Re: NOT ENFORCED constraint feature

On Thu, Mar 27, 2025 at 7:45 PM Álvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org> wrote:
>
> On 2025-Mar-27, Amul Sul wrote:
>
> > On Thu, Mar 27, 2025 at 6:28 PM Peter Eisentraut <peter(at)eisentraut(dot)org> wrote:
>
> > > That said, is there a simpler way? Patch 0003 appears to add a lot of
> > > complexity. Could we make this simpler by saying, if you have otherwise
> > > matching constraints with different enforceability, make this an error.
> > > Then users can themselves adjust the enforceability how they want to
> > > make it match.
> >
> > We can simply discard this patch, as it still reflects the correct
> > behavior. It creates a new constraint without affecting the existing
> > constraint with differing enforceability on the child. I noticed
> > similar behavior with deferrability -- when it differs, the
> > constraints are not merged, and a new constraint is created on the
> > child. Let me know your thoughts so I can avoid squashing patch 0006.
>
> I didn't read that patch and I don't know what level of complexity we're
> talking about, but the idea of creating a second constraint beside an
> existing one itches me. I'm pretty certain most users would rather not
> end up with redundant constraints that only differ in enforceability or
> whatever other properties. I failed to realize that this was happening
> when adding FKs on partitioned tables, and I now think it was a mistake.
> (As I said in some previous thread, I'd rather have this kind of
> situation raise an error so that the user can do something about it,
> rather than silently moving ahead with a worse solution like creating a
> redundant constraint.)
>

Okay, in the attached version, I’ve added an error in
tryAttachPartitionForeignKey() if the enforceability is different.
Please have a look at the 0005 patch and let me know if it looks good.
The rest of the patches remain unchanged.

I haven’t squashed the patches because, if we decide to keep the error
and avoid further complexity, we can simply discard the 0006 patch.

Regards,
Amul