Date: Thu, 28 Feb 2002 23:56:17 +0100
From: Alexander Steinert <stony8@gmx.de>
To: pgsql-sql@postgresql.org
Subject: Re: Large Objects
Message-ID: <20020228235617.A1403@tyche.svt.tu-harburg.de>
Mail-Followup-To: Alexander Steinert <stony8@gmx.de>, pgsql-sql@postgresql.org
References: <59358A738F45D51186A30008C74CE250E81448@slc-exc1.ctron.com>
	<GNELIHDDFBOCMGBFGEFOGEJDCBAA.chriskl@familyhealth.com.au>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
User-Agent: Mutt/1.2.5i
In-Reply-To: <GNELIHDDFBOCMGBFGEFOGEJDCBAA.chriskl@familyhealth.com.au>;
	from chriskl@familyhealth.com.au on Thu, Feb 28, 2002 at
	09:05:52AM +0800

>    Don't use them.  They were needed when Postgres only supported 8k
>    per row.  Now you can just use the 'text' datatype for text data
>    and the 'bytea' datatype for binary data.  You have a limit of a
>    few gigs per row with them.

The problem with the text or bytea type is, that inserting large amounts
of data causes a big performance loss because everything must go through
the SQL-Parser. I would be glad if someone would correct me.

So far I have found no way to grant integrity / use PG's transactions
for large objects with satisfying performance. lo_import/export for
bytea would be a very nice interface to transfer directly between the
database and files readable/writable by the client process.

Suggestions are welcome.

Stony