Using a script to implement a Java interface with variadic methods

[John Keeping]

I get surprising results using Nashorn to implement a Java interface via
a script when that interface contains variadic methods.

The easiest way to explain is probably with an example:

If I have a Java interface:

    public interface MyInterface {
        void test(int i, String... strings);
    }

and I implement this in a script:

    function test(i, strings) {
        print('i = ' + i);
        print('strings = ' + strings);
    }

which I then wrap from Java:

    ScriptEngine engine = new ScriptEngineManager().getEngineByName("nashorn");
    engine.eval(script);
    Invocable inv = (Invocable) engine;
    MyInterface obj = inv.getInterface(MyInterface.class);

Calling the "test" method:

    obj.test(10, "one", "two", "three");

gives:

    i = 10
    strings = one

when I would expect something like:

    i = 10
    strings = [Ljava.lang.String;@deadbeef

(not useful output, but it illustrates the point I hope).


At this point, I considered changing the script to use "arguments" to
access the passed in values:

    function test(i, strings) {
        for (var index = 0; index < arguments.length; index++)
            print('arg ' + index + ' = ' + arguments[index]);
    }

but that gives an error trying to get the interface:

    java.lang.invoke.WrongMethodTypeException: Parameter counts differ:
        (Object[])Object vs. (int,String[])void


Is this a deficiency in Nashorn's handling of variadic methods or am I
missing something here?


-- 
John