GH-49586: [C++][CI] StructToStructSubset test failure with libc++ 22.1.1

[tadeja]

Rationale for this change

Fixes #49586

std::multimap::extract(key) does not guarantee returning the first element when there are duplicate keys.

What changes are included in this PR?

Replace extract(key) with lower_bound(key) plus erase(iterator). lower_bound guarantees first matching element, then read it->second and then erase the node (extract is not needed as the node isn't reused here).

Are these changes tested?

Yes, CI jobs are passing.

Are there any user-facing changes?

No.

• GitHub Issue: [C++][CI] StructToStructSubset test failure with libc++ 22.1.1 #49586

[github-actions]

⚠️ GitHub issue #49586 has been automatically assigned in GitHub to PR creator.

[tadeja]

@pitrou

[zanmato1984]

+1

Thanks for fixing this!

[zanmato1984]

Thanks for the fix.

Switching from std::multimap::extract(key) to lower_bound(key) + erase(iterator) is the right change here. It makes duplicate-field selection deterministic (first matching field) and matches the expected behavior covered by the existing StructToStructSubset duplicate-name tests.

LGTM.

[Copilot]

Pull request overview

Fixes a libc++ 22.1.1 behavioral difference causing StructToStructSubset failures by making selection of duplicate struct field names deterministic when mapping input struct fields to output struct fields.

Changes:

• Replace std::multimap::extract(key) with lower_bound(key) + erase(it) to reliably pick the first matching entry for duplicate keys.
• Update inline comments to document why lower_bound is required for deterministic behavior across standard library implementations.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[kou]

+1

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 3 benchmarking runs that have been run so far on merge-commit 35fb62e.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 2 possible false positives for unstable benchmarks that are known to sometimes produce them.