Wrapping future

[jroper]

I'm wondering why FutureConverters.toJava creates a new CompletableFuture, rather than simply wrapping the Future with an implementation of CompletionStage that delegates to Future? Doing this would make conversion a little more efficient, and would mean conversions back and forth could be optimised so a conversion from a previously converted CompletionStage to Future would simply unwrap the Future.

The latter especially will be useful in Play, since we want to allow both Java and Scala code to intercept requests, and mixing these interceptors will potentially mean lots of conversions back and forth between the types. If it's just a simple wrapping/unwrapping, then we don't need to worry about the cost associated with it, otherwise we may end up with long chains of Futures and CompletionStages that add nothing to the actual value except overhead.

[retronym]

/cc @rkuhn, who contributed this part of the library.

[rkuhn]

The current approach was meant to be minimal, but we might have to bite the bullet. Writing a fully optimized version will be very painful, though.