[PR] time type conversion to timedelta if timedelta64 is given (#660)

* make sure time diff type is timedelta: in case it was timedelta64 convert it, so following `.seconds` line could run without error

* Update xcdat/bounds.py

Co-authored-by: Stephen Po-Chedley <[email protected]>

* proceed `pd.to_timedelta` only if diff type is `np.timedelta64`

* Update `_create_time_bounds()`
- Add comment about needing to convert dype=timedelta64[ns] to pandas timedelta object
- Ignore C901 flake8 error

* Add coverage for new code

* Update tests/test_bounds.py

---------

Co-authored-by: Stephen Po-Chedley <[email protected]>
Co-authored-by: Tom Vo <[email protected]>

tests/fixtures.py

@@ -92,6 +92,22 @@
         "standard_name": "time",
     },
 )
+time_hourly_dt = xr.DataArray(
+    data=np.array(
+        [
+            "2000-01-01T00:00:00.000000000",
+            "2000-01-01T01:00:00.000000000",
+            "2000-01-01T02:00:00.000000000",
+        ],
+        dtype="datetime64[ns]",
+    ),
+    dims=["time"],
+    attrs={
+        "axis": "T",
+        "long_name": "time",
+        "standard_name": "time",
+    },
+)
 time_subhourly = xr.DataArray(
     data=np.array(
         [
@@ -189,6 +205,30 @@
         "xcdat_bounds": "True",
     },
 )
+time_bnds_hourly_dt = xr.DataArray(
+    name="time_bnds",
+    data=np.array(
+        [
+            [
+                "2000-01-01T00:00:00.000000000",
+                "2000-01-01T01:00:00.000000000",
+            ],
+            [
+                "2000-01-01T01:00:00.000000000",
+                "2000-01-01T02:00:00.000000000",
+            ],
+            [
+                "2000-01-01T02:00:00.000000000",
+                "2000-01-01T03:00:00.000000000",
+            ],
+        ],
+        dtype="datetime64[ns]",
+    ),
+    dims=["time", "bnds"],
+    attrs={
+        "xcdat_bounds": "True",
+    },
+)
 time_bnds_subhourly = xr.DataArray(
     name="time_bnds",
     data=np.array(
@@ -495,7 +535,8 @@ def generate_dataset(
 
 
 def generate_dataset_by_frequency(
-    freq: Literal["subhour", "hour", "day", "month", "year"] = "month"
+    freq: Literal["subhour", "hour", "day", "month", "year"] = "month",
+    obj_type: Literal["cftime", "datetime"] = "cftime",
 ) -> xr.Dataset:
     """Generates a dataset for a given temporal frequency.
 
@@ -523,8 +564,15 @@ def generate_dataset_by_frequency(
         time = time_daily.copy()
         time_bnds = time_bnds_daily.copy()
     elif freq == "hour":
-        time = time_hourly.copy()
-        time_bnds = time_bnds_hourly.copy()
+        # Test cftime and datetime. datetime subtraction results in
+        # dtype=timedelta64[ns] objects, which need to be converted to Pandas
+        # TimeDelta objects to use the `.seconds` time component.
+        if obj_type == "cftime":
+            time = time_hourly.copy()
+            time_bnds = time_bnds_hourly.copy()
+        else:
+            time = time_hourly_dt.copy()
+            time_bnds = time_bnds_hourly_dt.copy()
     elif freq == "subhour":
         time = time_subhourly.copy()
         time_bnds = time_bnds_subhourly.copy()

tests/test_bounds.py

@@ -826,6 +826,21 @@ def test_add_bounds_for_time_coords_with_different_frequencies(self):
         assert monthly_bounds.identical(ds_monthly_with_bnds)
         assert yearly_bounds.identical(ds_yearly_with_bnds)
 
+    def test_add_bounds_for_hourly_time_coords_as_datetime_objects(self):
+        # get reference datasets
+        ds_hrly_with_bnds = generate_dataset_by_frequency("hour", "datetime")
+
+        # drop bounds for testing
+        ds_hrly_wo_bnds = ds_hrly_with_bnds.drop_vars("time_bnds")
+
+        # test adding bounds
+        hourly_bounds = ds_hrly_wo_bnds.bounds.add_time_bounds(
+            method="freq", freq="hour"
+        )
+
+        # ensure identical
+        assert hourly_bounds.identical(ds_hrly_with_bnds)
+
     def test_add_monthly_bounds_for_end_of_month_set_to_true(self):
         ds_with_bnds = self.ds_with_bnds.copy()
 

xcdat/bounds.py

@@ -504,7 +504,7 @@ def _get_bounds_keys(self, axis: CFAxisKey) -> List[str]:
 
         return list(set(keys))
 
-    def _create_time_bounds(
+    def _create_time_bounds(  # noqa: C901
         self,
         time: xr.DataArray,
         freq: Optional[Literal["year", "month", "day", "hour"]] = None,
@@ -601,9 +601,16 @@ def _create_time_bounds(
         elif freq == "day":
             time_bnds = self._create_daily_time_bounds(timesteps, obj_type)
         elif freq == "hour":
-            # Determine the daily frequency for generating time  bounds.
+            # Determine the daily frequency for generating time bounds.
             if daily_subfreq is None:
                 diff = time.values[1] - time.values[0]
+
+                # Arrays with `dtype="timedelta64[ns]"` must be converted to
+                # pandas timedelta objects in order to access the `.seconds`
+                # time component.
+                if isinstance(diff, np.timedelta64):
+                    diff = pd.to_timedelta(diff)
+
                 hrs = diff.seconds / 3600
                 daily_subfreq = int(24 / hrs)  # type: ignore