upgrade mypy and pyright. close 8 issues (#1087)

* update mypy and allow timestamp slicing

* fix for getitem with pythong 3.12 and mypy

* allow apply to return an offset

* allow loc setitem to accept a slice

* allow indexing with .loc using Timestamp

* define __bool__ for Series and DataFrame to be NoReturn

* support Path div

* change TypeVar to use Union where possible

* support dict keys in constructor to Series

* remove path overloads. add assert False for unreachable code

Author: Dr-Irv

pandas-stubs/_libs/tslibs/timestamps.pyi

@@ -11,6 +11,7 @@ from time import struct_time
 from typing import (
     ClassVar,
     Literal,
+    SupportsIndex,
     overload,
 )
 
@@ -48,7 +49,7 @@ _Nonexistent: TypeAlias = (
     Literal["raise", "NaT", "shift_backward", "shift_forward"] | Timedelta | timedelta
 )
 
-class Timestamp(datetime):
+class Timestamp(datetime, SupportsIndex):
     min: ClassVar[Timestamp]  # pyright: ignore[reportIncompatibleVariableOverride]
     max: ClassVar[Timestamp]  # pyright: ignore[reportIncompatibleVariableOverride]
 
@@ -309,3 +310,5 @@ class Timestamp(datetime):
     @property
     def unit(self) -> TimeUnit: ...
     def as_unit(self, unit: TimeUnit, round_ok: bool = ...) -> Self: ...
+    # To support slicing
+    def __index__(self) -> int: ...

pandas-stubs/_typing.pyi

@@ -57,7 +57,7 @@ Incomplete: TypeAlias = Any
 ArrayLike: TypeAlias = ExtensionArray | np.ndarray[Any, Any]
 AnyArrayLike: TypeAlias = Index[Any] | Series[Any] | np.ndarray[Any, Any]
 PythonScalar: TypeAlias = str | bool | complex
-DatetimeLikeScalar = TypeVar("DatetimeLikeScalar", Period, Timestamp, Timedelta)
+DatetimeLikeScalar = TypeVar("DatetimeLikeScalar", bound=Period | Timestamp | Timedelta)
 PandasScalar: TypeAlias = bytes | datetime.date | datetime.datetime | datetime.timedelta
 IntStrT = TypeVar("IntStrT", int, str)
 # Scalar: TypeAlias = PythonScalar | PandasScalar
@@ -490,7 +490,8 @@ AxisColumn: TypeAlias = Literal["columns", 1]
 Axis: TypeAlias = AxisIndex | AxisColumn
 DtypeNp = TypeVar("DtypeNp", bound=np.dtype[np.generic])
 KeysArgType: TypeAlias = Any
-ListLike = TypeVar("ListLike", Sequence, np.ndarray, Series, Index)
+ListLike: TypeAlias = Sequence | np.ndarray | Series | Index
+ListLikeT = TypeVar("ListLikeT", bound=ListLike)
 ListLikeExceptSeriesAndStr: TypeAlias = (
     MutableSequence[Any] | np.ndarray[Any, Any] | tuple[Any, ...] | Index[Any]
 )

pandas-stubs/core/dtypes/concat.pyi

@@ -6,7 +6,7 @@ from pandas import (
     Series,
 )
 
-_CatT = TypeVar("_CatT", Categorical, CategoricalIndex, Series)
+_CatT = TypeVar("_CatT", bound=Categorical | CategoricalIndex | Series)
 
 def union_categoricals(
     to_union: list[_CatT], sort_categories: bool = ..., ignore_order: bool = ...

pandas-stubs/core/frame.pyi

@@ -9,10 +9,12 @@ from collections.abc import (
 )
 import datetime as dt
 from re import Pattern
+import sys
 from typing import (
     Any,
     ClassVar,
     Literal,
+    NoReturn,
     overload,
 )
 
