Hi iainmerrick, just for info the measured per-file cost didn't include reading from the filesystem. Only the in-memory parsing was taken into account.