Re: Hm, table constraints aren't so unique as all that

Peter Geoghegan <peter(dot)geoghegan86(at)gmail(dot)com> writes:
> On 29 January 2013 00:25, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>> Of course this wouldn't be material for back-patching, but it seems to
>> me there's still time to fix this for 9.3, and we should do so if we
>> want to claim that the enhanced-errors patch uniquely identifies
>> constraints.

> I can see the case for fixing this, but I don't feel that it's
> particularly important that constraints be uniquely identifiable from
> the proposed new errdata fields.

I think that we'll soon be buried in gripes if they're not. Pretty much
the whole point of this patch is to allow applications to get rid of
ad-hoc, it-usually-works coding techniques. I'd argue that not checking
the entire constraint identity is about as fragile as trying to "sed"
the constraint name out of a potentially-localized error message.
In both cases, it often works fine, until the application's context
changes.

regards, tom lane