From: | Federico Di Gregorio <fog(at)dndg(dot)it> |
---|---|
To: | Daniele Varrazzo <daniele(dot)varrazzo(at)gmail(dot)com> |
Cc: | Marko Kreen <markokr(at)gmail(dot)com>, psycopg ML <psycopg(at)postgresql(dot)org> |
Subject: | Re: [snafu] isolation-level change in 2.4.2 |
Date: | 2011-12-15 13:01:02 |
Message-ID: | 4EE9EF8E.6070206@dndg.it |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | psycopg |
On 15/12/11 13:55, Daniele Varrazzo wrote:
> First note: even if currently postgres behaviour is the same at levels
> read committed and read uncommitted, they are still two distinct
> levels:
>
> test=> set default_transaction_isolation = 'read uncommitted';
> SET
> test=> SHOW default_transaction_isolation;
> default_transaction_isolation
> -------------------------------
> read uncommitted
> (1 row)
>
> I think psycopg shouldn't assume the two being equivalent, otherwise
> in the event they will become different in future postgres versions we
> would have older psycopg version not supporting them. Doing it now
> doesn't cost anything.
>
> I've also just noted that the check for pg version not supporting
> levels UNC/REPREAD is in two different points (unfortunately
> set_session and set_isolation_level have no common code path).
>
> I've made a patch against these points, but I'll be able to run the
> complete test grid only this evening.
Merged.
federico
From | Date | Subject | |
---|---|---|---|
Next Message | Ronan Dunklau | 2011-12-15 15:17:19 | Generic casters for composite types |
Previous Message | Daniele Varrazzo | 2011-12-15 12:55:57 | Re: [snafu] isolation-level change in 2.4.2 |