Re: Jdbc/postgres performance

On 10/18/06, Bucky Jordan <bjordan@lumeta.com> wrote:
> > On 10/17/06, Rohit_Behl <Rohit_Behl@infosys.com> wrote:
> > > Select events.event_id, ctrl.real_name, events.tsds, events.value,
> > events.lds, events.correction, ctrl.type, ctrl.freq from table events,
> > iso_midw_control ctrl where events.obj_id = ctrl.obj_id and
> > events.event_id > ?::bigint order by events.event_id limit ?
> >

> After a quick search on the JDBC list, it looks like there's some recent
> discussion on the subject of how to give the planner better insight for
> prepared statements (the subject is "Blind Message" if you're
> looking...).
>
> So, I'm off to go read there and perhaps join the jdbc mailing list too.

this is not really a jdbc issue, just a practical problem with
prepared statements...except for the mechanism if any the jdbc driver
allows you to choose if a statement is prepared.

> But, a more general postgres question. I assume if I want to turn
> prepared statements off altogether (say I'm using a jdbc abstraction

you turn off prepared statements by not invoking sql prepare or
PQprepare.  (or, if jdbc implements its own protocol client, it's
version of PQprepare).

> layer that likes parameterized statements, and there's other benefits to
> parameterizing other than just saving on db parse/plan) can I set
> max_prepared_transactions to 0? Is there any other option outside of

this setting is for 2pc and is not relevent to the discussion :)  even
if it were, im not so sure about a setting designed to enforce a
partcular method of querying.

yes, you are correct this is not exactly the use case for hints being
discussed in -hackers. however, imho, this is much more important and
relevant so long as prepared statements continue to work the way they
do.

merlin