added unstack in apply

fix dataset test bug

fixed apply

xarray/core/groupby.py

@@ -101,31 +101,22 @@ def __init__(self, obj, group, squeeze=False, grouper=None):
         """
         from .dataset import as_dataset
 
+        if getattr(group, 'name', None) is None:
+            raise ValueError('`group` must have a name')
+        self._stacked_dim = None
         if group.ndim != 1:
             # try to stack the dims of the group into a single dim
             # TODO: figure out how to exclude dimensions from the stacking
             #       (e.g. group over space dims but leave time dim intact)
             orig_dims = group.dims
             stacked_dim_name = 'stacked_' + '_'.join(orig_dims)
-            # the copy is necessary here
+            # the copy is necessary here, otherwise read only array raises error
+            # in pandas: https://github.com/pydata/pandas/issues/12813
+            # Is there a performance penalty for calling copy?
             group = group.stack(**{stacked_dim_name: orig_dims}).copy()
-            # without it, an error is raised deep in pandas
-            ########################
-            # xarray/core/groupby.py
-            # ---> 31     inverse, values = pd.factorize(ar, sort=True)
-            # pandas/core/algorithms.pyc in factorize(values, sort, order, na_sentinel, size_hint)
-            # --> 196     labels = table.get_labels(vals, uniques, 0, na_sentinel, True)
-            # pandas/hashtable.pyx in pandas.hashtable.Float64HashTable.get_labels (pandas/hashtable.c:10302)()
-            # pandas/hashtable.so in View.MemoryView.memoryview_cwrapper (pandas/hashtable.c:29882)()
-            # pandas/hashtable.so in View.MemoryView.memoryview.__cinit__ (pandas/hashtable.c:26251)()
-            # ValueError: buffer source array is read-only
-            #######################
-            # seems related to
-            # https://github.com/pydata/pandas/issues/10043
-            # https://github.com/pydata/pandas/pull/10070
             obj = obj.stack(**{stacked_dim_name: orig_dims})
-        if getattr(group, 'name', None) is None:
-            raise ValueError('`group` must have a name')
+            self._stacked_dim = stacked_dim_name
+            self._unstacked_dims = orig_dims
         if not hasattr(group, 'dims'):
             raise ValueError("`group` must have a 'dims' attribute")
         group_dim, = group.dims
@@ -249,6 +240,13 @@ def _maybe_restore_empty_groups(self, combined):
             combined = combined.reindex(**indexers)
         return combined
 
+    def _maybe_unstack_array(self, arr):
+        """This gets called if we are applying on an array with a
+        multidimensional group."""
+        if self._stacked_dim is not None and self._stacked_dim in arr.dims:
+            arr = arr.unstack(self._stacked_dim)
+        return arr
+
     def fillna(self, value):
         """Fill missing values in this object by group.
 
@@ -358,6 +356,11 @@ def lookup_order(dimension):
         new_order = sorted(stacked.dims, key=lookup_order)
         return stacked.transpose(*new_order)
 
+    def _restore_multiindex(self, combined):
+        if self._stacked_dim is not None and self._stacked_dim in combined.dims:
+            combined[self._stacked_dim] = self.group[self._stacked_dim]
+        return combined
+
     def apply(self, func, shortcut=False, **kwargs):
         """Apply a function over each array in the group and concatenate them
         together into a new array.
@@ -399,23 +402,23 @@ def apply(self, func, shortcut=False, **kwargs):
             grouped = self._iter_grouped_shortcut()
         else:
             grouped = self._iter_grouped()
-        applied = (maybe_wrap_array(arr, func(arr, **kwargs)) for arr in grouped)
+        applied = (maybe_wrap_array(arr,func(arr, **kwargs)) for arr in grouped)
         combined = self._concat(applied, shortcut=shortcut)
-        result = self._maybe_restore_empty_groups(combined)
+        result = self._maybe_restore_empty_groups(
+                    self._maybe_unstack_array(combined))
         return result
 
     def _concat(self, applied, shortcut=False):
         # peek at applied to determine which coordinate to stack over
         applied_example, applied = peek_at(applied)
         concat_dim, positions = self._infer_concat_args(applied_example)
-
         if shortcut:
             combined = self._concat_shortcut(applied, concat_dim, positions)
         else:
             combined = concat(applied, concat_dim, positions=positions)
-
         if isinstance(combined, type(self.obj)):
             combined = self._restore_dim_order(combined)
+            combined = self._restore_multiindex(combined)
         return combined
 
     def reduce(self, func, dim=None, axis=None, keep_attrs=False,

xarray/test/test_dataarray.py

@@ -1255,6 +1255,17 @@ def test_groupby_multidim(self):
             actual_sum = array.groupby(dim).sum()
             self.assertDataArrayIdentical(expected_sum, actual_sum)
 
+    def test_groupby_multidim_apply(self):
+        array = DataArray([[0,1],[2,3]],
+                    coords={'lon': (['ny','nx'], [[30,40],[40,50]] ),
+                            'lat': (['ny','nx'], [[10,10],[20,20]] ),},
+                    dims=['ny','nx'])
+        actual = array.groupby('lon').apply(
+                lambda x : x - x.mean(), shortcut=False)
+        expected = DataArray([[0.,-0.5],[0.5,0.]],
+                    coords=array.coords, dims=array.dims)
+        self.assertDataArrayIdentical(expected, actual)
+
     def make_rolling_example_array(self):
         times = pd.date_range('2000-01-01', freq='1D', periods=21)
         values = np.random.random((21, 4))