GH28301 check for non-unique index in stack_multi_columns (pandas-dev#28336)

Author: christopherzimmerman

doc/source/whatsnew/v1.0.0.rst

@@ -205,7 +205,7 @@ Groupby/resample/rolling
 Reshaping
 ^^^^^^^^^
 
--
+- Bug in :meth:`DataFrame.stack` not handling non-unique indexes correctly when creating MultiIndex (:issue: `28301`)
 -
 
 Sparse

pandas/core/reshape/reshape.py

@@ -725,8 +725,9 @@ def _convert_level_number(level_num, columns):
         new_names = list(this.index.names)
         new_codes = [lab.repeat(levsize) for lab in this.index.codes]
     else:
-        new_levels = [this.index]
-        new_codes = [np.arange(N).repeat(levsize)]
+        old_codes, old_levels = _factorize_from_iterable(this.index)
+        new_levels = [old_levels]
+        new_codes = [old_codes.repeat(levsize)]
         new_names = [this.index.name]  # something better?
 
     new_levels.append(level_vals)

pandas/tests/frame/test_reshape.py

@@ -1002,6 +1002,27 @@ def test_stack_preserve_categorical_dtype_values(self):
         )
         tm.assert_series_equal(result, expected)
 
+    @pytest.mark.parametrize(
+        "index, columns",
+        [
+            ([0, 0, 1, 1], pd.MultiIndex.from_product([[1, 2], ["a", "b"]])),
+            ([0, 0, 2, 3], pd.MultiIndex.from_product([[1, 2], ["a", "b"]])),
+            ([0, 1, 2, 3], pd.MultiIndex.from_product([[1, 2], ["a", "b"]])),
+        ],
+    )
+    def test_stack_multi_columns_non_unique_index(self, index, columns):
+        # GH-28301
+        df = pd.DataFrame(index=index, columns=columns).fillna(1)
+        stacked = df.stack()
+        new_index = pd.MultiIndex.from_tuples(stacked.index.to_numpy())
+        expected = pd.DataFrame(
+            stacked.to_numpy(), index=new_index, columns=stacked.columns
+        )
+        tm.assert_frame_equal(stacked, expected)
+        stacked_codes = np.asarray(stacked.index.codes)
+        expected_codes = np.asarray(new_index.codes)
+        tm.assert_numpy_array_equal(stacked_codes, expected_codes)
+
     @pytest.mark.parametrize("level", [0, 1])
     def test_unstack_mixed_extension_types(self, level):
         index = pd.MultiIndex.from_tuples(