bool() operation on DataFrame should throw ValueError

[dtcaciuc]

DataFrame should behave consistently with Series and NumPy arrays when used in boolean comparison. Additionally, should provide consistent all/any behaviour.

>>> from pandas import *
>>> s = Series([1, 2, 3]) 
>>> bool(s)   
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()
>>> d = DataFrame([[1,2,3]])
>>> bool(d)
True  # should throw ValueError as well
>>> all(s)
True
>>> all(d)
False # should be True

[dtcaciuc]

To clarify, this allows for the following to succeed, which is quite counter-intuitive:

>>> from pandas import *
>>> assert DataFrame([1,2,3]) == DataFrame([4,5,6])

[echlebek]

I think part of the issue is that DataFrame borrows some of its semantics from list, and some from numpy.

bool(dataframe) returns True when it is non-empty, like list. But dataframe_a == dataframe_b returns a new DataFrame of boolean values, whereas list would return a scalar bool. The usual test for this returned data frame, if we were using numpy, would be to call any() on it or all() to observe if some or all of the values are true. But DataFrame does not provide this functionality.

[wesm]

I merged in this API change. Hopefully will not cause too many problems.