DEPR: DTI int64 values and tz interpreted as UTC (pandas-dev#30115)

doc/source/whatsnew/v1.0.0.rst

@@ -546,6 +546,7 @@ or ``matplotlib.Axes.plot``. See :ref:`plotting.formatters` for more.
 - Removed the previously deprecated "by" keyword from :meth:`DataFrame.sort_index`, use :meth:`DataFrame.sort_values` instead (:issue:`10726`)
 - Removed support for nested renaming in :meth:`DataFrame.aggregate`, :meth:`Series.aggregate`, :meth:`DataFrameGroupBy.aggregate`, :meth:`SeriesGroupBy.aggregate`, :meth:`Rolling.aggregate` (:issue:`18529`)
 - Passing ``datetime64`` data to :class:`TimedeltaIndex` or ``timedelta64`` data to ``DatetimeIndex`` now raises ``TypeError`` (:issue:`23539`, :issue:`23937`)
+- Passing ``int64`` values to :class:`DatetimeIndex` and a timezone now interprets the values as nanosecond timestamps in UTC, not wall times in the given timezone (:issue:`24559`)
 - A tuple passed to :meth:`DataFrame.groupby` is now exclusively treated as a single key (:issue:`18314`)
 - Addition and subtraction of ``int`` or integer-arrays is no longer allowed in :class:`Timestamp`, :class:`DatetimeIndex`, :class:`TimedeltaIndex`, use ``obj + n * obj.freq`` instead of ``obj + n`` (:issue:`22535`)
 - Removed :meth:`Series.from_array` (:issue:`18258`)

pandas/core/arrays/datetimes.py

@@ -55,21 +55,6 @@
 from pandas.tseries.offsets import Day, Tick
 
 _midnight = time(0, 0)
-# TODO(GH-24559): Remove warning, int_as_wall_time parameter.
-_i8_message = """
-    Passing integer-dtype data and a timezone to DatetimeIndex. Integer values
-    will be interpreted differently in a future version of pandas. Previously,
-    these were viewed as datetime64[ns] values representing the wall time
-    *in the specified timezone*. In the future, these will be viewed as
-    datetime64[ns] values representing the wall time *in UTC*. This is similar
-    to a nanosecond-precision UNIX epoch. To accept the future behavior, use
-
-        pd.to_datetime(integer_data, utc=True).tz_convert(tz)
-
-    To keep the previous behavior, use
-
-        pd.to_datetime(integer_data).tz_localize(tz)
-"""
 
 
 def tz_to_dtype(tz):
@@ -422,7 +407,6 @@ def _from_sequence(
         dayfirst=False,
         yearfirst=False,
         ambiguous="raise",
-        int_as_wall_time=False,
     ):
 
         freq, freq_infer = dtl.maybe_infer_freq(freq)
@@ -435,7 +419,6 @@ def _from_sequence(
             dayfirst=dayfirst,
             yearfirst=yearfirst,
             ambiguous=ambiguous,
-            int_as_wall_time=int_as_wall_time,
         )
 
         freq, freq_infer = dtl.validate_inferred_freq(freq, inferred_freq, freq_infer)
@@ -1811,7 +1794,6 @@ def sequence_to_dt64ns(
     dayfirst=False,
     yearfirst=False,
     ambiguous="raise",
-    int_as_wall_time=False,
 ):
     """
     Parameters
@@ -1824,13 +1806,6 @@ def sequence_to_dt64ns(
     yearfirst : bool, default False
     ambiguous : str, bool, or arraylike, default 'raise'
         See pandas._libs.tslibs.conversion.tz_localize_to_utc.
-    int_as_wall_time : bool, default False
-        Whether to treat ints as wall time in specified timezone, or as
-        nanosecond-precision UNIX epoch (wall time in UTC).
-        This is used in DatetimeIndex.__init__ to deprecate the wall-time
-        behaviour.
-
-        ..versionadded:: 0.24.0
 
