conversion and the unused true_count and false_count variables. The fix in unmaterialized_table_internal.h is unchanged.

Changes to the test:

• Added a static_cast and removed the unused counters
• Changed the batch size from 1024 to 1001, so that batches start at offsets which aren't byte-aligned
• Added nulls to the boolean column, and a check that they come through the merge correctly
• Removed the unused arrow/dataset includes and added <limits>
• Fixed a stale comment and failure message, and renamed the helper lambda to make_source

The other failures in the previous CI run (parquet-reader-test on MinGW, an S3 test timeout on macOS, and the macOS GLib build) were also failing on main, so they aren't caused by this PR.

[Copilot]

Copilot review overview

🟢 Approval recommended

The offset calculation is correct and the regression test covers values, validity, and non-byte-aligned slices.

Review effort: Balanced
Findings: None

What changed in this PR

Fixes Boolean column materialization for arrays with non-byte-aligned offsets.

Changes:

• Correctly calculates byte and bit offsets for Boolean values.
• Adds regression coverage for sliced, nullable Boolean columns during sorted merge.

File	Description
cpp/​src/​arrow/​acero/​unmaterialized_table_internal.h	Corrects bit-packed Boolean access.
cpp/​src/​arrow/​acero/​sorted_merge_node_test.cc	Tests Boolean materialization across unaligned batches.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[kou]

@zanmato1984 Do you want to review this before we merge this?

[zanmato1984]

Hi @kou , I'm taking a look. Thanks for letting me know.

[zanmato1984]

+1

Thanks for working on this.

[thisisnic]

CI failures are ones on main, so I'll merge

[kou]

@zanmato1984 Thanks for reviewing this!

@thisisnic Thanks for completing this!

[thisisnic]

Thanks both for looking at this! I'm going to see if there are any other similar PRs that are almost complete but look like they have been forgotten that I can complete with very minimal AI effort to get them over the line. If it starts getting annoying, given I don't quite have the C++ skills to totally understand what I'm producing, just let me know and I can change approach. Will try not to take on any which need more than minor changes anyway.