Add levels property for MultiIndex

databricks/koalas/indexes.py

@@ -571,6 +571,32 @@ def __getattr__(self, item: str) -> Any:
     def rename(self, name, inplace=False):
         raise NotImplementedError()
 
+    @property
+    def levels(self) -> list:
+        """
+        Names of index columns in list.
+
+        Examples:
+        --------
+        >>> mi = pd.MultiIndex.from_arrays((list('abc'), list('def')))
+        >>> mi.names = ['level_1', 'level_2']
+        >>> kdf = ks.DataFrame({'a': [1, 2, 3]}, index=mi)
+        >>> kdf.index.levels
+        [['a', 'b', 'c'], ['d', 'e', 'f']]
+
+        >>> mi = pd.MultiIndex.from_arrays((list('bac'), list('fee')))
+        >>> mi.names = ['level_1', 'level_2']
+        >>> kdf = ks.DataFrame({'a': [1, 2, 3]}, index=mi)
+        >>> kdf.index.levels
+        [['a', 'b', 'c'], ['e', 'f']]
+        """
+        scols = self._kdf._internal.index_scols
+        sdf = self._kdf._sdf.select([F.collect_set(scol) for scol in scols]).collect()[0]
+
+        # use sorting is because pandas doesn't care the appearance order of level
+        # names, so e.g. if ['b', 'd', 'a'] will return as ['a', 'b', 'd']
+        return [sorted(col) for col in sdf]
+
     def __repr__(self):
         max_display_count = get_option("display.max_rows")
         if max_display_count is None:

databricks/koalas/missing/indexes.py

@@ -128,7 +128,6 @@ class _MissingPandasLikeMultiIndex(object):
     T = unsupported_property('T')
     codes = unsupported_property('codes')
     is_all_dates = unsupported_property('is_all_dates')
-    levels = unsupported_property('levels')
     levshape = unsupported_property('levshape')
     shape = unsupported_property('shape')
 

databricks/koalas/tests/test_indexes.py

@@ -283,3 +283,15 @@ def test_multiindex_nlevel(self):
         kdf = ks.from_pandas(pdf)
 
         self.assertEqual(kdf.index.nlevels, 2)
+
+    def test_multiindex_levels(self):
+        tuples = [[list('abc'), list('def')], [list('aac'), list('fed')]]
+
+        for tup in tuples:
+            pdf = pd.DataFrame({'a': [1, 2, 3]}, index=tup)
+            kdf = ks.from_pandas(pdf)
+
+            # pandas returns FronzeList, so need to convert it to normal list
+            # for comparison
+            pdf_levels = [list(i) for i in pdf.index.levels]
+            self.assertEqual(pdf_levels, kdf.index.levels)