[Zope-CMF] DBObjects and CMF

[Shane Landrum]

On Mon, Jan 26, 2004 at 11:24:12AM -0600, alan runyan wrote:
> >Is anyone out there using the DBObjects relational-to-object
> >mapping product ( 
> >http://www.zope.org/Members/srichter/Products/DBObjects )?
> >Anyone using similar products? If so, how do they compare? I'm looking
> >for some relational-to-object mapping code that will work nicely with
> >my existing large-scale CMF instance.
> 
> I know Ben and Joel were talking about making Archetypes much nicer
> for doing efficient O/R tasks.  But I think it would require funding 
> for it
> to move forward.  Archetypes has a highly inefficient way of doing it 
> now.

I don't know about Archetypes; could you explain?

> You may also want to speak with Ian Bicking.  He wrote SQLObject which
> is a simple O/R system.   He was talking about using Zope2 with it but
> I am not sure how far he got.  We are using SQLObject with Zope3 with
> great results.

Okay, thanks. I'm limited to Zope2 here, so I may send him mail.

> >In particular, I'm looking for strategies people have used for
> >presenting relational-style data as objects within CMF. CMF usage 
> >presents
> >some interesting challenges for this sort of data, particularly
> >around workflow and change tracking, and I'd love to hear other 
> >people's
> >experiences. I've got some data feeds which are very relational,
> >and storing the data directly in Zope seems to be a Bad Idea if
> >I ever want to be able to modify the data in relational-style ways.
> 
> There are really two ways (that I know of):
> 
>   - On-the-fly traversal hooks.  I think something like 
> CMFSubversionBrowser
>   does stuff like this.  This is one level above the ZODB storage (I 
> believe).

Hm, that's not so good for me, because I'm using a placeless asset storage
model; I can't easily rely on traversal. (Unless of course I'm misunderstanding
what I'm reading in the CMFSubversionBrowser CVS code.)

>   - At the storage level - like apelib uses. I think this is really 
> cool.

http://cvs.zope.org/Products/Ape/doc/outline.txt answers some
of my questions about what Ape is and why it's interesting. However,
I'm not clear on what the advantages/drawbacks of its approach are, even
reading Joel Burton's talk notes on the topic.

>   - Having proxies in the ZODB that represent a SQL result.  I believe 
> this is
>   how DBObjects work.  And this is how it works in Archetypes at the 
> moment.
>   This is not a good approach for large scale systems.  I recommend 
> against
>   this approach.

Why doesn't it work nicely for large-scale systems? 

I should mention that I'm not using Zope to serve actual content; rather,
pages get served out from Zope and written to static HTML for scalability
reasons. (It's a lot easier to scale to 100+ million pageviews a month
if we don't have to worry about our Zopes falling over.)

> You might want to check out a talk that Joel Burton did at the plone 
> conference.
> 
> http://plone.org/events/conferences/1/video/
> 
> I do not believe anything he mentions is related to Plone specifically.

Thanks; the talk notes are useful. I'd watch the video of his talk
if I had a player on my Linux box that would handle it.

srl
-- 
Shane Landrum, Software Engineer    srl at boston.com
boston.com / NY Times Digital