Re: Index scan vs. Seq scan on timestamps

On Tue, Dec 07, 2004 at 09:25:20AM +0100, Per Jensen wrote:
> Den Tue, Dec 07, 2004 at 03:13:04AM -0000 eller der omkring skrev Andrew - Supernews:
> > On 2004-12-07, Stephen Frost <sfrost@snowman.net> wrote:
> > > Is there a way to say "just take the value of this function at the start
> > > of the transaction and then have it be constant" in a query?
> >
> > Why not use CURRENT_TIMESTAMP, etc., which do exactly that?
>
> Because when using transactions, CURRENT_TIMESTAMP does not advance, but is fixed
> to time of session start

CURRENT_TIMESTAMP is fixed to the time of transaction start, not
session start; this is documented and observable behavior.  Can you
demonstrate otherwise?  If so, on what version of PostgreSQL?

--
Michael Fuhr
http://www.fuhr.org/~mfuhr/