Re: FOL/HOL: is there a middle ground?

"Ralph Becket" <rafe_at_cs.mu.oz.au> wrote in message news:3638acfd.0407192151.8e4df4f_at_posting.google.com...
> "Marshall Spight" <mspight_at_dnai.com> wrote in message news:<vE%Kc.136839$Oq2.3490@attbi_s52>...
>
> Okay, say I'm interested in boolean relations (I'll call them
> `operators') of arity three. I could store all of my primitive
> operators in a single table, as I have above. Each row in the
> table gives the name of the boolean operator and one of the
> possible instantiations of its three parameters for which that
> boolean operator holds.

Oh, okay. Actually I think I said early on that relation-valued attributes didn't add any expressiveness, because they are informationally equivalent to a pair of tables, the second with a foreign key to the first. So I get this.

> (although without a polymorphic type system, you'd probably
> end up writing a lot of boiler-plate...)

It's clear dbmss need to get their type systems out of the 1960s. Bring on the parametric polymorphism, sum types, etc! Crap how about just ADTs even?!

> I think this approach gives you much of what you're looking
> for when you talk about "relation valued attributes".
>
> > Where's the RVA-equivalent?
>
> I'm afraid I'm not really a databases person - what does RVA
> stand for?

"relation-valued attribute."

> And I still don't understand your motivation for
> wanting these things (I'm not saying your motivation is invalid,
> I'm simply curious as to what you want your goal is.)

Convenience and expressiveness.

Marshall Received on Tue Jul 20 2004 - 17:24:37 CEST