Virgile Prevosto wrote:
> Le mer 27 jun 2007 12:12:34 CEST,
> Jeremy Yallop <jeremy.yallop@ed.ac.uk> a écrit :
>> OCaml seems a little inconsistent here, actually.  The application
>> `id 0' is only valid as the rhs of let rec because the compiler can 
> 
> Well, it just seems that Ocaml performs type inference before checking
> the validity of the recursive definition. If you give a well-typed
> term, it will complain about a forbidden rhs of let rec:

You're right; I'd overestimated the extent of the analysis.  Note that 
applications on the rhs of let rec are acceptable if there's no 
recursion at all involved in the application:

   let id x = x in
   let rec f = id id
       and g x = g (f x)
    in g

Jeremy.

> 
>         Objective Caml version 3.10.0
> 
> #  let rec id = fun x -> x and _foo = id 0 in id 1;;
> This kind of expression is not allowed as right-hand side of `let rec'