Backport PR #51966 on branch 2.0.x (CoW: __array__ not recognizing ea dtypes)

pandas/core/generic.py

@@ -101,6 +101,7 @@
     validate_inclusive,
 )
 
+from pandas.core.dtypes.astype import astype_is_view
 from pandas.core.dtypes.common import (
     ensure_object,
     ensure_platform_int,
@@ -1995,10 +1996,17 @@ def empty(self) -> bool_t:
     def __array__(self, dtype: npt.DTypeLike | None = None) -> np.ndarray:
         values = self._values
         arr = np.asarray(values, dtype=dtype)
-        if arr is values and using_copy_on_write():
-            # TODO(CoW) also properly handle extension dtypes
-            arr = arr.view()
-            arr.flags.writeable = False
+        if (
+            astype_is_view(values.dtype, arr.dtype)
+            and using_copy_on_write()
+            and self._mgr.is_single_block
+        ):
+            # Check if both conversions can be done without a copy
+            if astype_is_view(self.dtypes.iloc[0], values.dtype) and astype_is_view(
+                values.dtype, arr.dtype
+            ):
+                arr = arr.view()
+                arr.flags.writeable = False
         return arr
 
     @final

pandas/core/series.py

@@ -920,8 +920,7 @@ def __array__(self, dtype: npt.DTypeLike | None = None) -> np.ndarray:
         """
         values = self._values
         arr = np.asarray(values, dtype=dtype)
-        if arr is values and using_copy_on_write():
-            # TODO(CoW) also properly handle extension dtypes
+        if using_copy_on_write() and astype_is_view(values.dtype, arr.dtype):
             arr = arr.view()
             arr.flags.writeable = False
         return arr

pandas/tests/copy_view/test_array.py

@@ -4,6 +4,7 @@
 from pandas import (
     DataFrame,
     Series,
+    date_range,
 )
 import pandas._testing as tm
 from pandas.tests.copy_view.util import get_array
@@ -119,3 +120,66 @@ def test_ravel_read_only(using_copy_on_write, order):
     if using_copy_on_write:
         assert arr.flags.writeable is False
     assert np.shares_memory(get_array(ser), arr)
+
+
+def test_series_array_ea_dtypes(using_copy_on_write):
+    ser = Series([1, 2, 3], dtype="Int64")
+    arr = np.asarray(ser, dtype="int64")
+    assert np.shares_memory(arr, get_array(ser))
+    if using_copy_on_write:
+        assert arr.flags.writeable is False
+    else:
+        assert arr.flags.writeable is True
+
+    arr = np.asarray(ser)
+    assert not np.shares_memory(arr, get_array(ser))
+    assert arr.flags.writeable is True
+
+
+def test_dataframe_array_ea_dtypes(using_copy_on_write):
+    df = DataFrame({"a": [1, 2, 3]}, dtype="Int64")
+    arr = np.asarray(df, dtype="int64")
+    # TODO: This should be able to share memory, but we are roundtripping
+    # through object
+    assert not np.shares_memory(arr, get_array(df, "a"))
+    assert arr.flags.writeable is True
+
+    arr = np.asarray(df)
+    if using_copy_on_write:
+        # TODO(CoW): This should be True
+        assert arr.flags.writeable is False
+    else:
+        assert arr.flags.writeable is True
+
+
+def test_dataframe_array_string_dtype(using_copy_on_write, using_array_manager):
+    df = DataFrame({"a": ["a", "b"]}, dtype="string")
+    arr = np.asarray(df)
+    if not using_array_manager:
+        assert np.shares_memory(arr, get_array(df, "a"))
+    if using_copy_on_write:
+        assert arr.flags.writeable is False
+    else:
+        assert arr.flags.writeable is True
+
+
+def test_dataframe_multiple_numpy_dtypes():
+    df = DataFrame({"a": [1, 2, 3], "b": 1.5})
+    arr = np.asarray(df)
+    assert not np.shares_memory(arr, get_array(df, "a"))
+    assert arr.flags.writeable is True
+
+
+def test_values_is_ea(using_copy_on_write):
+    df = DataFrame({"a": date_range("2012-01-01", periods=3)})
+    arr = np.asarray(df)
+    if using_copy_on_write:
+        assert arr.flags.writeable is False
+    else:
+        assert arr.flags.writeable is True
+
+
+def test_empty_dataframe():
+    df = DataFrame()
+    arr = np.asarray(df)
+    assert arr.flags.writeable is True