PERF: slightly faster dtype.kind checks (pandas-dev#52387)

Author: jbrockmendel

pandas/_libs/lib.pyx

@@ -747,7 +747,7 @@ cpdef ndarray[object] ensure_string_array(
 
     if hasattr(arr, "to_numpy"):
 
-        if hasattr(arr, "dtype") and arr.dtype.kind in ["m", "M"]:
+        if hasattr(arr, "dtype") and arr.dtype.kind in "mM":
             # dtype check to exclude DataFrame
             # GH#41409 TODO: not a great place for this
             out = arr.astype(str).astype(object)
@@ -2641,7 +2641,7 @@ def maybe_convert_objects(ndarray[object] objects,
             dtype = dtype_if_all_nat
             if cnp.PyArray_DescrCheck(dtype):
                 # i.e. isinstance(dtype, np.dtype)
-                if dtype.kind not in ["m", "M"]:
+                if dtype.kind not in "mM":
                     raise ValueError(dtype)
                 else:
                     res = np.empty((<object>objects).shape, dtype=dtype)

pandas/_libs/tslibs/np_datetime.pyx

@@ -419,7 +419,7 @@ def compare_mismatched_resolutions(ndarray left, ndarray right, op):
     array([ True])
     """
 
-    if left.dtype.kind != right.dtype.kind or left.dtype.kind not in ["m", "M"]:
+    if left.dtype.kind != right.dtype.kind or left.dtype.kind not in "mM":
         raise ValueError("left and right must both be timedelta64 or both datetime64")
 
     cdef:

pandas/_libs/tslibs/timedeltas.pyx

@@ -775,7 +775,7 @@ def _binary_op_method_timedeltalike(op, name):
                 item = cnp.PyArray_ToScalar(cnp.PyArray_DATA(other), other)
                 return f(self, item)
 
-            elif other.dtype.kind in ["m", "M"]:
+            elif other.dtype.kind in "mM":
                 return op(self.to_timedelta64(), other)
             elif other.dtype.kind == "O":
                 return np.array([op(self, x) for x in other])
@@ -2028,7 +2028,7 @@ class Timedelta(_Timedelta):
                     result[mask] = np.nan
                 return result
 
-            elif other.dtype.kind in ["i", "u", "f"]:
+            elif other.dtype.kind in "iuf":
                 if other.ndim == 0:
                     return self // other.item()
                 else:

pandas/_libs/tslibs/timestamps.pyx

@@ -464,7 +464,7 @@ cdef class _Timestamp(ABCTimestamp):
             raise integer_op_not_supported(self)
 
         elif is_array(other):
-            if other.dtype.kind in ["i", "u"]:
+            if other.dtype.kind in "iu":
                 raise integer_op_not_supported(self)
             if other.dtype.kind == "m":
                 if self.tz is None:
@@ -496,7 +496,7 @@ cdef class _Timestamp(ABCTimestamp):
             return self + neg_other
 
         elif is_array(other):
-            if other.dtype.kind in ["i", "u"]:
+            if other.dtype.kind in "iu":
                 raise integer_op_not_supported(self)
             if other.dtype.kind == "m":
                 if self.tz is None:

pandas/_testing/asserters.py

@@ -530,7 +530,7 @@ def assert_interval_array_equal(
     _check_isinstance(left, right, IntervalArray)
 
     kwargs = {}
-    if left._left.dtype.kind in ["m", "M"]:
+    if left._left.dtype.kind in "mM":
         # We have a DatetimeArray or TimedeltaArray
         kwargs["check_freq"] = False
 

pandas/core/algorithms.py

@@ -568,7 +568,7 @@ def factorize_array(
     uniques : ndarray
     """
     original = values
-    if values.dtype.kind in ["m", "M"]:
+    if values.dtype.kind in "mM":
         # _get_hashtable_algo will cast dt64/td64 to i8 via _ensure_data, so we
         #  need to do the same to na_value. We are assuming here that the passed
         #  na_value is an appropriately-typed NaT.

