Comment author: administrator

Full_Name: Mikhail Zabelin
Version: OCaml 3.07pl2 native Win32 port (MinGW)
OS: Win2000
Submission from: www-proxy.ioffe.rssi.ru (194.85.224.35)

source file f.ml:

let a = fun x -> x = [] in
let b = ref a in
b := (fun x -> x = [1.0]);
b := a;
ignore (a [1]);
ignore (!b [1])

compile:

ocamlc f.ml
File "f.ml", line 6, characters 12-13:
This expression has type int but is here used with type float
======================

This is the normal and well known behaviour of Caml polymorphism.
Identifier b has never been defined as a polymorphic value, due to its
definition as a function application. Also, after the first assigment
to b, the type of b is entirely monomorphic and hence cannot be used
both with float lists and int lists.

The following toplevel session may enlight the type assignments that
occur in your example :

let a = fun x -> x = [];;

val a : 'a list -> bool =

let b = ref a;;

val b : ('_a list -> bool) ref = {contents = }

b := (fun x -> x = [1.0]);;

• : unit = ()

b;;

• : (float list -> bool) ref = {contents = }

From here !b can be applied to float lists and to float lists only.

The next assignment does not change the types neither b's nor a's:

b := a;;

• : unit = ()

b;;

• : (float list -> bool) ref = {contents = }

a;;

• : 'a list -> bool =

Hence the error reported by the typechecker.

Best regards,

Pierre Weis

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

Comment author: administrator

Normal behavior (monomorphic refs).