array(string) tokenize_and_normalize(string what)
This can be optimized quite significantly when compared to tokenize( normalize( x ) ) in the future, currently it's not all that much faster, but still faster.