Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.refactron.dev/llms.txt

Use this file to discover all available pages before exploring further.

Transform ID: lru_cache_to_cache Language: Python

​
What it does

Rewrites the unbounded-cache idiom @functools.lru_cache(maxsize=None) to the equivalent @functools.cache decorator added in Python 3.9. Handles both the attribute form (@functools.lru_cache(...)) and the bare-name form (@lru_cache(...) after from functools import lru_cache), and updates the from functools import … line so it imports cache (renamed if lru_cache is no longer used, augmented alongside it if a finite-cap site survives). The sidecar (src/transform/transforms/python/_py/lru_cache_to_cache.py) drives the LibCST rewrite.

​
Detector pattern

The detector at src/analyze/detectors/python/lru-cache-maxsize-none.ts matches decorator call expressions whose function is lru_cache or functools.lru_cache AND whose argument list is exactly one keyword argument maxsize=None. Bare @lru_cache(), @lru_cache(128), and @lru_cache(maxsize=None, typed=True) are intentionally not flagged.

​
Preconditions

  1. python_version_too_low — refuses when the project’s resolved Python version is < 3.9 (or unknown). functools.cache is a 3.9 addition; pin via the pythonVersion config key.
  2. The decorator call must be exactly lru_cache(maxsize=None) — no typed=True, no finite maxsize, no zero-arg form (@lru_cache() defaults to maxsize=128, not unbounded, so it is not a safe @cache substitute).
  3. Bare-name decorators only resolve when from functools import lru_cache (or the no-op as lru_cache) is in scope. A genuine rebinding (from functools import lru_cache as cache_dec) is left untouched — rewriting under an alias would invalidate the existing call sites.

​
Before / after

import functools


@functools.lru_cache(maxsize=None)
def fib(n: int) -> int:
    if n < 2:
        return n
    return fib(n - 1) + fib(n - 2)
For the bare-name form, the from functools import … line is also updated: 
from functools import lru_cache


@lru_cache(maxsize=None)
def expensive(x: int) -> int:
    return x * x
When a finite-cap @lru_cache(128) site survives in the same file, the import is augmented instead of renamed: from functools import lru_cache, cache.

​
Edge cases NOT handled (skip via precondition)

  • @lru_cache() (zero args) — semantically equivalent to @lru_cache(maxsize=128) per CPython docs; not unbounded, so swapping in @cache would change behaviour.
  • @lru_cache(128) / @lru_cache(maxsize=1024) — finite caps; @cache would grow without bound.
  • @lru_cache(maxsize=None, typed=True) — type-discriminated caching has no @cache equivalent.
  • Decorators using a renamed import (from functools import lru_cache as cache_dec) — left alone because we can’t rewrite the alias’s call sites.
Version gating is hard — if pythonVersion is null and not detectable from pyproject.toml’s requires-python, the transform refuses rather than guess. Set "pythonVersion": "3.9" (or higher) in .refactronrc.json to unlock the rewrite.