BUG: DataFrameGroupBy.__getitem__ fails to propagate dropna (#35078)

Author: arw2019

doc/source/whatsnew/v1.2.0.rst

@@ -128,7 +128,7 @@ Indexing
 Missing
 ^^^^^^^
 
--
+- Bug in :meth:`SeriesGroupBy.transform` now correctly handles missing values for `dropna=False` (:issue:`35014`)
 -
 
 MultiIndex

pandas/core/groupby/generic.py

@@ -35,11 +35,11 @@
 from pandas.util._decorators import Appender, Substitution, doc
 
 from pandas.core.dtypes.cast import (
+    find_common_type,
     maybe_cast_result,
     maybe_cast_result_dtype,
     maybe_convert_objects,
     maybe_downcast_numeric,
-    maybe_downcast_to_dtype,
 )
 from pandas.core.dtypes.common import (
     ensure_int64,
@@ -513,7 +513,6 @@ def _transform_general(
         """
         Transform with a non-str `func`.
         """
-
         if maybe_use_numba(engine):
             numba_func, cache_key = generate_numba_func(
                 func, engine_kwargs, kwargs, "groupby_transform"
@@ -535,24 +534,23 @@ def _transform_general(
             if isinstance(res, (ABCDataFrame, ABCSeries)):
                 res = res._values
 
-            indexer = self._get_index(name)
-            ser = klass(res, indexer)
-            results.append(ser)
+            results.append(klass(res, index=group.index))
 
         # check for empty "results" to avoid concat ValueError
         if results:
             from pandas.core.reshape.concat import concat
 
-            result = concat(results).sort_index()
+            concatenated = concat(results)
+            result = self._set_result_index_ordered(concatenated)
         else:
             result = self.obj._constructor(dtype=np.float64)
-
         # we will only try to coerce the result type if
         # we have a numeric dtype, as these are *always* user-defined funcs
         # the cython take a different path (and casting)
-        dtype = self._selected_obj.dtype
-        if is_numeric_dtype(dtype):
-            result = maybe_downcast_to_dtype(result, dtype)
+        if is_numeric_dtype(result.dtype):
+            common_dtype = find_common_type([self._selected_obj.dtype, result.dtype])
+            if common_dtype is result.dtype:
+                result = maybe_downcast_numeric(result, self._selected_obj.dtype)
 
         result.name = self._selected_obj.name
         result.index = self._selected_obj.index

pandas/tests/groupby/test_groupby_dropna.py

@@ -162,6 +162,40 @@ def test_groupby_dropna_series_by(dropna, expected):
     tm.assert_series_equal(result, expected)
 
 
+@pytest.mark.parametrize(
+    "dropna,df_expected,s_expected",
+    [
+        pytest.param(
+            True,
+            pd.DataFrame({"B": [2, 2, 1]}),
+            pd.Series(data=[2, 2, 1], name="B"),
+            marks=pytest.mark.xfail(raises=ValueError),
+        ),
+        (
+            False,
+            pd.DataFrame({"B": [2, 2, 1, 1]}),
+            pd.Series(data=[2, 2, 1, 1], name="B"),
+        ),
+    ],
+)
+def test_slice_groupby_then_transform(dropna, df_expected, s_expected):
+    # GH35014
+
+    df = pd.DataFrame({"A": [0, 0, 1, None], "B": [1, 2, 3, None]})
+    gb = df.groupby("A", dropna=dropna)
+
+    res = gb.transform(len)
+    tm.assert_frame_equal(res, df_expected)
+
+    gb_slice = gb[["B"]]
+    res = gb_slice.transform(len)
+    tm.assert_frame_equal(res, df_expected)
+
+    gb_slice = gb["B"]
+    res = gb["B"].transform(len)
+    tm.assert_series_equal(res, s_expected)
+
+
 @pytest.mark.parametrize(
     "dropna, tuples, outputs",
     [