BUG: unique on an empty MI

[jreback]

In [13]: pd.MultiIndex.from_arrays([[], []])
Out[13]: 
MultiIndex(levels=[[], []],
           labels=[[], []])

In [14]: pd.MultiIndex.from_arrays([[], []]).unique()
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-14-d5be3b23b8c3> in <module>()
----> 1 pd.MultiIndex.from_arrays([[], []]).unique()

~/pandas/pandas/core/indexes/multi.py in unique(self, level)
   1071 
   1072         if level is None:
-> 1073             return super(MultiIndex, self).unique()
   1074         else:
   1075             level = self._get_level_number(level)

~/pandas/pandas/core/indexes/base.py in unique(self, level)
   4366             self._validate_index_level(level)
   4367         result = super(Index, self).unique()
-> 4368         return self._shallow_copy(result)
   4369 
   4370     def drop_duplicates(self, keep='first'):

~/pandas/pandas/core/indexes/multi.py in _shallow_copy(self, values, **kwargs)
    558             # discards freq
    559             kwargs.pop('freq', None)
--> 560             return MultiIndex.from_tuples(values, **kwargs)
    561         return self.view()
    562 

~/pandas/pandas/core/indexes/multi.py in from_tuples(cls, tuples, sortorder, names)
   1315             if names is None:
   1316                 msg = 'Cannot infer number of levels from empty list'
-> 1317                 raise TypeError(msg)
   1318             arrays = [[]] * len(names)
   1319         elif isinstance(tuples, (np.ndarray, Index)):

TypeError: Cannot infer number of levels from empty list

[jreback]

xref #20544 (comment)

cc @WillAyd
@toobaz

[toobaz]

Duplicate of #20308