[Zope3-dev] Re: a new zcml directive?

[Martijn Faassen]

Jim Fulton wrote:
> Philipp von Weitershausen wrote:
> 
>> Marius Gedminas wrote:
>>
>>> I'd prefer
>>>
>>>    from zope.annotation.adapter import AnnotationAdapter
>>>
>>>    getFoo = AnnotationAdapter(for_=IBar,
>>>                               interface=IFoo,
>>>                               factory=Foo,
>>>                               key=FOO_KEY)
>>>    # I suppose the key could be optional; you could use a dotted
>>>    # interface name by default
>>>
>>> and then the ordinary
>>>
>>>    <zope:adapter factory=".foo.getFoo" />
>>
>>
>>
>> +10
>  
> Likewise.

I just tried to write AnnotationAdapter. Unfortunately, I haven't gotten 
very far yet, as it turns out I need to understand the internals of the 
implementation of the zope:adapter ZCML statement.

zope:adapter has a bunch of automation to look for the interface of 
something and what it adapts on the factory. This works if the factory 
is the class itself, but in this case, the factory is a class which has 
instances which can provide an instance of the annotation class when called.

So, I need to come up with trickery so that instances of 
AnnotationAdapter return the 'for_' argument when asked for by 
zope.interface.implementedBy() and the 'factory' argument when called 
with component.adaptedBy(). Since the instances of AnnotationAdapter are 
actually not doing any adaptation or implementing any of the interface, 
but are actually just providing a factory, it seems like this trickery 
would amount to actually lying.

This is hairy enough for me to give up for now.

Unless I'm missing something, I conclude that implementing 
AnnotationAdapter is currently at least as difficult as just 
implementing a new directive. The other conclusion is that figuring out 
what is happening in case something goes wrong is also at least as hard 
to understand as the innards of a directive, due to the trickery 
necessary to make it work.

Using <zope:adapter..> with an explicit 'for' and 'provides' would avoid 
the problem, but this seems to be going against the suggestions given in 
this thread.

Regards,

Martijn