@@ -112,6 +114,7 @@ from pandas._typing import (
     ReplaceMethod,
     Scalar,
     ScalarT,
+    SequenceNotStr,
     SeriesByT,
     SortKind,
     StataDateFormat,
@@ -193,7 +196,11 @@ class _LocIndexerFrame(_LocIndexer):
     def __getitem__(  # type: ignore[overload-overlap] # pyright: ignore[reportOverlappingOverload]
         self,
         idx: tuple[
-            int | StrLike | tuple[Scalar, ...] | Callable[[DataFrame], ScalarT],
+            int
+            | StrLike
+            | Timestamp
+            | tuple[Scalar, ...]
+            | Callable[[DataFrame], ScalarT],
             int | StrLike | tuple[Scalar, ...],
         ],
     ) -> Scalar: ...
@@ -206,6 +213,7 @@ class _LocIndexerFrame(_LocIndexer):
                 IndexType
                 | MaskType
                 | _IndexSliceTuple
+                | SequenceNotStr[float | str | Timestamp]
                 | Callable[
                     [DataFrame], ScalarT | list[HashableT] | IndexType | MaskType
                 ],
@@ -219,7 +227,9 @@ class _LocIndexerFrame(_LocIndexer):
     @overload
     def __setitem__(
         self,
-        idx: MaskType | StrLike | _IndexSliceTuple | list[ScalarT] | IndexingInt,
+        idx: (
+            MaskType | StrLike | _IndexSliceTuple | list[ScalarT] | IndexingInt | slice
+        ),
         value: Scalar | NAType | NaTType | ArrayLike | Series | DataFrame | list | None,
     ) -> None: ...
     @overload
@@ -229,8 +239,32 @@ class _LocIndexerFrame(_LocIndexer):
         value: Scalar | NAType | NaTType | ArrayLike | Series | list | None,
     ) -> None: ...
 
-class DataFrame(NDFrame, OpsMixin):
-    __hash__: ClassVar[None]  # type: ignore[assignment]
+# With mypy 1.14.1 and python 3.12, the second overload needs a type-ignore statement
+if sys.version_info >= (3, 12):
+    class _GetItemHack:
+        @overload
+        def __getitem__(self, key: Scalar | tuple[Hashable, ...]) -> Series: ...  # type: ignore[overload-overlap] # pyright: ignore[reportOverlappingOverload]
+        @overload
+        def __getitem__(  # type: ignore[overload-overlap] # pyright: ignore[reportOverlappingOverload]
+            self, key: Iterable[Hashable] | slice
+        ) -> DataFrame: ...
+        @overload
+        def __getitem__(self, key: Hashable) -> Series: ...
+
+else:
+    class _GetItemHack:
+        @overload
+        def __getitem__(self, key: Scalar | tuple[Hashable, ...]) -> Series: ...  # type: ignore[overload-overlap] # pyright: ignore[reportOverlappingOverload]
+        @overload
+        def __getitem__(  # pyright: ignore[reportOverlappingOverload]
+            self, key: Iterable[Hashable] | slice
+        ) -> DataFrame: ...
+        @overload
+        def __getitem__(self, key: Hashable) -> Series: ...
+
+class DataFrame(NDFrame, OpsMixin, _GetItemHack):
+
+    __hash__: ClassVar[None]  # type: ignore[assignment] # pyright: ignore[reportIncompatibleMethodOverride]
 
     @overload
     def __new__(
@@ -607,14 +641,6 @@ class DataFrame(NDFrame, OpsMixin):
     @property
     def T(self) -> DataFrame: ...
     def __getattr__(self, name: str) -> Series: ...
-    @overload
-    def __getitem__(self, key: Scalar | tuple[Hashable, ...]) -> Series: ...  # type: ignore[overload-overlap] # pyright: ignore[reportOverlappingOverload]
-    @overload
-    def __getitem__(  # pyright: ignore[reportOverlappingOverload]
-        self, key: Iterable[Hashable] | slice
-    ) -> DataFrame: ...
-    @overload
-    def __getitem__(self, key: Hashable) -> Series: ...
     def isetitem(
         self, loc: int | Sequence[int], value: Scalar | ArrayLike | list[Any]
     ) -> None: ...
@@ -2453,6 +2479,7 @@ class DataFrame(NDFrame, OpsMixin):
     ) -> Self: ...
     def __truediv__(self, other: float | DataFrame | Series | Sequence) -> Self: ...
     def __rtruediv__(self, other: float | DataFrame | Series | Sequence) -> Self: ...
+    def __bool__(self) -> NoReturn: ...
 
