BUG: DataFrame(ndarray[dt64], dtype=object) (#41599)

Author: jbrockmendel

doc/source/whatsnew/v1.3.0.rst

@@ -819,6 +819,8 @@ Conversion
 - Bug in :func:`factorize` where, when given an array with a numeric numpy dtype lower than int64, uint64 and float64, the unique values did not keep their original dtype (:issue:`41132`)
 - Bug in :class:`DataFrame` construction with a dictionary containing an arraylike with ``ExtensionDtype`` and ``copy=True`` failing to make a copy (:issue:`38939`)
 - Bug in :meth:`qcut` raising error when taking ``Float64DType`` as input (:issue:`40730`)
+- Bug in :class:`DataFrame` and :class:`Series` construction with ``datetime64[ns]`` data and ``dtype=object`` resulting in ``datetime`` objects instead of :class:`Timestamp` objects (:issue:`41599`)
+- Bug in :class:`DataFrame` and :class:`Series` construction with ``timedelta64[ns]`` data and ``dtype=object`` resulting in ``np.timedelta64`` objects instead of :class:`Timedelta` objects (:issue:`41599`)
 
 Strings
 ^^^^^^^

pandas/core/dtypes/cast.py

@@ -31,7 +31,6 @@
     Timedelta,
     Timestamp,
     conversion,
-    ints_to_pydatetime,
 )
 from pandas._libs.tslibs.timedeltas import array_to_timedelta64
 from pandas._typing import (
@@ -1679,18 +1678,13 @@ def maybe_cast_to_datetime(
                             raise
                         pass
 
-        # coerce datetimelike to object
-        elif is_datetime64_dtype(vdtype) and not is_datetime64_dtype(dtype):
-            if is_object_dtype(dtype):
-                value = cast(np.ndarray, value)
-
-                if value.dtype != DT64NS_DTYPE:
-                    value = value.astype(DT64NS_DTYPE)
-                ints = np.asarray(value).view("i8")
-                return ints_to_pydatetime(ints)
-
-            # we have a non-castable dtype that was passed
-            raise TypeError(f"Cannot cast datetime64 to {dtype}")
+        elif getattr(vdtype, "kind", None) in ["m", "M"]:
+            # we are already datetimelike and want to coerce to non-datetimelike;
+            #  astype_nansafe will raise for anything other than object, then upcast.
+            #  see test_datetimelike_values_with_object_dtype
+            # error: Argument 2 to "astype_nansafe" has incompatible type
+            # "Union[dtype[Any], ExtensionDtype]"; expected "dtype[Any]"
+            return astype_nansafe(value, dtype)  # type: ignore[arg-type]
 
     elif isinstance(value, np.ndarray):
         if value.dtype.kind in ["M", "m"]:

pandas/core/generic.py

@@ -267,7 +267,7 @@ def _init_mgr(
             if (
                 isinstance(mgr, BlockManager)
                 and len(mgr.blocks) == 1
-                and mgr.blocks[0].values.dtype == dtype
+                and is_dtype_equal(mgr.blocks[0].values.dtype, dtype)
             ):
                 pass
             else:

pandas/tests/frame/test_constructors.py

@@ -24,6 +24,7 @@
 from pandas.core.dtypes.dtypes import (
     DatetimeTZDtype,
     IntervalDtype,
+    PandasDtype,
     PeriodDtype,
 )
 
@@ -111,6 +112,40 @@ def test_array_of_dt64_nat_with_td64dtype_raises(self, frame_or_series):
         with pytest.raises(ValueError, match=msg):
             frame_or_series(arr, dtype="m8[ns]")
 
+    @pytest.mark.parametrize("kind", ["m", "M"])
+    def test_datetimelike_values_with_object_dtype(self, kind, frame_or_series):
+        # with dtype=object, we should cast dt64 values to Timestamps, not pydatetimes
+        if kind == "M":
+            dtype = "M8[ns]"
+            scalar_type = Timestamp
+        else:
+            dtype = "m8[ns]"
+            scalar_type = Timedelta
+
+        arr = np.arange(6, dtype="i8").view(dtype).reshape(3, 2)
+        if frame_or_series is Series:
+            arr = arr[:, 0]
+
+        obj = frame_or_series(arr, dtype=object)
+        assert obj._mgr.arrays[0].dtype == object
+        assert isinstance(obj._mgr.arrays[0].ravel()[0], scalar_type)
+
+        # go through a different path in internals.construction
+        obj = frame_or_series(frame_or_series(arr), dtype=object)
+        assert obj._mgr.arrays[0].dtype == object
+        assert isinstance(obj._mgr.arrays[0].ravel()[0], scalar_type)
+
+        obj = frame_or_series(frame_or_series(arr), dtype=PandasDtype(object))
+        assert obj._mgr.arrays[0].dtype == object
+        assert isinstance(obj._mgr.arrays[0].ravel()[0], scalar_type)
+
+        if frame_or_series is DataFrame:
+            # other paths through internals.construction
+            sers = [Series(x) for x in arr]
+            obj = frame_or_series(sers, dtype=object)
+            assert obj._mgr.arrays[0].dtype == object
+            assert isinstance(obj._mgr.arrays[0].ravel()[0], scalar_type)
+
     def test_series_with_name_not_matching_column(self):
         # GH#9232
         x = Series(range(5), name=1)