bench: borrow inputs in the benches that only read them

[jlucaso1]

Follow-up to #837, found while flamegraphing bench_get_children_by_tag (57.1 us in Simulation): 66% of that time is drop_in_place of the input Node tree, because the bench consumed its pre-built input with bench_values and the drop ran inside the measured window. The actual traversal costs about 2 us.

Eight benches had the pattern: get_children_by_tag, jid_to_owned_access, the six reporting-token benches (content extraction, full generation, message encoding). All of them only read the input, so they now use bench_refs, and values produced inside the closure are returned so divan drops them outside the timing. process_history_sync keeps bench_values: that operation genuinely takes ownership of its blob, so the consume is part of the measured contract (and its result was already returned).

The CodSpeed report on this PR will show large "improvements" on these eight benchmarks; that is measurement correction, not a performance change. No library code is touched.

[coderabbitai]

📝 Walkthrough

Summary by CodeRabbit

Release Notes

• Tests
  • Updated internal benchmark methodology for improved performance measurement accuracy.
  • Refined benchmarking approach across multiple test suites.

Note: These are internal testing infrastructure improvements with no impact on end-user functionality.

Walkthrough

These changes convert benchmark harnesses across two files from divan's .bench_values(...) to .bench_refs(...) pattern, shifting measurement focus onto references instead of owned values. Reporting token benchmarks now measure content extraction, token generation, and message encoding directly; binary node benchmarks similarly shift for children iteration and JID attribute access.

Changes

Benchmark Harness Migration

Layer / File(s)	Summary
Reporting token benchmark harness updates wacore/benches/reporting_token_benchmark.rs	Content extraction, full token generation, and message encoding benchmarks convert from .bench_values(...) to .bench_refs(...), wrapping results in divan::black_box to shield computed values from optimization.
Binary node benchmark harness updates wacore/binary/benches/binary_benchmark.rs	Children-by-tag iteration and JID attribute access benchmarks switch to .bench_refs(...) input handling. The JID access benchmark preserves unmarshal/to_owned flow. The attribute parser benchmark now evaluates the final owned node before returning.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related PRs

• oxidezap/whatsapp-rust#247: Both PRs touch bench_get_children_by_tag, with this one refining the divan harness pattern built on the iterator refactor from #247.
• oxidezap/whatsapp-rust#837: Related divan benchmark input conversion effort, adjusting how benchmark closure inputs are handled across the codebase.
• oxidezap/whatsapp-rust#828: Part of the broader divan framework migration in wacore benchmarks; this PR refines the harness measurement patterns introduced in that conversion.

Suggested labels

performance

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely describes the main change: refactoring benchmarks to borrow inputs instead of consuming them, which is the core objective of this PR.
Description check	✅ Passed	The description is directly related to the changeset, explaining the motivation (measurement noise from drop operations), which benchmarks were affected, and why the changes were made.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch bench/get-children-borrow-input

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[codspeed-hq]

Merging this PR will improve performance by 54.01%

⚡ 9 improved benchmarks
✅ 131 untouched benchmarks

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	bench_get_children_by_tag	57.1 µs	19.3 µs	×3
⚡	Simulation	bench_message_encoding_simple	17.4 µs	9.1 µs	+91.25%
⚡	Simulation	bench_content_extraction_simple	19.1 µs	10.9 µs	+74.9%
⚡	Simulation	bench_message_encoding_extended	26.3 µs	15.9 µs	+65.3%
⚡	Simulation	bench_content_extraction_extended	29.9 µs	19.7 µs	+52%
⚡	Simulation	bench_full_token_generation_simple	42.3 µs	34.1 µs	+24.19%
⚡	Simulation	bench_full_token_generation_extended	53.2 µs	43.3 µs	+23.07%
⚡	Simulation	bench_unpack_uncompressed	212.5 ns	183.3 ns	+15.91%
⚡	Simulation	bench_unpad_message_ref	205.8 ns	186.4 ns	+10.43%

Tip

Curious why this is faster? Comment @codspeedbot explain why this is faster on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing bench/get-children-borrow-input (e6d8ff6) with main (be3d255)}

[cubic-dev-ai]

No issues found across 2 files

_{Re-trigger cubic}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@wacore/benches/reporting_token_benchmark.rs`:
- Line 54: The closure passed to .bench_refs currently calls
black_box(generate_reporting_token_content(msg)) but doesn't return the
generated value, so destruction happens inside the timed section; change these
closures (the ones invoking generate_reporting_token_content via .bench_refs at
the occurrences noted and the similar cases at the other two locations) to
return the produced token (e.g., return the Vec or token value) instead of only
black_box-ing it, so the benchmark measures creation cost while drop happens
outside the timed window—mirror the approach used in bench_jid_to_owned_access
where the produced value is returned.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: e87596b7-f23a-4c6a-8c6a-90e2bdf0055e

📥 Commits

Reviewing files that changed from the base of the PR and between be3d255 and e6d8ff6.

📒 Files selected for processing (2)

• wacore/benches/reporting_token_benchmark.rs
• wacore/binary/benches/binary_benchmark.rs