On Thu, 2006-08-03 at 18:07 +0100, Richard Huxton wrote:
> Rafal Pietrak wrote:
> > When working on a new database application I quite frequently:
> > -------------------------------
> > test_xx# \c template1
> > template1# DROP DATABASE test_xx;
> > template1# CREATE DATABASE test_xx TEMPLATE = earlier_version;
> > template1# \c test_xx
> > template1# \i some_cooked_dataset.sql-dump
> > -------------------------------
>
> The quickest solution is to put the whole thing in one script and
> execute that with \i or -f from the command-prompt.
Well, the trick is, that when I refere back to previous CREATE DATABASE,
I quite often adjust it: name or owner, or template.
To put it the other way around: I don't remember me creating a database
and *needing* to stay within the initiator (like template1) - that's why
I've ventured the feature request. If not a 'CONNECT' option, may be z
"SET parameter" for a session user? This way pg_dumpall, which does
CREATE DATABASE and stays with the old one, since it is supposed to be
run as user postgres will continue to work correctly.
Frankly, when posting the original e-mail I feared to receive: "NO!!
that'll break a lot of scripts/applications". But if that's not the
case, I'd vote to put this request into (event very low priority)
to-be-implementes list. postgres is awesome to work with, part of the
quality of a 'product' is good set of defaults - having "CREATE DB"
connect to the new instance is GoodThing(tm) :).
just my 2c. not that I'm eager to elaborate this case any further.
Thenx.
--
-R
