On 01/31/2012 03:02 PM, Martijn van Oosterhout wrote:
> On Mon, Jan 30, 2012 at 11:18:31PM -0500, Tom Lane wrote:
>> I don't recall that we thought very hard about what should happen when
>> pg_dump switches are used to produce a selective dump, but ISTM
>> reasonable that if it's "user data" then it should be dumped only if
>> data in a regular user table would be. So I agree it's pretty broken
>> that "pg_dump -t foo" will dump data belonging to a config table not
>> selected by the -t switch. I think this should be changed in both HEAD
>> and 9.1 (note that HEAD will presumably return to the 9.1 behavior once
>> that --exclude-table-data patch gets fixed).
> Perhaps a better way of dealing with this is providing a way of dumping
> extensions explicitly. Then you could say:
>
> pg_dump --extension=postgis -s
>
> to get the data. And you can use all the normal pg_dump options for
> controlling the output. The flag currently used to seperate the table
> schema from the table content could then interact logically. Another
> way perhaps:
>
> pg_dump --extension-postgis=data-only
> pg_dump --extension-postgis=schema
> pg_dump --extension-postgis=all
> pg_dump --extension-postgis=none
>
This one won't fly. We use option processing that requires that the
option names to be known at compile time, so you can't embed an
arbitrary extension name in there.
cheers
andrew
