BUG: allow empty multiindex (fixes .isin regression, GH16777) (pandas…

…-dev#16782)

(cherry picked from commit 500cd0f)

doc/source/whatsnew/v0.20.3.txt

@@ -55,6 +55,7 @@ Indexing
 ^^^^^^^^
 
 - Bug in ``Float64Index`` causing an empty array instead of ``None`` to be returned from ``.get(np.nan)`` on a Series whose index did not contain any ``NaN`` s (:issue:`8569`)
+- Bug in ``MultiIndex.isin`` causing an error when passing an empty iterable (:issue:`16777`)
 
 I/O
 ^^^

pandas/core/indexes/multi.py

@@ -1134,10 +1134,11 @@ def from_tuples(cls, tuples, sortorder=None, names=None):
                                   of iterables
         """
         if len(tuples) == 0:
-            # I think this is right? Not quite sure...
-            raise TypeError('Cannot infer number of levels from empty list')
-
-        if isinstance(tuples, (np.ndarray, Index)):
+            if names is None:
+                msg = 'Cannot infer number of levels from empty list'
+                raise TypeError(msg)
+            arrays = [[]] * len(names)
+        elif isinstance(tuples, (np.ndarray, Index)):
             if isinstance(tuples, Index):
                 tuples = tuples._values
 
@@ -2626,8 +2627,9 @@ def _wrap_joined_index(self, joined, other):
     @Appender(Index.isin.__doc__)
     def isin(self, values, level=None):
         if level is None:
-            return algos.isin(self.values,
-                              MultiIndex.from_tuples(values).values)
+            values = MultiIndex.from_tuples(values,
+                                            names=self.names).values
+            return algos.isin(self.values, values)
         else:
             num = self._get_level_number(level)
             levs = self.levels[num]

pandas/tests/indexes/test_multi.py

@@ -1720,6 +1720,13 @@ def test_from_tuples(self):
         idx = MultiIndex.from_tuples(((1, 2), (3, 4)), names=['a', 'b'])
         assert len(idx) == 2
 
+    def test_from_tuples_empty(self):
+        # GH 16777
+        result = MultiIndex.from_tuples([], names=['a', 'b'])
+        expected = MultiIndex.from_arrays(arrays=[[], []],
+                                          names=['a', 'b'])
+        tm.assert_index_equal(result, expected)
+
     def test_argsort(self):
         result = self.index.argsort()
         expected = self.index.values.argsort()