Re: Making joins involving ctid work for the benefit of UPSERT

On Fri, Jul 18, 2014 at 11:32 AM, Peter Geoghegan <pg(at)heroku(dot)com> wrote:
> Well, maybe. If the genericity of this syntax isn't what people want,
> I may go with something else.

By the way, I didn't mention that there is now also the optional
ability to specify a "merge on" unique index directly in DML. It would
be much nicer to specify a sort key instead, but that might be tricky
in the general case. I imagine that other systems solve the problem by
being very restrictive in what will be (implicitly) accepted as a
merge-on index. Seemingly there are problems with all major SQL MERGE
implementations, so I don't necessarily expect to be able to draw
lessons from them in any way here.

--
Peter Geoghegan