Re: [GENERAL] get the previous assigned sequence value

Orakle's fault :-) I remembered there was a thorough (very thorough)
discussion months before. This raises a question: because the web
based search engine is slow, why can't setup a text file achive?
so that people can ftp the big file and use vi do whatever they
want locally? does using the search engine increase potential value
to the site? if in that case, I'd better shut up ;-), but I can not
see any, it is rather a load burden. perhaps some commercialization
plan for future?



On Thu, 9 Dec 1999, Ed Loehr wrote:

> "Ross J. Reedstrom" wrote:
>
> > Gah, this is getting a bit annoying, correcting this bit of
> > mis-information.
> >
> > currval() is kept as part of the session context, as is completely
> > multi-user safe. That's why the person you quoted said 'in the same
> > session'.
> >
> > Where is everyone coming up with the wrong idea on this? Is there a
> > major commercial DB with sequence objects that gets this wrong?
>
> I think it is primarily because (a) it is intuitive to think there is a race
> condition, which fortunately psgql folks have gone to significant efforts to
> avoid, and (b) the documentation at
>
>     http://www.postgresql.org/docs/postgres/sql-createsequence.htm
>
> is less than concise/clear on this issue.
>
> Same thing is true for the question of how to retrieve the newly-inserted
> serial, which is why people keep asking the same question over and over (that,
> and because people don't know about www.deja.com or the mailing list search
> engine...or the engine is broken).
>
> Cheers.
> Ed
>
>
>
>
>
> ************
>