On Sun, 2 May 2004, Jon Harrop wrote:

> Could Set (and others) be implemented polymorphically by using a comparison
> function passed as an argument?
>
> If so, what would be the implications of this approach? I think: you couldn't
> use the type checker to enforce consistent comparison functions between two
> different sets which were, say, being merged. I think you can enforce this
> using the functor approach provided the Sets came from the same functor
> "instantiation".

With my words: I understand functors as a way to parametrize a type with a
value.
They allow you to detect inconsistencies at compile time (e.g. 2 sets
parametrized with different "compare" functions).
Without functors, as far as I know, you can only test inconsistencies at
run time (e.g. by putting all the parameters in a record an check physical
equality of the records).


> Functors appear to be somewhat similar to templates in C++. Does the functor
> approach produce more efficient code as it is partially specialised over the
> comparison function?

Unfortunately no.
Julien Signoles wrote a defunctorizer (works from the source code of the
functor):

  http://www.lri.fr/~signoles/ocamldefun/manual.html


Maybe some user manual should tell us more about when to use functors
rather than just how to use them.



Martin


-------------------
To unsubscribe, mail caml-list-request@inria.fr Archives: http://caml.inria.fr
Bug reports: http://caml.inria.fr/bin/caml-bugs FAQ: http://caml.inria.fr/FAQ/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners