Does that mean you think the tokeniser produces a <"/" operator> token? It doesn't. There's no such thing. The tokeniser produces either a regexp pattern token (which is a term) or a division operator token when faced with "/" in the input stream.