fixes pydata#1027

xarray/core/combine.py

@@ -189,8 +189,8 @@ def differs(vname):
     concat_over.update(process_subset_opt(coords, 'coords'))
     if dim in datasets[0]:
         concat_over.add(dim)
-
-    return concat_over
+    # return a list to keep the variables order
+    return [vn for vn in datasets[0].variables if vn in concat_over]
 
 
 def _dataset_concat(datasets, dim, data_vars, coords, compat, positions):

xarray/test/test_dataset.py

@@ -1858,6 +1858,21 @@ def test_groupby_nan(self):
         expected = Dataset({'foo': ('bar', [1.5, 3]), 'bar': [1, 2]})
         self.assertDatasetIdentical(actual, expected)
 
+    def test_groupby_order(self):
+        # groupby should preserve variables order
+
+        ds = Dataset()
+        for vn in ['a', 'b', 'c']:
+            ds[vn] = DataArray(np.arange(10), dims=['t'])
+        all_vars_ref = list(ds.variables.keys())
+        data_vars_ref = list(ds.data_vars.keys())
+        ds = ds.groupby('t').mean()
+        all_vars = list(ds.variables.keys())
+        data_vars = list(ds.data_vars.keys())
+        self.assertEqual(data_vars, data_vars_ref)
+        # coords are now at the end of the list, so the test below fails
+        # self.assertEqual(all_vars, all_vars_ref)
+
     def test_resample_and_first(self):
         times = pd.date_range('2000-01-01', freq='6H', periods=10)
         ds = Dataset({'foo': (['time', 'x', 'y'], np.random.randn(10, 5, 3)),