Re: On view updating

On Mon, 20 Sep 2004 06:03:24 -0700, Tony Andrews wrote:
>
> In this particular case I think the theorists have been over-scornful
> of the SQL vendors (perhaps out of habit!)

If you mean by "the theorists" those people that work as scientists in the field of database theory and publish in ICDT, SIGMOD, ICDE et cetera, then you couldn't be further from the truth. Those people think by no means that this is a solved problem.

> Some clear problems with updating views:
> 1) the transformation made by a view is often one way: you can't be
> sure what is in the base tables to achieve the result you see from the
> view.
> 2) views, unlike base tables, are often not even in 2NF - hence
> updating views could lead to "update anomalies": if you delete the last
> v_employee then you also delete his department.

These are not inherent problems of updating views. You can define perfectly when a view does not have such problems and then you can attempt to find the corresponding syntactic rules. It's easy to find a set of rules that are so strict that this is guaranteed, and it is also well known that finding an exact syntactic characterization is impossible, but finding the optimal set of rules inbetween those two extremes is an unsolved problem.

• Jan Hidders