 class _PandasNamedTuple(tuple[Any, ...]):
     def __getattr__(self, field: str) -> Scalar: ...

pandas-stubs/core/indexes/accessors.pyi

@@ -42,7 +42,7 @@ from pandas._typing import (
 
 class Properties(PandasDelegate, NoNewAttributesMixin): ...
 
-_DTFieldOpsReturnType = TypeVar("_DTFieldOpsReturnType", Series[int], Index[int])
+_DTFieldOpsReturnType = TypeVar("_DTFieldOpsReturnType", bound=Series[int] | Index[int])
 
 class _DayLikeFieldOps(Generic[_DTFieldOpsReturnType]):
     @property
@@ -84,7 +84,9 @@ class _DatetimeFieldOps(
     _DayLikeFieldOps[_DTFieldOpsReturnType], _MiniSeconds[_DTFieldOpsReturnType]
 ): ...
 
-_DTBoolOpsReturnType = TypeVar("_DTBoolOpsReturnType", Series[bool], np_ndarray_bool)
+_DTBoolOpsReturnType = TypeVar(
+    "_DTBoolOpsReturnType", bound=Series[bool] | np_ndarray_bool
+)
 
 class _IsLeapYearProperty(Generic[_DTBoolOpsReturnType]):
     @property
@@ -106,7 +108,7 @@ class _DatetimeBoolOps(
     @property
     def is_year_end(self) -> _DTBoolOpsReturnType: ...
 
-_DTFreqReturnType = TypeVar("_DTFreqReturnType", str, BaseOffset)
+_DTFreqReturnType = TypeVar("_DTFreqReturnType", bound=str | BaseOffset)
 
 class _FreqProperty(Generic[_DTFreqReturnType]):
     @property
@@ -121,10 +123,10 @@ class _DatetimeObjectOps(
 ): ...
 
 _DTOtherOpsDateReturnType = TypeVar(
-    "_DTOtherOpsDateReturnType", Series[dt.date], np.ndarray
+    "_DTOtherOpsDateReturnType", bound=Series[dt.date] | np.ndarray
 )
 _DTOtherOpsTimeReturnType = TypeVar(
-    "_DTOtherOpsTimeReturnType", Series[dt.time], np.ndarray
+    "_DTOtherOpsTimeReturnType", bound=Series[dt.time] | np.ndarray
 )
 
 class _DatetimeOtherOps(Generic[_DTOtherOpsDateReturnType, _DTOtherOpsTimeReturnType]):
@@ -157,11 +159,7 @@ class _DatetimeLikeOps(
 
 _DTTimestampTimedeltaReturnType = TypeVar(
     "_DTTimestampTimedeltaReturnType",
-    Series,
-    TimestampSeries,
-    TimedeltaSeries,
-    DatetimeIndex,
-    TimedeltaIndex,
+    bound=Series | TimestampSeries | TimedeltaSeries | DatetimeIndex | TimedeltaIndex,
 )
 
 class _DatetimeRoundingMethods(Generic[_DTTimestampTimedeltaReturnType]):
@@ -199,8 +197,10 @@ class _DatetimeRoundingMethods(Generic[_DTTimestampTimedeltaReturnType]):
 _DTNormalizeReturnType = TypeVar(
     "_DTNormalizeReturnType", TimestampSeries, DatetimeIndex
 )
-_DTStrKindReturnType = TypeVar("_DTStrKindReturnType", Series[str], Index)
-_DTToPeriodReturnType = TypeVar("_DTToPeriodReturnType", PeriodSeries, PeriodIndex)
+_DTStrKindReturnType = TypeVar("_DTStrKindReturnType", bound=Series[str] | Index)
+_DTToPeriodReturnType = TypeVar(
+    "_DTToPeriodReturnType", bound=PeriodSeries | PeriodIndex
+)
 
