Fix Series.__new__ #722
Fix Series.__new__ #722
Conversation
| @@ -208,55 +209,52 @@ class Series(IndexOpsMixin, NDFrame, Generic[S1]): | |||
| _ListLike: TypeAlias = ArrayLike | dict[_str, np.ndarray] | list | tuple | Index | |||
| __hash__: ClassVar[None] | |||
|
|
|||
| # TODO: can __new__ be converted to __init__? Pandas implements __init__ | |||
There was a problem hiding this comment.
Can probably be done if #718 is possible
pandas-stubs/core/series.pyi
Outdated
| index: Axes | None = ..., | ||
| *, | ||
| dtype: Literal["datetime64[ns]"], |
There was a problem hiding this comment.
Probably should be TimestampDtypeArg
There was a problem hiding this comment.
I think all the other overloads would also need that: one overload with a specific data and one for _ListLike but a specific dtype?
There was a problem hiding this comment.
My concern there would be that it would be too wide. In other words, there would be contents of lists that wouldn't work with certain dtypes. I think the one is in there for TimestampSeries because it handles any kind of list.
There was a problem hiding this comment.
Probably should be
TimestampDtypeArg
I think this is the only change left to make.
pandas-stubs/core/series.pyi
Outdated
| ) -> TimestampSeries: ... | ||
| @overload | ||
| def __new__( | ||
| cls, | ||
| data: PeriodIndex, | ||
| index: Axes | None = ..., | ||
| dtype=..., | ||
| dtype: Dtype = ..., |
There was a problem hiding this comment.
Seems like we need to create a PeriodDtypeArg in _typing.pyi with the valid values here
| data: IntervalIndex[Interval[Timedelta]] | ||
| | Interval[Timedelta] | ||
| | Sequence[Interval[Timedelta]], | ||
| data: IntervalIndex[Interval[_OrderableT]] |
There was a problem hiding this comment.
Needed to switch to bound to simplify these overloads. All the other changes are related to this change.
Series.__new__()to be consistent in argument ordering. #717 (Replace xxxx with the Github issue number)assert_type()to assert the type of any return valueI only added
*for one overload as the implementation does not requires keyword-only arguments.I also removed
fastpathas it is not documented.