Re: pg_dump versus views and opclasses

On Sun, Jan 18, 2009 at 1:41 AM, Brendan Jurd <direvus@gmail.com> wrote:
> ... but I'm confused as to why 8.4
> doesn't freak out when told to create a view it doesn't actually know
> how to execute.  Have the rules for evaluating views changed?
>

Update.  Turns out that 8.4 DOES know how to execute the view.  If you
try to group on a user-defined composite type, 8.4 just goes ahead and
groups it, rather than giving the old "could not identify an ordering
operator" error.

Which renders the whole idea of creating my own opclass for the type
very much bunk.

What sorcery is this?  Does 8.4 add some kind of default ordering
strategy for composite types?

Cheers,
BJ