Re: Large binary data [was: contrib: hack for psycopg2 performance]

Daniele Varrazzo <[email protected]> Wed, 12 May 2010 14:32:29 +0100
Newsgroups gmane.comp.python.db.psycopg.devel
Message-ID <[email protected]>
On Wed, May 12, 2010 at 2:27 PM, Karsten Hilbert
<[email protected]> wrote:
> On Wed, May 12, 2010 at 02:12:20PM +0100, Daniele Varrazzo wrote:
>
>> > Not really (beyond the one the OP posted). Common
>> > (PostgreSQL) wisdom holds that it is advantageous when
>> > dealing with significant amounts of truly binary data (such
>> > as bytea) - which I am (medical documents).
>>
>> Aren't large objects more suitable to deal with them?
>
> It much depends. You wanted bytea if you wanted
> transactional control. That has been improving somewhat
> recently with PG 9.0.

I thought they were under transaction control. Reading [1] "All large
object manipulation using [libpq lo_*()] functions must take place
within an SQL transaction block".

[1] http://www.postgresql.org/docs/8.4/static/lo-interfaces.html

-- Daniele