[PEAK] sequenceOf should inherit from type?

Phillip J. Eby pje at telecommunity.com
Wed Jun 2 01:00:43 EDT 2004


At 12:33 AM 6/2/04 -0400, Bob Ippolito wrote:
>Logically, wouldn't it make sense if sequenceOf(n) returned an object that 
>inherited from type?
>
>I say this because I had some (ugly, hackish) code that did 
>issubclass(foo.typeObject, SomeClass) and it worked just fine in every 
>case I had, except for sequenceOf, which would throw an exception since 
>it's just an instance (not even a new style one, at that).

Hm.  Well, if I understand your use of typeObject correctly, you may be in 
violation of its contract anyway.  typeObjects are supposed to implement 
one of the model.IType* interfaces (I don't recall the name of the 
appropriate one right off).  So, 'protocols.sequenceOf()' isn't really 
going to cut it for most of what a feature expects of its typeObject.

With respect to protocols being type-ish, I'd have to point out that it's 
not at all a requirement and don't see a reason to complicate Protocol, 
Variation, SequenceProtocol, etc. by mixing in type-ishness or even forcing 
them to be new-style.  I intentionally made the base Protocol class 
"classic" so that you could mix it in to another classic class if you had 
need to, without forcing the result to be new-style.




More information about the PEAK mailing list