Python optimizations

[samuelcolvin]

Summary by cubic

Speeds up Python parsing by locking the string cache once per parse instead of per string, and enables the same release optimizations for wheel builds that the Rust benchmarks already had.

Also makes the Python benchmark script more correct: it now validates each parser's output against json.loads and shows a cross-case summary.

• The cache lock is acquired with try_lock, so concurrent parses fall back to uncached strings instead of blocking on the mutex.
• Re-entering the cache from a parse on the same thread panics with a clear message instead of deadlocking; other threads still block.
• Single-character ASCII strings use CPython's interned singletons, avoiding allocation for each short string.
• profile.release now uses lto = "fat" and codegen-units = 1 for compiled wheels.
• Benchmark rows that don't match stdlib output are shown as invalid and excluded from the summary; parsers that raise are shown as errors and not timed.

^{Written for commit fc620f4. Summary will update on new commits.}

[cubic-dev-ai]

All reported issues were addressed across 4 files

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

Re-trigger cubic</sub>

[codecov]

Codecov Report

❌ Patch coverage is 93.42105% with 5 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
crates/jiter/src/py_string_cache.rs	93.05%	5 Missing ⚠️

📢 Thoughts on this report? Let us know!

[codspeed-hq]

Merging this PR will regress 3 benchmarks

⚡ 7 improved benchmarks
❌ 3 regressed benchmarks
✅ 51 untouched benchmarks
⏩ 71 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	WallTime	python_parse_x100	137 ns	198 ns	-30.81%
❌	WallTime	python_parse_other	369 ns	429 ns	-13.99%
❌	WallTime	python_parse_x100_not_cached	132 ns	145 ns	-8.97%
⚡	WallTime	python_parse_string_array	13.3 µs	10.4 µs	+28.09%
⚡	WallTime	python_parse_string_array_unique	1.6 ms	1.3 ms	+26.7%
⚡	WallTime	python_parse_medium_response	25.4 µs	22 µs	+15.82%
⚡	WallTime	python_parse_true_object	26.3 µs	22.9 µs	+15.11%
⚡	WallTime	short_numbers_jiter_value	57.1 µs	51.5 µs	+10.86%
⚡	WallTime	string_array_jiter_value_owned	17.5 µs	16 µs	+9.94%
⚡	WallTime	python_parse_numeric	1.3 µs	1.3 µs	+6.5%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing python-optimizations (fc620f4) with main (d2e138b)}

Footnotes

1. 71 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. ↩

[cubic-dev-ai]

All reported issues were addressed across 5 files

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

Re-trigger cubic</sub>