Merging this PR will improve performance by 5.67%

⚡ 23 improved benchmarks
✅ 102 untouched benchmarks

Performance Changes

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

_{Comparing micha/parser-shrink-to-fit (3ade505) with main (572e4b5)}

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

Formatter (stable)

✅ ecosystem check detected no format changes.

Formatter (preview)

✅ ecosystem check detected no format changes.

Typing conformance results

No changes detected ✅

Memory usage report

Summary

Significant changes

ecosystem-analyzer results

No diagnostic changes detected ✅

Full report with detailed diff (timing results)

IMO the memory gains here are clearly worth a 1-2% linter regression.

I scoped this PR down to only cover statements to get a better sense for the performance trade-off when adding the same treatment to some other nodes.

I also tried a few different techniques to mitigate the performance regression with very limited success:

• Use a pool of Suite buffers that ruff parses into. This allows us to have a few vecs with large capacity. The downside is that it's always necessary to copy all elements (minus module level where I did not apply this optimization). This was maybe marginally faster, but not worth the complexity.
• Special case a body with a single statement because they are more common than I thought. This was maybe marginally faster, but not worth the complexity
• Use a SmallVec of size 8

In short, Vec is pretty fast