Re: RM and abstract syntax trees

On Nov 2, 9:10 am, paul c <toledobythe..._at_ooyah.ac> wrote:
> Marshall wrote:
>
> ...
>
> > Your conception of the RM is too narrow. There is nothing in
> > the RM that precludes nested structures, nor union types.
>
> By "union types", I presume you mean "unions of types that a db may use"
> and also that the RM doesn't preclude *nor define* unions of types
> except for unions of relation types as long as what some people call an
> identity function (IIRC) and others call an equality operator is
> provided, somehow, for the union type.

I like the way Date associates a union type with a supertype over the subtypes of which it is taking the union.

This seems an elegant way to define discriminated unions. I've wondered whether a language could support retrospective addition of supertypes of a given type. This could make sense if one drops the conventional OO idea of bundling methods with classes. Received on Fri Nov 02 2007 - 03:29:22 CET