> -----Original Message-----
> From: Ross J. Reedstrom
>
> On Wed, Sep 13, 2000 at 03:44:25PM +1000, Philip Warner wrote:
> > At 13:07 13/09/00 +0900, Tatsuo Ishii wrote:
> > >It seems current source is broken if MB is enabled.
>
> Gah, I Was afraid of this. My patch, I'm afraid.
>
> > >
> >
> > I haven't looked at the code yet, but isViewRule is going to
> change to use
> > the new reltype for views. Maybe this will side-step the problem?
> >
>
> Yes, it should. However, I've just done a quick test in a non-MB
> compile, and
> it looks like char(n) = 'a string constant' returns true if the
> first n chars
> match. If this is correct behavior, and holds in the multibyte case, then
> you can strip out _all_ the rulename truncation from pg_dump.
Isn't it a problem of backend side ?
It seems quite strange to me that clients should/could assume
such a complicated rule. I was surprized to see how many
applications have used complicated(but incomplete in some
cases) definiton(criterion ?)s of views to see if a table is a
view.
Now we have a relkind for views and in addtion haven't we
already had pg_views view to encapsulate the definition of
views.
Regards.
Hiroshi Inoue