[WIP] Support nano second time encoding.

xarray/coding/times.py

@@ -26,6 +26,7 @@
 _STANDARD_CALENDARS = {"standard", "gregorian", "proleptic_gregorian"}
 
 _NS_PER_TIME_DELTA = {
+    "ns": int(1e0),
     "us": int(1e3),
     "ms": int(1e6),
     "s": int(1e9),
@@ -35,7 +36,15 @@
 }
 
 TIME_UNITS = frozenset(
-    ["days", "hours", "minutes", "seconds", "milliseconds", "microseconds"]
+    [
+        "days",
+        "hours",
+        "minutes",
+        "seconds",
+        "milliseconds",
+        "microseconds",
+        "nanoseconds",
+    ]
 )
 
 
@@ -44,6 +53,7 @@ def _netcdf_to_numpy_timeunit(units):
     if not units.endswith("s"):
         units = "%ss" % units
     return {
+        "nanoseconds": "ns",
         "microseconds": "us",
         "milliseconds": "ms",
         "seconds": "s",

xarray/tests/test_backends.py

@@ -4601,3 +4601,40 @@ def test_extract_zarr_variable_encoding():
         actual = backends.zarr.extract_zarr_variable_encoding(
             var, raise_on_invalid=True
         )
+
+
+@requires_netCDF4
+def test_cftime_nanosecond(tmp_path):
+    import netCDF4
+
+    time_data = xr.DataArray(
+        np.asarray(
+            ["2020-09-01T00:00:00.000000", "2020-09-01T00:00:00.586001"],
+            dtype=np.datetime64,
+        ),
+        name="time",
+    )
+    filename = str(tmp_path / "time.nc")
+
+    time_data.to_netcdf(filename)
+    with netCDF4.Dataset(filename, mode="r") as loaded:
+        # Prior to 0.10.9, the time encoding for objects with nano second
+        # deltas was in seconds, but use floating point to attempt to maintain
+        # the difference
+        assert loaded["time"].units.startswith("seconds")
+        assert loaded["time"].dtype == np.float64
+
+    time_data.encoding = {
+        "units": "nanoseconds since 2020-09-01",
+        "calendar": "proleptic_gregorian",
+    }
+    time_data.to_netcdf(filename)
+    with netCDF4.Dataset(filename, mode="r") as loaded:
+        # Prior to 0.10.9, the time encoding for objects with nano second
+        # deltas was in seconds, but use floating point to attempt to maintain
+        # the difference
+        assert loaded["time"].units.startswith("nanoseconds")
+        assert loaded["time"].dtype == np.int64
+
+    time_data_loaded = xr.load_dataarray(filename)
+    assert_equal(time_data, time_data_loaded)

xarray/tests/test_coding_times.py

@@ -58,6 +58,8 @@
     ([0.5, 1.5], "hours since 1900-01-01T00:00:00"),
     (0, "milliseconds since 2000-01-01T00:00:00"),
     (0, "microseconds since 2000-01-01T00:00:00"),
+    # 2020/09/01: hmaarrfk: cftime/_cftime.pyx doesn't support nanosecond
+    # (0, "nanoseconds since 2000-01-01T00:00:00"),
     (np.int32(788961600), "seconds since 1981-01-01"),  # GH2002
     (12300 + np.arange(5), "hour since 1680-01-01 00:00:00.500000"),
 ]
@@ -537,6 +539,7 @@ def test_infer_cftime_datetime_units(calendar, date_args, expected):
         ("1h", "hours", np.int64(1)),
         ("1ms", "milliseconds", np.int64(1)),
         ("1us", "microseconds", np.int64(1)),
+        ("1ns", "nanoseconds", np.int64(1)),
         (["NaT", "0s", "1s"], None, [np.nan, 0, 1]),
         (["30m", "60m"], "hours", [0.5, 1.0]),
         ("NaT", "days", np.nan),