Daniel Nouri <[EMAIL PROTECTED]>
writes:
> Daniel Nouri wrote:
>> Ross Patterson wrote:
>>> Daniel Nouri <[EMAIL PROTECTED]>
>>> writes:
>>>
>>>> Martin Aspeli wrote:
>>>>
>>>>> Unfortunately, acquisition doesn't work properly inside Python
>>>>> property __get__ accessors. I've seen this pattern before, but
>>>>> I've seen people have various problems with it (that I can't
>>>>> remember the details of right now).
>>>> Hmm, I've been using @property now for quite a long time for
>>>> handling context without any problems. It would be good to have a
>>>> minimal failing example.
>>>>
>>>> I know that formlib actions, which are applied via a decorator,
>>>> have problems with acquisition. But then, they do much more than
>>>> @property.
>>> Here's a doctest showing what I think Martin was talking about:
>>>
>>> <snip>
>>>
>>> >>> foo.aq_parent
>>> <Application at >
>>> >>> foo.has_aq
>>> True
>>> >>> foo.has_aq_descr
>>> True
>>>
>>> Those last two "True"s come out "False" if you actually run this
>>> test.
>>>
>>> It's a problem in all descriptors as far as I can tell. I'd be very
>>> interested in any work arounds for this as I run into it all the
>>> time.
>>>
>>> Also, is there someplace this test might be included? Or is this
>>> considered not to be a bug?
>>
>> Thanks for providing this clear example. I agree that it's a bug.
>>
>> I added a bit to the doctest, just to make sure that the acquisition
>> wrapping of an attribute that's accessed in a descriptor (or
>> property) is preserved:
>>
>> >>> foo.bar = SimpleItem('bar').__of__(foo)
>> >>> foo.bar_property.aq_parent
>> <Foo at />
>> >>> foo.bar_descr.aq_parent
>> <Foo at />
>>
>> For a view, this means that when returning context in a property
>> 'context', the wrapping would be okay. This base class would thus do
>> the "right thing" and also work if Acquisition would be fixed to also
>> work in descriptors:
>
> Some further experiments showed that this is not true. The call to
> 'self.context' will aq wrap the value returned by the context
> descriptor in self. Thus making a context descriptor proves to be
> totally pointless. :)
What I'd really like to know is why acquisition wrappers get stripped
from self in descriptors in the first place. Is it an acquisition
extension class issue? Is ther any way to work around it? Is this
problem known about or otherwise documented by those who guard the
darker secrets of acquisition? :)
Ross
_______________________________________________
Product-Developers mailing list
Product-Developers@lists.plone.org
http://lists.plone.org/mailman/listinfo/product-developers
