CLN: NDFrame._where (pandas-dev#38742)

Author: jbrockmendel

doc/source/whatsnew/v1.3.0.rst

@@ -295,7 +295,7 @@ Sparse
 
 ExtensionArray
 ^^^^^^^^^^^^^^
-
+- Bug in :meth:`DataFrame.where` when ``other`` is a :class:`Series` with ExtensionArray dtype (:issue:`38729`)
 -
 -
 

pandas/core/generic.py

@@ -89,9 +89,10 @@
 import pandas as pd
 from pandas.core import arraylike, indexing, missing, nanops
 import pandas.core.algorithms as algos
+from pandas.core.arrays import ExtensionArray
 from pandas.core.base import PandasObject, SelectionMixin
 import pandas.core.common as com
-from pandas.core.construction import create_series_with_explicit_dtype
+from pandas.core.construction import create_series_with_explicit_dtype, extract_array
 from pandas.core.flags import Flags
 from pandas.core.indexes import base as ibase
 from pandas.core.indexes.api import (
@@ -8788,6 +8789,9 @@ def _where(
         """
         inplace = validate_bool_kwarg(inplace, "inplace")
 
+        if axis is not None:
+            axis = self._get_axis_number(axis)
+
         # align the cond to same shape as myself
         cond = com.apply_if_callable(cond, self)
         if isinstance(cond, NDFrame):
@@ -8827,22 +8831,39 @@ def _where(
             if other.ndim <= self.ndim:
 
                 _, other = self.align(
-                    other, join="left", axis=axis, level=level, fill_value=np.nan
+                    other,
+                    join="left",
+                    axis=axis,
+                    level=level,
+                    fill_value=np.nan,
+                    copy=False,
                 )
 
                 # if we are NOT aligned, raise as we cannot where index
-                if axis is None and not all(
-                    other._get_axis(i).equals(ax) for i, ax in enumerate(self.axes)
-                ):
+                if axis is None and not other._indexed_same(self):
                     raise InvalidIndexError
 
+                elif other.ndim < self.ndim:
+                    # TODO(EA2D): avoid object-dtype cast in EA case GH#38729
+                    other = other._values
+                    if axis == 0:
+                        other = np.reshape(other, (-1, 1))
+                    elif axis == 1:
+                        other = np.reshape(other, (1, -1))
+
+                    other = np.broadcast_to(other, self.shape)
+
             # slice me out of the other
             else:
                 raise NotImplementedError(
                     "cannot align with a higher dimensional NDFrame"
                 )
 
-        if isinstance(other, np.ndarray):
+        if not isinstance(other, (MultiIndex, NDFrame)):
+            # mainly just catching Index here
+            other = extract_array(other, extract_numpy=True)
+
+        if isinstance(other, (np.ndarray, ExtensionArray)):
 
             if other.shape != self.shape:
 
@@ -8887,10 +8908,10 @@ def _where(
         else:
             align = self._get_axis_number(axis) == 1
 
-        if align and isinstance(other, NDFrame):
-            other = other.reindex(self._info_axis, axis=self._info_axis_number)
         if isinstance(cond, NDFrame):
-            cond = cond.reindex(self._info_axis, axis=self._info_axis_number)
+            cond = cond.reindex(
+                self._info_axis, axis=self._info_axis_number, copy=False
+            )
 
         block_axis = self._get_block_manager_axis(axis)
 

pandas/core/internals/blocks.py

@@ -1064,9 +1064,7 @@ def putmask(self, mask, new, axis: int = 0) -> List["Block"]:
             # If the default repeat behavior in np.putmask would go in the
             # wrong direction, then explicitly repeat and reshape new instead
             if getattr(new, "ndim", 0) >= 1:
-                if self.ndim - 1 == new.ndim and axis == 1:
-                    new = np.repeat(new, new_values.shape[-1]).reshape(self.shape)
-                new = new.astype(new_values.dtype)
+                new = new.astype(new_values.dtype, copy=False)
 
             # we require exact matches between the len of the
             # values we are setting (or is compat). np.putmask
@@ -1104,13 +1102,6 @@ def putmask(self, mask, new, axis: int = 0) -> List["Block"]:
                     new = new.T
                 axis = new_values.ndim - axis - 1
 
-            # Pseudo-broadcast
-            if getattr(new, "ndim", 0) >= 1:
-                if self.ndim - 1 == new.ndim:
-                    new_shape = list(new.shape)
-                    new_shape.insert(axis, 1)
-                    new = new.reshape(tuple(new_shape))
-
             # operate column-by-column
             def f(mask, val, idx):
 

pandas/tests/frame/indexing/test_where.py

@@ -653,3 +653,22 @@ def test_where_categorical_filtering(self):
         expected.loc[0, :] = np.nan
 
         tm.assert_equal(result, expected)
+
+    def test_where_ea_other(self):
+        # GH#38729/GH#38742
+        df = DataFrame({"A": [1, 2, 3], "B": [4, 5, 6]})
+        arr = pd.array([7, pd.NA, 9])
+        ser = Series(arr)
+        mask = np.ones(df.shape, dtype=bool)
+        mask[1, :] = False
+
+        # TODO: ideally we would get Int64 instead of object
+        result = df.where(mask, ser, axis=0)
+        expected = DataFrame({"A": [1, pd.NA, 3], "B": [4, pd.NA, 6]}).astype(object)
+        tm.assert_frame_equal(result, expected)
+
+        ser2 = Series(arr[:2], index=["A", "B"])
+        expected = DataFrame({"A": [1, 7, 3], "B": [4, pd.NA, 6]})
+        expected["B"] = expected["B"].astype(object)
+        result = df.where(mask, ser2, axis=1)
+        tm.assert_frame_equal(result, expected)