Fix CategoricalIndex ordering in Groupby.agg when pandas-compatib…

…ility mode is enabled (#13978)

closes #13974 

This PR re-calculates the `CategoricalIndex`'s `categories` order to match the order in which the grouping has been done for the `CategoricalColumn`. This fix is being done only when pandas-compatibility mode is enabled.

Authors:
  - GALI PREM SAGAR (https://github.com/galipremsagar)

Approvers:
  - Matthew Roeschke (https://github.com/mroeschke)

URL: #13978

python/cudf/cudf/core/groupby/groupby.py

@@ -613,6 +613,15 @@ def agg(self, func):
                     how="left",
                 )
                 result = result.take(indices)
+                if isinstance(result._index, cudf.CategoricalIndex):
+                    # Needs re-ordering the categories in the order
+                    # they are after grouping.
+                    result._index = cudf.Index(
+                        result._index._column.reorder_categories(
+                            result._index._column._get_decategorized_column()
+                        ),
+                        name=result._index.name,
+                    )
 
         if not self._as_index:
             result = result.reset_index()

python/cudf/cudf/tests/test_groupby.py

@@ -3440,3 +3440,19 @@ def test_groupby_consecutive_operations():
     expected = pg.cumsum()
 
     assert_groupby_results_equal(actual, expected, check_dtype=False)
+
+
+def test_categorical_grouping_pandas_compatibility():
+    gdf = cudf.DataFrame(
+        {
+            "key": cudf.Series([2, 1, 3, 1, 1], dtype="category"),
+            "a": [0, 1, 3, 2, 3],
+        }
+    )
+    pdf = gdf.to_pandas()
+
+    with cudf.option_context("mode.pandas_compatible", True):
+        actual = gdf.groupby("key", sort=False).sum()
+    expected = pdf.groupby("key", sort=False).sum()
+
+    assert_eq(actual, expected)