Re: pg_views.definition

On Wed, 2002-07-17 at 09:56, Jan Wieck wrote:
> Joe Conway wrote:
> > The problem is that you would still need to keep a copy of your view
> > around to recreate it if you wanted to drop and recreate a table it
> > depends on. I really like the idea about keeping the original view
> > source handy in the system catalogs.
>
> This has been the case all the time. I only see an attempt to
> make this impossible with the new dependency system. If I *must*
> specify CASCADE to drop an object, my view depends on, my view
> will be gone. If I don't CASCADE, I cannot drop the object.
>
> So there is no way left to break the view temporarily (expert
> mode here, I know what I do so please let me) and fix it later by
> just reparsing the views definition.

As somebody said, this is the place where CREATE OR REPLACE TABLE could
be useful. (IMHO it should recompile dependent views/rules/...
automatically or mark them as broken if compilation fails)

-------------
Hannu