(fix): clean up datetime extension handling

xarray/core/dataset.py

@@ -99,7 +99,6 @@
     parse_dims_as_set,
 )
 from xarray.core.variable import (
-    UNSUPPORTED_EXTENSION_ARRAY_TYPES,
     IndexVariable,
     Variable,
     as_variable,
@@ -7268,7 +7267,7 @@ def from_dataframe(cls, dataframe: pd.DataFrame, sparse: bool = False) -> Self:
         extension_arrays = []
         for k, v in dataframe.items():
             if not is_extension_array_dtype(v) or isinstance(
-                v.array, UNSUPPORTED_EXTENSION_ARRAY_TYPES
+                v.array, pd.arrays.NumpyExtensionArray
             ):
                 arrays.append((k, np.asarray(v)))
             else:

xarray/core/extension_array.py

@@ -93,9 +93,6 @@ class PandasExtensionArray(Generic[T_ExtensionArray], NDArrayMixin):
     def __post_init__(self):
         if not isinstance(self.array, pd.api.extensions.ExtensionArray):
             raise TypeError(f"{self.array} is not an pandas ExtensionArray.")
-        # This does not use the UNSUPPORTED_EXTENSION_ARRAY_TYPES whitelist because
-        # we do support extension arrays from datetime, for example, that need
-        # duck array support internally via this class.
         if isinstance(self.array, pd.arrays.NumpyExtensionArray):
             raise TypeError(
                 "`NumpyExtensionArray` should be converted to a numpy array in `xarray` internally."

xarray/core/variable.py

@@ -63,11 +63,6 @@
 )
 # https://github.com/python/mypy/issues/224
 BASIC_INDEXING_TYPES = integer_types + (slice,)
-UNSUPPORTED_EXTENSION_ARRAY_TYPES = (
-    pd.arrays.DatetimeArray,
-    pd.arrays.TimedeltaArray,
-    pd.arrays.NumpyExtensionArray,  # type: ignore[attr-defined]
-)
 
 if TYPE_CHECKING:
     from xarray.core.types import (
@@ -195,7 +190,7 @@ def _maybe_wrap_data(data):
     """
     if isinstance(data, pd.Index):
         return PandasIndexingAdapter(data)
-    if isinstance(data, UNSUPPORTED_EXTENSION_ARRAY_TYPES):
+    if isinstance(data, pd.arrays.NumpyExtensionArray):
         return data.to_numpy()
     if isinstance(data, pd.api.extensions.ExtensionArray):
         return PandasExtensionArray(data)
@@ -261,7 +256,7 @@ def convert_non_numpy_type(data):
         if (
             isinstance(data, pd.Series)
             and pd.api.types.is_extension_array_dtype(data)
-            and not isinstance(data.array, UNSUPPORTED_EXTENSION_ARRAY_TYPES)
+            and not isinstance(data.array, pd.arrays.NumpyExtensionArray)
         ):
             pandas_data = data.array
         else:

xarray/tests/test_variable.py

@@ -15,6 +15,7 @@
 from xarray import DataArray, Dataset, IndexVariable, Variable, set_options
 from xarray.core import dtypes, duck_array_ops, indexing
 from xarray.core.common import full_like, ones_like, zeros_like
+from xarray.core.extension_array import PandasExtensionArray
 from xarray.core.indexing import (
     BasicIndexer,
     CopyOnWriteArray,
@@ -2757,15 +2758,15 @@ def test_tz_datetime(self) -> None:
             warnings.simplefilter("ignore")
             actual: T_DuckArray = as_compatible_data(times_s)
         assert actual.array == times_s
-        assert actual.array.dtype == pd.DatetimeTZDtype("s", tz)  # type: ignore[arg-type]
+        assert actual.array.dtype == times_s.dtype  # type: ignore[arg-type]
 
         series = pd.Series(times_s)
         with warnings.catch_warnings():
             warnings.simplefilter("ignore")
             actual2: T_DuckArray = as_compatible_data(series)
 
-        np.testing.assert_array_equal(actual2, np.asarray(series.values))
-        assert actual2.dtype == np.dtype("datetime64[s]")
+        np.testing.assert_array_equal(actual2, np.asarray(series.array))
+        assert actual2.dtype == times_s.dtype
 
     def test_full_like(self) -> None:
         # For more thorough tests, see test_variable.py
@@ -3096,8 +3097,13 @@ def test_datetime_conversion(values, unit) -> None:
     else:
         # The only case where a non-datetime64 dtype can occur currently is in
         # the case that the variable is backed by a timezone-aware
-        # DatetimeIndex, and thus is hidden within the PandasIndexingAdapter class.
-        assert isinstance(var._data, PandasIndexingAdapter)
+        # DatetimeIndex/DateTimeArray, and thus is hidden within the PandasIndexingAdapter/PandasExtensionArray class.
+        assert isinstance(
+            var._data,
+            PandasIndexingAdapter
+            if isinstance(values, pd.DatetimeIndex)
+            else PandasExtensionArray,
+        )
         assert var._data.array.dtype == pd.DatetimeTZDtype(
             "ns", pytz.timezone("America/New_York")
         )
@@ -3132,19 +3138,9 @@ def test_pandas_two_only_datetime_conversion_warnings(
 ) -> None:
     # todo: check for redundancy (suggested per review)
     var = Variable(["time"], data.astype(dtype))  # type: ignore[arg-type]
-
-    # we internally convert series to numpy representations to avoid too much nastiness with extension arrays
-    # when calling data.array e.g., with NumpyExtensionArrays
-    if isinstance(data, pd.Series):
-        assert var.dtype == np.dtype("datetime64[s]")
-    elif var.dtype.kind == "M":
-        assert var.dtype == dtype
-    else:
-        # The only case where a non-datetime64 dtype can occur currently is in
-        # the case that the variable is backed by a timezone-aware
-        # DatetimeIndex, and thus is hidden within the PandasIndexingAdapter class.
+    assert var.dtype == dtype
+    if isinstance(data, pd.DatetimeIndex):
         assert isinstance(var._data, PandasIndexingAdapter)
-        assert var._data.array.dtype == pd.DatetimeTZDtype("s", tz_ny)
 
 
 @pytest.mark.parametrize(