Lazily create builtin bindings

[charliermarsh]

Summary

The linter previously inserted every version-specific builtin, magic global, notebook builtin, and configured custom builtin into the module scope before walking the AST, even when a file never referenced most of those names.

This moves builtin identification into SemanticModel and creates concrete builtin bindings only when a load or binding operation needs them.

[codspeed-hq]

Merging this PR will improve performance by 9.79%

⚠️ 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

⚡ 10 improved benchmarks
✅ 50 untouched benchmarks
⏩ 93 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	linter/default-rules[numpy/globals.py]	179.1 µs	148.7 µs	+20.42%
⚡	Memory	linter/default-rules[pydantic/types.py]	348.9 KB	306.7 KB	+13.78%
⚡	Memory	linter/default-rules[numpy/ctypeslib.py]	164.2 KB	146.2 KB	+12.31%
⚡	Simulation	linter/all-with-preview-rules[numpy/globals.py]	832.5 µs	749.1 µs	+11.13%
⚡	Simulation	linter/all-rules[numpy/globals.py]	717 µs	664.6 µs	+7.88%
⚡	Memory	linter/all-rules[pydantic/types.py]	648.8 KB	606.6 KB	+6.97%
⚡	Simulation	linter/default-rules[unicode/pypinyin.py]	370.1 µs	346.3 µs	+6.87%
⚡	Memory	linter/all-with-preview-rules[pydantic/types.py]	667.2 KB	625 KB	+6.76%
⚡	Memory	linter/all-rules[numpy/ctypeslib.py]	289.3 KB	271.3 KB	+6.64%
⚡	Memory	linter/all-with-preview-rules[numpy/ctypeslib.py]	317.3 KB	299.3 KB	+6.01%

Tip

Curious why this is faster? Use the CodSpeed MCP and ask your agent.

---

_{Comparing charlie/codex-codspeed-speedup (634be4d) with main (71d151f)}

Footnotes

1. 93 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. ↩

[ntBre]

Nice! The performance improvement looks great, but I'm a bit wary of the API. It seems like it would be easy to forget to call ensure_builtin_binding everywhere it's necessary. Is there any way to use a Cell or something to initialize the builtins automatically when we try to access them?

As one example, Codex claims that this snippet:

from airflow.operators.python import PythonOperator

task = PythonOperator(
    task_id="task",
    op_args="{{ ti.xcom_pull(task_ids='len') }}",
)

ruff check --preview --select AIR201 --unsafe-fixes repro.py

has no fix on this branch but offers a fix on main. The result of the fix actually looks like a bug on main (playground), but it's still an example of a behavior change that I think is accidental.

In general Codex claims "every caller must remember whether lookup_symbol alone is sufficient or whether it also needs has_builtin_binding," so it seems nicer if we could hide the builtin construction inside lookup_symbol somehow.

[astral-sh-bot]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[charliermarsh]

Good feedback -- how does this look?

[ntBre]

Looks great to me, thank you!