[MichaReiser]

Summary

Several lint rules can resolve the same callee while checking a single call. For example, with import numpy as np, they can each look up np.random.randint in np.random.randint(10). The call checker now resolves the callee once if any rule in the group is enabled, skips the group if resolution fails, and passes the qualified name to each enabled handler.

The first three rows count native arm64 instructions spent linting Ruff’s five benchmark files, with parsing and setup excluded. The last row counts the whole ruff check process over 686 Python 3.12 standard-library files with default rules and caching disabled.

Workload	Before (million instructions)	After (million instructions)	Change
Lint, default rules	29.02	27.41	−5.53%
Lint, all stable rules	58.34	53.06	−9.05%
Lint, all rules with preview	71.65	66.24	−7.54%
Full standard-library check	3,308.72	3,236.85	−2.17%

Test Plan

The linter and semantic-model tests passed, and 1,711 rule fixtures produced the same diagnostics and applied fixes, allowing for display order at tied source offsets.

[astral-automations-bot]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

Linter (stable) · Linter (preview)

---

Results for 44142d2 · Workflow run

[codspeed]

Merging this PR will improve performance by 7.28%

⚡ 5 improved benchmarks
✅ 55 untouched benchmarks
⏩ 242 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	CPU Simulation	linter/all-rules[large/dataset.py]	33.2 ms	29.7 ms	+11.66%
⚡	CPU Simulation	linter/all-with-preview-rules[large/dataset.py]	39.5 ms	36.2 ms	+9.33%
⚡	CPU Simulation	linter/default-rules[large/dataset.py]	16.9 ms	16.1 ms	+5.48%
⚡	CPU Simulation	linter/all-rules[pydantic/types.py]	15.7 ms	14.9 ms	+5.44%
⚡	CPU Simulation	linter/all-with-preview-rules[pydantic/types.py]	18.4 ms	17.6 ms	+4.65%

Tip

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

---

_{Comparing micha/reuse-callee-qualified-name (44142d2) with main (08cab22)}

Footnotes

1. 242 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. ↩