(fix): don't handle time-dtypes as extension arrays in from_dataframe (#9042)

* (fix): don't handle time-dtypes as extension arrays

* (fix): check series type

* Add whats-new

---------

Co-authored-by: Deepak Cherian <[email protected]>
Co-authored-by: Deepak Cherian <[email protected]>

Author: 2 people

doc/whats-new.rst

@@ -40,6 +40,9 @@ Deprecations
 
 Bug fixes
 ~~~~~~~~~
+- Preserve conversion of timezone-aware pandas Datetime arrays to numpy object arrays
+  (:issue:`9026`, :pull:`9042`).
+  By `Ilan Gold <https://github.com/ilan-gold>`_.
 
 - :py:meth:`DataArrayResample.interpolate` and :py:meth:`DatasetResample.interpolate` method now
   support aribtrary kwargs such as ``order`` for polynomial interpolation. (:issue:`8762`).

properties/test_pandas_roundtrip.py

@@ -30,6 +30,16 @@
 )
 
 
+datetime_with_tz_strategy = st.datetimes(timezones=st.timezones())
+dataframe_strategy = pdst.data_frames(
+    [
+        pdst.column("datetime_col", elements=datetime_with_tz_strategy),
+        pdst.column("other_col", elements=st.integers()),
+    ],
+    index=pdst.range_indexes(min_size=1, max_size=10),
+)
+
+
 @st.composite
 def datasets_1d_vars(draw) -> xr.Dataset:
     """Generate datasets with only 1D variables
@@ -98,3 +108,15 @@ def test_roundtrip_pandas_dataframe(df) -> None:
     roundtripped = arr.to_pandas()
     pd.testing.assert_frame_equal(df, roundtripped)
     xr.testing.assert_identical(arr, roundtripped.to_xarray())
+
+
+@given(df=dataframe_strategy)
+def test_roundtrip_pandas_dataframe_datetime(df) -> None:
+    # Need to name the indexes, otherwise Xarray names them 'dim_0', 'dim_1'.
+    df.index.name = "rows"
+    df.columns.name = "cols"
+    dataset = xr.Dataset.from_dataframe(df)
+    roundtripped = dataset.to_dataframe()
+    roundtripped.columns.name = "cols"  # why?
+    pd.testing.assert_frame_equal(df, roundtripped)
+    xr.testing.assert_identical(dataset, roundtripped.to_xarray())

xarray/core/dataset.py

@@ -7420,7 +7420,9 @@ def from_dataframe(cls, dataframe: pd.DataFrame, sparse: bool = False) -> Self:
         arrays = []
         extension_arrays = []
         for k, v in dataframe.items():
-            if not is_extension_array_dtype(v):
+            if not is_extension_array_dtype(v) or isinstance(
+                v.array, (pd.arrays.DatetimeArray, pd.arrays.TimedeltaArray)
+            ):
                 arrays.append((k, np.asarray(v)))
             else:
                 extension_arrays.append((k, v))