BUG: DataFrame.stack(..., dropna=False) with partial MultiIndex.

doc/source/whatsnew/v0.15.2.txt

@@ -113,7 +113,8 @@ Bug Fixes
 
 - Bug where ``get_data_google``returned object dtypes (:issue:`3995`)
 
-
+- Bug in ``DataFrame.stack(..., dropna=False)`` when the DataFrame's ``columns`` is a ``MultiIndex``
+  whose ``labels`` do not reference all its ``levels``. (:issue:`8844`)
 
 
 

pandas/core/reshape.py

@@ -648,7 +648,9 @@ def _convert_level_number(level_num, columns):
     # time to ravel the values
     new_data = {}
     level_vals = this.columns.levels[-1]
-    levsize = len(level_vals)
+    level_labels = sorted(set(this.columns.labels[-1]))
+    level_vals_used = level_vals[level_labels]
+    levsize = len(level_labels)
     drop_cols = []
     for key in unique_groups:
         loc = this.columns.get_loc(key)
@@ -661,7 +663,7 @@ def _convert_level_number(level_num, columns):
         elif slice_len != levsize:
             chunk = this.ix[:, this.columns[loc]]
             chunk.columns = level_vals.take(chunk.columns.labels[-1])
-            value_slice = chunk.reindex(columns=level_vals).values
+            value_slice = chunk.reindex(columns=level_vals_used).values
         else:
             if frame._is_mixed_type:
                 value_slice = this.ix[:, this.columns[loc]].values
@@ -685,7 +687,7 @@ def _convert_level_number(level_num, columns):
         new_names = [this.index.name]  # something better?
 
     new_levels.append(frame.columns.levels[level_num])
-    new_labels.append(np.tile(np.arange(levsize), N))
+    new_labels.append(np.tile(level_labels, N))
     new_names.append(frame.columns.names[level_num])
 
     new_index = MultiIndex(levels=new_levels, labels=new_labels,

pandas/tests/test_frame.py

@@ -12266,6 +12266,56 @@ def test_stack_datetime_column_multiIndex(self):
         expected = DataFrame([1, 2, 3, 4], index=eidx, columns=ecols)
         assert_frame_equal(result, expected)
 
+    def test_stack_partial_multiIndex(self):
+        # GH 8844
+        def _test_stack_with_multiindex(multiindex):
+            df = DataFrame(np.arange(3 * len(multiindex)).reshape(3, len(multiindex)),
+                           columns=multiindex)
+            for level in (-1, 0, 1, [0, 1], [1, 0]):
+                result = df.stack(level=level, dropna=False)
+
+                if isinstance(level, int):
+                    # Stacking a single level should not make any all-NaN rows,
+                    # so df.stack(level=level, dropna=False) should be the same
+                    # as df.stack(level=level, dropna=True).
+                    expected = df.stack(level=level, dropna=True)
+                    if isinstance(expected, Series):
+                        assert_series_equal(result, expected)
+                    else:
+                        assert_frame_equal(result, expected)
+
+                df.columns = MultiIndex.from_tuples(df.columns.get_values(),
+                                                    names=df.columns.names)
+                expected = df.stack(level=level, dropna=False)
+                if isinstance(expected, Series):
+                    assert_series_equal(result, expected)
+                else:
+                    assert_frame_equal(result, expected)
+
+        full_multiindex = MultiIndex.from_tuples([('B', 'x'), ('B', 'z'),
+                                                  ('A', 'y'),
+                                                  ('C', 'x'), ('C', 'u')],
+                                                 names=['Upper', 'Lower'])
+        for multiindex_columns in ([0, 1, 2, 3, 4],
+                                   [0, 1, 2, 3], [0, 1, 2, 4],
+                                   [0, 1, 2], [1, 2, 3], [2, 3, 4],
+                                   [0, 1], [0, 2], [0, 3],
+                                   [0], [2], [4]):
+            _test_stack_with_multiindex(full_multiindex[multiindex_columns])
+            if len(multiindex_columns) > 1:
+                multiindex_columns.reverse()
+                _test_stack_with_multiindex(full_multiindex[multiindex_columns])
+
+        df = DataFrame(np.arange(6).reshape(2, 3), columns=full_multiindex[[0, 1, 3]])
+        result = df.stack(dropna=False)
+        expected = DataFrame([[0, 2], [1, nan], [3, 5], [4, nan]],
+                             index=MultiIndex(levels=[[0, 1], ['u', 'x', 'y', 'z']],
+                                              labels=[[0, 0, 1, 1], [1, 3, 1, 3]],
+                                              names=[None, 'Lower']),
+                             columns=Index(['B', 'C'], name='Upper'),
+                             dtype=df.dtypes[0])
+        assert_frame_equal(result, expected)
+
     def test_repr_with_mi_nat(self):
         df = DataFrame({'X': [1, 2]},
                        index=[[pd.NaT, pd.Timestamp('20130101')], ['a', 'b']])