STYL: a few cleanups in pyi files #46921
Conversation
| @@ -163,7 +163,7 @@ class HashTable: | |||
| na_sentinel: int = ..., | |||
| na_value: object = ..., | |||
| mask=..., | |||
| ) -> tuple[np.ndarray, npt.NDArray[np.intp],]: ... # np.ndarray[subclass-specific] | |||
| ) -> tuple[np.ndarray, npt.NDArray[np.intp]]: ... # np.ndarray[subclass-specific] | |||
There was a problem hiding this comment.
_unique above only returns tuple if return_inverse is True, which is how it's called from factorize. The return type of _unique is a Union, but these are more easily avoided with overloads in stubs.
I assume that _unique is intended to be private so we probably shouldn't be calling it from outside, so rather than adding overloads could just remove _unique from the stub.
There was a problem hiding this comment.
I removed _unique but haven't yet added overloads to unique as I embarrassingly couldn't find the implementation.
There was a problem hiding this comment.
it's created from a template. pandas/_libs/hashtable_class_helper.pxi.in
There was a problem hiding this comment.
Thanks! Would need to first deprecate return_inverse as a positional argument. Can I just use the normal deprecation decorator in cython (or would that have an unintended performance hit).
There was a problem hiding this comment.
I doubt HashTable is part of the public API of pandas but is obviously not private to the module as it is used elsewhere. Can probably just make the arguments keyword only without deprecation. (OK as follow-up, for our internal consistency checking, if we do not have any errors reported then not a priority to add overloads if not public)
|
Thanks @twoertwein |
Might make sense to run flake8 on pyi files (add flake8-pyi to pre-commit)