BUG: sort_index throws IndexError for some permutations (#26053)

    - (lev - i) will be out of bounds for pop if the first index dim is to be sorted last

Author: jayanthkaturi

pandas/core/indexes/multi.py

@@ -2085,8 +2085,14 @@ def sortlevel(self, level=0, ascending=True, sort_remaining=True):
             shape = list(self.levshape)
 
             # partition codes and shape
-            primary = tuple(codes.pop(lev - i) for i, lev in enumerate(level))
-            primshp = tuple(shape.pop(lev - i) for i, lev in enumerate(level))
+            primary = tuple(codes[lev] for lev in level)
+            primshp = tuple(shape[lev] for lev in level)
+
+            # Reverse sorted to retain the order of
+            # smaller indices that needs to be removed
+            for lev in sorted(level, reverse=True):
+                codes.pop(lev)
+                shape.pop(lev)
 
             if sort_remaining:
                 primary += primary + tuple(codes)

pandas/tests/frame/test_sorting.py

@@ -496,11 +496,28 @@ def test_sort_index_duplicates(self):
     def test_sort_index_level(self):
         mi = MultiIndex.from_tuples([[1, 1, 3], [1, 1, 1]], names=list('ABC'))
         df = DataFrame([[1, 2], [3, 4]], mi)
-        res = df.sort_index(level='A', sort_remaining=False)
-        assert_frame_equal(df, res)
 
-        res = df.sort_index(level=['A', 'B'], sort_remaining=False)
-        assert_frame_equal(df, res)
+        result = df.sort_index(level='A', sort_remaining=False)
+        expected = df
+        assert_frame_equal(result, expected)
+
+        result = df.sort_index(level=['A', 'B'], sort_remaining=False)
+        expected = df
+        assert_frame_equal(result, expected)
+
+        # Error thrown by sort_index when
+        # first index is sorted last (#26053)
+        result = df.sort_index(level=['C', 'B', 'A'])
+        expected = df.iloc[[1, 0]]
+        assert_frame_equal(result, expected)
+
+        result = df.sort_index(level=['B', 'C', 'A'])
+        expected = df.iloc[[1, 0]]
+        assert_frame_equal(result, expected)
+
+        result = df.sort_index(level=['C', 'A'])
+        expected = df.iloc[[1, 0]]
+        assert_frame_equal(result, expected)
 
     def test_sort_index_categorical_index(self):