In spark I want to create one class that extend RDD ,but compute method will return iterator or conv subclass

Question

I hope create class extend from RDD(spark). but spark RDD compute method return Iterator In RDD class that method define: def compute(split: Partition, context: TaskContext): Iterator[T] but I hope return tuple. In this situation, I do not want change RDD too much. Likewise use Either type replace Iterator, this cause too many code change. Does someone have better solution? thanks.