Backport PR #39071 on branch 1.2.x (Regression in loc.setitem raising ValueError with unordered MultiIndex columns and scalar indexer)

doc/source/whatsnew/v1.2.1.rst

@@ -22,6 +22,7 @@ Fixed regressions
 - Fixed regression in :meth:`DataFrame.any` and :meth:`DataFrame.all` not returning a result for tz-aware ``datetime64`` columns (:issue:`38723`)
 - Fixed regression in :meth:`DataFrame.__setitem__` raising ``ValueError`` when expanding :class:`DataFrame` and new column is from type ``"0 - name"`` (:issue:`39010`)
 - Fixed regression in :meth:`.GroupBy.sem` where the presence of non-numeric columns would cause an error instead of being dropped (:issue:`38774`)
+- Fixed regression in :meth:`DataFrame.loc.__setitem__` raising ``ValueError`` when :class:`DataFrame` has unsorted :class:`MultiIndex` columns and indexer is a scalar (:issue:`38601`)
 - Fixed regression in :func:`read_excel` with non-rawbyte file handles (:issue:`38788`)
 - Bug in :meth:`read_csv` with ``float_precision="high"`` caused segfault or wrong parsing of long exponent strings. This resulted in a regression in some cases as the default for ``float_precision`` was changed in pandas 1.2.0 (:issue:`38753`)
 - Fixed regression in :meth:`DataFrameGroupBy.diff` raising for ``int8`` and ``int16`` columns (:issue:`39050`)

pandas/core/indexing.py

@@ -13,6 +13,7 @@
 
 from pandas.core.dtypes.common import (
     is_array_like,
+    is_bool_dtype,
     is_hashable,
     is_integer,
     is_iterator,
@@ -1925,12 +1926,14 @@ def _ensure_iterable_column_indexer(self, column_indexer):
         """
         Ensure that our column indexer is something that can be iterated over.
         """
-        # Ensure we have something we can iterate over
         if is_integer(column_indexer):
             ilocs = [column_indexer]
         elif isinstance(column_indexer, slice):
-            ri = Index(range(len(self.obj.columns)))
-            ilocs = ri[column_indexer]
+            ilocs = np.arange(len(self.obj.columns))[column_indexer]
+        elif isinstance(column_indexer, np.ndarray) and is_bool_dtype(
+            column_indexer.dtype
+        ):
+            ilocs = np.arange(len(column_indexer))[column_indexer]
         else:
             ilocs = column_indexer
         return ilocs

pandas/tests/frame/indexing/test_indexing.py

@@ -1682,6 +1682,21 @@ def test_getitem_interval_index_partial_indexing(self):
         res = df.loc[:, 0.5]
         tm.assert_series_equal(res, expected)
 
+    @pytest.mark.parametrize("indexer", ["A", ["A"], ("A", slice(None))])
+    def test_setitem_unsorted_multiindex_columns(self, indexer):
+        # GH#38601
+        mi = MultiIndex.from_tuples([("A", 4), ("B", "3"), ("A", "2")])
+        df = DataFrame([[1, 2, 3], [4, 5, 6]], columns=mi)
+        obj = df.copy()
+        obj.loc[:, indexer] = np.zeros((2, 2), dtype=int)
+        expected = DataFrame([[0, 2, 0], [0, 5, 0]], columns=mi)
+        tm.assert_frame_equal(obj, expected)
+
+        df = df.sort_index(1)
+        df.loc[:, indexer] = np.zeros((2, 2), dtype=int)
+        expected = expected.sort_index(1)
+        tm.assert_frame_equal(df, expected)
+
 
 class TestDataFrameIndexingUInt64:
     def test_setitem(self, uint64_frame):