Honor level and sort_remaining in sort_index(axis=1)

python/cudf/cudf/core/indexed_frame.py

@@ -2921,7 +2921,23 @@ def sort_index(
             if ignore_index:
                 out = out.reset_index(drop=True)
         else:
-            labels = sorted(self._column_names, reverse=not ascending)
+            # The column labels are host-side pandas metadata, so have
+            # pandas compute the sorted column order: a Series holding the
+            # original positions, indexed by the column labels, sorted by
+            # its index yields the positional indexer while inheriting
+            # pandas' exact level resolution and validation, per-level
+            # ascending, sort_remaining, and na_position semantics.
+            pd_columns = self._data.to_pandas_index
+            indexer = pd.Series(
+                range(len(pd_columns)), index=pd_columns
+            ).sort_index(
+                level=level,
+                ascending=ascending,
+                sort_remaining=sort_remaining,
+                na_position=na_position,
+                kind="stable",
+            )
+            labels = [self._column_names[i] for i in indexer]
             result_columns = (self._data[label] for label in labels)
             if ignore_index:
                 ca = ColumnAccessor(

python/cudf/cudf/pandas/scripts/pandas-testing-plugin.py

@@ -1713,7 +1713,6 @@ def pytest_unconfigure(config):
     "tests/frame/test_stack_unstack.py::TestStackUnstackMultiLevel::test_stack[True]": "TODO: Add a reason for failure",
     "tests/frame/test_stack_unstack.py::TestStackUnstackMultiLevel::test_stack_level_name[False]": "TODO: Add a reason for failure",
     "tests/frame/test_stack_unstack.py::TestStackUnstackMultiLevel::test_stack_level_name[True]": "TODO: Add a reason for failure",
-    "tests/frame/test_stack_unstack.py::TestStackUnstackMultiLevel::test_stack_mixed_dtype[True]": "TODO: Add a reason for failure",
     "tests/frame/test_stack_unstack.py::TestStackUnstackMultiLevel::test_stack_multiple_bug[False]": "TODO: Add a reason for failure",
     "tests/frame/test_stack_unstack.py::TestStackUnstackMultiLevel::test_stack_multiple_bug[True]": "TODO: Add a reason for failure",
     "tests/frame/test_stack_unstack.py::TestStackUnstackMultiLevel::test_stack_multiple_out_of_bounds[False]": "TODO: Add a reason for failure",

python/cudf/cudf/tests/dataframe/methods/test_sort_index.py

@@ -1,4 +1,4 @@
-# SPDX-FileCopyrightText: Copyright (c) 2025-2026, NVIDIA CORPORATION.
+# SPDX-FileCopyrightText: Copyright (c) 2025-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 import itertools
 
@@ -163,3 +163,125 @@ def test_df_cat_sort_index():
     expect = df.to_pandas().set_index("a").sort_index()
 
