First draft of translation document - contravariance

[Rémi Forax]

Le 19/05/2010 02:32, Reinier Zwitserloot a écrit :
> Now I'm a bit confused. That sounds like this:
>
> #void(Integer) vi = #void(Integer i) { };
> #void(String) vs = vi;
>
> would be legal as well which should be wrong, no?
>
> --Reinier Zwitserloot

It's not legal,
MethodHandle are runtime time not compiler one.

List<String> is erased to List,
#void(String)  is erased to MethodHandle

like List<String> and List<Integer> aren't compatible,
#void(String) and #void(Integer) aren't compatible too.

cheers,
Rémi


>
>
>
> On Tue, May 18, 2010 at 11:10 AM, Rémi Forax <forax at univ-mlv.fr 
> <mailto:forax at univ-mlv.fr>> wrote:
>
>     Le 18/05/2010 11:01, Howard Lovatt a écrit :
>     > I am guessing that the following will fail with the suggested
>     > MethodHandle translation (it is hard to say from the JSR292
>     > literature):
>     >
>     > #String(Object) so = #String(final Object o) (o.toString());
>     > #Object(String) os = so;
>     >
>     > Assuming it does fail, this failure may not be important since
>     the use
>     > case is rare. Therefore this is post is a request for clarification
>     > rather than a criticism.
>     >
>     >    -- Howard.
>     >
>     >
>
>     #String(Object) and #Object(String) will be erased to
>     java.dyn.MethodHandle
>     so it doesn't fail.
>
>     BTW, jdk7 beta binaries already contain a preview of JSR 292 API,
>     this is not the final API and there are some known bugs
>     but you can already play with it.
>
>     cheers,
>     Rémi
>
>
>
>
>