     Returns
     -------
@@ -1891,10 +1866,6 @@ def sequence_to_dt64ns(
                 data, dayfirst=dayfirst, yearfirst=yearfirst
             )
             tz = maybe_infer_tz(tz, inferred_tz)
-            # When a sequence of timestamp objects is passed, we always
-            # want to treat the (now i8-valued) data as UTC timestamps,
-            # not wall times.
-            int_as_wall_time = False
 
     # `data` may have originally been a Categorical[datetime64[ns, tz]],
     # so we need to handle these types.
@@ -1928,12 +1899,6 @@ def sequence_to_dt64ns(
 
         if data.dtype != _INT64_DTYPE:
             data = data.astype(np.int64, copy=False)
-        if int_as_wall_time and tz is not None and not timezones.is_utc(tz):
-            warnings.warn(_i8_message, FutureWarning, stacklevel=4)
-            data = conversion.tz_localize_to_utc(
-                data.view("i8"), tz, ambiguous=ambiguous
-            )
-            data = data.view(_NS_DTYPE)
         result = data.view(_NS_DTYPE)
 
     if copy:

pandas/core/indexes/datetimes.py

@@ -295,7 +295,6 @@ def __new__(
             dayfirst=dayfirst,
             yearfirst=yearfirst,
             ambiguous=ambiguous,
-            int_as_wall_time=True,
         )
 
         subarr = cls._simple_new(dtarr, name=name, freq=dtarr.freq, tz=dtarr.tz)

pandas/tests/indexes/datetimes/test_construction.py

@@ -122,15 +122,11 @@ def test_construction_with_alt_tz_localize(self, kwargs, tz_aware_fixture):
         i = pd.date_range("20130101", periods=5, freq="H", tz=tz)
         kwargs = {key: attrgetter(val)(i) for key, val in kwargs.items()}
 
-        if str(tz) in ("UTC", "tzutc()", "UTC+00:00"):
-            warn = None
-        else:
-            warn = FutureWarning
-
-        with tm.assert_produces_warning(warn, check_stacklevel=False):
-            result = DatetimeIndex(i.tz_localize(None).asi8, **kwargs)
-        expected = DatetimeIndex(i, **kwargs)
-        tm.assert_index_equal(result, expected)
+        if "tz" in kwargs:
+            result = DatetimeIndex(i.asi8, tz="UTC").tz_convert(kwargs["tz"])
+
+            expected = DatetimeIndex(i, **kwargs)
+            tm.assert_index_equal(result, expected)
 
         # localize into the provided tz
         i2 = DatetimeIndex(i.tz_localize(None).asi8, tz="UTC")
@@ -485,11 +481,13 @@ def test_construction_with_ndarray(self):
         expected = DatetimeIndex(["2013-10-07", "2013-10-08", "2013-10-09"], freq="B")
         tm.assert_index_equal(result, expected)
 
-    def test_integer_values_and_tz_deprecated(self):
+    def test_integer_values_and_tz_interpreted_as_utc(self):
         # GH-24559
-        values = np.array([946684800000000000])
-        with tm.assert_produces_warning(FutureWarning):
-            result = DatetimeIndex(values, tz="US/Central")
+        val = np.datetime64("2000-01-01 00:00:00", "ns")
+        values = np.array([val.view("i8")])
+
+        result = DatetimeIndex(values).tz_localize("US/Central")
+
         expected = pd.DatetimeIndex(["2000-01-01T00:00:00"], tz="US/Central")
         tm.assert_index_equal(result, expected)
 
