On 10 January 2017 at 14:31, Fabien COELHO <coelho@cri.ensmp.fr> wrote:
> I do not like Pavel's feature, this is a subjective opinion. This feature
> does not provide a correct solution for the use case, this is an objective
> fact. The presented feature does not have a real use case, this is too bad.
Oh, also, you might want to tell Oracle and the many people who use
package variables that.
Now, that said, huge numbers of people blindly do all sorts of unsafe
things and mostly get away with it. Using MERGE in concurrent OLTP
workloads. Racey upserts. Blindly assuming xacts will succeed and not
keeping the info around to retry them until confirmation of commit is
received. That sort of business.
Nonetheless, it's pretty clear they're far from having a "real use case".
I'd like to see transactional vars. I think it's worthwhile and you've
made a reasonable argument that they're useful, and should probably
even be the default. Your unwillingness to listen to anyone else isn't
doing your argument any favours though.
-- Craig Ringer http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training & Services