keep everything in method for minimal change

Author: jorisvandenbossche

pandas/core/arrays/arrow/array.py

@@ -442,7 +442,7 @@ def _cast_pointwise_result(self, values) -> ArrayLike:
             # e.g. test_by_column_values_with_same_starting_value with nested
             #  values, one entry of which is an ArrowStringArray
             #  or test_agg_lambda_complex128_dtype_conversion for complex values
-            return values
+            return super()._cast_pointwise_result(values)
 
         if pa.types.is_null(arr.type):
             if lib.infer_dtype(values) == "decimal":
@@ -498,7 +498,7 @@ def _cast_pointwise_result(self, values) -> ArrayLike:
             if self.dtype.na_value is np.nan:
                 # ArrowEA has different semantics, so we return numpy-based
                 #  result instead
-                return values
+                return super()._cast_pointwise_result(values)
             return ArrowExtensionArray(arr)
         return self._from_pyarrow_array(arr)
 

pandas/core/arrays/base.py

@@ -356,6 +356,36 @@ def _from_sequence_of_strings(
         """
         raise AbstractMethodError(cls)
 
+    @classmethod
+    def _from_factorized(cls, values, original):
+        """
+        Reconstruct an ExtensionArray after factorization.
+
+        Parameters
+        ----------
+        values : ndarray
+            An integer ndarray with the factorized values.
+        original : ExtensionArray
+            The original ExtensionArray that factorize was called on.
+
+        See Also
+        --------
+        factorize : Top-level factorize method that dispatches here.
+        ExtensionArray.factorize : Encode the extension array as an enumerated type.
+
+        Examples
+        --------
+        >>> interv_arr = pd.arrays.IntervalArray(
+        ...     [pd.Interval(0, 1), pd.Interval(1, 5), pd.Interval(1, 5)]
+        ... )
+        >>> codes, uniques = pd.factorize(interv_arr)
+        >>> pd.arrays.IntervalArray._from_factorized(uniques, interv_arr)
+        <IntervalArray>
+        [(0, 1], (1, 5]]
+        Length: 2, dtype: interval[int64, right]
+        """
+        raise AbstractMethodError(cls)
+
     @classmethod
     def _from_scalars(cls, scalars, *, dtype: DtypeObj) -> Self:
         """
@@ -374,7 +404,7 @@ def _from_scalars(cls, scalars, *, dtype: DtypeObj) -> Self:
         Notes
         -----
         This is called in a try/except block when casting the result of a
-        pointwise operation.
+        pointwise operation in ExtensionArray._cast_pointwise_result.
         """
         try:
             return cls._from_sequence(scalars, dtype=dtype, copy=False)
@@ -388,37 +418,7 @@ def _from_scalars(cls, scalars, *, dtype: DtypeObj) -> Self:
             )
             raise
 
-    @classmethod
-    def _from_factorized(cls, values, original):
-        """
-        Reconstruct an ExtensionArray after factorization.
-
-        Parameters
-        ----------
-        values : ndarray
-            An integer ndarray with the factorized values.
-        original : ExtensionArray
-            The original ExtensionArray that factorize was called on.
-
-        See Also
-        --------
-        factorize : Top-level factorize method that dispatches here.
-        ExtensionArray.factorize : Encode the extension array as an enumerated type.
-
-        Examples
-        --------
-        >>> interv_arr = pd.arrays.IntervalArray(
-        ...     [pd.Interval(0, 1), pd.Interval(1, 5), pd.Interval(1, 5)]
-        ... )
-        >>> codes, uniques = pd.factorize(interv_arr)
-        >>> pd.arrays.IntervalArray._from_factorized(uniques, interv_arr)
-        <IntervalArray>
-        [(0, 1], (1, 5]]
-        Length: 2, dtype: interval[int64, right]
-        """
-        raise AbstractMethodError(cls)
-
-    def _cast_pointwise_result(self, values: ArrayLike) -> ArrayLike:
+    def _cast_pointwise_result(self, values) -> ArrayLike:
         """
         Construct an ExtensionArray after a pointwise operation.
 
@@ -437,7 +437,8 @@ def _cast_pointwise_result(self, values: ArrayLike) -> ArrayLike:
         try:
             return type(self)._from_scalars(values, dtype=self.dtype)
         except (ValueError, TypeError):
-            return values
+            values = np.asarray(values, dtype=object)
+            return lib.maybe_convert_objects(values, convert_non_numeric=True)
 
     # ------------------------------------------------------------------------
     # Must be a Sequence

pandas/core/arrays/sparse/array.py

@@ -622,6 +622,7 @@ def _from_factorized(cls, values, original) -> Self:
         return cls(values, dtype=original.dtype)
 
     def _cast_pointwise_result(self, values):
+        values = np.asarray(values, dtype=object)
         result = lib.maybe_convert_objects(values, convert_non_numeric=True)
         if result.dtype.kind == self.dtype.kind:
             try:

pandas/core/dtypes/cast.py

@@ -414,42 +414,6 @@ def maybe_upcast_numeric_to_64bit(arr: NumpyIndexT) -> NumpyIndexT:
         return arr
 
