BUG: Consistent result ordering for groupbys with categorical groupings

doc/source/whatsnew/v2.0.0.rst

@@ -277,6 +277,7 @@ Categorical
 ^^^^^^^^^^^
 - Bug in :meth:`Categorical.set_categories` losing dtype information (:issue:`48812`)
 - Bug in :meth:`DataFrame.groupby` and :meth:`Series.groupby` would reorder categories when used as a grouper (:issue:`48749`)
+- Bug in :class:`GroupBy` when a categorical column was used as a grouper with a range index, ordering of the result would depend on the `sort` argument but if the index was categorical it would depend on the `order` attribute of the index (:issue:`49223`)
 
 Datetimelike
 ^^^^^^^^^^^^

pandas/core/groupby/categorical.py

@@ -53,7 +53,7 @@ def recode_for_groupby(
         unique_codes = unique1d(c.codes)
 
         take_codes = unique_codes[unique_codes != -1]
-        if c.ordered:
+        if sort:
             take_codes = np.sort(take_codes)
 
         # we recode according to the uniques

pandas/core/groupby/grouper.py

@@ -502,6 +502,17 @@ def __init__(
                 self._group_index,
             ) = index._get_grouper_for_level(mapper, level=ilevel, dropna=dropna)
 
+            if is_categorical_dtype(self.grouping_vector):
+                self.grouping_vector: Categorical
+
+                self._passed_categorical = True
+                self._orig_cats = self.grouping_vector.categories
+
+                # Should the sort arg be just `sort` or `sort or self.grouping_vector.ordered`?
+                self.grouping_vector, self._all_grouper = recode_for_groupby(
+                    self.grouping_vector, sort, observed
+                )
+
         # a passed Grouper like, directly get the grouper in the same way
         # as single grouper groupby, use the group_info to get codes
         elif isinstance(self.grouping_vector, Grouper):
@@ -529,6 +540,8 @@ def __init__(
             self._passed_categorical = True
 
             self._orig_cats = self.grouping_vector.categories
+
+            # Should the sort arg be just `sort` or `sort or self.grouping_vector.ordered`?
             self.grouping_vector, self._all_grouper = recode_for_groupby(
                 self.grouping_vector, sort, observed
             )

pandas/tests/groupby/test_categorical.py

@@ -2046,3 +2046,34 @@ def test_many_categories(as_index, sort, index_kind, ordered):
         expected = DataFrame({"a": Series(index), "b": data})
 
     tm.assert_frame_equal(result, expected)
+
+
+def test_categorical_vs_range_index_sorting():
+    categories = np.arange(4, -1, -1)
+    for range_index in [True, False]:
+        df_ordered = DataFrame(
+            {
+                "a": Categorical([2, 1, 2, 3], categories=categories, ordered=True),
+                "b": range(4),
+            }
+        )
+        df_unordered = DataFrame(
+            {
+                "a": Categorical([2, 1, 2, 3], categories=categories, ordered=False),
+                "b": range(4),
+            }
+        )
+
+        if not range_index:
+            df_ordered = df_ordered.set_index("a")
+            df_unordered = df_unordered.set_index("a")
+
+        gb_ordered_sort = df_ordered.groupby("a", sort=True, observed=True)
+        gb_ordered_nosort = df_ordered.groupby("a", sort=False, observed=True)
+        gb_unordered_sort = df_unordered.groupby("a", sort=True, observed=True)
+        gb_unordered_nosort = df_unordered.groupby("a", sort=False, observed=True)
+
+        assert gb_ordered_sort.sum()["b"].tolist() == [3, 2, 1]
+        assert gb_ordered_nosort.sum()["b"].tolist() == [2, 1, 3]
+        assert gb_unordered_sort.sum()["b"].tolist() == [3, 2, 1]
+        assert gb_unordered_nosort.sum()["b"].tolist() == [2, 1, 3]