pandas/core/array_algos/datetimelike_accumulations.py

@@ -50,7 +50,7 @@ def _cum_func(
     result = func(y)
     result[mask] = iNaT
 
-    if values.dtype.kind in ["m", "M"]:
+    if values.dtype.kind in "mM":
         return result.view(values.dtype.base)
     return result
 

pandas/core/array_algos/putmask.py

@@ -142,7 +142,7 @@ def setitem_datetimelike_compat(values: np.ndarray, num_set: int, other):
     if values.dtype == object:
         dtype, _ = infer_dtype_from(other, pandas_dtype=True)
 
-        if isinstance(dtype, np.dtype) and dtype.kind in ["m", "M"]:
+        if isinstance(dtype, np.dtype) and dtype.kind in "mM":
             # https://github.com/numpy/numpy/issues/12550
             #  timedelta64 will incorrectly cast to int
             if not is_list_like(other):

pandas/core/array_algos/quantile.py

@@ -180,7 +180,7 @@ def _nanpercentile(
     quantiles : scalar or array
     """
 
-    if values.dtype.kind in ["m", "M"]:
+    if values.dtype.kind in "mM":
         # need to cast to integer to avoid rounding errors in numpy
         result = _nanpercentile(
             values.view("i8"),

pandas/core/arrays/datetimelike.py

@@ -740,7 +740,7 @@ def isin(self, values) -> npt.NDArray[np.bool_]:
         if not hasattr(values, "dtype"):
             values = np.asarray(values)
 
-        if values.dtype.kind in ["f", "i", "u", "c"]:
+        if values.dtype.kind in "fiuc":
             # TODO: de-duplicate with equals, validate_comparison_value
             return np.zeros(self.shape, dtype=bool)
 
@@ -769,7 +769,7 @@ def isin(self, values) -> npt.NDArray[np.bool_]:
             except ValueError:
                 return isin(self.astype(object), values)
 
-        if self.dtype.kind in ["m", "M"]:
+        if self.dtype.kind in "mM":
             self = cast("DatetimeArray | TimedeltaArray", self)
             values = values.as_unit(self.unit)
 
@@ -1205,7 +1205,7 @@ def _sub_nat(self):
         # For period dtype, timedelta64 is a close-enough return dtype.
         result = np.empty(self.shape, dtype=np.int64)
         result.fill(iNaT)
-        if self.dtype.kind in ["m", "M"]:
+        if self.dtype.kind in "mM":
             # We can retain unit in dtype
             self = cast("DatetimeArray| TimedeltaArray", self)
             return result.view(f"timedelta64[{self.unit}]")

pandas/core/arrays/masked.py

@@ -694,7 +694,7 @@ def _arith_method(self, other, op):
         else:
             # Make sure we do this before the "pow" mask checks
             #  to get an expected exception message on shape mismatch.
-            if self.dtype.kind in ["i", "u"] and op_name in ["floordiv", "mod"]:
+            if self.dtype.kind in "iu" and op_name in ["floordiv", "mod"]:
                 # TODO(GH#30188) ATM we don't match the behavior of non-masked
                 #  types with respect to floordiv-by-zero
                 pd_op = op

pandas/core/arrays/numeric.py

@@ -80,7 +80,7 @@ def __from_arrow__(
             # test_from_arrow_type_error raise for string, but allow
             #  through itemsize conversion GH#31896
             rt_dtype = pandas_dtype(array.type.to_pandas_dtype())
-            if rt_dtype.kind not in ["i", "u", "f"]:
+            if rt_dtype.kind not in "iuf":
                 # Could allow "c" or potentially disallow float<->int conversion,
                 #  but at the moment we specifically test that uint<->int works
                 raise TypeError(

pandas/core/arrays/numpy_.py

@@ -217,7 +217,7 @@ def _validate_scalar(self, fill_value):
         return fill_value
 
