Python caching with type hints

Back to all notes

The functools.cache function in the Python standard library is amazing, as I’ve previously noted. But there is one annoying aspect (that I’ve noticed so far): when you use the @cache decorator, you lose the type hints on the now-cached function, which your IDE will now interpret as a lru_cache_wrapper instead of whatever signature it had before.

So if you have:

from functools import cache

@cache
def my_function(a: int, b: int) -> float:
    return a/b

And you write somewhere else in your code:

c = my_function("abc", 0.5)

Your IDE will not give you any warning that you’re doing something wrong. This is a well-known issue, which means that there is also a reasonably easy solution.

We can write our own decorator:

from functools import cache
import inspect

def cache_hint(func):
    new_func = cache(func)
    new_func.__signature__ = inspect.signature(func)
    return new_func

This simply takes the computed signature of the original function, and bluntly assign it to the new function. It doesn’t sound like it should work, but it does! Now if we do:

@cache_hint
def my_function(a: int, b: int) -> float:
    return a/b

c = my_function("abc", 0.5)

It will be flagged by the IDE, and you can notice it before you try to divide a string by a float, which is the kind of things an IDE should be able to do.