bench: cover the receive path — plaintext decode and appstate index-MAC dedup

[jlucaso1]

Bench-only PR (same flow as #854): map the baseline with CodSpeed before touching any code. Two inbound-side coverage gaps:

1. decode_plaintext — unpad + prost decode (wacore/src/messages.rs)

Runs on every received message and is the inbound mirror of the already-benched encode_and_pad, closing the send/receive coverage asymmetry. It's also where the #853 submessage boxing lands on the receive side, so the flamegraph doubles as a check on that.

Shapes match the send-side bench (text_reply, media_refs, large_text) plus group_skdm_text — the inline-SKDM group first-message.

Local numbers (median): text_reply 591 ns, large_text 644 ns, media_refs 1.11 µs, group_skdm_text 500 ns.

2. App-state index-MAC dedup — O(N²) scan (wacore/src/appstate_sync.rs)

The linear-scan dedup feeding the batched previous-value-MAC lookup, extracted from its two inline copies (inbound process_patch_list and outbound build_patch) into a pure collect_unique_index_macs so it can be measured. Byte-identical logic, no behavior change.

This is the measurement the known HashSet fix is blocked on: HashSet-vs-scan measured 6–120% worse at small N in this project, so both ends are pinned — N=10 (typical incremental patch) and N=1000 (resume-sync upper bound).

Local numbers (median): N=10 → 288 ns, N=1000 → 1.48 ms. The quadratic blowup is real at the upper bound (~500k Vec<u8> compares per resume patch); whether the fix pays without regressing small N is now CodSpeed's call.

Verification

• cargo fmt / cargo clippy -p wacore --benches --tests clean (only the pre-existing clippy.toml chrono warning)
• cargo test -p wacore: 987 + auxiliary suites, all green
• cargo check --all passes
• Both benches run locally (numbers above)

https://claude.ai/code/session_01EoJjbyorpCARCBTZSmUMRr

---

Generated by Claude Code

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: ec605cb6-0235-49f0-ae9a-5e7be5fec7b2

📥 Commits

Reviewing files that changed from the base of the PR and between 38cc31d and a179b51.

📒 Files selected for processing (4)

• wacore/Cargo.toml
• wacore/benches/appstate_sync_benchmark.rs
• wacore/benches/message_utils_benchmark.rs
• wacore/src/appstate_sync.rs

---

📝 Walkthrough

Summary by CodeRabbit

• Tests

  • Added new performance benchmarks for app state synchronization and message processing operations to measure and optimize internal hot paths.

• Chores

  • Updated build configuration to support additional benchmark targets.

Walkthrough

This PR extracts duplicate index MAC collection logic from two patch-processing sites into a single reusable helper, collect_unique_index_macs, then benchmarks that helper and message decoding costs under realistic load patterns. The changes are surgical: a focused helper reduces maintenance surface, and the benchmarks establish performance baselines for the orchestration layer's hot paths.

Changes

Appstate sync extraction and benchmarking

Layer / File(s)	Summary
Extract and refactor deduplication helper wacore/src/appstate_sync.rs	New public function collect_unique_index_macs extracts and deduplicates mutation index MAC blobs using first-seen order and linear-scan membership checks. Inbound patch processing and build_patch both refactored to call this helper instead of duplicating the deduplication loop.
Benchmark appstate_sync helper wacore/Cargo.toml, wacore/benches/appstate_sync_benchmark.rs	Cargo.toml registers the new benchmark target with harness disabled. Benchmark file defines setup_mutations to generate n distinct Set mutations with 32-byte index MACs, then measures collect_unique_index_macs throughput for n in [10, 1000] under black_box isolation.
Benchmark message decode flow wacore/benches/message_utils_benchmark.rs	Module docs updated to describe both encode (pad + prost) and decode (unpad) costs. New recv_shape helper generates receive message variants, including a group_skdm_text case that embeds a sender key distribution message. New bench_decode_plaintext benchmark measures pad + encode + decode round-trip for each shape.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related PRs

• oxidezap/whatsapp-rust#821: Directly aligns with batching previous-MAC prefetch in build_patch by collecting and deduplicating index MACs upfront.
• oxidezap/whatsapp-rust#828: Complements the Divan benchmark migration by adding CodSpeed CI infrastructure.

Suggested labels

performance

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately and specifically describes the main changes: adding benchmarks for plaintext decode and appstate index-MAC deduplication on the receive path.
Description check	✅ Passed	The description is comprehensive and directly related to the changeset, clearly explaining both benchmarks, their purpose, local performance numbers, and verification steps performed.
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 claude/receive-path-benchmarks-l786xn

---

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.}

[cubic-dev-ai]

No issues found across 4 files

_{Re-trigger cubic}

[codspeed-hq]

Merging this PR will not alter performance

✅ 145 untouched benchmarks
🆕 12 new benchmarks

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
🆕	Memory	bench_collect_unique_index_macs[10]	N/A	560 B	N/A
🆕	Memory	bench_collect_unique_index_macs[1000]	N/A	54.7 KB	N/A
🆕	Memory	bench_decode_plaintext[group_skdm_text]	N/A	767 B	N/A
🆕	Memory	bench_decode_plaintext[large_text]	N/A	4 KB	N/A
🆕	Memory	bench_decode_plaintext[media_refs]	N/A	12.8 KB	N/A
🆕	Memory	bench_decode_plaintext[text_reply]	N/A	3.4 KB	N/A
🆕	Simulation	bench_collect_unique_index_macs[10]	N/A	4.4 µs	N/A
🆕	Simulation	bench_collect_unique_index_macs[1000]	N/A	5.7 ms	N/A
🆕	Simulation	bench_decode_plaintext[group_skdm_text]	N/A	14 µs	N/A
🆕	Simulation	bench_decode_plaintext[large_text]	N/A	15.9 µs	N/A
🆕	Simulation	bench_decode_plaintext[media_refs]	N/A	25.3 µs	N/A
🆕	Simulation	bench_decode_plaintext[text_reply]	N/A	15.1 µs	N/A

---

_{Comparing claude/receive-path-benchmarks-l786xn (a179b51) with main (38cc31d)}