BUG: Bug in multi-index slicing with missing indexers (GH7866)

Author: jreback

doc/source/v0.15.0.txt

@@ -274,7 +274,7 @@ Bug Fixes
 - Bug in ``DatetimeIndex`` and ``PeriodIndex`` in-place addition and subtraction cause different result from normal one (:issue:`6527`)
 - Bug in adding and subtracting ``PeriodIndex`` with ``PeriodIndex`` raise ``TypeError`` (:issue:`7741`)
 - Bug in ``combine_first`` with ``PeriodIndex`` data raises ``TypeError`` (:issue:`3367`)
-
+- Bug in multi-index slicing with missing indexers (:issue:`7866`)
 
 
 - Bug in pickles contains ``DateOffset`` may raise ``AttributeError`` when ``normalize`` attribute is reffered internally (:issue:`7748`)

pandas/core/index.py

@@ -3638,9 +3638,17 @@ def _convert_indexer(r):
                 ranges.append(k)
             elif com.is_list_like(k):
                 # a collection of labels to include from this level (these are or'd)
-                ranges.append(reduce(
-                    np.logical_or,[ _convert_indexer(self._get_level_indexer(x, level=i)
-                                                     ) for x in k ]))
+                indexers = []
+                for x in k:
+                    try:
+                        indexers.append(_convert_indexer(self._get_level_indexer(x, level=i)))
+                    except (KeyError):
+
+                        # ignore not founds
+                        continue
+
+                ranges.append(reduce(np.logical_or,indexers))
+
             elif _is_null_slice(k):
                 # empty slice
                 pass

pandas/tests/test_indexing.py

@@ -1941,6 +1941,39 @@ def f():
             df.val['X']
         self.assertRaises(KeyError, f)
 
+
+        # GH 7866
+        # multi-index slicing with missing indexers
+        s = pd.Series(np.arange(9),
+                      index=pd.MultiIndex.from_product([['A','B','C'],['foo','bar','baz']],
+                                                       names=['one','two'])
+                      ).sortlevel()
+
+        expected = pd.Series(np.arange(3),
+                             index=pd.MultiIndex.from_product([['A'],['foo','bar','baz']],
+                                                              names=['one','two'])
+                             ).sortlevel()
+
+        result = s.loc[['A']]
+        assert_series_equal(result,expected)
+        result = s.loc[['A','D']]
+        assert_series_equal(result,expected)
+
+        # empty series
+        result = s.loc[['D']]
+        expected = s.loc[[]]
+        assert_series_equal(result,expected)
+
+        idx = pd.IndexSlice
+        expected = pd.Series([0,3,6],
+                             index=pd.MultiIndex.from_product([['A','B','C'],['foo']],
+                                                              names=['one','two'])
+                             ).sortlevel()
+        result = s.loc[idx[:,['foo']]]
+        assert_series_equal(result,expected)
+        result = s.loc[idx[:,['foo','bah']]]
+        assert_series_equal(result,expected)
+
     def test_setitem_dtype_upcast(self):
 
         # GH3216