BUG: Let MultiIndex.set_levels accept any iterable (#23273) (#23291)

Author: imankulov

doc/source/whatsnew/v0.24.0.txt

@@ -1091,6 +1091,7 @@ Indexing
 - Bug in :meth:`DataFrame.loc` when indexing with an :class:`IntervalIndex` (:issue:`19977`)
 - :class:`Index` no longer mangles ``None``, ``NaN`` and ``NaT``, i.e. they are treated as three different keys. However, for numeric Index all three are still coerced to a ``NaN`` (:issue:`22332`)
 - Bug in `scalar in Index` if scalar is a float while the ``Index`` is of integer dtype (:issue:`22085`)
+- Bug in `MultiIndex.set_levels` when levels value is not subscriptable (:issue:`23273`)
 
 Missing
 ^^^^^^^

pandas/core/indexes/multi.py

@@ -389,6 +389,9 @@ def set_levels(self, levels, level=None, inplace=False,
                    labels=[[0, 0, 1, 1], [0, 1, 0, 1]],
                    names=[u'foo', u'bar'])
         """
+        if is_list_like(levels) and not isinstance(levels, Index):
+            levels = list(levels)
+
         if level is not None and not is_list_like(level):
             if not is_list_like(levels):
                 raise TypeError("Levels must be list-like")

pandas/tests/indexes/multi/test_get_set.py

@@ -414,3 +414,17 @@ def test_set_value_keeps_names():
     df.at[('grethe', '4'), 'one'] = 99.34
     assert df._is_copy is None
     assert df.index.names == ('Name', 'Number')
+
+
+def test_set_levels_with_iterable():
+    # GH23273
+    sizes = [1, 2, 3]
+    colors = ['black'] * 3
+    index = pd.MultiIndex.from_arrays([sizes, colors], names=['size', 'color'])
+
+    result = index.set_levels(map(int, ['3', '2', '1']), level='size')
+
+    expected_sizes = [3, 2, 1]
+    expected = pd.MultiIndex.from_arrays([expected_sizes, colors],
+                                         names=['size', 'color'])
+    tm.assert_index_equal(result, expected)