From: Tom Lane <tgl@sss.pgh.pa.us>
To: "David G. Johnston" <david.g.johnston@gmail.com>
cc: Gaurav Pant <gauravpant145@gmail.com>, pgsql-general@lists.postgresql.org
Subject: Re: Introduction of a new field in pg_class indicating presence of a large object in a table
In-reply-to: <CAKFQuwY4xifDbf8EZKwYtrECQiCbYsK_RLo+vnnWN-MuSSaqLQ@mail.gmail.com>
References: <CAG53p5L7=K7qOCneELDCyLQ4oZcd3d8EWhrEHpzav-zHNM+RZg@mail.gmail.com> <CAKFQuwY4xifDbf8EZKwYtrECQiCbYsK_RLo+vnnWN-MuSSaqLQ@mail.gmail.com>
Comments: In-reply-to "David G. Johnston" <david.g.johnston@gmail.com>
	message dated "Tue, 30 Apr 2024 12:13:52 -0700"
MIME-Version: 1.0
Content-Type: text/plain; charset="UTF-8"
Content-ID: <1800616.1714505285.1@sss.pgh.pa.us>
Content-Transfer-Encoding: 8bit
Date: Tue, 30 Apr 2024 15:28:05 -0400
Message-ID: <1800617.1714505285@sss.pgh.pa.us>
Archived-At: <https://www.postgresql.org/message-id/1800617.1714505285%40sss.pgh.pa.us>
Precedence: bulk

"David G. Johnston" <david.g.johnston@gmail.com> writes:
> On Tue, Apr 30, 2024 at 11:57 AM Gaurav Pant <gauravpant145@gmail.com>
> wrote:
>> I wanted to know if there is any such system table that we can use to
>> identify and map the fields containing large objects and the respective
>> tables and if it is not already there, do we have any plans to incorporate
>> the same in pg_class like we have for pg_toast?

> Large Objects are nothing like TOAST.  There is no system level association
> between large objects and tables.  Sure, the DBA can choose to store a
> large object OID in a table, but how you'd go about figuring out which
> columns contain those is going to be installation specific.

Yeah.  You might want to look at contrib/vacuumlo, but realize that
that's fairly heuristic.

> Though
> hopefully they used a bigint data type and maybe added "oid" to the column
> name...I suppose it would be interesting if one could define a FK on a
> table and point it at pg_largeobject_metadata but that I suspect would be
> the extent to which we'd do something along the lines of your request.

That would solve the opposite problem, of preventing a column from
containing any OIDs that *weren't* large object OIDs.  Given that
recording a large object OID elsewhere in the database is purely
an application decision, I don't think there's a reasonable way
for the system to track it.

			regards, tom lane