TYP: pandas/core/dtypes/base.py #31352
Conversation
pandas/core/dtypes/base.py
Outdated
| f"'construct_from_string' expects a string, got {type(string)}" | ||
| ) | ||
|
|
||
| if hasattr(cls, "_match"): |
There was a problem hiding this comment.
some repeated logic in pandas/core/dtypes/dtypes.py can be removed with this. will remove for now til needed. (will also be adding a test in the base extension tests for consistent error messages for wrong type to construct_from_string.)
I'll open a PR to break off changes to pandas/core/dtypes/dtypes.py without the deduplication for now.
| @@ -119,11 +121,11 @@ def __eq__(self, other: Any) -> bool: | |||
| def __hash__(self) -> int: | |||
| return hash(tuple(getattr(self, attr) for attr in self._metadata)) | |||
|
|
|||
| def __ne__(self, other) -> bool: | |||
| def __ne__(self, other: Any) -> bool: | |||
There was a problem hiding this comment.
Did we ever decide on consistent use of Any vs object here? Thought we were using object but maybe missed that
There was a problem hiding this comment.
in this class we already have def __eq__(self, other: Any) -> bool: and in #29947, Any was applied throughout.
I think it should be object in order to check the body of the __eq__/__ne__ methods, but since we want consistency these will need to be changed as well...
__eq__
pandas/core/arrays/interval.py: def __eq__(self, other):
pandas/core/arrays/sparse/dtype.py: def __eq__(self, other: Any) -> bool:
pandas/core/arrays/_arrow_utils.py: def __eq__(self, other):
pandas/core/arrays/_arrow_utils.py: def __eq__(self, other):
pandas/core/dtypes/base.py: def __eq__(self, other: Any) -> bool:
pandas/core/dtypes/dtypes.py: def __eq__(self, other: Any) -> bool:
pandas/core/dtypes/dtypes.py: def __eq__(self, other: Any) -> bool:
pandas/core/dtypes/dtypes.py: def __eq__(self, other: Any) -> bool:
pandas/core/dtypes/dtypes.py: def __eq__(self, other: Any) -> bool:
pandas/core/indexes/frozen.py: def __eq__(self, other: Any) -> bool:
pandas/io/pytables.py: def __eq__(self, other: Any) -> bool:
pandas/io/pytables.py: def __eq__(self, other: Any) -> bool:
pandas/io/stata.py: def __eq__(self, other: Any) -> bool:
pandas/tests/groupby/test_function.py: def __eq__(self, other):
pandas/tests/indexing/test_indexing.py: def __eq__(self, other) -> bool:
pandas/tests/scalar/timedelta/test_timedelta.py: def __eq__(self, other):
pandas/tests/scalar/timestamp/test_comparisons.py: def __eq__(self, other) -> bool:
pandas/tests/series/test_ufunc.py: def __eq__(self, other) -> bool:
pandas/tests/test_algos.py: def __eq__(self, other) -> bool:
pandas/tseries/offsets.py: def __eq__(self, other: Any) -> bool:
pandas/_libs/tslibs/offsets.pyx: def __eq__(self, other: Any) -> bool:
__ne__
pandas/core/arrays/interval.py: def __ne__(self, other):
pandas/io/pytables.py: def __ne__(self, other) -> bool:
pandas/tests/series/test_duplicates.py: def __ne__(self, other):
pandas/tseries/offsets.py: def __ne__(self, other):
pandas/_libs/tslibs/offsets.pyx: def __ne__(self, other):
However, doing this raises issues with Extension Array equalities, see #31646. So should be done in a separate PR, I think though that this does affirm the need to use object instead of Any for more thorough checking.
pandas/core/dtypes/dtypes.py
Outdated
| @@ -74,7 +74,7 @@ def register(self, dtype: Type[ExtensionDtype]) -> None: | |||
|
|
|||
| def find( | |||
| self, dtype: Union[Type[ExtensionDtype], str] | |||
| ) -> Optional[Type[ExtensionDtype]]: | |||
| ) -> Optional[Union[ExtensionDtype, Type[ExtensionDtype]]]: | |||
There was a problem hiding this comment.
construct_from_string on L98 returns an instance. However, since the type annotations for construct_from_string were removed following #31352 (comment), this can also be removed for now.
…ns/pandas into pandas/core/dtypes/base.py
|
This looks fine, any sticking points? |
|
Thanks @simonjayhawkins |
No description provided.