(no subject)

[Nikolaj Bjorner]

[-- Attachment #1: Type: text/plain, Size: 3060 bytes --]

I am pretty sure this is not related to references, but to let-rec.

Try:

let rec f x = x and g y = f y + 1;;

you will see that the type inferred is:

val f : int -> int
val g : int -> int

Now try:

let f x = x and g y = f y + 1;;

this time:

val f : 'a -> 'a
val g : int -> int

You may still be wondering why, of course.
The basic answer (I hope I am not too rusty in this)  
is that type inference for polymorphic let-rec is not decidable
(you have to solve a semi-unification problem).
Hence, the "solution" is not to abstract type variables  
until all type inference for all  
let-rec-and clauses have been analyzed.

So:

let rec f x = x and g y = f y + 1 and h z = not (h z);;

does not type check at all, but again:  
let f x = x and g y = f y + 1 and h z = not (h z);;
does work.


Nikolaj




From: Chris Hecker [mailto:checker@d6.com]
Sent: Thu 2/6/2003 5:31 PM
To: caml-list@inria.fr
Subject: [Caml-list] streams and value restriction




I assume I'm running into the polymorphism restriction thing here, but
I'm not sure why (I kind of understand it when references are in play,
but this is baffling me).

Here's a simple version of some stream parser code.  I'd like
parse_opt_comma_list to be polymorphic so I can use it to parse lists
of any of my values.  It works fine as a separate let rec, but if I
put it in the overall let rec of the main parser (with and) it won't
generalize and gets tagged as string-only since that's the first way
it's used.

Can somebody explain why this is a problem?  Sorry for being dense.

Thanks,
Chris


type t =
    Int of int
  | String of string

(* this works: *)
let rec parse_opt_comma_list parse list = parser
    [< 'Genlex.Kwd ",";
       a = parse ;
       optlist = parse_opt_comma_list parse (a::list) >] -> optlist
  | [< >] -> list

let rec parse_values = parser
    [< 'Genlex.String s;
       optlist = parse_opt_comma_list (parser [< 'Genlex.String s >] -> s) [] >] ->
         (String s) :: List.map (fun s -> String s) optlist
  | [< 'Genlex.Int i;
       optlist = parse_opt_comma_list (parser [< 'Genlex.Int i >] -> i) [] >] ->
         (Int i) :: List.map (fun i -> Int i) optlist

(* this doesn't: *)
let rec parse_values = parser
    [< 'Genlex.String s;
       optlist = parse_opt_comma_list (parser [< 'Genlex.String s >] -> s) [] >] ->
         (String s) :: List.map (fun s -> String s) optlist
  | [< 'Genlex.Int i;
       optlist = parse_opt_comma_list (parser [< 'Genlex.Int i >] -> i) [] >] ->
         (Int i) :: List.map (fun i -> Int i) optlist

and parse_opt_comma_list parse list = parser
    [< 'Genlex.Kwd ",";
       a = parse ;
       optlist = parse_opt_comma_list parse (a::list) >] -> optlist
  | [< >] -> list


-------------------
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/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners

[-- Attachment #2: Type: text/html, Size: 6062 bytes --]

[Caml-list] Re: your mail

[Mike Potanin]

On Thu, 6 Feb 2003, Nikolaj Bjorner wrote:

> I am pretty sure this is not related to references, but to let-rec.
>
> Try:
>
> let rec f x = x and g y = f y + 1;;
>
> you will see that the type inferred is:
>
> val f : int -> int
> val g : int -> int
>
> Now try:
>
> let f x = x and g y = f y + 1;;
      ^                 ^
This is different functions. First "f" defined new function. Second "f"
referred to old function "f".

>
> this time:
>
> val f : 'a -> 'a
> val g : int -> int
>
> You may still be wondering why, of course.
> The basic answer (I hope I am not too rusty in this)
> is that type inference for polymorphic let-rec is not decidable
> (you have to solve a semi-unification problem).
> Hence, the "solution" is not to abstract type variables
> until all type inference for all
> let-rec-and clauses have been analyzed.
>
> So:
>
> let rec f x = x and g y = f y + 1 and h z = not (h z);;
>
> does not type check at all, but again:
> let f x = x and g y = f y + 1 and h z = not (h z);;
> does work.
>
>
> Nikolaj
>
>
>
>
> From: Chris Hecker [mailto:checker@d6.com]
> Sent: Thu 2/6/2003 5:31 PM
> To: caml-list@inria.fr
> Subject: [Caml-list] streams and value restriction
>
>
>
>
> I assume I'm running into the polymorphism restriction thing here, but
> I'm not sure why (I kind of understand it when references are in play,
> but this is baffling me).
>
> Here's a simple version of some stream parser code.  I'd like
> parse_opt_comma_list to be polymorphic so I can use it to parse lists
> of any of my values.  It works fine as a separate let rec, but if I
> put it in the overall let rec of the main parser (with and) it won't
> generalize and gets tagged as string-only since that's the first way
> it's used.
>
> Can somebody explain why this is a problem?  Sorry for being dense.
>
> Thanks,
> Chris
>
>
> type t =
>     Int of int
>   | String of string
>
> (* this works: *)
> let rec parse_opt_comma_list parse list = parser
>     [< 'Genlex.Kwd ",";
>        a = parse ;
>        optlist = parse_opt_comma_list parse (a::list) >] -> optlist
>   | [< >] -> list
>
> let rec parse_values = parser
>     [< 'Genlex.String s;
>        optlist = parse_opt_comma_list (parser [< 'Genlex.String s >] -> s) [] >] ->
>          (String s) :: List.map (fun s -> String s) optlist
>   | [< 'Genlex.Int i;
>        optlist = parse_opt_comma_list (parser [< 'Genlex.Int i >] -> i) [] >] ->
>          (Int i) :: List.map (fun i -> Int i) optlist
>
> (* this doesn't: *)
> let rec parse_values = parser
>     [< 'Genlex.String s;
>        optlist = parse_opt_comma_list (parser [< 'Genlex.String s >] -> s) [] >] ->
>          (String s) :: List.map (fun s -> String s) optlist
>   | [< 'Genlex.Int i;
>        optlist = parse_opt_comma_list (parser [< 'Genlex.Int i >] -> i) [] >] ->
>          (Int i) :: List.map (fun i -> Int i) optlist
>
> and parse_opt_comma_list parse list = parser
>     [< 'Genlex.Kwd ",";
>        a = parse ;
>        optlist = parse_opt_comma_list parse (a::list) >] -> optlist
>   | [< >] -> list
>
>
> -------------------
> 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/
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
>

-------------------
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/
Beginner's list: http://groups.yahoo.com/group/ocaml_beginners