     def _values_for_factorize(self) -> tuple[np.ndarray, float | None]:
-        if self.dtype.kind in ["i", "u", "b"]:
+        if self.dtype.kind in "iub":
             fv = None
         else:
             fv = np.nan

pandas/core/arrays/sparse/array.py

@@ -228,7 +228,7 @@ def _sparse_array_op(
         if (
             name in ["floordiv", "mod"]
             and (right == 0).any()
-            and left.dtype.kind in ["i", "u"]
+            and left.dtype.kind in "iu"
         ):
             # Match the non-Sparse Series behavior
             opname = f"sparse_{name}_float64"

pandas/core/construction.py

@@ -756,7 +756,7 @@ def _try_cast(
         return ensure_wrapped_if_datetimelike(arr).astype(dtype, copy=copy)
 
     elif dtype.kind == "U":
-        # TODO: test cases with arr.dtype.kind in ["m", "M"]
+        # TODO: test cases with arr.dtype.kind in "mM"
         if is_ndarray:
             arr = cast(np.ndarray, arr)
             shape = arr.shape
@@ -768,7 +768,7 @@ def _try_cast(
             shape
         )
 
-    elif dtype.kind in ["m", "M"]:
+    elif dtype.kind in "mM":
         return maybe_cast_to_datetime(arr, dtype)
 
     # GH#15832: Check if we are requesting a numeric dtype and

pandas/core/dtypes/astype.py

@@ -86,7 +86,7 @@ def _astype_nansafe(
     elif not isinstance(dtype, np.dtype):  # pragma: no cover
         raise ValueError("dtype must be np.dtype or ExtensionDtype")
 
-    if arr.dtype.kind in ["m", "M"]:
+    if arr.dtype.kind in "mM":
         from pandas.core.construction import ensure_wrapped_if_datetimelike
 
         arr = ensure_wrapped_if_datetimelike(arr)

pandas/core/dtypes/cast.py

@@ -210,7 +210,7 @@ def _maybe_unbox_datetimelike(value: Scalar, dtype: DtypeObj) -> Scalar:
 
     Notes
     -----
-    Caller is responsible for checking dtype.kind in ["m", "M"]
+    Caller is responsible for checking dtype.kind in "mM"
     """
     if is_valid_na_for_dtype(value, dtype):
         # GH#36541: can't fill array directly with pd.NaT
@@ -295,7 +295,7 @@ def maybe_downcast_to_dtype(result: ArrayLike, dtype: str | np.dtype) -> ArrayLi
 
     # a datetimelike
     # GH12821, iNaT is cast to float
-    if dtype.kind in ["M", "m"] and result.dtype.kind in ["i", "f"]:
+    if dtype.kind in "mM" and result.dtype.kind in "if":
         result = result.astype(dtype)
 
     elif dtype.kind == "m" and result.dtype == _dtype_obj:
@@ -544,7 +544,7 @@ def ensure_dtype_can_hold_na(dtype: DtypeObj) -> DtypeObj:
         return _dtype_obj
     elif dtype.kind == "b":
         return _dtype_obj
-    elif dtype.kind in ["i", "u"]:
+    elif dtype.kind in "iu":
         return np.dtype(np.float64)
     return dtype
 
@@ -623,8 +623,7 @@ def _maybe_promote(dtype: np.dtype, fill_value=np.nan):
         dtype = _dtype_obj
         return dtype, fill_value
 
-    kinds = ["i", "u", "f", "c", "m", "M"]
-    if is_valid_na_for_dtype(fill_value, dtype) and dtype.kind in kinds:
+    if is_valid_na_for_dtype(fill_value, dtype) and dtype.kind in "iufcmM":
         dtype = ensure_dtype_can_hold_na(dtype)
         fv = na_value_for_dtype(dtype)
         return dtype, fv
@@ -1196,7 +1195,7 @@ def maybe_cast_to_datetime(
     from pandas.core.arrays.datetimes import DatetimeArray
     from pandas.core.arrays.timedeltas import TimedeltaArray
 
