API: Series([pydate, pydatetime]) (pandas-dev#49446)

Author: jbrockmendel

doc/source/whatsnew/v2.0.0.rst

@@ -144,6 +144,7 @@ Other API changes
 - The ``other`` argument in :meth:`DataFrame.mask` and :meth:`Series.mask` now defaults to ``no_default`` instead of ``np.nan`` consistent with :meth:`DataFrame.where` and :meth:`Series.where`. Entries will be filled with the corresponding NULL value (``np.nan`` for numpy dtypes, ``pd.NA`` for extension dtypes). (:issue:`49111`)
 - When creating a :class:`Series` with a object-dtype :class:`Index` of datetime objects, pandas no longer silently converts the index to a :class:`DatetimeIndex` (:issue:`39307`, :issue:`23598`)
 - :meth:`Series.unique` with dtype "timedelta64[ns]" or "datetime64[ns]" now returns :class:`TimedeltaArray` or :class:`DatetimeArray` instead of ``numpy.ndarray`` (:issue:`49176`)
+- Passing a sequence containing ``datetime`` objects and ``date`` objects to :class:`Series` constructor will return with ``object`` dtype instead of ``datetime64[ns]`` dtype, consistent with :class:`Index` behavior (:issue:`49341`)
 - Passing strings that cannot be parsed as datetimes to :class:`Series` or :class:`DataFrame` with ``dtype="datetime64[ns]"`` will raise instead of silently ignoring the keyword and returning ``object`` dtype (:issue:`24435`)
 -
 

pandas/_libs/lib.pyx

@@ -1640,8 +1640,11 @@ def infer_datetimelike_array(arr: ndarray[object]) -> tuple[str, bool]:
             return "interval"
         return "mixed"
 
-    if seen_date and not (seen_datetime or seen_timedelta):
-        return "date"
+    if seen_date:
+        if not seen_datetime and not seen_timedelta:
+            return "date"
+        return "mixed"
+
     elif seen_datetime and not seen_timedelta:
         return "datetime"
     elif seen_timedelta and not seen_datetime:
@@ -2570,10 +2573,15 @@ def maybe_convert_objects(ndarray[object] objects,
     if seen.datetimetz_:
         if is_datetime_with_singletz_array(objects):
             from pandas import DatetimeIndex
-            dti = DatetimeIndex(objects)
 
-            # unbox to DatetimeArray
-            return dti._data
+            try:
+                dti = DatetimeIndex(objects)
+            except OutOfBoundsDatetime:
+                # e.g. test_to_datetime_cache_coerce_50_lines_outofbounds
+                pass
+            else:
+                # unbox to DatetimeArray
+                return dti._data
         seen.object_ = True
 
     elif seen.datetime_:

pandas/core/dtypes/cast.py

@@ -1224,71 +1224,37 @@ def maybe_infer_to_datetimelike(
 
     v = np.array(value, copy=False)
 
-    shape = v.shape
     if v.ndim != 1:
         v = v.ravel()
 
     if not len(v):
         return value
 
-    def try_datetime(v: np.ndarray) -> np.ndarray | DatetimeArray:
-        # Coerce to datetime64, datetime64tz, or in corner cases
-        #  object[datetimes]
-        from pandas.core.arrays.datetimes import sequence_to_datetimes
-
-        try:
-            dta = sequence_to_datetimes(v)
-        except (ValueError, OutOfBoundsDatetime):
-            # ValueError for e.g. mixed tzs
-            # GH#19761 we may have mixed timezones, in which cast 'dta' is
-            #  an ndarray[object].  Only 1 test
-            #  relies on this behavior, see GH#40111
-            return v.reshape(shape)
-        else:
-            return dta.reshape(shape)
-
-    def try_timedelta(v: np.ndarray) -> np.ndarray:
-        # safe coerce to timedelta64
-
-        # will try first with a string & object conversion
-        try:
-            # bc we know v.dtype == object, this is equivalent to
-            #  `np.asarray(to_timedelta(v))`, but using a lower-level API that
-            #  does not require a circular import.
-            td_values = array_to_timedelta64(v).view("m8[ns]")
-        except OutOfBoundsTimedelta:
-            return v.reshape(shape)
-        else:
-            return td_values.reshape(shape)
-
-    # TODO: this is _almost_ equivalent to lib.maybe_convert_objects,
-    #  the main differences are described in GH#49340 and GH#49341
-    #  and maybe_convert_objects doesn't catch OutOfBoundsDatetime
     inferred_type = lib.infer_datetimelike_array(ensure_object(v))
 