 class _DatetimeLikeNoTZMethods(
     _DatetimeRoundingMethods[_DTTimestampTimedeltaReturnType],
@@ -289,9 +289,11 @@ class DatetimeProperties(
     def as_unit(self, unit: TimeUnit) -> _DTTimestampTimedeltaReturnType: ...
 
 _TDNoRoundingMethodReturnType = TypeVar(
-    "_TDNoRoundingMethodReturnType", Series[int], Index
+    "_TDNoRoundingMethodReturnType", bound=Series[int] | Index
+)
+_TDTotalSecondsReturnType = TypeVar(
+    "_TDTotalSecondsReturnType", bound=Series[float] | Index
 )
-_TDTotalSecondsReturnType = TypeVar("_TDTotalSecondsReturnType", Series[float], Index)
 
 class _TimedeltaPropertiesNoRounding(
     Generic[_TDNoRoundingMethodReturnType, _TDTotalSecondsReturnType]
@@ -318,11 +320,15 @@ class TimedeltaProperties(
     def unit(self) -> TimeUnit: ...
     def as_unit(self, unit: TimeUnit) -> TimedeltaSeries: ...
 
-_PeriodDTReturnTypes = TypeVar("_PeriodDTReturnTypes", TimestampSeries, DatetimeIndex)
-_PeriodIntReturnTypes = TypeVar("_PeriodIntReturnTypes", Series[int], Index[int])
-_PeriodStrReturnTypes = TypeVar("_PeriodStrReturnTypes", Series[str], Index)
-_PeriodDTAReturnTypes = TypeVar("_PeriodDTAReturnTypes", DatetimeArray, DatetimeIndex)
-_PeriodPAReturnTypes = TypeVar("_PeriodPAReturnTypes", PeriodArray, PeriodIndex)
+_PeriodDTReturnTypes = TypeVar(
+    "_PeriodDTReturnTypes", bound=TimestampSeries | DatetimeIndex
+)
+_PeriodIntReturnTypes = TypeVar("_PeriodIntReturnTypes", bound=Series[int] | Index[int])
+_PeriodStrReturnTypes = TypeVar("_PeriodStrReturnTypes", bound=Series[str] | Index)
+_PeriodDTAReturnTypes = TypeVar(
+    "_PeriodDTAReturnTypes", bound=DatetimeArray | DatetimeIndex
+)
+_PeriodPAReturnTypes = TypeVar("_PeriodPAReturnTypes", bound=PeriodArray | PeriodIndex)
 
 class _PeriodProperties(
     Generic[

pandas-stubs/core/series.pyi

@@ -1,3 +1,4 @@
+from collections import dict_keys  # type: ignore[attr-defined]
 from collections.abc import (
     Callable,
     Hashable,
@@ -13,11 +14,13 @@ from datetime import (
     time,
     timedelta,
 )
+from pathlib import Path
 from typing import (
     Any,
     ClassVar,
     Generic,
     Literal,
+    NoReturn,
     overload,
 )
 
