Re: ordered pg_dump

On Mon, 2008-11-10 at 17:05 -0800, Jeff Davis wrote:
> Is there any interest in an optional mode for pg_dump to order the
> output so that it's easier to use diff?
>
> I don't think it would make the output 100% deterministic, but it would
> make it easier to at least compare the data for small databases.

That'd be cool.  I'd done some poking around on the topic a little while
back.  The goal was to make the output more predictable so that backups
would be more efficient, specifically with a product that does binary
diffs of some sort.

I may still have some notes somewhere if you're interested.  But I
believe the idea was to use COPY with a SELECT statement.  The
non-trivial part was to figure out a proper ordering to use.

Or did you plan on combining it with -t, where you could then specify
the ordering for each table?

> I think this has been brought up before, but I couldn't find the thread,
> so I don't know what conclusion was reached.
>
> Regards,
>     Jeff Davis

(... Plus, you potentially get a free CLUSTER on a reload.)

- Josh Williams