In the example, why do you use two string tokenizers instead of one 2D string tokenizer?