I’ve recently tried to parse an array with a fixed size in antlr4:

array[int size]
    : {size > 0}? thing array[$size - 1]
    | 
    ;

This allows me to have arrays with a size determined by my data, e.g.

foo: number array[$number.value]; // array with $number amount of elements

My problem is that this approach has massive overhead (stack overflow for big arrays).
Is there a good way to linearize this approach?

I’ve tried to “optimize” it: (TARGET: Java)

array[int length] returns [ThingContext[] objects]:
        {
            _localctx.objects = new ThingContext[length];
        }
        ( {$length > 0}?
         firstObject=thing {
            _localctx.objects[0] = _localctx.firstObject;
            for(int i=1; i<length; i++){
                _localctx.objects[i] = thing();
            }
        }
        |
        );

This seems to mess with the adaptive prediction, though (I’m sometimes getting an infinite loop).
Any ideas?