Speed up multipart callback dispatch

[Kludex]

Summary

• normalize omitted and explicit-None multipart callbacks to shared no-op functions once during parser initialization
• cache callbacks in private dispatch slots and keep them synchronized through set_callback()
• dispatch callbacks directly instead of constructing callback names and calling BaseParser.callback() for every event
• pass resolved data callbacks into the existing mark/lookbehind helper

Performance

CodSpeed reports a 24.15% efficiency improvement for the corrected 100-field complete-parse benchmark, from 3.7 ms to 3.0 ms. The other four corrected benchmarks are unchanged. CodSpeed notes that the compared runs used different runtime environments, so the same-environment local results below provide additional context.

Same-environment local wall-time benchmarks on Python 3.13:

Benchmark	Base	Head	Improvement
Simple multipart form	15.53 µs	13.63 µs	12.2%
100-field multipart form	403.28 µs	334.28 µs	17.1%
1,000-field multipart form	3,988.07 µs	3,341.27 µs	16.2%
8 MiB file upload	605.73 µs	599.75 µs	effectively unchanged
1 MiB CR/LF-dense upload	57.96 µs	55.52 µs	4.2%

A Starlette-shaped 100-field multipart parse improved from approximately 661 µs to 616 µs (about 6.8%).

Validation

• 162 tests pass
• 100% statement coverage
• Ruff passes
• mypy passes
• source distribution check passes

[codspeed-hq]

Merging this PR will improve performance by 24.15%

⚠️Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 1 improved benchmark
✅ 4 untouched benchmarks
⏩ 5 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	test_parse_multipart_large_form	3.7 ms	3 ms	+24.15%

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 speed-up-multipart-callback-dispatch (d74fe57) with main (d9cb4c6)}

Footnotes

1. 5 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit:0c875e1e81

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Kludex]

After rebasing onto the corrected benchmark lifecycle from #318, CodSpeed now reports a 24.15% efficiency improvement for the 100-field complete-parse benchmark (3.7 ms to 3.0 ms), with the other four corrected benchmarks unchanged. The report warns that the runs used different environments, while same-environment local measurements show a 17.1% improvement for the equivalent workload. The Cubic finding is marked addressed in d74fe57 and its thread is resolved; no further code change is needed.

[cubic-dev-ai]

All reported issues were addressed across 2 files (changes from recent commits).

<sub>Reply with feedback, questions, or to request a fix.

Re-trigger cubic</sub>

[Kludex]

The current CodSpeed comparison is not representative because the benchmark fixture reuses an already-finished parser across measured iterations. I opened #318 to create and finalize a fresh parser per benchmark invocation. PR #317 should be rebased and evaluated again after #318 merges; the corrected baseline will provide a meaningful performance comparison.

[Kludex]

Rebased onto main after #318 merged. CodSpeed will now compare PR #317 against fresh-parser, complete-request benchmark iterations. Also addressed the Cubic review in d74fe57 by testing set_callback("part_data", None) and resolving the thread.