Feature wish: to omit explicit coercion for mli's

Feature wish: to omit explicit coercion for mli's

[Keiko Nakata]

Hello,

it would be nice if I can omit explicit type coercion 
to match .ml files with .mli's.

For instance, the following obj.ml does not match with obj.mli.

in obj.ml: 
 let o = object  method f1 = 1  method f2 = 2 end 

in obj.mli:
 val o : < f1 : int >
	

Explicit coercion would make the type system efficient.
However, I suppose that only inferring coercion when checking 
that .ml's meet .mli 's does not have much effect.

Regards,
Keiko NAKATA

Re: [Caml-list] Feature wish: to omit explicit coercion for mli's

[Oliver Bandel]

On Mon, Nov 21, 2005 at 06:35:30PM +0900, Keiko Nakata wrote:
> Hello,
> 
> it would be nice if I can omit explicit type coercion 
> to match .ml files with .mli's.
> 
> For instance, the following obj.ml does not match with obj.mli.

Your feature wish is a bug-wish. ;-)

If *.ml and *.mli do not match, then you have written
code that better would be rewritten, to say it nicely. ;-)

Ciao,
   Oliver

Re: [Caml-list] Feature wish: to omit explicit coercion for mli's

[Keiko Nakata]

From: Oliver Bandel <oliver@first.in-berlin.de>
> If *.ml and *.mli do not match, then you have written
> code that better would be rewritten, to say it nicely. ;-)

I do not think this is a bug-wish.

ob.ml would match with ob.mli by inserting explicit coercion as in :

in ob.ml: 
 let o = (object  method f1 = 1  method f2 = 2 end :> < f1 : int >)

in ob.mli:
 val o : < f1 : int >


I thought that this explicit coercion can be suppressed  
with little inefficiency of the type checker.
I occasionally use immediate objects to build polymorphic objects 
without writing type annotations.
(I mean, if I build polymorphic objects through class definitions,
I may need to define the classes with many type parameters.)
For the same reason, I thought that it would be nice 
if I could omit explicit coercion in some cases
when it does not cause inefficiency a lot.

Regards,
Keiko