Some questions about type inference

Some questions about type inference

[Benedikt Rosenau]

I have played with covariance. Namely, I tried to define a class
of objects with a method "eat" who would only eat objects defining
a method "eat" themselves.

The direct approach fails:

  class virtual broken_carnivore =
    object
      method virtual eat: <eat: _; ..> -> unit
    end

since there is an unbound type variable in the ellipsis.


However, the following works (more or less):

  class virtual ['a] carnivore =
    object
      constraint 'a = ('a) #carnivore
      method virtual eat: <eat: _; ..> -> unit
    end

and I can only speculate how this binds the type variable.


Another piece de resistence for my understanding is the following:
  match []::[] with a::b -> a = [] & b = []
works, while
  match []::[] with a::b -> a = b
gives a type error.
   Benedikt

Re: Some questions about type inference

[John Max Skaller]

Benedikt Rosenau wrote:
> Another piece de resistence for my understanding is the following:
>   match []::[] with a::b -> a = [] & b = []

	What does the '&' mean? Do you mean '@' ??
Assuming [] @ b, then this is a 'x list for some 'x,
and thus can be compared with [] which has type 'y list
for some 'y, and we can set 'x = 'y.

> works, while

>   match []::[] with a::b -> a = b
> gives a type error.

That one is easy: the type of a is 'x list, and that of b is 'x list
list.
You cannot compare a 'x list with a 'x list list.

-- 
John (Max) Skaller, mailto:skaller@maxtal.com.au
10/1 Toxteth Rd Glebe NSW 2037 Australia voice: 61-2-9660-0850
checkout Vyper http://Vyper.sourceforge.net
download Interscript http://Interscript.sourceforge.net

Re: Some questions about type inference

[Didier Remy]

> Namely, I tried to define a class
> of objects with a method "eat" who would only eat objects defining
> a method "eat" themselves.
...
>  However, the following works (more or less):

>    class virtual ['a] carnivore =
>      object
>        constraint 'a = ('a) #carnivore
>        method virtual eat: <eat: _; ..> -> unit
>      end

This is correct but maybe cumbersome. You may also write: 

    class virtual ['a] carnivore =
      object
        constraint 'a = < eat: _; .. >
        method virtual eat: 'a  -> unit
      end
        ;;

which may be clearer.

BTW, if you are trying variations on the well-known cow/animal example
introduced by David Shang ["Is a cow an animal?", Unpublished note January
1996] you may find a discussion of this example together with a version in
Ocaml at:

        http://cristal.inria.fr/~remy/work/virtual/
        >> [Section] Expressiveness and abstraction

-Didier