Re: [Caml-list] mutability analysis too strict?

["Ohad Rodeh" <ORODEH@il.ibm.com>]

First of all, sorry my mail was sent twice, this was due to mail delivery
problems
from my site.

Perhaps I was not specific enough about what I wanted to acheive. What I
need
is a repository that has the following interface:

module type Repos = sig
  val put : 'a -> 'b -> unit
  val get : 'a -> 'b
end

An  implementation that looks like this:

module S : Repos = struct
   let h = Hashtbl.create 10
   let put key data = Hashtbl.add h key data
   let get key = Hashtbl.find h key
end

Does not work. Compilation error:

Signature mismatch:
Modules do not match:
  sig
    val h : ('_a, '_b) Hashtbl.t
    val put : '_a -> '_b -> unit
    val get : '_a -> '_b
  end
is not included in
  Repos
Values do not match:
  val put : '_a -> '_b -> unit
is not included in
  val put : 'a -> 'b -> unit

I tried also using the Map and Set modules, but they don't really allow
building
a repository of immutable values either. Is there any deep reason for this
behavior? Could you expound on this line:

> A monomorphic, mutable
> structure that contains polymorphic data is sound, but cannot be
expressed
> in ML's type system where universal quantification must be prenex.

Ohad.



Francois Pottier <francois.pottier@inria.fr>@pauillac.inria.fr on
10/12/2001 10:13:31

Please respond to Francois.Pottier@inria.fr

Sent by:  owner-caml-list@pauillac.inria.fr


To:   Ohad Rodeh/Haifa/IBM@IBMIL
cc:   caml-list@inria.fr
Subject:  Re: [Caml-list] mutability analysis too strict?




On Sun, Dec 09, 2001 at 05:43:41PM +0200, Ohad Rodeh wrote:
>
>       let h = Hashtbl.create 10;;
>       h : ('_a, '_b) Hashtbl.t
>
> The objects and keys in the table are infered to be mutable. However,
> in my case, they are immutable and I have to coerce them using Obj.magic
> from '_b to 'b.

You are slightly wrong here: the analysis infers the table itself (not the
keys or objects in it) to be mutable, which it indeed is. If the table was
given a polymorphic type, you would be able to store objects of a certain
type and to retrieve them at another type (by taking different instances of
'b), which would be unsound.

Furthermore, I'm surprised to hear that using Obj.magic helps; indeed, any
application of Obj.magic is itself deemed `dangerous' by O'Caml, leading
to the following behavior:

  # let h = Hashtbl.create 10;;
  val h : ('_a, '_b) Hashtbl.t = <abstr>
  # let h = (Obj.magic h : ('a, 'b) Hashtbl.t);;
  val h : ('_a, '_b) Hashtbl.t = <abstr>

That is, Obj.magic doesn't help at all in this case.

Perhaps you could tell us what you are trying to achieve? Any polymorphic,
mutable structure is unsound and rightly rejected. A monomorphic, mutable
structure that contains polymorphic data is sound, but cannot be expressed
in ML's type system where universal quantification must be prenex.

--
François Pottier
Francois.Pottier@inria.fr
http://pauillac.inria.fr/~fpottier/
-------------------
Bug reports: http://caml.inria.fr/bin/caml-bugs  FAQ:
http://caml.inria.fr/FAQ/
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/
To unsubscribe, mail caml-list-request@inria.fr  Archives: http://caml.inria.fr