Thread: PHP and autocommit of PostgreSQL 7.4
Dear all, Please vote for this http://bugs.php.net/26592 So that in next release of PHP we can set autocommit = off from with in PHP.ini Regards, V Kashyap
On Thu, 2003-12-11 at 17:17, Peter Eisentraut wrote: > Sai Hertz And Control Systems wrote: > > Please vote for this > > http://bugs.php.net/26592 > > Please don't. See my comment. > given you can override php.ini setting via php commands, ISTM application developers could still control this even if it was an ini setting. how do other languages handle this? IIRC in perl/dbi it is a parameter of the connection string ie. pg_connect(host=foo,user=bar,autocommit=on) in php-ish speak, which doesn't seem like a bad way to do it. (personally i'm down with Scott, application developers should just use begin/commit, but i've met some folks who really look at it the other way) Robert Treat -- Build A Brighter Lamp :: Linux Apache {middleware} PostgreSQL
El Jue 11 Dic 2003 20:25, Robert Treat escribió:
>
> (personally i'm down with Scott, application developers should just use
> begin/commit, but i've met some folks who really look at it the other
> way)
$db = DB::Connect("pgsql://user:pass@host/db");
$db->autocommit(false);
$db->query(query1);
$db->query(query2);
$db->commit();
Changing the first line would make it work on Oracle, or Informix. :-)
--
select 'mmarques' || '@' || 'unl.edu.ar' AS email;
-----------------------------------------------------------------
Martín Marqués | mmarques@unl.edu.ar
Programador, Administrador, DBA | Centro de Telemática
Universidad Nacional
del Litoral
-----------------------------------------------------------------
El Jue 11 Dic 2003 21:36, Martin Marques escribió:
> El Jue 11 Dic 2003 20:25, Robert Treat escribió:
> > (personally i'm down with Scott, application developers should just use
> > begin/commit, but i've met some folks who really look at it the other
> > way)
>
> $db = DB::Connect("pgsql://user:pass@host/db");
> $db->autocommit(false);
> $db->query(query1);
> $db->query(query2);
> $db->commit();
Replying to myself isn't a good thing, but I found something interesting here.
By default PEAR::DB has autocommit set to true (it's defined at the begining
on the class), but with the changes in PG 7.4 the default should be false,
which would make it work right.
Now, the problem that I additionally find is that when autocommit is set to
true there isn't a BEGIN / END before and after the query. This should also
change, unless a new function is added to have autocommit changed in the DB
conection and PEAR::DB uses it to change the autocommit setting.
P.D.: I'm adding the pgsql.php authors to the mail to see that they get a copy
--
09:21:01 up 16 days, 15:37, 2 users, load average: 0.61, 0.67, 0.45
-----------------------------------------------------------------
Martín Marqués | select 'mmarques' || '@' || 'unl.edu.ar'
Centro de Telematica | DBA, Programador, Administrador
Universidad Nacional
del Litoral
-----------------------------------------------------------------
Sai Hertz And Control Systems wrote: > Please vote for this > http://bugs.php.net/26592 Please don't. See my comment.
Robert Treat wrote: > given you can override php.ini setting via php commands, ISTM > application developers could still control this even if it was an ini > setting. Think about which entity determines what setting is required for autocommit. It's not the database administrator, it's not the database end user, it's not some tuning guy, it's not the binding library -- it's the particular client application. Because only the client application knows how it will issue SQL commands and what autocommit mode it expects for proper functioning. Therefore, there is never a legitimate reason to control this setting anywhere else. > how do other languages handle this? IIRC in perl/dbi it is a > parameter of the connection string ie. > pg_connect(host=foo,user=bar,autocommit=on) in php-ish speak, which > doesn't seem like a bad way to do it. In JDBC there is a method on the connection that controls this setting, and the default setting is fixed in the spec (not determined by external configuration). In ECPG, autocommit is off. Those are the only interfaces I know that have sufficient standardization in this area. The Perl way you cite seems reasonable as well.