Split fields with sv_split_next() and words with sv_token_next(), and see how they treat empty pieces.