[Tutor] Is there a test for hashability?

[Richard D. Moores]

On Thu, Sep 1, 2011 at 18:08, Steven D'Aprano <steve at pearwood.info> wrote:
> Richard D. Moores wrote:
>>
>> I'm trying to write a general test for hashability. How can I test if
>> an object has both a  __hash__() method and an __eq__() method?
>
>
> Just because an object has a __hash__ method doesn't mean it is guaranteed
> to be hashable. The method might (deliberately, or accidentally) fail and
> raise an exception.
>
>>>> t = (1, 2, 3)
>>>> t.__hash__
> <method-wrapper '__hash__' of tuple object at 0xb7f09fcc>
>>>> hash(t)
> -378539185
>
> But:
>
>>>> t = (1, 2, [3])
>>>> t.__hash__
> <method-wrapper '__hash__' of tuple object at 0xb7f0c6e4>
>>>> hash(t)
> Traceback (most recent call last):
>  File "<stdin>", line 1, in <module>
> TypeError: list objects are unhashable
>
>
> The only effective way to find out if an object is hashable is to try it and
> see.
>
> Even more effective is to avoid trying to find out whether it is hashable,
> and just use it, as required, and deal with the error if it isn't. This is
> the "Easy to get forgiveness than to ask permission" model of error
> handling, as opposed to "Look before you leap".
>
> For various reasons, LBYL can be risky... like in Quantum Mechanics, the
> very act of *looking* at an object might change its behaviour. (Methods can
> have side-effects.) So, even if is_hashable(obj) returns True, you can't
> *quite* be 100% certain that mydict[obj] will succeed until you try it.

Steven,

Do you have an actual example of an actual object for which "my"
is_hashable(object) function would return the wrong answer?

>>> def is_hashable(object):
...     try:
...         hash(object)
...     except TypeError:
...         return False
...     return True
...
>>> is_hashable((1, 2, [3]))
False
>>>

Is that one?

Dick