Re: On view updating
Date: Mon, 20 Sep 2004 19:04:52 +0200
Message-ID: <414f0dbd$0$65124$e4fe514c_at_news.xs4all.nl>
Laconic2 wrote:
> Tony Andrews wrote:
>
>>My feeling is that views are NOT equivalent to base tables in all >>cases, and therefore it is unsurprising that they can't be used as if >>they were base tables in all cases.
>
> You make several good points. With regard to this last point, I think that
> it goes to the heart of what a "view" is.
> It's a query, and it's a virtual table. It's not surprising that some
> queries work for update, and some do not.
>
> Maybe views need more definition... In addition to providing a "SELECT"
> statement, maybe it should provide "UPDATE", "INSERT", and "DELETE"
> statements.
This is beginning to sound an awful lot like IDMS' (A Codasyl DBMS by Cullinet, now CA) LRF: The Logical Record Facility. It's been 10 years since I last saw it, so I don't know if it developed further (but I doubt it). Subschema's would specify so-called Logical Records and what to do on Obtain, Modify, Store and Erase actions on such a LR.
Instead-of triggers come to mind. A condition here would of course be that the base tables should *not* be visible directly at all.
> These statements could work somewhat like triggers, in the
> sense that they would be activated by circumstances. In some cases, the
> "UPDATE", "INSERT" and "DELETE" are redundant, because they can be inferred
> from the SELECT. In other cases, they can't be inferred, and have to be
> explicit.
> I'm just playing around with this idea. Maybe I'll reject it after further
> consideration.
It would be nice if you could do some of your pondering out in the open :-) Received on Mon Sep 20 2004 - 19:04:52 CEST