Re: the case for machine-readable error fields

On Thu, Aug 06, 2009 at 11:41:55AM +0200, Pavel Stehule wrote:
> typically in SQL/PSM (stored procedures - look on GET DIAGNOSTICS
> statement in plpgsql doc), maybe in ecpg. Other's environments raise
> exception - so you can get some data from exception or from special
> structures related to environment - php, ruby, .NET etc

Sorry, I should have said that I saw how it was used in stored
procedures. My interest was in getting the client doing something
interesting, if you've already got the complexity of a stored procedure
it shouldn't be to hard to teach it where the problem is.

One thing I didn't see any comment on was on the fact that I think
CREATE UNIQUE INDEX is really creating a constraint--it's just not
showing up as one. For the constraint name to be sent back in the case
of an error I think this needs to be changed.

Triggers (and other domain specific code) seem less important here as
they can always fail with whatever error is appropriate.

--
Sam http://samason.me.uk /