BUG: OutOfBoundsDatetime with non-nano dt64tz dtype

doc/source/whatsnew/v2.2.0.rst

@@ -330,6 +330,7 @@ Datetimelike
 - Bug in addition or subtraction of :class:`BusinessDay` offset with ``offset`` attribute to non-nanosecond :class:`Index`, :class:`Series`, or :class:`DataFrame` column giving incorrect results (:issue:`55608`)
 - Bug in addition or subtraction of :class:`DateOffset` objects with microsecond components to ``datetime64`` :class:`Index`, :class:`Series`, or :class:`DataFrame` columns with non-nanosecond resolution (:issue:`55595`)
 - Bug in addition or subtraction of very large :class:`Tick` objects with :class:`Timestamp` or :class:`Timedelta` objects raising ``OverflowError`` instead of ``OutOfBoundsTimedelta`` (:issue:`55503`)
+- Bug in creating a :class:`Index`, :class:`Series`, or :class:`DataFrame` with a non-nanosecond :class:`DatetimeTZDtype` and inputs that would be out of bounds with nanosecond resolution incorrectly raising ``OutOfBoundsDatetime`` (:issue:`54620`)
 - Bug in creating a :class:`Index`, :class:`Series`, or :class:`DataFrame` with a non-nanosecond ``datetime64`` dtype and inputs that would be out of bounds for a ``datetime64[ns]`` incorrectly raising ``OutOfBoundsDatetime`` (:issue:`55756`)
 -
 

pandas/_libs/tslib.pyi

@@ -29,5 +29,5 @@ def array_to_datetime(
 # returned ndarray may be object dtype or datetime64[ns]
 
 def array_to_datetime_with_tz(
-    values: npt.NDArray[np.object_], tz: tzinfo
+    values: npt.NDArray[np.object_], tz: tzinfo, creso: int
 ) -> npt.NDArray[np.int64]: ...

pandas/_libs/tslib.pyx

@@ -671,7 +671,7 @@ cdef _array_to_datetime_object(
     return oresult_nd, None
 
 
-def array_to_datetime_with_tz(ndarray values, tzinfo tz):
+def array_to_datetime_with_tz(ndarray values, tzinfo tz, NPY_DATETIMEUNIT creso):
     """
     Vectorized analogue to pd.Timestamp(value, tz=tz)
 
@@ -707,7 +707,7 @@ def array_to_datetime_with_tz(ndarray values, tzinfo tz):
                 else:
                     # datetime64, tznaive pydatetime, int, float
                     ts = ts.tz_localize(tz)
-                ts = ts.as_unit("ns")
+                ts = (<_Timestamp>ts)._as_creso(creso)
                 ival = ts._value
 
         # Analogous to: result[i] = ival

pandas/core/arrays/datetimes.py

@@ -355,7 +355,7 @@ def _from_sequence_not_strict(
                 # DatetimeTZDtype
                 unit = dtype.unit
 
-        subarr, tz, inferred_freq = _sequence_to_dt64ns(
+        subarr, tz, inferred_freq = _sequence_to_dt64(
             data,
             copy=copy,
             tz=tz,
@@ -2179,7 +2179,7 @@ def std(
 # Constructor Helpers
 
 
-def _sequence_to_dt64ns(
+def _sequence_to_dt64(
     data,
     *,
     copy: bool = False,
@@ -2205,7 +2205,8 @@ def _sequence_to_dt64ns(
     Returns
     -------
     result : numpy.ndarray
-        The sequence converted to a numpy array with dtype ``datetime64[ns]``.
+        The sequence converted to a numpy array with dtype ``datetime64[unit]``.
+        Where `unit` is "ns" unless specified otherwise by `out_unit`.
     tz : tzinfo or None
         Either the user-provided tzinfo or one inferred from the data.
     inferred_freq : Tick or None
@@ -2228,9 +2229,9 @@ def _sequence_to_dt64ns(
     data, copy = maybe_convert_dtype(data, copy, tz=tz)
     data_dtype = getattr(data, "dtype", None)
 