-    assert dtype.kind in ["m", "M"]
+    assert dtype.kind in "mM"
     if not is_list_like(value):
         raise TypeError("value must be listlike")
 
@@ -1251,7 +1250,7 @@ def _ensure_nanosecond_dtype(dtype: DtypeObj) -> None:
         # i.e. datetime64tz
         pass
 
-    elif dtype.kind in ["m", "M"]:
+    elif dtype.kind in "mM":
         reso = get_unit_from_dtype(dtype)
         if not is_supported_unit(reso):
             # pre-2.0 we would silently swap in nanos for lower-resolutions,
@@ -1294,7 +1293,7 @@ def find_result_type(left: ArrayLike, right: Any) -> DtypeObj:
 
     if (
         isinstance(left, np.ndarray)
-        and left.dtype.kind in ["i", "u", "c"]
+        and left.dtype.kind in "iuc"
         and (lib.is_integer(right) or lib.is_float(right))
     ):
         # e.g. with int8 dtype and right=512, we want to end up with
@@ -1335,7 +1334,7 @@ def common_dtype_categorical_compat(
     # GH#38240
 
     # TODO: more generally, could do `not can_hold_na(dtype)`
-    if isinstance(dtype, np.dtype) and dtype.kind in ["i", "u"]:
+    if isinstance(dtype, np.dtype) and dtype.kind in "iu":
         for obj in objs:
             # We don't want to accientally allow e.g. "categorical" str here
             obj_dtype = getattr(obj, "dtype", None)
@@ -1429,7 +1428,7 @@ def construct_2d_arraylike_from_scalar(
 ) -> np.ndarray:
     shape = (length, width)
 
-    if dtype.kind in ["m", "M"]:
+    if dtype.kind in "mM":
         value = _maybe_box_and_unbox_datetimelike(value, dtype)
     elif dtype == _dtype_obj:
         if isinstance(value, (np.timedelta64, np.datetime64)):
@@ -1486,13 +1485,13 @@ def construct_1d_arraylike_from_scalar(
         if length and is_integer_dtype(dtype) and isna(value):
             # coerce if we have nan for an integer dtype
             dtype = np.dtype("float64")
-        elif isinstance(dtype, np.dtype) and dtype.kind in ("U", "S"):
+        elif isinstance(dtype, np.dtype) and dtype.kind in "US":
             # we need to coerce to object dtype to avoid
             # to allow numpy to take our string as a scalar value
             dtype = np.dtype("object")
             if not isna(value):
                 value = ensure_str(value)
-        elif dtype.kind in ["M", "m"]:
+        elif dtype.kind in "mM":
             value = _maybe_box_and_unbox_datetimelike(value, dtype)
 
         subarr = np.empty(length, dtype=dtype)
@@ -1504,7 +1503,7 @@ def construct_1d_arraylike_from_scalar(
 
 
 def _maybe_box_and_unbox_datetimelike(value: Scalar, dtype: DtypeObj):
-    # Caller is responsible for checking dtype.kind in ["m", "M"]
+    # Caller is responsible for checking dtype.kind in "mM"
 
     if isinstance(value, dt.datetime):
         # we dont want to box dt64, in particular datetime64("NaT")
@@ -1642,7 +1641,7 @@ def maybe_cast_to_integer_array(arr: list | np.ndarray, dtype: np.dtype) -> np.n
             f"To cast anyway, use pd.Series(values).astype({dtype})"
         )
 
