From a design standpoint, CHECK constraints are better than triggers in
general, and if it is possible to do what you want in a CHECK constraint, you
should choose that option, and leave triggers for other tasks that you can't use
CHECK for.
Or to put this a different way, triggers are the proper way to do
non-deterministic things such as constraints or actions that depend on the
current system time, while CHECK is better for deterministic things such as
constraints a database should always have that only require looking at the
contents of the database itself.
-- Darren Duncan
Gauthier, Dave wrote:
> Ya, I know I could pass them in. Just wanted to know if they were already floating around out there as globals or
something. That way I wouldn't have to drop/recreate the constraint itself.
>
> These would be constraint violations where the name of the constraint being violated is in the returned message
(shouldthere be a violation). I named the constraints to identify the column and describe the nature of the problem.
Notsure how to capture that info if the trigger disallowed the value.
>
> Also, I need to defer constraint checking for some transactions. Making the trigger sensitive to this deferral is
notsomething that I wanted to devote a lot of time to. It just seemed more natural to keep this as a check constraint.
>
> -----Original Message-----
> From: Jerry Sievers [mailto:gsievers19@comcast.net]
> Sent: Thursday, March 31, 2011 4:14 PM
> To: Gauthier, Dave
> Cc: pgsql-general@postgresql.org
> Subject: Re: [GENERAL] Access to NEW.column outside of a trigger function.
>
> "Gauthier, Dave" <dave.gauthier@intel.com> writes:
>
>> I have a check constraint that runs a PlPgsql function which returns a
>> pass/ fail status which the constraint uses to allow or disallow the
>> value. This is not a trigger function. It's just a plain-ole
>> PlPgsql. Is there a way I can read (not write, just read) the
>> NEW.column values that a trigger function would normally have access
>> to?
>
> Sure. Just pass them into your validator func as parameters.
>
> But why are you avoiding use of a trigger here?
>
>> Thanks in Advance for any help.
>>
>