 
-def cast_pointwise_result(
-    result: ArrayLike,
-    original_array: ArrayLike,
-) -> ArrayLike:
-    """
-    Try casting result of a pointwise operation back to the original dtype if
-    appropriate.
-
-    Parameters
-    ----------
-    result : array-like
-        Result to cast.
-    original_array : array-like
-        Input array from which result was calculated.
-
-    Returns
-    -------
-    array-like
-    """
-    if isinstance(original_array.dtype, ExtensionDtype):
-        try:
-            result = original_array._cast_pointwise_result(result)
-        except (TypeError, ValueError):
-            pass
-
-        if isinstance(result.dtype, ExtensionDtype):
-            return result
-
-    if not isinstance(result, np.ndarray):
-        result = np.asarray(result, dtype=object)
-
-    if result.dtype != object:
-        return result
-    return lib.maybe_convert_objects(result, convert_non_numeric=True)
-
-
 @overload
 def ensure_dtype_can_hold_na(dtype: np.dtype) -> np.dtype: ...
 

pandas/core/frame.py

@@ -84,7 +84,6 @@
 from pandas.core.dtypes.cast import (
     LossySetitemError,
     can_hold_element,
-    cast_pointwise_result,
     construct_1d_arraylike_from_scalar,
     construct_2d_arraylike_from_scalar,
     find_common_type,
@@ -11201,7 +11200,7 @@ def _append_internal(
         if isinstance(self.index.dtype, ExtensionDtype):
             # GH#41626 retain e.g. CategoricalDtype if reached via
             #  df.loc[key] = item
-            row_df.index = cast_pointwise_result(row_df.index._values, self.index.array)
+            row_df.index = self.index.array._cast_pointwise_result(row_df.index._values)
 
         # infer_objects is needed for
         #  test_append_empty_frame_to_series_with_dateutil_tz

pandas/core/groupby/ops.py

@@ -35,7 +35,6 @@
 from pandas.util._decorators import cache_readonly
 
 from pandas.core.dtypes.cast import (
-    cast_pointwise_result,
     maybe_downcast_to_dtype,
 )
 from pandas.core.dtypes.common import (
@@ -964,7 +963,7 @@ def agg_series(
         np.ndarray or ExtensionArray
         """
         result = self._aggregate_series_pure_python(obj, func)
-        return cast_pointwise_result(result, obj.array)
+        return obj.array._cast_pointwise_result(result)
 
     @final
     def _aggregate_series_pure_python(

pandas/core/indexes/base.py

@@ -89,7 +89,6 @@
 from pandas.core.dtypes.cast import (
     LossySetitemError,
     can_hold_element,
-    cast_pointwise_result,
     common_dtype_categorical_compat,
     find_result_type,
     infer_dtype_from,
@@ -6544,7 +6543,7 @@ def map(self, mapper, na_action: Literal["ignore"] | None = None):
             # e.g. if we are floating and new_values is all ints, then we
             #  don't want to cast back to floating.  But if we are UInt64
             #  and new_values is all ints, we want to try.
-            new_values = cast_pointwise_result(new_values, arr)
+            new_values = arr._cast_pointwise_result(new_values)
             dtype = new_values.dtype
         return Index(new_values, dtype=dtype, copy=False, name=self.name)
 

pandas/core/series.py

@@ -69,7 +69,6 @@
 from pandas.core.dtypes.astype import astype_is_view
 from pandas.core.dtypes.cast import (
     LossySetitemError,
-    cast_pointwise_result,
     construct_1d_arraylike_from_scalar,
     find_common_type,
     infer_dtype_from,
@@ -3257,7 +3256,7 @@ def combine(
                 new_values[:] = [func(lv, other) for lv in self._values]
             new_name = self.name
 
-        res_values = cast_pointwise_result(new_values, self.array)
+        res_values = self.array._cast_pointwise_result(new_values)
         return self._constructor(
             res_values,
             dtype=res_values.dtype,

pandas/tests/extension/decimal/array.py

@@ -111,14 +111,15 @@ def _from_sequence_of_strings(cls, strings, *, dtype: ExtensionDtype, copy=False
     def _from_factorized(cls, values, original):
         return cls(values)
 
-    def _cast_pointwise_result(self, values):
-        try:
-            # If this were ever made a non-test EA, special-casing could
-            #  be avoided by handling Decimal in maybe_convert_objects
-            res = type(self)._from_sequence(values, dtype=self.dtype)
-        except (ValueError, TypeError):
-            return values
-        return res
+    # test to ensure that the base class _cast_pointwise_result works as expected
+    # def _cast_pointwise_result(self, values):
+    #     try:
+    #         # If this were ever made a non-test EA, special-casing could
+    #         #  be avoided by handling Decimal in maybe_convert_objects
+    #         res = type(self)._from_sequence(values, dtype=self.dtype)
+    #     except (ValueError, TypeError):
+    #         return values
+    #     return res
 
     _HANDLED_TYPES = (decimal.Decimal, numbers.Number, np.ndarray)