From mboxrd@z Thu Jan  1 00:00:00 1970
Received: (from weis@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id PAA29871 for caml-red; Fri, 22 Dec 2000 15:25:34 +0100 (MET)
Received: (from weis@localhost) by pauillac.inria.fr (8.7.6/8.7.3) id PAA27513; Fri, 22 Dec 2000 15:22:50 +0100 (MET)
From: Pierre Weis <weis@pauillac.inria.fr>
Message-Id: <200012221422.PAA27513@pauillac.inria.fr>
Subject: Re: Ref syntax
In-Reply-To: <20001222183020R.garrigue@kurims.kyoto-u.ac.jp> from Jacques Garrigue at "Dec 22, 100 06:30:20 pm"
To: garrigue@kurims.kyoto-u.ac.jp (Jacques Garrigue)
Date: Fri, 22 Dec 2000 15:22:50 +0100 (MET)
Cc: caml-list@pauillac.inria.fr
X-Mailer: ELM [version 2.4ME+ PL28 (25)]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: weis@pauillac.inria.fr

> Really, I don't think it would be useful at toplevel.
> I view let mutable .. in as a way to provide some state, but
> immediately cleanly wrapped, either by only being used locally, or
> in exported functions.
> This is completely similar to mutable object fields; both the goal and
> the method.

I hope this is effectively simple to type check, and ensure correct.

> > For instance, what do we
> > do if such a letref variable is assigned to, from within the body of a
> > function (that could be exported) ?
> 
> This is just syntactic sugar for references, which is why I said it
> was easy. Similarly typing is just the typing of references.

So you mean, that if we define x with let mutable x = 1, the variable
x has in fact type int ref ?

> > Furthermore, this construct would add an entirely
> > new notion to Caml: lvalues.
> 
> As stated above: they are already here, object fields.
> You may think of it as a good or bad idea, but the distinction
> between it and the fact a.x behaves differently when there is a <- and
> when there is none is subtle.

Sorry, but for field mutations the syntactic construct is not
 expr <- new_val
but
 expr.label <- new_val
which is semantically assign_to_label (expr, new_val), as
 expr.(i) <- new_val
is a short-hand for assign_to_vector (expr, i, new_val).

In both constructs there is no notion of left values.

I really don't know if the corresponding notion for objects needs
lvalues or not.

> But well, this was only first in my wish list because I was answering
> to a message related to that. My personal priority is much lower.

I think exactly the same: the let mutable (or var construct as we used
to call it) is desirable, but may not be worth the implementation
effort.

[...]
> > So you are already satisfied, since you can write
> > 
> >   ocamlc -c -I +camltk -I +camlimages ocaml_unreal_t.ml
> > 
> > in the current development version!
> 
> Great! I had seen the introduction of -where, but didn't catch this
> one.
> Thanks, Papa Weis !
> 
> Jacques

Hey, this is your ``Santa Xavier's'' work. Thanks to him.

Cheers,

Pierre Weis

INRIA, Projet Cristal, Pierre.Weis@inria.fr, http://pauillac.inria.fr/~weis/