On Tue, 2007-01-30 at 17:05 -0500, Fernando Alegre wrote:

> There are two separate notions of equality mixed together.
> My point is, should "OCaml polymorphic equality" be the
> same concept at "IEEE equality"?

No, this is not possible. x <> x is the IEEE test for a NaN,
and cannot possibly be used in a structural equality
test which is primarily required for inserting data into
containers: you'd have a Set in which after inserting
a value the value wasn't in the Set!


> Why not wish for three separate operators (=,== and =.)?

What about abstract types?

The bottom line is: you can always write your own comparison
functions and use them with modular functors (extensionally
polymorphic), you can't get a single coherent intensionally
polymorphic equality concept because programs are always using
representations of value rather than the mathematical values.

It is a pity we don't have G'Caml around: it may provide
a convenient way to mix user defined equality with
Ocaml's run time driven polymorphic structural equality.


-- 
John Skaller <skaller at users dot sf dot net>
Felix, successor to C++: http://felix.sf.net