caml-list - the Caml user's mailing list
 help / color / mirror / Atom feed
* Recursive module and class
@ 2008-07-09 23:32 Jun Furuse
  2008-07-12  7:34 ` [Caml-list] " Keiko Nakata
  0 siblings, 1 reply; 2+ messages in thread
From: Jun Furuse @ 2008-07-09 23:32 UTC (permalink / raw)
  To: caml-list

Hi,

Sorry if the following is already well known issue.

The following program trying to define a type and a class
mutually recursive is rejected by ocamlc 
(3.11+dev12 Private_abbrevs+natdynlink (2008-02-29)):

	module rec M : sig 
	  type t = Foo of N.c 
	end = struct
	  type t = Foo of N.c
	end and N : sig 
	  class c : object method x : M.t end 
	end = struct
	  class c = object (self)
	    method x = M.Foo (self :> c)
	  end
	end

The error message is puzzling:

	Error: Signature mismatch:
	       Modules do not match:
	         sig class c : object method x : M.t end end
	       is not included in
	         sig class c : object method x : M.t end end
	       Type declarations do not match:
	         type c = N.c
	       is not included in
	         type c = < x : M.t >

These signatures are literaly same, but do not match. 

The above code is compiled if I write

	    method x = M.Foo (self :> N.c)

to coerce the object to the outer class N.c instead of c.

Is it a bug of typing? Or it is ok but I was just confused by 
the error message ?

--
Jun FURUSE


^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: [Caml-list] Recursive module and class
  2008-07-09 23:32 Recursive module and class Jun Furuse
@ 2008-07-12  7:34 ` Keiko Nakata
  0 siblings, 0 replies; 2+ messages in thread
From: Keiko Nakata @ 2008-07-12  7:34 UTC (permalink / raw)
  To: jun.furuse; +Cc: caml-list

Hello.

To the best of my knowledge, there is subtle interaction 
between typing of recursive modules and unfolding of recursive structural types,
i.e, classes and polymorphic variants. 
I remember Jacques has hit a similar problem:
http://caml.inria.fr/pub/ml-archives/caml-list/2008/02/de4c1ddaa8bf3c0ef9350d8fb4f1e8f7.en.html

My understanding is that with recursive modules you may define two types 
that are equivalent but do not join together by simple unfolding 
of abbreviations; this seems to confuse the type checker. 
I believe this is a known bug of the current implementation
of the type checker. 

With best regards,
Keiko



^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2008-07-12  7:34 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2008-07-09 23:32 Recursive module and class Jun Furuse
2008-07-12  7:34 ` [Caml-list] " Keiko Nakata

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).