On Tue, Apr 7, 2015 at 12:01 PM, Peter Geoghegan <pg@heroku.com> wrote:
> I still don't like the idea of
> supporting this, though. I'm not aware of any other system allowing
> something like this for either MERGE or a non-standard UPSERT.
That includes MySQL, BTW. Only their REPLACE statement (which is a
disaster for various reasons) can do something like this. Their INSERT
... ON DUPLICATE KEY UPDATE statement (which is roughly comparable to
the proposed UPSERT patch) cannot update the entire row using a terse
expression that references a row excluded from insertion (following
the implementation taking the UPDATE path).
--
Peter Geoghegan