-    if arr.dtype.kind in ["m", "M"]:
+    if arr.dtype.kind in "mM":
         # test_constructor_maskedarray_nonfloat
         raise TypeError(
             f"Constructing a Series or DataFrame from {arr.dtype} values and "
@@ -1667,7 +1666,7 @@ def can_hold_element(arr: ArrayLike, element: Any) -> bool:
     bool
     """
     dtype = arr.dtype
-    if not isinstance(dtype, np.dtype) or dtype.kind in ["m", "M"]:
+    if not isinstance(dtype, np.dtype) or dtype.kind in "mM":
         if isinstance(dtype, (PeriodDtype, IntervalDtype, DatetimeTZDtype, np.dtype)):
             # np.dtype here catches datetime64ns and timedelta64ns; we assume
             #  in this case that we have DatetimeArray/TimedeltaArray
@@ -1715,7 +1714,7 @@ def np_can_hold_element(dtype: np.dtype, element: Any) -> Any:
 
     tipo = _maybe_infer_dtype_type(element)
 
-    if dtype.kind in ["i", "u"]:
+    if dtype.kind in "iu":
         if isinstance(element, range):
             if _dtype_can_hold_range(element, dtype):
                 return element
@@ -1731,7 +1730,7 @@ def np_can_hold_element(dtype: np.dtype, element: Any) -> Any:
             raise LossySetitemError
 
         if tipo is not None:
-            if tipo.kind not in ["i", "u"]:
+            if tipo.kind not in "iu":
                 if isinstance(element, np.ndarray) and element.dtype.kind == "f":
                     # If all can be losslessly cast to integers, then we can hold them
                     with np.errstate(invalid="ignore"):
@@ -1783,7 +1782,7 @@ def np_can_hold_element(dtype: np.dtype, element: Any) -> Any:
 
         if tipo is not None:
             # TODO: itemsize check?
-            if tipo.kind not in ["f", "i", "u"]:
+            if tipo.kind not in "iuf":
                 # Anything other than float/integer we cannot hold
                 raise LossySetitemError
             if not isinstance(tipo, np.dtype):
@@ -1819,7 +1818,7 @@ def np_can_hold_element(dtype: np.dtype, element: Any) -> Any:
             raise LossySetitemError
 
         if tipo is not None:
-            if tipo.kind in ["c", "f", "i", "u"]:
+            if tipo.kind in "iufc":
                 return element
             raise LossySetitemError
         raise LossySetitemError

pandas/core/dtypes/common.py

@@ -1376,7 +1376,7 @@ def is_ea_or_datetimelike_dtype(dtype: DtypeObj | None) -> bool:
     Checks only for dtype objects, not dtype-castable strings or types.
     """
     return isinstance(dtype, ExtensionDtype) or (
-        isinstance(dtype, np.dtype) and dtype.kind in ["m", "M"]
+        isinstance(dtype, np.dtype) and dtype.kind in "mM"
     )
 
 

pandas/core/dtypes/concat.py

@@ -155,7 +155,7 @@ def concat_compat(
     # "Sequence[Union[ExtensionArray, ndarray[Any, Any]]]"; expected
     # "Union[_SupportsArray[dtype[Any]], _NestedSequence[_SupportsArray[dtype[Any]]]]"
     result: np.ndarray = np.concatenate(to_concat, axis=axis)  # type: ignore[arg-type]
-    if "b" in kinds and result.dtype.kind in ["i", "u", "f"]:
+    if "b" in kinds and result.dtype.kind in "iuf":
         # GH#39817 cast to object instead of casting bools to numeric
         result = result.astype(object, copy=False)
     return result

pandas/core/dtypes/dtypes.py

@@ -1138,7 +1138,7 @@ def _can_hold_na(self) -> bool:
             raise NotImplementedError(
                 "_can_hold_na is not defined for partially-initialized IntervalDtype"
             )
-        if subtype.kind in ["i", "u"]:
+        if subtype.kind in "iu":
             return False
         return True
 
@@ -1437,7 +1437,7 @@ def from_numpy_dtype(cls, dtype: np.dtype) -> BaseMaskedDtype:
             from pandas.core.arrays.boolean import BooleanDtype
 
             return BooleanDtype()
-        elif dtype.kind in ["i", "u"]:
+        elif dtype.kind in "iu":
             from pandas.core.arrays.integer import INT_STR_TO_DTYPE
 
             return INT_STR_TO_DTYPE[dtype.name]