Add support for cftime.datetime coordinates with coarsen (pydata#2778)

doc/whats-new.rst

@@ -45,6 +45,10 @@ Enhancements
   See :ref:`comput.coarsen` for details.
   (:issue:`2525`)
   By `Keisuke Fujii <https://github.com/fujiisoup>`_.
+- Taking the mean of arrays of :py:class:`cftime.datetime` objects, and
+  by extension, use of :py:meth:`~xarray.DataArray.coarsen` with
+  :py:class:`cftime.datetime` coordinates is now possible. By `Spencer Clark
+  <https://github.com/spencerkclark>`_. 
 - Upsampling an array via interpolation with resample is now dask-compatible,
   as long as the array is not chunked along the resampling dimension.
   By `Spencer Clark <https://github.com/spencerkclark>`_.

xarray/core/common.py

@@ -997,15 +997,15 @@ def is_np_datetime_like(dtype):
             np.issubdtype(dtype, np.timedelta64))
 
 
-def contains_cftime_datetimes(var):
-    """Check if a variable contains cftime datetime objects"""
+def _contains_cftime_datetimes(array):
+    """Check if an array contains cftime.datetime objects"""
     try:
         from cftime import datetime as cftime_datetime
     except ImportError:
         return False
     else:
-        if var.dtype == np.dtype('O') and var.data.size > 0:
-            sample = var.data.ravel()[0]
+        if array.dtype == np.dtype('O') and array.size > 0:
+            sample = array.ravel()[0]
             if isinstance(sample, dask_array_type):
                 sample = sample.compute()
                 if isinstance(sample, np.ndarray):
@@ -1015,6 +1015,11 @@ def contains_cftime_datetimes(var):
             return False
 
 
+def contains_cftime_datetimes(var):
+    """Check if an xarray.Variable contains cftime.datetime objects"""
+    return _contains_cftime_datetimes(var.data)
+
+
 def _contains_datetime_like_objects(var):
     """Check if a variable contains datetime like objects (either
     np.datetime64, np.timedelta64, or cftime.datetime)"""

xarray/core/duck_array_ops.py

@@ -294,7 +294,7 @@ def datetime_to_numeric(array, offset=None, datetime_unit=None, dtype=float):
 
     Parameters
     ----------
-    da : array
+    da : np.array
         Input data
     offset: Scalar with the same type of array or None
         If None, subtract minimum values to reduce round off error
@@ -306,6 +306,7 @@ def datetime_to_numeric(array, offset=None, datetime_unit=None, dtype=float):
     -------
     array
     """
+    # TODO: make this function dask-compatible?
     if offset is None:
         offset = array.min()
     array = array - offset
@@ -326,15 +327,34 @@ def datetime_to_numeric(array, offset=None, datetime_unit=None, dtype=float):
     return array.astype(dtype)
 
 
+def _to_pytimedelta(array, unit='us'):
+    index = pd.TimedeltaIndex(array.ravel(), unit=unit)
+    return index.to_pytimedelta().reshape(array.shape)
+
+
 def mean(array, axis=None, skipna=None, **kwargs):
-    """ inhouse mean that can handle datatime dtype """
+    """inhouse mean that can handle np.datetime64 or cftime.datetime
+    dtypes"""
+    from .common import _contains_cftime_datetimes
+
     array = asarray(array)
     if array.dtype.kind in 'Mm':
         offset = min(array)
-        # xarray always uses datetime[ns] for datetime
+        # xarray always uses np.datetime64[ns] for np.datetime64 data
         dtype = 'timedelta64[ns]'
         return _mean(datetime_to_numeric(array, offset), axis=axis,
                      skipna=skipna, **kwargs).astype(dtype) + offset
+    elif _contains_cftime_datetimes(array):
+        if isinstance(array, dask_array_type):
+            raise NotImplementedError(
+                'Computing the mean of an array containing '
+                'cftime.datetime objects is not yet implemented on '
+                'dask arrays.')
+        offset = min(array)
+        timedeltas = datetime_to_numeric(array, offset, datetime_unit='us')
+        mean_timedeltas = _mean(timedeltas, axis=axis, skipna=skipna,
+                                **kwargs)
+        return _to_pytimedelta(mean_timedeltas, unit='us') + offset
     else:
         return _mean(array, axis=axis, skipna=skipna, **kwargs)
 

xarray/tests/test_dataset.py

@@ -23,7 +23,7 @@
     InaccessibleArray, UnexpectedDataAccess, assert_allclose,
     assert_array_equal, assert_equal, assert_identical, has_cftime, has_dask,
     raises_regex, requires_bottleneck, requires_dask, requires_scipy,
-    source_ndarray)
+    source_ndarray, requires_cftime)
 
 try:
     import dask.array as da
@@ -4530,6 +4530,15 @@ def test_coarsen_coords(ds, dask):
     actual = da.coarsen(time=2).mean()
 
 
+@requires_cftime
+def test_coarsen_coords_cftime():
+    times = xr.cftime_range('2000', periods=6)
+    da = xr.DataArray(range(6), [('time', times)])
+    actual = da.coarsen(time=3).mean()
+    expected_times = xr.cftime_range('2000-01-02', freq='3D', periods=2)
+    np.testing.assert_array_equal(actual.time, expected_times)
+
+
 def test_rolling_properties(ds):
     # catching invalid args
     with pytest.raises(ValueError) as exception:

xarray/tests/test_duck_array_ops.py

@@ -270,6 +270,31 @@ def test_datetime_reduce(dask):
     assert da['time'][0].mean() == da['time'][:1].mean()
 
 
+@requires_cftime
+def test_cftime_datetime_mean():
+    times = cftime_range('2000', periods=4)
+    da = DataArray(times, dims=['time'])
+
+    assert da.isel(time=0).mean() == da.isel(time=0)
+
+    expected = DataArray(times.date_type(2000, 1, 2, 12))
+    result = da.mean()
+    assert_equal(result, expected)
+
+    da_2d = DataArray(times.values.reshape(2, 2))
+    result = da_2d.mean()
+    assert_equal(result, expected)
+
+
+@requires_cftime
+@requires_dask
+def test_cftime_datetime_mean_dask_error():
+    times = cftime_range('2000', periods=4)
+    da = DataArray(times, dims=['time']).chunk()
+    with pytest.raises(NotImplementedError):
+        da.mean()
+
+
 @pytest.mark.parametrize('dim_num', [1, 2])
 @pytest.mark.parametrize('dtype', [float, int, np.float32, np.bool_])
 @pytest.mark.parametrize('dask', [False, True])