Fwd: odd transaction behavior

Oops - I only sent this to Federico.

---------- Forwarded message ----------
From: Kerri Reno <[email protected]>
Date: Mon, Mar 22, 2010 at 11:43 AM
Subject: Re: [Psycopg] odd transaction behavior
To: Federico Di Gregorio <fog-NGVKUo/i/[email protected]>


Federico,

Thanks for your response.  Do you suggest that I change the isolation level
back to READ COMMIT if I want to do multiple lines in a transaction?

Kerri

On Mon, Mar 22, 2010 at 11:39 AM, Federico Di Gregorio <fog-NGVKUo/i/[email protected]>wrote:

> On 22/03/2010 17:43, Kerri Reno wrote:
> > Hello all!  I'm a newbie to this mailing list but I've been using
> > psycopg for quite a while.  I hope you can help me!
> >
> > I recently changed our isolation level from 'READ COMMIT' to
> > 'AUTOCOMMIT' and have seen some odd behavior when I want to post a
> > multiple line transaction.  I am using
> >
> 'conn.set_isolation_level(psycopg2.extensions.ISOLATION_LEVEL_AUTOCOMMIT)'
> >
> >
> in the connection class to make all transactions auto commit.  Then I am
> > using cursor.execute('BEGIN') to start a longer transaction.  Is this
> > what I should be doing?  Because conn.commit() doesn't seem to work
> > on these longer transactions.  BUT, if I do cursor.execute('commit'),
> > then the transactions are posted. Should conn.commit() work?  If so,
> > what am I doing wrong?  This is a viable way to commit the data, but
> > I'm just wondering if it is the 'correct' way.
>
> If you switch transactions off (by ISOLATION_LEVEL_AUTOCOMMIT) then you
> have two options:
>
> 1) to not use transactions
> 2) as you discovered, manage transactions by yourself issuing BEGIN,
>   COMMIT and ABORT commands using execute()
>
> Sincerely, I don't like (2). If you want transactions, choose the
> appropriate transaction isolation level, be pythonic and let psycopg do
> its work (it will start a transaction before your first query and just
> after you call .commit() or .rollback() on the connection).
>
> If you decide to manage transactions by yourself then avoid to call
> .commit() or .rollback(), because psycopg is not supposed to know what's
> you're doing and won't oblige to your command.
>
> federico
>
> --
> Federico Di Gregorio                                       fog-NGVKUo/i/[email protected]
>   I filosofi son come i sociologi: il mondo non lo capiscono. -- A.R.M.
>
>
> _______________________________________________
> Psycopg mailing list
> Psycopg-IAPFreCvJWPBWskQ1e/[email protected]
> http://lists.initd.org/mailman/listinfo/psycopg
>
>


-- 
If you have an emergency, and can't reach me, please call Craig at
928-257-4980 or the help desk at 928-502-4202.
.·:*¨¨*:·.   .·:*¨¨*:·.   .·:*¨¨*:·.
Yuma Educational Computer Consortium
Compass Development Team
Kerri Reno
[email protected]      (928) 502-4240
.·:*¨¨*:·.   .·:*¨¨*:·.   .·:*¨¨*:·.




-- 
If you have an emergency, and can't reach me, please call Craig at
928-257-4980 or the help desk at 928-502-4202.
.·:*¨¨*:·.   .·:*¨¨*:·.   .·:*¨¨*:·.
Yuma Educational Computer Consortium
Compass Development Team
Kerri Reno
[email protected]      (928) 502-4240
.·:*¨¨*:·.   .·:*¨¨*:·.   .·:*¨¨*:·.

_______________________________________________
Psycopg mailing list
Psycopg-IAPFreCvJWPBWskQ1e/[email protected]
http://lists.initd.org/mailman/listinfo/psycopg