So it occurs to me there's an OBVIOUS reason why this regex might be the most costly: as other commenters have mentioned, I'm doing this on a bunch of stuff that might be embedded inside of strings, and what have you. So even if I leave all the regexes as they are, I think I can just reorder them and make things potentially a lot faster -- even if I do still have multiple passes through the string. I'll take a look at a strategy of "first, throw away as much as I can, then work with what's left."