From mboxrd@z Thu Jan  1 00:00:00 1970
Received: (from majordomo@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id WAA04461; Tue, 14 Jan 2003 22:06:43 +0100 (MET)
X-Authentication-Warning: pauillac.inria.fr: majordomo set sender to owner-caml-list@pauillac.inria.fr using -f
Received: (from weis@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id WAA04543 for caml-list@pauillac.inria.fr; Tue, 14 Jan 2003 22:06:42 +0100 (MET)
Received: from nez-perce.inria.fr (nez-perce.inria.fr [192.93.2.78]) by pauillac.inria.fr (8.7.6/8.7.3) with ESMTP id UAA01902 for <caml-list@pauillac.inria.fr>; Tue, 14 Jan 2003 20:15:03 +0100 (MET)
Received: from wptx49.physik.uni-wuerzburg.de (wptx49.physik.uni-wuerzburg.de [132.187.40.49])
	by nez-perce.inria.fr (8.11.1/8.11.1) with ESMTP id h0EJF2v10404
	for <caml-list@inria.fr>; Tue, 14 Jan 2003 20:15:02 +0100 (MET)
Received: from wptx47.physik.uni-wuerzburg.de (wptx47.physik.uni-wuerzburg.de [132.187.40.47])
	by wptx49.physik.uni-wuerzburg.de (8.10.0/8.10.0.Beta6) with ESMTP id h0EJEx507332;
	Tue, 14 Jan 2003 20:14:59 +0100 (MET)
Received: (from ohl@localhost)
	by wptx47.physik.uni-wuerzburg.de (8.10.0.Beta12/8.10.0.Beta12) id h0EJExL28392;
	Tue, 14 Jan 2003 20:14:59 +0100
From: Thorsten Ohl <ohl@physik.uni-wuerzburg.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <15908.25011.52235.354636@wptx47.physik.uni-wuerzburg.de>
Date: Tue, 14 Jan 2003 20:14:59 +0100
To: caml-list@inria.fr
Cc: Julien.Signoles@lri.fr
Subject: [Caml-list] Legality of using module types from .mli in .ml
X-Mailer: VM 6.84 under Emacs 20.7.1
Reply-To: ohl@physik.uni-wuerzburg.de
Sender: owner-caml-list@pauillac.inria.fr
Precedence: bulk

I'm often referring to module types defined in an interface file in
order to reduce redundancy.  For example

    $ cat a.mli
    module type T = sig val n : int end
    module M : T

    $ cat a.ml
    module type T = A.T
    module M : T = struct let n = 42 end

where T can be rather long.  Is this legal?  The manual says in
section 6.12

    A compilation unit behaves roughly as the module definition
    module unit-name : sig  unit-interface end =  struct unit-implementation end 

and refrains from defining `roughly' :-).  I came up with conflicting
evidence:

O'Caml accepts it
    
    $ ocamlopt a.mli a.ml 

However, it is not equivalent to

    module B :
      sig
	module type T = sig val n : int end
      end =
      struct
	module type T = B.T
	module M = struct let n = 42 end
      end
    
which is obviously rejected by O'Caml, since B.T is not bound.  The
pair (a.mli, a.ml) more closely resembles

    module Aux_C =
      struct
	module type T = sig val n : int end
      end
    
    module type C =
      sig
	module type T = Aux_C.T
	module M : T
      end
    
    module C : C =
      struct
	module type T = Aux_C.T
	module M = struct let n = 42 end
      end

without binding Aux.T.  My question is now: is the pair (a.mli, a.ml)
intentionally legal or do I need to expand the reference A.T, as in

    $ cat pedantic_a.mli
    module type T = sig val n : int end
    module M : T

    $ cat pedantic_a.ml
    module type T = sig val n : int end
    module M : T = struct let n = 42 end

because I've have taken advantage of a bug in the compiler.

The reason I'm asking this, is that my shorthand [as in (a.mli, aml)]
breaks Julien Signoles' defunctorizer.

Cheers,
-Thorsten
-- 
Thorsten Ohl, Physics Dept., Wuerzburg Univ. -- ohl@physik.uni-wuerzburg.de
http://theorie.physik.uni-wuerzburg.de/~ohl/     [<=== PGP public key here]
-------------------
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