BUG: fix groupby.transform rename bug (#23461)

Author: Diego Torres Quintanilla

doc/source/whatsnew/v0.24.0.txt

@@ -1299,6 +1299,7 @@ Groupby/Resample/Rolling
 - :func:`RollingGroupby.agg` and :func:`ExpandingGroupby.agg` now support multiple aggregation functions as parameters (:issue:`15072`)
 - Bug in :meth:`DataFrame.resample` and :meth:`Series.resample` when resampling by a weekly offset (``'W'``) across a DST transition (:issue:`9119`, :issue:`21459`)
 - Bug in :meth:`DataFrame.expanding` in which the ``axis`` argument was not being respected during aggregations (:issue:`23372`)
+- Bug in :meth:`pandas.core.groupby.DataFrameGroupBy.transform` which caused missing values when the input function can accept a :class:`DataFrame` but renames it (:issue:`23455`).
 
 Reshaping
 ^^^^^^^^^

pandas/core/groupby/generic.py

@@ -583,14 +583,18 @@ def _choose_path(self, fast_path, slow_path, group):
         try:
             res_fast = fast_path(group)
 
-            # compare that we get the same results
+            # verify fast path does not change columns (and names), otherwise
+            # its results cannot be joined with those of the slow path
+            if (getattr(res_fast, 'columns', None)
+                    != getattr(group, 'columns', None)):
+                return path, res
+            # verify numerical equality with the slow path
             if res.shape == res_fast.shape:
                 res_r = res.values.ravel()
                 res_fast_r = res_fast.values.ravel()
                 mask = notna(res_r)
-            if (res_r[mask] == res_fast_r[mask]).all():
-                path = fast_path
-
+                if (res_r[mask] == res_fast_r[mask]).all():
+                    path = fast_path
         except Exception:
             pass
         return path, res

pandas/tests/groupby/test_transform.py

@@ -808,3 +808,24 @@ def test_any_all_np_func(func):
 
     res = df.groupby('key')['val'].transform(func)
     tm.assert_series_equal(res, exp)
+
+
+def test_groupby_transform_rename():
+    # https://github.com/pandas-dev/pandas/issues/23461
+    def demean_rename(x):
+        result = x - x.mean()
+
+        if isinstance(x, pd.Series):
+            return result
+
+        result = result.rename(
+            columns={c: '{}_demeaned'.format(c) for c in result.columns})
+
+        return result
+
+    df = pd.DataFrame({'group': list('ababa'),
+                       'value': [1, 1, 1, 2, 2]})
+    expected = pd.DataFrame({'value': [-1. / 3, -0.5, -1. / 3, 0.5, 2. / 3]})
+
+    result = df.groupby('group').transform(demean_rename)
+    tm.assert_frame_equal(result, expected)