On Wed, May 20, 2015 at 11:26 AM, Andres Freund <andres@anarazel.de> wrote:
> Even if maybe not directly under the guise of exclusion constraints
> themselves, but I do think it's an interesting way to more easily allow
> to implement unique constraints on !amcanunique type indexes. Or, more
> interestingly, for unique keys spanning partitions
Alright, then. It's just that at one point people seemed to think that
upsert should support exclusion constraints, and that position was, at
the time, lacking a good justification IMV. What you talk about here
seems much more practical than generalizing upsert to work with
exclusion constraints. You're talking about exclusion constraints as
an implementation detail of something interesting, which I had not
considered.
--
Peter Geoghegan