OK, that was what I thought at first, but then I read this note in `man
pg_dump`:
When -t is specified, pg_dump makes no attempt to dump any other
database objects that the selected table(s) might depend upon.
so I supposed that that dependency information was *not* required. So I
posted the bug.
Is that note then outdated/incorrect? Or am I mistaken in finding it at
odds with your explanation?
Regards,
Gulli
On Mon, Jun 10, 2013 at 2:04 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Gunnlaugur Thor Briem <gunnlaugur@gmail.com> writes:
> > pg_dump takes O(N) time dumping just one table (or a few) explicitly
> > specified with a -t parameter. It thus becomes painfully slow on a
> database
> > with very many tables.
>
> This is not a bug. It needs information about all the tables anyway
> to deal with dependencies (possible inheritance and similar situations).
>
> Having said that, it does look like getTables is pulling back a lot of
> info that we don't need *yet*, and would never need if we conclude we
> don't need to dump the table. Possibly some of this work could usefully
> be postponed to, say, getTableAttrs. OTOH, if that makes the normal
> dump-everything case noticeably slower, it's unlikely such a patch would
> get accepted.
>
> regards, tom lane
>