Sounds to me like you should do proper tokenization instead, perhaps
with a helper function written in C if it gets too slow otherwise.

Reply via email to