Comment by n1ghtm4n

Why not a 2-word struct with the rune count instead of the byte count? There's zero performance cost for many strings because the rune count is known at compile time. For the rest, most strings are too short for Big-O analysis to be relevant and I would guess (enlighten me if I'm wrong) that the cost of computing the bounds of each character is negligible on a modern processor. Multi-byte chars in a string are going to be adjacent in memory, adjacent in cache, and therefore trivial for today's not-at-all-instruction-bound CPUs. Again, correct me if I'm wrong.