Re: Ambiguos OPERATOR items in pg_restore manifest file

On Wed, Jul 20, 2011 at 8:24 AM, Renzo Kottmann <rkottman(at)mpi-bremen(dot)de> wrote:
> On 07/18/2011 05:25 PM, Robert Haas wrote:
>> On Tue, Jul 12, 2011 at 9:18 AM, Sandro Santilli <strk(at)keybit(dot)net> wrote:
>>> Trying to exclude items from dumps of postgis-enabled databases
>>> we use pg_restore -l output and strip what we think belong to PostGIS.
>>>
>>> In doing so, Renzo found that for OPERATOR there are not enough
>>> informations to unambiguosly find it being part of PostGIS (see
>>> included mail snippet).
>>>
>>> Do you think this could be improved on the pg_restore side ?
>> In 9.1, we've added the concept of EXTENSIONs.  I'm not sure whether
>> PostGIS is planning to take advantage of this mechanism, but it's
>> designed to solve exactly this problem.
>>
> The extensions concept will be a big  step forward, no doubt.
>
> I just do not think that it solves the aforementioned problem. AFAIK
> the extension system will be available for 9.1 only. Then it would not
> be available for all kinds of dump/restore scenarios <9.1 (in my case
> from 8.4 to 9.0).

Yeah, that's a problem. The mechanism does include some magic that's
supposed to help deal with this. The idea is -- first, you upgrade to
9.1 -- next, you say something like CREATE EXTENSION ename FROM
unpackaged -- finally, you upgrade the extension using ALTER EXTENSION
UPDATE. However, I'm not sure whether PostGIS is planning to support
this mechanism, or whether it meets their needs.

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company