Re: Updatable views
Date: Tue, 13 Dec 2005 09:37:26 +0200
Message-ID: <dnltnr$uek$1_at_domitilla.aioe.org>
"Rick Elbers" <rick.elbers_at_chello.nl> wrote in message news:c2trp1ljiugm4j06btj9tpebsimeo91ifk_at_4ax.com...
> >> My question was and is: can somebody prove that this can't be done
> >> in general or can't be done in certain circumstances ?
> >
> >The problems are abiguous actions and circular dependencies.
> Ok. Finally mister x reveals. Ok. Autojoins. Do we agree that an
> autojoin is not really circular like in a recursive algorithm ? That
> an autojoin is the RDBMS imperfect way of talking about a tree of
> trees or about a simple 1-n-n or composite relation ? We express it
> like its relationship infinity but it really never was ment to be ?
> The database way to go 3d ?
There are more problems not just autojoins.
A relation can be constrained by itself in more than one way by multiple
circular paths.
The difference from a recursive algorithm is that there are a finit number
of finite relations.
You must find a way to update those relations in a way that the constraints still stands after the update and the DBMS will not throw errors at you. Keep in mind that some constraints can be enforced in triggers. This is the first problem.
The second one is ambiguous actions. If there are more ways to fulfill an update request on your view, the request is ambiguous and you risk to peek the wrong one. If you know some formal language theory maybe you have heard of ambiguous grammars.
Other problem is that the simple algorithms you might find to solve those problems might be complex. You know: combinatorial explosion ...
As for the database way to go 3d, I only heard of this "3d" thing in geometry. What is the geometry of data ? Do you have a definition ? If the data in a database is finite, then it always can be put in one to one corespondence with a finite subset of natural numbers. Received on Tue Dec 13 2005 - 08:37:26 CET