[TransWarp] Re: What happens/happened with AOP in PEAK

[Phillip J. Eby]

At 01:01 PM 11/26/02 +0100, jan.van.hees at nl.abnamro.com wrote:
>I was very enthousisatic about AOP. Do you still apply the concepts from
>AOP within PEAK but do you keep it under the hood or don't you support the
>AOP concepts anymore?

It's still there: module inheritance.  See the docs for 
peak.config.modules.  I've actually been working on it quite a bit lately, 
trying to improve the performance.  It currently takes around 3-4 seconds 
to process merging the UML metamodel module with another module to add in a 
domain logic aspect, and that's too bloody long in my opinion.  So I think 
I'm going to end up having to rewrite some of it in C (or at least 
Pyrex).  Interestingly, this'll be only the second time in five years of 
Python programming that I've needed to go to C for performance reasons.


>I alsoo miss the XMI interface.

See 'peak.model.xmi'.  Although that's about to move to 'peak.storage.xmi', 
possibly today.  I've added support for XMI 1.1, and think perhaps I'll 
have support for writing out XMI files by the end of the year.  The XMI 
support will be for arbitrary metamodels, not just UML and MOF, so you'll 
be able to save or load application databases from XMI, if you create their 
metamodel using 'peak.model' Element and Feature classes.