Address lingering generic np.timedelta64("NaT") warning; defensively

switch to non-generic np.datetime64("NaT") values.

Author: spencerkclark

xarray/coding/times.py

@@ -830,7 +830,7 @@ def cftime_to_nptime(
                     f"standard calendar.  Reason: {e}."
                 ) from e
             else:
-                dt = np.datetime64("NaT")
+                dt = np.datetime64("NaT", time_unit)
         new.append(dt)
     return np.asarray(new).reshape(times.shape)
 

xarray/tests/test_accessor_dt.py

@@ -266,7 +266,7 @@ def test_seasons(self) -> None:
         dates = xr.date_range(
             start="2000/01/01", freq="ME", periods=12, use_cftime=False
         )
-        dates = dates.append(pd.Index([np.datetime64("NaT")]))
+        dates = dates.append(pd.Index([np.datetime64("NaT", "us")]))
         dates = xr.DataArray(dates)
         seasons = xr.DataArray(
             [

xarray/tests/test_coding_times.py

@@ -1391,7 +1391,11 @@ def test_roundtrip_datetime64_nanosecond_precision(
 ) -> None:
     # test for GH7817
     time = np.datetime64(timestr, format)
-    times = [np.datetime64("1970-01-01T00:00:00", format), np.datetime64("NaT"), time]
+    times = [
+        np.datetime64("1970-01-01T00:00:00", format),
+        np.datetime64("NaT", format),
+        time,
+    ]
 
     if use_encoding:
         encoding = dict(dtype=dtype, _FillValue=fill_value)

xarray/tests/test_dataset.py

@@ -1335,7 +1335,7 @@ def test_chunk_by_frequency(self, freq: str, calendar: str, add_gap: bool) -> No
         ).to_numpy(copy=True)
         if add_gap:
             # introduce an empty bin
-            time[31 : 31 + ΔN] = np.datetime64("NaT")
+            time[31 : 31 + ΔN] = np.datetime64("NaT", "us")
             time = time[~np.isnat(time)]
         else:
             time = time[:N]

xarray/tests/test_duck_array_ops.py

@@ -723,11 +723,11 @@ def test_argmin_max_error():
     ["array", "expected"],
     [
         (
-            np.array([np.datetime64("2000-01-01"), np.datetime64("NaT")]),
+            np.array([np.datetime64("2000-01-01"), np.datetime64("NaT", "D")]),
             np.array([False, True]),
         ),
         (
-            np.array([np.timedelta64(1, "h"), np.timedelta64("NaT")]),
+            np.array([np.timedelta64(1, "h"), np.timedelta64("NaT", "h")]),
             np.array([False, True]),
         ),
         (

xarray/tests/test_groupby.py

@@ -726,7 +726,7 @@ def test_groupby_drops_nans(shuffle: bool, chunk: Literal[False] | dict) -> None
         [
             np.datetime64("2001-01-01"),
             np.datetime64("2001-01-01"),
-            np.datetime64("NaT"),
+            np.datetime64("NaT", "D"),
         ],
     )
     expected5 = xr.DataArray(3, [("t", [np.datetime64("2001-01-01")])])
@@ -2018,7 +2018,7 @@ def test_resample_first_last(self, use_cftime) -> None:
         # missing periods, GH10169
         actual = array.isel(time=[0, 1, 2, 3, 8, 9]).resample(time="1D").last()
         expected = DataArray(
-            np.array([times[3], np.datetime64("NaT"), times[9]]),
+            np.array([times[3], np.datetime64("NaT", "us"), times[9]]),
             dims="time",
             coords={"time": times[::4]},
             name="time",
@@ -3836,7 +3836,7 @@ def test_mean_datetime_edge_cases():
     # Test with NaT values
     dates_with_nat = pd.date_range("2021-01-01", periods=4, freq="D")
     dates_with_nat_array = dates_with_nat.values.copy()
-    dates_with_nat_array[1] = np.datetime64("NaT")
+    dates_with_nat_array[1] = np.datetime64("NaT", "us")
 
     ds = xr.Dataset(
         {