TYP: mostly DataFrame return overloads #56739
Conversation
| @@ -11538,9 +11538,42 @@ def _reduce_axis1(self, name: str, func, skipna: bool) -> Series: | |||
| res_ser = self._constructor_sliced(result, index=self.index, copy=False) | |||
| return res_ser | |||
|
|
|||
| @doc(make_doc("any", ndim=2)) | |||
| # error: Signature of "any" incompatible with supertype "NDFrame" | |||
There was a problem hiding this comment.
I would be fine removing # error: ... when we do not expect that they can be fixed (it doesn't make sense for NDFrame to have those overloads, so there will always be a mismatch between them.)
pandas/io/formats/info.py
Outdated
| @@ -392,7 +392,7 @@ def dtype_counts(self) -> Mapping[str, int]: | |||
|
|
|||
| @property | |||
| @abstractmethod | |||
| def non_null_counts(self) -> Sequence[int]: | |||
| def non_null_counts(self) -> list[int] | Series: | |||
There was a problem hiding this comment.
Series is not a Sequence as Series.__hash__ = None
|
The keyword-only changes do not seem to be needed by mypy/pyright: can revert those changes (DataFrame.any is already keyword-only, but DataFrame.all/min//sum/... aren't) |
mroeschke
added
DataFrame
xref #57087 (I'm also in favor of keyword-only) |
|
@mroeschke Let me know if I should try to outsource some of the changes to a separate PR to make it easier to review. The keyword-only changes caused many changes: typing overloads and fixing positional usage in the tests and docs. |
|
Yeah could you break out the keyword-only deprecation into it's own PR separate from the misc typing changes? |
I removed the keyword-deprecation decorators |
| @@ -11805,6 +11805,7 @@ def _logical_func( | |||
|
|
|||
| def any( | |||
| self, | |||
| *, | |||
There was a problem hiding this comment.
Should these be in your follow up PR? Does it change the current behavior of e.g. DataFrame/Series.any()?
There was a problem hiding this comment.
I would be fine having those changes: NDFrame is private and all those methods are overwritten by DataFrame and Series, so these changes are not user-facing. Let me revert them to keep it less controversial.
There was a problem hiding this comment.
If you don't mind, I will revert this change back - see failing mypy&stubtest
pandas/core/series.py
Outdated
|
|
||
| @deprecate_nonkeyword_arguments(version="3.0", allowed_args=["self"], name="std") |
There was a problem hiding this comment.
These should be in your follow up too?
|
Thanks @twoertwein |
* TYP: mostly DataFrame return overloads * remove keyword-only enforcement * revert keyword-only in NDFrame * revert revert * remove decorators in Series
DataFrame.any/boolreturnboolwhenaxis=NoneandSeriesotherwiseDataFrame.sum/prodcurrently always returns aSeries(but will return anything whenaxis=Nonein the future)DataFrame.mean/median/var/std/skew/kurtreturnSerieswhenaxisis notNoneand potentially anything whenaxis=NoneI assume it is fine to directly use keyword-only arguments in NDFrame as it is private(?), and Series+DataFrame overwrite the methods in question.