TST: Extend datetime64 arith tests to array classes, fix several brok…

…en cases (#23771)

pandas/core/arrays/datetimelike.py

@@ -727,6 +727,10 @@ def __add__(self, other):
             else:  # pragma: no cover
                 return NotImplemented
 
+            if is_timedelta64_dtype(result) and isinstance(result, np.ndarray):
+                from pandas.core.arrays import TimedeltaArrayMixin
+                # TODO: infer freq?
+                return TimedeltaArrayMixin(result)
             return result
 
         cls.__add__ = __add__
@@ -791,6 +795,10 @@ def __sub__(self, other):
             else:  # pragma: no cover
                 return NotImplemented
 
+            if is_timedelta64_dtype(result) and isinstance(result, np.ndarray):
+                from pandas.core.arrays import TimedeltaArrayMixin
+                # TODO: infer freq?
+                return TimedeltaArrayMixin(result)
             return result
 
         cls.__sub__ = __sub__

pandas/core/arrays/datetimes.py

@@ -222,6 +222,11 @@ def __new__(cls, values, freq=None, tz=None, dtype=None):
         # if dtype has an embedded tz, capture it
         tz = dtl.validate_tz_from_dtype(dtype, tz)
 
+        if is_object_dtype(values):
+            # kludge; dispatch until the DatetimeArray constructor is complete
+            from pandas import DatetimeIndex
+            values = DatetimeIndex(values, freq=freq, tz=tz)
+
         if isinstance(values, ABCSeries):
             # extract to ndarray or DatetimeIndex
             values = values._values

pandas/core/arrays/timedeltas.py

@@ -300,6 +300,11 @@ def _evaluate_with_timedelta_like(self, other, op):
 
         return NotImplemented
 
+    def __neg__(self):
+        if self.freq is not None:
+            return type(self)(-self._data, freq=-self.freq)
+        return type(self)(-self._data)
+
     # ----------------------------------------------------------------
     # Conversion Methods - Vectorized analogues of Timedelta methods
 

pandas/core/ops.py

@@ -947,6 +947,10 @@ def should_series_dispatch(left, right, op):
         # numpy integer dtypes as timedelta64 dtypes in this scenario
         return True
 
+    if is_datetime64_dtype(ldtype) and is_object_dtype(rdtype):
+        # in particular case where right is an array of DateOffsets
+        return True
+
     return False
 
 

pandas/tests/arithmetic/conftest.py

@@ -57,14 +57,6 @@ def numeric_idx(request):
     return request.param
 
 
-@pytest.fixture
-def tdser():
-    """
-    Return a Series with dtype='timedelta64[ns]', including a NaT.
-    """
-    return pd.Series(['59 Days', '59 Days', 'NaT'], dtype='timedelta64[ns]')
-
-
 # ------------------------------------------------------------------
 # Scalar Fixtures