[Caml-list] strange typechecking result

[Caml-list] strange typechecking result

[Matej Kosik]

Hello,

For one of my modules, the typechecker started to raise strange 
complaints. I was not able to figure out exactly why, but at least I 
wanted to narrow down the problem.

This small program:

   type r1 = {l1 : unit list}

   and r2 = {l2 : int64 list}

   let rec f1 _ =
     ()

   and _ r1 =
     f1 r1.l1

   and _ r2 =
     f1 r2.l2

is rejected by the typechecker with a following error message:

   File "test.ml", line 12, characters 5-10:
   Error: This expression has type int64 list
          but an expression was expected of type unit list

I do not understand why the given program was rejected.

Thanks in advance for any help.
--
Matej Kosik

Re: [Caml-list] strange typechecking result

[Leo White]

On Feb 22 2013, Matej Kosik wrote:

>Hello,
>
>For one of my modules, the typechecker started to raise strange 
>complaints. I was not able to figure out exactly why, but at least I 
>wanted to narrow down the problem.
>
>This small program:
>
>   type r1 = {l1 : unit list}
>
>   and r2 = {l2 : int64 list}
>
>   let rec f1 _ =
>     ()
>
>   and _ r1 =
>     f1 r1.l1
>
>   and _ r2 =
>     f1 r2.l2
>
>is rejected by the typechecker with a following error message:
>
>   File "test.ml", line 12, characters 5-10:
>   Error: This expression has type int64 list
>          but an expression was expected of type unit list
>
>I do not understand why the given program was rejected.
>

I think that by default recursive uses of a function are monomorphic. You 
can fix this with an explicit polymorphic annotation:

   let rec f1: 'a. 'a -> unit = fun _ -> ()
   
   and f2 r1 = 
     f1 r1.l1
   
   and f3 r2 = 
     f1 r2.l2;;

Regards,

Leo

Re: [Caml-list] strange typechecking result

[Matej Kosik]

On 22/02/13 18:33, Leo White wrote:
>>
>
> I think that by default recursive uses of a function are monomorphic.
> You can fix this with an explicit polymorphic annotation:
>
> let rec f1: 'a. 'a -> unit = fun _ -> ()
> and f2 r1 = f1 r1.l1
> and f3 r2 = f1 r2.l2;;
>
> Regards,
>
> Leo
>

Yuri, Leo, thanks.

Now I see that there is even simpler program, that demonstrates the problem:

   let rec f1 _ =
     ()

   and f2 (value:int) =
     f1 value

   and f3 (value:char) =
     f1 value

The typechecker rejects it for similar reasons.

   File "test.ml", line 8, characters 5-10:
   Error: This expression has type char but an expression
          was expected of type int.

That's quite strange.

I would like to ask:

- Where is that restriction explained?
   (I've searched Ocaml reference manual for "monomorphic" but nothing
    relevant seemed to come up)

- Where are things like:

      'a. 'a -> unit

   described?

   I've already seen it in some document, but now I fail
   to find it again.

Re: [Caml-list] strange typechecking result

[Leo White]

>I would like to ask:
>
>- Where is that restriction explained?
>   (I've searched Ocaml reference manual for "monomorphic" but nothing
>    relevant seemed to come up)
>
>- Where are things like:
>
>      'a. 'a -> unit
>
>   described?
>

I'm not sure if the restriction is explained anywhere but the solution is 
described in:

  http://caml.inria.fr/pub/docs/manual-ocaml-4.00/manual021.html#toc79

Regards,

Leo

Re: [Caml-list] strange typechecking result

[Jeff Meister]

[-- Attachment #1: Type: text/plain, Size: 1101 bytes --]

As I understand it, polymorphic recursive functions require type
annotations because inferring their types is undecidable in the general
case.


On Fri, Feb 22, 2013 at 1:06 PM, Leo White <lpw25@cam.ac.uk> wrote:

> I would like to ask:
>>
>> - Where is that restriction explained?
>>   (I've searched Ocaml reference manual for "monomorphic" but nothing
>>    relevant seemed to come up)
>>
>> - Where are things like:
>>
>>      'a. 'a -> unit
>>
>>   described?
>>
>>
> I'm not sure if the restriction is explained anywhere but the solution is
> described in:
>
>  http://caml.inria.fr/pub/docs/**manual-ocaml-4.00/manual021.**html#toc79<http://caml.inria.fr/pub/docs/manual-ocaml-4.00/manual021.html#toc79>
>
> Regards,
>
> Leo
>
>
> --
> Caml-list mailing list.  Subscription management and archives:
> https://sympa.inria.fr/sympa/**arc/caml-list<https://sympa.inria.fr/sympa/arc/caml-list>
> Beginner's list: http://groups.yahoo.com/group/**ocaml_beginners<http://groups.yahoo.com/group/ocaml_beginners>
> Bug reports: http://caml.inria.fr/bin/caml-**bugs<http://caml.inria.fr/bin/caml-bugs>
>

[-- Attachment #2: Type: text/html, Size: 1836 bytes --]