On Tue, 2011-04-19 at 12:52 +0000, Daniel Migowski wrote:
> I would love to give pg_dump an option to produce ordered output. This would
> make it much easier to create dump files that are comparable with a textual
> diff tools.
>
> The following orderings are required:
>
> * Dump objects ordered by name. I know this cannot be done with all objects,
> but at least remain a reproducable, deterministic order.
It does do that.
> Also nice would be:
>
> * Dump data ordered by primary key, or if not available, dump data ordered
> by all columns.
Hmm, that would probably make pg_dump 5000 times slower on average.
There are various tools out there that can compare data between
databases that might do a better job.
> In addition it would be nice to have an option to omit oid comments, because
> oids are usually not used in databases.
It doesn't do that anymore in recent versions.