BUG: loc.setitem with expansion expanding rows

pandas/core/indexing.py

@@ -1684,6 +1684,13 @@ def _setitem_with_indexer_split_path(self, indexer, value):
                 for loc, v in zip(ilocs, value):
                     self._setitem_single_column(loc, v, pi)
 
+            elif len(ilocs) == 1 and com.is_null_slice(pi) and len(self.obj) == 0:
+                # This is a setitem-with-expansion, see
+                #  test_loc_setitem_empty_append_expands_rows
+                # e.g. df = DataFrame(columns=["x", "y"])
+                #  df.loc[:, "x"] = [1, 2, 3]
+                self._setitem_single_column(ilocs[0], value, pi)
+
             else:
                 raise ValueError(
                     "Must have equal len keys and value "

pandas/tests/indexing/test_loc.py

@@ -942,7 +942,7 @@ def test_loc_uint64(self):
         result = s.loc[[np.iinfo("uint64").max - 1, np.iinfo("uint64").max]]
         tm.assert_series_equal(result, s)
 
-    def test_loc_setitem_empty_append(self):
+    def test_loc_setitem_empty_append_expands_rows(self):
         # GH6173, various appends to an empty dataframe
 
         data = [1, 2, 3]
@@ -953,6 +953,13 @@ def test_loc_setitem_empty_append(self):
         df.loc[:, "x"] = data
         tm.assert_frame_equal(df, expected)
 
+        # same thing, but with mixed dtypes
+        df = DataFrame(columns=["x", "y"])
+        df["x"] = df["x"].astype(np.int64)
+        df.loc[:, "x"] = data
+        tm.assert_frame_equal(df, expected)
+
+    def test_loc_setitem_empty_append_single_value(self):
         # only appends one value
         expected = DataFrame({"x": [1.0], "y": [np.nan]})
         df = DataFrame(columns=["x", "y"], dtype=float)