Inconsistent comparisons between Series and 2-D ndarray

[cpcloud]

related: #4585

In [5]: randn(10, 2) == Series(randn(2))
Out[5]:
array([[False, False],
       [False, False],
       [False, False],
       [False, False],
       [False, False],
       [False, False],
       [False, False],
       [False, False],
       [False, False],
       [False, False]], dtype=bool)

but

In [6]: randn(10, 2) <= Series(randn(2))

throws an Exception.

got a PR in the works

[jreback]

maybe just do in top of #4585?

[jreback]

@jtratner you might need to add this as a case in your ops refactor....

[jtratner]

Is this just because of the lack of an __rle__ method?

[jreback]

since one doesn't exist currently could be

[jtratner]

In 0.12.0 they both raise a ValueError (same one):

ValueError                                Traceback (most recent call last)
<ipython-input-9-ca6a5964cdf5> in <module>()
----> 1 randn(10, 2) <= Series(randn(2))

/usr/local/lib/python2.7/site-packages/pandas/core/series.pyc in wrapper(self, other)
    224         elif isinstance(other, pa.Array):
    225             if len(self) != len(other):
--> 226                 raise ValueError('Lengths must match to compare')
    227             return Series(na_op(self.values, np.asarray(other)),
    228                           index=self.index, name=self.name)
ValueError: Lengths must match to compare

In current master, one (probably) works correctly [randn(10, 2) == Series(randn(2))], but the other one raises this exception:

Exception                                 Traceback (most recent call last)
<ipython-input-8-ca6a5964cdf5> in <module>()
----> 1 randn(10, 2) <= Series(randn(2))

../pandas/pandas/core/series.pyc in __array_wrap__(self, result)
    789         """
    790         return self._constructor(result, index=self.index, name=self.name,
--> 791                                  copy=False)
    792
    793     def __contains__(self, key):

../pandas/pandas/core/series.pyc in __init__(self, data, index, dtype, name, copy, fastpath)
    655             else:
    656                 data = _sanitize_array(data, index, dtype, copy,
--> 657                                        raise_cast_failure=True)
    658
    659                 data = SingleBlockManager(data, index, fastpath=True)

../pandas/pandas/core/series.pyc in _sanitize_array(data, index, dtype, copy, raise_cast_failure)
   3307     elif subarr.ndim > 1:
   3308         if isinstance(data, pa.Array):
-> 3309             raise Exception('Data must be 1-dimensional')
   3310         else:
   3311             subarr = _asarray_tuplesafe(data, dtype=dtype)

[jreback]

numpy prob let's these go thru, but we have to stop these....

In [5]: Series(range(2)) == np.arange(20).reshape(2,10)
Out[5]: 
0    False
1    False
dtype: bool

[cpcloud]

pushing this to first thing in 0.14

[jbrockmendel]

In master (and 0.23.3) neither of the operations in the OP work; both raise with "Lengths must match to compare".

Is this something we want to support?

[mroeschke]

I think we've tightened up comparison ops with pandas object to ensure similar shape a while back to be more strict and not as flexible as this. Going to close but happy to reopen if there's more interest for this