Re: views of binary operations

Cimode wrote:
> [snipped]
> erk wrote:
> > Views and snapshots aren't unique to SQL. They're fundamental to
> > relational.
> I understand and agree but views and snapshots are not *always*
> necessary to understand RM issues. Marshall's question seems to
> directly point out directly to inter relation assignment rather than
> views or snapshots.

I think there are certainly relationships between the concepts. The effect of an assignment in a relational languages hinges on views and snapshots. A view can at any time, in any statement or expression, be replaced by the relational expression it designates. It's nothing but a shortcut. Snapshots in assignments are different, of course.

> > A simple rule of thumb might be this: A view is an expression,
> > unevaluated, while an assignment necessarily requires a value, thus a
> > snapshot. Special syntax is needed to differentiate a view creation
> > ("view(r) = A op B" or perhaps "view r = A op B") from a simple
> > assignment of a relval to a relvar ("r = A op B").
>
> As stated previously, I am not convinced that r = A op B can not
> logically lead to view(r) = A op B because it supposes r = view(r)
> which is a shorcut leading to an absurdity which is that a relation is
> the same as a view defined on the relation. OTOH view(r) = view(A op
> B) would be acceptable.

I agree, but I don't think Marshall was saying that. I think part of the confusion is that r and view(r) are different types entirely; view(r) is not a view OF r, but simply a different variable r that happens to be a view. It would probably have been better to call it v to avoid the confusion.

Marshall, am I mistaken in the above?

• erk