Then it breaks. I didn't even pretend that the regex as given would parse HTML. It just alters a string which happens to have some HTML of a known format in it.