Dear Christopher,
> > "TABLESPACE" appears in a basic pg_dump SQL output. If the test tablespace
> > does not exist, the command will fail, and so my whole restoration.
> >
> > Thus I still stick to my opinion;-)
>
> Your complaint was that you need a way of continuing a restore if the
> _tablespace_ cannot be created. ie. If the directory does not exist.
Indeed it is possible that I was not clear enough!
The issue I feel should be addressed is the ability to restore a database
while ignoring tablespace issues, not only their creation but also their
uses.
> If you have objects in a tablespace, then too bad.
Well, ISTM that it is the problem I'm discussing...
If I cannot restore a base I see that as a problem, which is indeed a lack
of humour from my side.
> It's no different to if the schema the object in doesn't exist.
> Or the table the data is in doesn't exist.
> Or the functin the view references doesn't exist.
It is a little bit different because a schema, a table or a function are
database application issues and are normally addressed by pg_dump and
pg_restore, although tablespaces are more an administration issue wrt disk
layout and the like, which are likely to be different from one machine to
another (compare with I obviously want the same schema/table/function for
my application). So the notion of dump/restore of a tablespace need
some careful thinking.
But maybe I'm just stupid to dream that I could restore or transfer my
data even if I used a tablespace somewhere? ;-)
It looks that we don't have the same perspective about database
administration.
Anyway, have a nice day,
--
Fabien Coelho - coelho@cri.ensmp.fr