On Tue, Apr 12, 2005 at 06:38:41PM +0200, Michael Paesold wrote:
> Alvaro Herrera wrote:
>
> >On Tue, Apr 12, 2005 at 08:39:09AM +0200, Michael Paesold wrote:
> >>Tom Lane wrote:
> >
> >>>The other possible solution that comes to mind is to invent the
> >>>notion that a cast has a specific owner (which arguably it should
> >>>have anyway) and then say that "system casts" are those whose owner
> >>>is the original superuser.
> >>
> >>Just my toughts: I believe it's better when cast selection does not
> >>depend on the search_path. It seems dangerous for objects that you
> >>don't usually qualify with a schema. With all other objects in
> >>schemas I can think of, you can easily write the full-qualified
> >>name.
> >>
> >>So I vote for the latter.
> >
> >So casts created by the original superuser don't get dumped? That's
> >not good IMHO.
>
> Well perhaps there is an even better solution?
What about the simple one of having a bool "pg_cast.castissystem"
column, or something similar?
> >But yes, schema-qualifying casts seems weird:
> >'123'::someschema.user_type
> >
> >Is that even accepted by the grammar?
>
> It's the type you qualify here, not the cast, isn't it?
Yes, sorry. I'm low on caffeine apparently. Point on implicit casts
taken too.
--
Alvaro Herrera (<alvherre[@]dcc.uchile.cl>)
"I personally became interested in Linux while I was dating an English major
who wouldn't know an operating system if it walked up and bit him."
(Val Henson)