Re: Overloaded unification

On Jan 28, 9:01=A0pm, "David B. Benson" <dben...@[EMAIL PROTECTED]
> wrote:
>
> > Right, the more interesting question is: can it be used for anything
> > good?
>
> Well, SML/NJ somehow manages to stumble through '=3D' overloading in a
> fairly useful fa****on.

Actually, equality is not overloaded in SML. It is polymorphic, but
restricted to equality types (of which there are infinitely many).

> The obvious applications are in
> overloading '+' and '*' to treat, in the same scope, both
> real and complex arithmetic, for example. =A0Worse yet, one
> might also have pointwise addition of complex arrays, '+',
> and also scalar multiplication, '*'. =A0If all of the
> variables involved are previously defined (and type annotated if
> necessary to resolve genuine ambiguities), the
> resulting code ought to be about as clear as the textbook or
> paper from which the algorithms arose.
> [...]

Sure, but AFAICS, all the examples you enumerate can be readily dealt
with by more principled approaches like plain single parameter type
cl*****. My question referred to the wildly ad-hoc cases you also
wanted to be able to handle.

- Andreas