Re: Stored procedures and out parameters

On Mon, Sep 17, 2018 at 7:45 AM Jonathan S. Katz <jkatz(at)postgresql(dot)org> wrote:
>
> Hi,
>
> On 9/2/18 4:32 PM, Robert Haas wrote:
> > On Thu, Aug 30, 2018 at 4:14 PM, Dave Cramer <pg(at)fastcrypt(dot)com> wrote:
> >> Reading this from the (JDBC) drivers perspective, which is probably a fairly
> >> popular one,
> >> We now have a standard that we can't really support. Either the driver will
> >> have to support
> >> the new PROCEDURE with the {call } mechanism or stay with the existing
> >> FUNCTIONS.
> >> This puts the drivers in a no win situation.
> >
> > I understand and I agree.
> >
> >> Undoubtedly.. surely the opportunity to do something about this has not
> >> passed as this has not been
> >> officially released ?
> >
> > I agree with that, too, but I can't make other people do things they
> > don't want to do, and then there's the question of time. On the basis
> > of previous experience, there is going to be little appetite for
> > holding up the release to address this issue no matter how badly
> > busted it is. Ultimately, it's the RMT that must decide what to do in
> > cases like this. I have confidence that they are watching this
> > thread, but I don't know what they will decide to do about it.
> >
>
> First, I want everyone to know that the RMT took this very seriously and
> took time collect feedback and consult with as many people as we could
> in order to make the best possible decision for v11 and ensure that any
> decision we made did not hinder any future implementation for stored
> procedures nor introduced something that would break backwards
> compatibility.
>
> Ultimately, the decision came down to one of four options:
>
> #1: Do nothing and remove the open item
> #2: Introduce nonstandard syntax for calling functions / procedures
> #3: Have CALL support both functions & procedures (or SELECT support
> calling functions)
> #4: Disable CALL
>
> The RMT has decided to go with option #1 and will be removing the open
> item for the PostgreSQL 11 release.

Hooray for making the right choice. This is exhibit "Z" as to why
abstracting the SQL language behind a programming API can lead to
problems. The workaround is to simply not do that and you can get
precise control of behavior. It's a little unfortunate that API
{call} maps to a select but such is life.

merlin