@@ -718,17 +716,7 @@ def test_constructor_timestamp_near_dst(self):
     @pytest.mark.parametrize("box", [np.array, partial(np.array, dtype=object), list])
     @pytest.mark.parametrize(
         "tz, dtype",
-        [
-            pytest.param(
-                "US/Pacific",
-                "datetime64[ns, US/Pacific]",
-                marks=[
-                    pytest.mark.xfail(),
-                    pytest.mark.filterwarnings("ignore:\\n    Passing:FutureWarning"),
-                ],
-            ),
-            [None, "datetime64[ns]"],
-        ],
+        [("US/Pacific", "datetime64[ns, US/Pacific]"), (None, "datetime64[ns]")],
     )
     def test_constructor_with_int_tz(self, klass, box, tz, dtype):
         # GH 20997, 20964

pandas/tests/indexes/multi/test_integrity.py

@@ -49,9 +49,8 @@ def test_values_multiindex_datetimeindex():
     # Test to ensure we hit the boxing / nobox part of MI.values
     ints = np.arange(10 ** 18, 10 ** 18 + 5)
     naive = pd.DatetimeIndex(ints)
-    # TODO(GH-24559): Remove the FutureWarning
-    with tm.assert_produces_warning(FutureWarning, check_stacklevel=False):
-        aware = pd.DatetimeIndex(ints, tz="US/Central")
+
+    aware = pd.DatetimeIndex(ints, tz="US/Central")
 
     idx = pd.MultiIndex.from_arrays([naive, aware])
     result = idx.values

pandas/tests/indexes/test_base.py

@@ -453,9 +453,9 @@ def test_constructor_dtypes_to_timedelta(self, cast_index, vals):
             index = Index(vals)
             assert isinstance(index, TimedeltaIndex)
 
-    @pytest.mark.parametrize("attr, utc", [["values", False], ["asi8", True]])
+    @pytest.mark.parametrize("attr", ["values", "asi8"])
     @pytest.mark.parametrize("klass", [pd.Index, pd.DatetimeIndex])
-    def test_constructor_dtypes_datetime(self, tz_naive_fixture, attr, utc, klass):
+    def test_constructor_dtypes_datetime(self, tz_naive_fixture, attr, klass):
         # Test constructing with a datetimetz dtype
         # .values produces numpy datetimes, so these are considered naive
         # .asi8 produces integers, so these are considered epoch timestamps
@@ -466,30 +466,27 @@ def test_constructor_dtypes_datetime(self, tz_naive_fixture, attr, utc, klass):
         index = index.tz_localize(tz_naive_fixture)
         dtype = index.dtype
 
-        if (
-            tz_naive_fixture
-            and attr == "asi8"
-            and str(tz_naive_fixture) not in ("UTC", "tzutc()", "UTC+00:00")
-        ):
-            ex_warn = FutureWarning
+        if attr == "asi8":
+            result = pd.DatetimeIndex(arg).tz_localize(tz_naive_fixture)
         else:
-            ex_warn = None
-
-        # stacklevel is checked elsewhere. We don't do it here since
-        # Index will have an frame, throwing off the expected.
-        with tm.assert_produces_warning(ex_warn, check_stacklevel=False):
             result = klass(arg, tz=tz_naive_fixture)
         tm.assert_index_equal(result, index)
 
-        with tm.assert_produces_warning(ex_warn, check_stacklevel=False):
+        if attr == "asi8":
+            result = pd.DatetimeIndex(arg).astype(dtype)
+        else:
             result = klass(arg, dtype=dtype)
         tm.assert_index_equal(result, index)
 
-        with tm.assert_produces_warning(ex_warn, check_stacklevel=False):
+        if attr == "asi8":
+            result = pd.DatetimeIndex(list(arg)).tz_localize(tz_naive_fixture)
+        else:
             result = klass(list(arg), tz=tz_naive_fixture)
         tm.assert_index_equal(result, index)
 
-        with tm.assert_produces_warning(ex_warn, check_stacklevel=False):
+        if attr == "asi8":
+            result = pd.DatetimeIndex(list(arg)).astype(dtype)
+        else:
             result = klass(list(arg), dtype=dtype)
         tm.assert_index_equal(result, index)