-    out_dtype = DT64NS_DTYPE
-    if out_unit is not None:
-        out_dtype = np.dtype(f"M8[{out_unit}]")
+    if out_unit is None:
+        out_unit = "ns"
+    out_dtype = np.dtype(f"M8[{out_unit}]")
 
     if data_dtype == object or is_string_dtype(data_dtype):
         # TODO: We do not have tests specific to string-dtypes,
@@ -2241,8 +2242,10 @@ def _sequence_to_dt64ns(
         elif tz is not None and ambiguous == "raise":
             # TODO: yearfirst/dayfirst/etc?
             obj_data = np.asarray(data, dtype=object)
-            i8data = tslib.array_to_datetime_with_tz(obj_data, tz)
-            return i8data.view(DT64NS_DTYPE), tz, None
+            i8data = tslib.array_to_datetime_with_tz(
+                obj_data, tz, abbrev_to_npy_unit(out_unit)
+            )
+            return i8data.view(out_dtype), tz, None
         else:
             # data comes back here as either i8 to denote UTC timestamps
             #  or M8[ns] to denote wall times

pandas/tests/indexes/datetimes/test_constructors.py

@@ -1013,16 +1013,19 @@ def test_dti_convert_datetime_list(self, tzstr):
         dr2 = DatetimeIndex(list(dr), name="foo", freq="D")
         tm.assert_index_equal(dr, dr2)
 
-    def test_dti_constructor_with_non_nano_dtype(self):
-        # GH#55756
+    @pytest.mark.parametrize("tz", [None, "UTC", "US/Pacific"])
+    def test_dti_constructor_with_non_nano_dtype(self, tz):
+        # GH#55756, GH#54620
         ts = Timestamp("2999-01-01")
         dtype = "M8[us]"
+        if tz is not None:
+            dtype = f"M8[us, {tz}]"
         # NB: the 2500 is interpreted as nanoseconds and rounded *down*
         # to 2 microseconds
         vals = [ts, "2999-01-02 03:04:05.678910", 2500]
         result = DatetimeIndex(vals, dtype=dtype)
-        exp_arr = np.array([ts.asm8, vals[1], 2], dtype=dtype)
-        expected = DatetimeIndex(exp_arr, dtype=dtype)
+        exp_arr = np.array([ts.asm8, vals[1], 2], dtype="M8[us]")
+        expected = DatetimeIndex(exp_arr, dtype="M8[us]").tz_localize(tz)
         tm.assert_index_equal(result, expected)
 
         result2 = DatetimeIndex(np.array(vals, dtype=object), dtype=dtype)

pandas/tests/series/methods/test_astype.py

@@ -107,18 +107,21 @@ def test_astype_dict_like(self, dtype_class):
 
 
 class TestAstype:
-    def test_astype_object_to_dt64_non_nano(self):
-        # GH#55756
+    @pytest.mark.parametrize("tz", [None, "UTC", "US/Pacific"])
+    def test_astype_object_to_dt64_non_nano(self, tz):
+        # GH#55756, GH#54620
         ts = Timestamp("2999-01-01")
         dtype = "M8[us]"
+        if tz is not None:
+            dtype = f"M8[us, {tz}]"
         # NB: the 2500 is interpreted as nanoseconds and rounded *down*
         # to 2 microseconds
         vals = [ts, "2999-01-02 03:04:05.678910", 2500]
         ser = Series(vals, dtype=object)
         result = ser.astype(dtype)
 
-        exp_arr = np.array([ts.asm8, vals[1], 2], dtype=dtype)
-        expected = Series(exp_arr, dtype=dtype)
+        exp_arr = np.array([ts.asm8, vals[1], 2], dtype="M8[us]")
+        expected = Series(exp_arr, dtype="M8[us]").dt.tz_localize(tz)
         tm.assert_series_equal(result, expected)
 
     def test_astype_mixed_object_to_dt64tz(self):