REGR: Performance of DataFrame reduction ops with axis=1

pandas/core/frame.py

@@ -106,7 +106,10 @@
     needs_i8_conversion,
     pandas_dtype,
 )
-from pandas.core.dtypes.dtypes import ExtensionDtype
+from pandas.core.dtypes.dtypes import (
+    BaseMaskedDtype,
+    ExtensionDtype,
+)
 from pandas.core.dtypes.missing import (
     isna,
     notna,
@@ -3602,9 +3605,14 @@ def transpose(self, *args, copy: bool = False) -> DataFrame:
             # We have EAs with the same dtype. We can preserve that dtype in transpose.
             dtype = dtypes[0]
             arr_type = dtype.construct_array_type()
-            values = self.values
-
-            new_values = [arr_type._from_sequence(row, dtype=dtype) for row in values]
+            if isinstance(dtype, BaseMaskedDtype):
+                data, mask = self._mgr.as_array_masked()
+                new_values = [arr_type(data[i], mask[i]) for i in range(self.shape[0])]
+            else:
+                values = self.values
+                new_values = [
+                    arr_type._from_sequence(row, dtype=dtype) for row in values
+                ]
             result = type(self)._from_arrays(
                 new_values, index=self.columns, columns=self.index
             )

pandas/core/internals/managers.py

@@ -1749,6 +1749,40 @@ def as_array(
 
         return arr.transpose()
 
+    def as_array_masked(self) -> tuple[np.ndarray, np.ndarray]:
+        """
+        Convert the blockmanager data into an numpy array.
+
+        Returns
+        -------
+        arr : ndarray
+        """
+        # # TODO(CoW) handle case where resulting array is a view
+        # if len(self.blocks) == 0:
+        #     arr = np.empty(self.shape, dtype=float)
+        #     return arr.transpose()
+
+        # TODO we already established we only have a single dtype, but this
+        # could be generalized to be a mix of all masked dtypes
+        dtype = self.blocks[0].dtype.numpy_dtype
+
+        result_data = np.empty(self.shape, dtype=dtype)
+        result_mask = np.empty(self.shape, dtype="bool")
+
+        # itemmask = np.zeros(self.shape[0])
+
+        for blk in self.blocks:
+            rl = blk.mgr_locs
+            arr = blk.values
+            result_data[rl.indexer] = arr._data
+            result_mask[rl.indexer] = arr._mask
+            # itemmask[rl.indexer] = 1
+
+        # if not itemmask.all():
+        #     raise AssertionError("Some items were not contained in blocks")
+
+        return result_data.transpose(), result_mask.transpose()
+
     def _interleave(
         self,
         dtype: np.dtype | None = None,