Tom Lane wrote:
> Martijn van Oosterhout <kleptog@svana.org> writes:
>
>>On Tue, May 20, 2003 at 01:01:01PM +0300, Kaarel wrote:
>>
>>>I'm relatively new to PostgreSQL. I was wondering why were
>>>INSERT/UPDATE/DELETE view diabled in PostgreSQL?
>>
>
>>You can define rules to make insert, update and delete work on views. I'm
>>surprised it was ever enabled by default since in the general case it is
>>impossible to workout what the rules should be.
>
>
> It never was; Kaarel misunderstands the history.
>
> A view used to be an actual table that happened to have an ON SELECT DO
> INSTEAD rule attached to it. That rule did not prevent you from doing
> an INSERT ... but the inserted row went into the physical table. Where
> you could not see it, because the ON SELECT rule would redirect any
> attempt to select it. Likewise you could update and delete such rows,
> but you were flying blind all the time.
This is not entirely accurate ... you forgot that the RIR rules applied
did prevent UPDATE and DELETE to actually find back the rows :-)
>
> This was obviously useless, so we got rid of the physical table
> associated with views.
>
> What should happen, but no one has got round to yet, is to automatically
> generate appropriate ON INSERT, UPDATE, DELETE rules for cases where the
> view structure is simple enough that we can work out exactly what those
> rules ought to be. I believe that such cases correspond pretty closely
> to the rules laid down in the SQL spec for whether a view is updatable.
Which was what - one table select of attributes only, no computation, no
qualification, no aggregates, no groups, no leather, no chrome? Maybe
some day I get bored enough to do that, but it doesn't look like that's
gonna happen within the next two or three years.
Jan
--
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#================================================== JanWieck@Yahoo.com #
