[ZPatterns] ZPatterns and...CMF???

Phillip J. Eby pje@telecommunity.com
Sat, 10 Nov 2001 21:19:32 -0500


At 07:25 PM 11/10/01 -0500, Gary Poster wrote:
>Hello all.  So, I've become somewhat enamored of CMF.  Integrating my RDBMS
>with my CMF via ZPatterns suddenly seemed like a fantastic idea in theory,
>so I've been looking around to see what the Zope world has been doing in
>this general direction before I, in a fit of enthusiasm, wasted days of
>thought and work that people smarter than me have already done and perhaps
>discarded.  It seemed all I would have to do is make the ZPatterns ZClasses
>have the correct CMF mix-in classes and build the mapping to view, edit, and
>metadata forms.
>
>But nobody has said a darn thing I can find about doing ZPatterns with CMF,
>except ...
>http://zope.nipltd.com/public/lists/dev-archive.nsf/ByKey/E6526C76DA28C391
>... Shane:
> > Let's say you want CMF objects to be stored in an RDBMS.  You don't want
> > to use ZODB stubs and you want a tree of persistent objects.  ZPatterns
> > can't help you.  Can SmartObjects?
>
>(no answer I can see to the final SmartObjects question, btw).  Why does
>Shane write off the ZPatterns solution?  I suppose I could ask him,
>but...I'm asking here first. :-)

Because ZPatterns as it exists today would require ZODB stubs to do what is 
described, at least based on what you've quoted above.

Actually, it's not 100% true that ZPatterns can't help.  You *could* do it, 
but you'd need to hack the base classes (primarily ObjectManager or the 
equivalent).  It's messy.  At one point I contemplated providing the 
necessary machinery for this in DataSkins, but it's just continuing the 
direction of kludging and hacking things that I feel has made ZPatterns an 
evolutionary dead-end.   I've had enough trouble keeping ZPatterns' hacks 
current to the ever-more-frequent Zope releases, without trying to add any 
new capabilities.

The new Zope component architecture is supposed to fix all this; in 
practice it will not address something like storing CMF objects in an RDBMS 
for a long while if ever, so long as the issue is one of supporting all the 
possible third-party classes deriving from CMF bases.  To fix a problem 
like this requires aspect-weaving or other "open implementation" 
approaches, such as TransWarp provides.  (Which is another reason I gave up 
some time ago on doing any extensive improvements to ZPatterns.)