     assert_eq(got, expect)
+
+
+@pytest.mark.parametrize("level", [0, 1, "c1", ["c2"], ["c2", "c1"], [1, 0]])
+@pytest.mark.parametrize("sort_remaining", [True, False])
+@pytest.mark.parametrize("ascending", [True, False])
+@pytest.mark.parametrize("na_position", ["first", "last"])
+def test_sort_index_axis_1_multiindex_level(
+    level, sort_remaining, ascending, na_position
+):
+    # axis=1 previously ignored level= and sort_remaining= silently
+    pdf = pd.DataFrame(
+        [[1, 2, 3, 4]],
+        columns=pd.MultiIndex.from_tuples(
+            [("b", 2), ("a", 1), (np.nan, 3), ("a", 4)], names=["c1", "c2"]
+        ),
+    )
+    gdf = cudf.DataFrame(pdf)
+
+    expect = pdf.sort_index(
+        axis=1,
+        level=level,
+        sort_remaining=sort_remaining,
+        ascending=ascending,
+        na_position=na_position,
+    )
+    got = gdf.sort_index(
+        axis=1,
+        level=level,
+        sort_remaining=sort_remaining,
+        ascending=ascending,
+        na_position=na_position,
+    )
+    assert_eq(expect, got)
+
+
+def test_sort_index_axis_1_level_out_of_bounds():
+    pdf = pd.DataFrame(
+        [[1, 2]],
+        columns=pd.MultiIndex.from_tuples([("a", 1), ("b", 2)]),
+    )
+    gdf = cudf.DataFrame(pdf)
+    with pytest.raises(IndexError):
+        gdf.sort_index(axis=1, level=5)
+
+
+@pytest.mark.parametrize(
+    "level, ascending",
+    [
+        (["c1", "c2"], [True, False]),
+        (["c2", "c1"], [False, True]),
+        ([0, 1], [False, False]),
+    ],
+)
+def test_sort_index_axis_1_per_level_ascending(level, ascending):
+    pdf = pd.DataFrame(
+        [[1, 2, 3, 4]],
+        columns=pd.MultiIndex.from_tuples(
+            [("b", 2), ("a", 1), ("b", 3), ("a", 4)], names=["c1", "c2"]
+        ),
+    )
+    gdf = cudf.DataFrame(pdf)
+
+    expect = pdf.sort_index(axis=1, level=level, ascending=ascending)
+    got = gdf.sort_index(axis=1, level=level, ascending=ascending)
+    assert_eq(expect, got)
+
+
+def test_sort_index_axis_1_ascending_length_mismatch_raises():
+    pdf = pd.DataFrame(
+        [[1, 2]],
+        columns=pd.MultiIndex.from_tuples([("a", 1), ("b", 2)]),
+    )
+    gdf = cudf.DataFrame(pdf)
+    with pytest.raises(ValueError):
+        pdf.sort_index(axis=1, level=[0, 1], ascending=[True])
+    with pytest.raises(ValueError):
+        gdf.sort_index(axis=1, level=[0, 1], ascending=[True])
+
+
+def test_sort_index_axis_1_unknown_level_name_raises():
+    pdf = pd.DataFrame(
+        [[1, 2]],
+        columns=pd.MultiIndex.from_tuples(
+            [("a", 1), ("b", 2)], names=["c1", "c2"]
+        ),
+    )
+    gdf = cudf.DataFrame(pdf)
+    with pytest.raises(KeyError):
+        pdf.sort_index(axis=1, level="nope")
+    with pytest.raises(KeyError):
+        gdf.sort_index(axis=1, level="nope")
+
+
+@pytest.mark.parametrize("level", [0, "cols"])
+@pytest.mark.parametrize("ascending", [True, False])
+def test_sort_index_axis_1_flat_columns_level(level, ascending):
+    # a flat columns axis accepts level 0 / its own name like pandas
+    pdf = pd.DataFrame(
+        [[1, 2, 3]], columns=pd.Index(["b", "c", "a"], name="cols")
+    )
+    gdf = cudf.DataFrame(pdf)
+
+    expect = pdf.sort_index(axis=1, level=level, ascending=ascending)
+    got = gdf.sort_index(axis=1, level=level, ascending=ascending)
+    assert_eq(expect, got)
+
+
+@pytest.mark.parametrize("na_position", ["first", "last"])
+@pytest.mark.parametrize("ascending", [True, False])
+def test_sort_index_axis_1_flat_nan_labels_na_position(na_position, ascending):
+    # the plain (no level) axis=1 path previously used python sorted(),
+    # which cannot honor na_position for NaN labels
+    pdf = pd.DataFrame(
+        [[1, 2, 3]], columns=pd.Index(["b", np.nan, "a"], dtype="object")
+    )
+    gdf = cudf.DataFrame(pdf)
+
+    expect = pdf.sort_index(
+        axis=1, ascending=ascending, na_position=na_position
+    )
+    got = gdf.sort_index(axis=1, ascending=ascending, na_position=na_position)
+    assert_eq(expect, got)