-    if inferred_type in ["period", "interval"]:
+    if inferred_type in ["period", "interval", "timedelta", "datetime"]:
         # Incompatible return value type (got "Union[ExtensionArray, ndarray]",
         # expected "Union[ndarray, DatetimeArray, TimedeltaArray, PeriodArray,
         # IntervalArray]")
         return lib.maybe_convert_objects(  # type: ignore[return-value]
-            v, convert_period=True, convert_interval=True
+            v,
+            convert_period=True,
+            convert_interval=True,
+            convert_timedelta=True,
+            convert_datetime=True,
+            dtype_if_all_nat=np.dtype("M8[ns]"),
         )
 
-    if inferred_type == "datetime":
-        # Incompatible types in assignment (expression has type
-        # "Union[ndarray[Any, Any], DatetimeArray]", variable has type
-        # "ndarray[Any, Any]")
-        value = try_datetime(v)  # type: ignore[assignment]
-    elif inferred_type == "timedelta":
-        value = try_timedelta(v)
     elif inferred_type == "nat":
         # if all NaT, return as datetime
         # only reached if we have at least 1 NaT and the rest (NaT or None or np.nan)
+        # This is slightly different from what we'd get with maybe_convert_objects,
+        #  which only converts of all-NaT
+        from pandas.core.arrays.datetimes import sequence_to_datetimes
 
-        # Incompatible types in assignment (expression has type
-        # "Union[ndarray[Any, Any], DatetimeArray]", variable has type
-        # "ndarray[Any, Any]")
-        value = try_datetime(v)  # type: ignore[assignment]
+        # Incompatible types in assignment (expression has type "DatetimeArray",
+        # variable has type "ndarray[Any, Any]")
+        value = sequence_to_datetimes(v)  # type: ignore[assignment]
         assert value.dtype == "M8[ns]"
 
     return value

pandas/tests/dtypes/test_inference.py

@@ -1340,19 +1340,26 @@ def test_infer_dtype_period_with_na(self, na_value):
                 Timestamp("20170612", tz="US/Eastern"),
                 Timestamp("20170311", tz="US/Eastern"),
             ],
-            [date(2017, 6, 12), Timestamp("20170311", tz="US/Eastern")],
             [np.datetime64("2017-06-12"), np.datetime64("2017-03-11")],
             [np.datetime64("2017-06-12"), datetime(2017, 3, 11, 1, 15)],
         ],
     )
     def test_infer_datetimelike_array_datetime(self, data):
         assert lib.infer_datetimelike_array(data) == "datetime"
 
+    def test_infer_datetimelike_array_date_mixed(self):
+        # GH49341 pre-2.0 we these were inferred as "datetime" and "timedelta",
+        #  respectively
+        data = [date(2017, 6, 12), Timestamp("20170311", tz="US/Eastern")]
+        assert lib.infer_datetimelike_array(data) == "mixed"
+
+        data = ([timedelta(2017, 6, 12), date(2017, 3, 11)],)
+        assert lib.infer_datetimelike_array(data) == "mixed"
+
     @pytest.mark.parametrize(
         "data",
         [
             [timedelta(2017, 6, 12), timedelta(2017, 3, 11)],
-            [timedelta(2017, 6, 12), date(2017, 3, 11)],
             [np.timedelta64(2017, "D"), np.timedelta64(6, "s")],
             [np.timedelta64(2017, "D"), timedelta(2017, 3, 11)],
         ],

pandas/tests/series/test_constructors.py

@@ -52,6 +52,20 @@
 
 
 class TestSeriesConstructors:
+    def test_infer_with_date_and_datetime(self):
+        # GH#49341 pre-2.0 we inferred datetime-and-date to datetime64, which
+        #  was inconsistent with Index behavior
+        ts = Timestamp(2016, 1, 1)
+        vals = [ts.to_pydatetime(), ts.date()]
+
+        ser = Series(vals)
+        expected = Series(vals, dtype=object)
+        tm.assert_series_equal(ser, expected)
+
+        idx = Index(vals)
+        expected = Index(vals, dtype=object)
+        tm.assert_index_equal(idx, expected)
+
     def test_unparseable_strings_with_dt64_dtype(self):
         # pre-2.0 these would be silently ignored and come back with object dtype
         vals = ["aa"]