Charles Bouillaguet wrote:
> I would like to write a type which describe some kind of term in a toy 
> programming language. It has sets, but not sets of sets
> [...]
> a) is it possible to write that with polymorphic variants, and how ?
> b) Is it possible to wrote that with Recursive modules, and how ?

It reminds me of the following:

------------------------------------------------------------------------
** Norman Ramsey asked and Jacques Garrigue answered:

  > I'm trying to write a small, extensible interpreter, and I'd like to
  > use polymorphic variants as the extension mechanism.  But I'm getting
  > stuck on very simple things.  For example, I would like the value
type
  > to include a few simple values, but I would also like it to be
  > extensible, thus:
  >
  >   type value = [ `Nil
  >                | `Number   of float
  >                | `String   of string
  >                | `Function of [>value] list -> [>value]
  >                | `Table    of ([>value], [>value]) Hashtbl.t
  >                ]
  >
  > However, when I do this, the compiler complains that
  >
  >   The type constructor value is not yet completely defined
  >
  > Is there some way to define a recursive, *extensible* type using
  > polymorphic variants?

Have a look at "Private rows: abstracting the unnamed" and
"Code reuse through polymorphic variants" at
<http://www.math.nagoya-u.ac.jp/~garrigue/papers/>

They both give examples of how to define extensible languages using
polymorphic variants. The first one relies on an experimental feature
only available in the CVS version of ocaml.

========================================================================

I haven't read the papers yet, so I cannot go further.

> I'm CC-ing my classmate Romain (Hi, Romain !), as he may be interested...

And what about me? :-(

-- 
Stephane Glondu
ENS de Cachan, currently visiting Keio Universty (Tokyo)