MultiIndex Support for DataFrame.pivot

doc/source/user_guide/reshaping.rst

@@ -90,6 +90,19 @@ You can then select subsets from the pivoted ``DataFrame``:
 Note that this returns a view on the underlying data in the case where the data
 are homogeneously-typed.
 
+Now :meth:`DataFrame.pivot` method also supports multiple columns as indexes.
+
+.. ipython:: python
+
+   df1 = pd.DataFrame({'variable1': ['A', 'A', 'A', 'A', 'B', 'B', 'B', 'B'],
+                       'variable2': ['a', 'a', 'b', 'b', 'a', 'a', 'b', 'b'],
+                       'variable3': ['C', 'D', 'C', 'D', 'C', 'D', 'C', 'D'],
+                       'value': np.arange(8)})
+   df1
+
+   df1.pivot(index=['variable1', 'variable2'], columns='variable3',
+            values='value')
+
 .. note::
    :func:`~pandas.pivot` will error with a ``ValueError: Index contains duplicate
    entries, cannot reshape`` if the index/column pair is not unique. In this

doc/source/whatsnew/v0.25.0.rst

@@ -28,6 +28,7 @@ Other Enhancements
 - Indexing of ``DataFrame`` and ``Series`` now accepts zerodim ``np.ndarray`` (:issue:`24919`)
 - :meth:`Timestamp.replace` now supports the ``fold`` argument to disambiguate DST transition times (:issue:`25017`)
 - :meth:`DataFrame.at_time` and :meth:`Series.at_time` now support :meth:`datetime.time` objects with timezones (:issue:`24043`)
+- :meth:`DataFrame.pivot` now supports multiple column indexes by accepting a list of columns (:issue:`21425`)
 - ``Series.str`` has gained :meth:`Series.str.casefold` method to removes all case distinctions present in a string (:issue:`25405`)
 - :meth:`DataFrame.set_index` now works for instances of ``abc.Iterator``, provided their output is of the same length as the calling frame (:issue:`22484`, :issue:`24984`)
 - :meth:`DatetimeIndex.union` now supports the ``sort`` argument. The behaviour of the sort parameter matches that of :meth:`Index.union` (:issue:`24994`)

pandas/core/reshape/pivot.py

@@ -368,18 +368,34 @@ def _convert_by(by):
 @Appender(_shared_docs['pivot'], indents=1)
 def pivot(data, index=None, columns=None, values=None):
     if values is None:
-        cols = [columns] if index is None else [index, columns]
+        # Make acceptable for multiple column indexes.
+        # cols = []
+        if is_list_like(index):
+            cols = index  # cols.extend(index)
+        elif index is not None:
+            cols = [index]
+        else:
+            cols = []
+        cols.append(columns)
+
         append = index is None
         indexed = data.set_index(cols, append=append)
+
     else:
         if index is None:
-            index = data.index
+            index = MultiIndex.from_arrays([data.index, data[columns]])
+        elif is_list_like(index):
+            # Iterating through the list of multiple columns of an index.
+            indexes = [data[column] for column in index]
+            indexes.append(data[columns])
+            index = MultiIndex.from_arrays(indexes)
         else:
+            # Build multi-indexes if index is not None and not a list.
             index = data[index]
-        index = MultiIndex.from_arrays([index, data[columns]])
+            index = MultiIndex.from_arrays([index, data[columns]])
 
         if is_list_like(values) and not isinstance(values, tuple):
-            # Exclude tuple because it is seen as a single column name
+            # Exclude tuple because it is seen as a single column name.
             indexed = data._constructor(data[values].values, index=index,
                                         columns=values)
         else:

pandas/tests/reshape/test_pivot.py

@@ -301,6 +301,33 @@ def test_pivot_multi_functions(self):
         expected = concat([means, stds], keys=['mean', 'std'], axis=1)
         tm.assert_frame_equal(result, expected)
 
+    @pytest.mark.parametrize('method', [True, False])
+    def test_pivot_multiple_columns_as_index(self, method):
+        # adding the test case for multiple columns as index (#21425)
+        df = DataFrame({'lev1': [1, 1, 1, 1, 2, 2, 2, 2],
+                        'lev2': [1, 1, 2, 2, 1, 1, 2, 2],
+                        'lev3': [1, 2, 1, 2, 1, 2, 1, 2],
+                        'values': [0, 1, 2, 3, 4, 5, 6, 7]})
+        data = [[0, 1], [2, 3], [4, 5], [6, 7]]
+        exp_index = pd.MultiIndex.from_product([[1, 2], [1, 2]],
+                                               names=['lev1', 'lev2'])
+        if method:
+            result = df.pivot(index=['lev1', 'lev2'],
+                              columns='lev3',
+                              values='values')
+            exp_columns = Index([1, 2], name='lev3')
+
+        else:
+            result = df.pivot(index=['lev1', 'lev2'],
+                              columns='lev3')
+            exp_columns = MultiIndex(levels=[['values'], [1, 2]],
+                                     codes=[[0, 0], [0, 1]],
+                                     names=[None, 'lev3'])
+
+        expected = DataFrame(data=data, index=exp_index,
+                             columns=exp_columns)
+        tm.assert_frame_equal(result, expected)
+
     @pytest.mark.parametrize('method', [True, False])
     def test_pivot_index_with_nan(self, method):
         # GH 3588