CompletableFuture

[Remi Forax]

On 11/29/2012 04:19 PM, Doug Lea wrote:
> On 11/29/12 09:53, Remi Forax wrote:
>>
>> again Future<? extends U>.
>
> Thanks.
>
>>
>> I think that CompletableFuture<?> is better as return type than 
>> CompletableFuture<Void>
>> otherwise users may want to declare a local variable Void.
>
> We end up using explicit Void params in FJ stuff a lot
> and people seem OK with it, so we might as well make these
> use the same conventions unless anyone can think of a reason
> otherwise.

j.u.c.ExecutorService use the <?> convention, see submit(Runnable),
so now I let you choose if you perfer to be inconsistent with 
ExecutorService or FJ :)

>
>>>     /**
>>>      * Creates and returns a CompletableFuture that is completed with
>>>      * the result of the given function of either this or the other
>>>      * given CompletableFuture's results if/when either complete
>>>      * normally.
>>>      *
>>>      * @return the new CompletableFuture
>>>      */
>>>     public <U> CompletableFuture<U> or(CompletableFuture<? extends 
>>> T> x,
>>>                                        BiFunction<? super T,? super 
>>> T, ?
>>> extends U> fn);
>>
>> Thinking a little more about this one, fn should be a Function not a 
>> BiFunction
>> given that only one of the two completables needs to complete.
>
> Yes, thanks. (I fixed this before but then overwrote with old version; 
> oops.)
>
>>>     /**
>>>      * Returns the value if completed, else the given valueIfAbsent.
>>>      */
>>>     public T getNow(T valueIfAbsent);
>
>> still think it should use a supplier as parameter and not the value.
>
>
> The motivating use case is to allow simplification of
> polling-like constructions where you'd otherwise need to
> first check isDone, and then get(), but could replace with:
>   while(t.getNow(NOTREADY) == NOTREADY) { doSomethingElse(); }
> in which case trapping into a function is not at all helpful.

you mean,
    XXX value;
    while((value = t.getNow(NOTREADY)) == NOTREADY) { doSomethingElse(); }

> And I can't think of other cases not involving plain get()
> where this method is of much use.

Ok, given that I can write getNow(supplier) from getNow(valueIfAbsent), 
your signature is better than mine.

public static <T> T getNow(CompletableFuture<Object> future, Supplier<? 
extends T> supplier) {
   Object value;
   if ((value = future.getNow(NOT_READY) == NOT_READY) {
     return supplier.supply();
   }
   return (T)value;
}

>
> -Doug
>
>

Rémi