Incidentally, that second capture won't waste memory beyond what's required to note some offsets. Once you have any captures at all, the original string is copied and then all captures are just implemented like substr() lookups into that string.