@@ -139,6 +142,7 @@ from pandas._typing import (
     ReplaceMethod,
     Scalar,
     ScalarT,
+    SequenceNotStr,
     SeriesByT,
     SortKind,
     StrDtypeArg,
@@ -195,8 +199,7 @@ class _LocIndexerSeries(_LocIndexer, Generic[S1]):
         idx: (
             MaskType
             | Index
-            | Sequence[float]
-            | list[str]
+            | SequenceNotStr[float | str | Timestamp]
             | slice
             | _IndexSliceTuple
             | Sequence[_IndexSliceTuple]
@@ -208,7 +211,7 @@ class _LocIndexerSeries(_LocIndexer, Generic[S1]):
     @overload
     def __setitem__(
         self,
-        idx: Index | MaskType,
+        idx: Index | MaskType | slice,
         value: S1 | ArrayLike | Series[S1] | None,
     ) -> None: ...
     @overload
@@ -352,7 +355,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
     @overload
     def __new__(
         cls,
-        data: S1 | _ListLike[S1] | dict[HashableT1, S1],
+        data: S1 | _ListLike[S1] | dict[HashableT1, S1] | dict_keys[S1, Any],
         index: Axes | None = ...,
         *,
         dtype: Dtype = ...,
@@ -1030,6 +1033,14 @@ class Series(IndexOpsMixin[S1], NDFrame):
         **kwds,
     ) -> Series: ...
     @overload
+    def apply(
+        self,
+        func: Callable[..., BaseOffset],
+        convertDType: _bool = ...,
+        args: tuple = ...,
+        **kwds,
+    ) -> OffsetSeries: ...
+    @overload
     def apply(
         self,
         func: Callable[..., Series],
@@ -1640,7 +1651,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
         self, other: int | np_ndarray_anyint | Series[int]
     ) -> Series[int]: ...
     def __rsub__(self, other: num | _ListLike | Series[S1]) -> Series: ...
-    def __rtruediv__(self, other: num | _ListLike | Series[S1]) -> Series: ...
+    def __rtruediv__(self, other: num | _ListLike | Series[S1] | Path) -> Series: ...
     # ignore needed for mypy as we want different results based on the arguments
     @overload  # type: ignore[override]
     def __rxor__(  # pyright: ignore[reportOverlappingOverload]
@@ -1666,7 +1677,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
     ) -> TimedeltaSeries: ...
     @overload
     def __sub__(self, other: num | _ListLike | Series) -> Series: ...
-    def __truediv__(self, other: num | _ListLike | Series[S1]) -> Series: ...
+    def __truediv__(self, other: num | _ListLike | Series[S1] | Path) -> Series: ...
     # ignore needed for mypy as we want different results based on the arguments
     @overload  # type: ignore[override]
     def __xor__(  # pyright: ignore[reportOverlappingOverload]
@@ -2144,6 +2155,7 @@ class Series(IndexOpsMixin[S1], NDFrame):
         level: Level | None = ...,
         drop_level: _bool = ...,
     ) -> Self: ...
+    def __bool__(self) -> NoReturn: ...
 
 class TimestampSeries(Series[Timestamp]):
     @property

pandas-stubs/core/strings.pyi

@@ -28,11 +28,11 @@ from pandas._typing import (
 )
 
 # The _TS type is what is used for the result of str.split with expand=True
-_TS = TypeVar("_TS", DataFrame, MultiIndex)
+_TS = TypeVar("_TS", bound=DataFrame | MultiIndex)
 # The _TS2 type is what is used for the result of str.split with expand=False
-_TS2 = TypeVar("_TS2", Series[list[str]], Index[list[str]])
+_TS2 = TypeVar("_TS2", bound=Series[list[str]] | Index[list[str]])
 # The _TM type is what is used for the result of str.match
-_TM = TypeVar("_TM", Series[bool], np_ndarray_bool)
+_TM = TypeVar("_TM", bound=Series[bool] | np_ndarray_bool)
 
 class StringMethods(NoNewAttributesMixin, Generic[T, _TS, _TM, _TS2]):
     def __init__(self, data: T) -> None: ...

pyproject.toml

@@ -33,11 +33,11 @@ types-pytz = ">= 2022.1.1"
 numpy = ">= 1.23.5"
 
 [tool.poetry.group.dev.dependencies]
-mypy = "1.13.0"
+mypy = "1.14.1"
 pandas = "2.2.3"
 pyarrow = ">=10.0.1"
 pytest = ">=7.1.2"
-pyright = ">= 1.1.390"
+pyright = ">= 1.1.391"
 poethepoet = ">=0.16.5"
 loguru = ">=0.6.0"
 typing-extensions = ">=4.4.0"