Re: avoid circular references - Mailing list pgsql-sql

From Jerome Alet
Subject Re: avoid circular references
Date
Msg-id 20031106061807.GF17294@mail.librelogiciel.com
Whole thread Raw
In response to Re: avoid circular references  (Josh Berkus <josh@agliodbs.com>)
List pgsql-sql
On Wed, Nov 05, 2003 at 09:15:39PM -0800, Josh Berkus wrote:
> Jerome,
> 
> > Each object can contain one or many other objects from the same
> > table, so I guess this relationship can be expressed with something
> > like :
> 
> This is called a "tree structure".

Thanks !

> >   CREATE TABLE contains (id SERIAL PRIMARY KEY NOT NULL,
> >     containerid INTEGER REFERENCES objtable(idobj),
> >       contentid INTEGER REFERENCES objtable(idobj) );
> 
> This is overcomplicating things, unless a single object can belong to multiple 
> containers, which would strike me as peculiar.  Easier just to have a 
> "containerid" in the objtable, which is your basic Proximity List.
> 
> But of course, maybe I'm not understanding you and you do have multiple 
> inheritance.

Yes an object could theorically belong to multiple containers, so
that's not really a tree I suppose.

> > What I would like is to be able to avoid circular references :
> >
> >         - an object can't contain itself.
> >
> >         - an object can't contain one of its containers.
> >
> > So I'd like to know how to create additionnal integrity constraints
> > to solve this problem, and especially what syntax I should use.
> 
> Best to use some kind of recursive function.  I do this for a calendaring 
> setup with event templates, where events can have multiple (possible) parents 
> and multiple children.  Just write a pl/pgSQL function which reverse-traces 
> the parentage of the new object, looking for copies of itself.

Any example for this, especially on how to raise an integrity 
exception ? 

> > I thought about creating a rule but maybe the best is a trigger
> > (I'm not sure I really understand the difference) if I see
> > what happens when I create new tables. But what can I do in
> > my trigger to have PostgreSQL understand there's an integrity
> > violation ?
> 
> A trigger.   Just use a BEFORE trigger and raise an exception if a self-parent 
> is found.

Thanks for your help !

Jerome Alet
-- 
"A non-free program is a predatory social system that keeps people 
in a state of domination and division, and uses the spoils to 
dominate more." - RMS


pgsql-sql by date:

Previous
From: Josh Berkus
Date:
Subject: Re: avoid circular references
Next
From: "sTesting"
Date:
Subject: Now() in a function