[Zope3-dev] Re: Recursive containment constraints

[Jim Fulton]

Tres Seaver wrote:
> Stephan Richter wrote:
> 
>> On Wednesday 15 September 2004 09:01, Dominik Huber wrote:
>>
>>> Month ago I read the contenttype part of stephan's book. at that time he
>>> used
>>> setTaggedValue to solve recursive containment constraints:
>>>
>>>    43  IMessage['__setitem__'].setTaggedValue('precondition',
>>>    44
>>> ItemTypePrecondition(IMessage))
>>>    45  IMessage.setTaggedValue('__parent__', Field(
>>>    46      constraint=ContainerTypesConstraint(IMessageBoard,
>>> IMessage)))
>>>
>>> Today I wanted to use the setTaggedValue('__parent__', Field(...))
>>> approach but it
>>> seems not work anymore. Is that right?
>>
>>
>>
>> No, it should still work. I did not change to another pattern for 
>> technical reasons.
>>
>>
>>>    >>> '__parent__' in IMessage.names('all')
>>>    False
>>>
>>> The new version of the book solves this problem with two indepentent
>>> interfaces IMessage
>>> and IMessageContained. Is that the 'official' pattern people should
>>> follow?
>>
>>
>>
>> Yeah, I think this is the recommended pattern. I changed to this one, 
>> since it is easier to read and understand. Also, all sample 
>> applications use this pattern as well.
> 
> 
> In reality, I think we would be better off moving the constraint 
> declarations out of the classes;

You mean "interfaces" or "class statements" :)

 >  that would both remove the "forward
> reference" problem and make it easy to move to a "containment constraint 
> registry" solution when that became available.
> 
> My biggest reason for wanting this is that the constraints are *policy*, 

They often are, but aren't always.

But, I agree, that we almost certainly want to move the constraints out
of the interfaces.

Jim

-- 
Jim Fulton           mailto:jim at zope.com       Python Powered!
CTO                  (540) 361-1714            http://www.python.org
Zope Corporation     http://www.zope.com       http://www.zope.org