Re: More on view updates and inverse views

Walter Mitty wrote:
...
>
> PMFJI. I'm trying to combine this latest response with what Brian said in
> his latest response (about permitting NULLs in some attributes and
> providing default values in others.) Let me duck the issue of NULLS at
> least for the moment. Let's say that all attributes the base relvars and
> not in the projection have a default value. This makes the projection view
> "invertable" even though without the default value it would not be
> invertable.
>
>
> So now my question is, can business rules such as default values (in the
> case of insert into projection) or triggers that implement a delete from
> join predictably and uniquely be expressed as "equations"? Is it possible
> that a system of equations that describe views in terms of base relvars and
> also describe business rules, including constraints, might always be
> "invertable". Well, it's more complicated than what I ve expressed, but I
> don't want to get all tied up in knots. The point is to ask whether there's
> always a way to make a non invertable system invertable by adding more
> rules.
> ...

Well I think if one wants to avoid nulls (good!), using default values might need so much catering in apps that what Vadim says would make much more sense, if I understand him, define base projections.

I'm not against that because it is so apt, ie., if there is a need to insert to projections, then there is a need for base projections.

For me, the big updating problems are join delete and union insert, the only reason I brought up projection insert was that I'm curious about other forms besides conventional tables and relvars. Also, if the world ever sees join delete and union insert straight, which I don't think it does now, there might be interesting implications (I mean interesting in the sense of applications profiting). Received on Wed Sep 09 2009 - 00:55:23 CEST