Re: ON CONFLICT with constraint name doesn't work

On 2017-03-16 12:44:23 -0700, Peter Geoghegan wrote:
> On Thu, Mar 16, 2017 at 12:42 PM, Heikki Linnakangas <hlinnaka(at)iki(dot)fi> wrote:
> > We debated this for a long time when the ON CONFLICT feature was being
> > developed. In the end, we settled on this behavior, on the grounds that a
> > constraint is a logical concept, while an index is a physical implementation
> > detail. Note that the SQL standard also doesn't say anything about indexes,
> > but constraints are in the standard.
>
> Right. Besides, you really are only supposed to use the ON CONSTRAINT
> syntax when inference won't work, as an escape hatch. This doesn't
> look like an example of where inference won't work. That's limited to
> ON CONFLICT DO NOTHING with exclusion constraints, which is fairly
> limited.

FWIW, I never was completely on board with this design goal, and I think
we should have (and still should) support using indexes directly.

- Andres