the reversing bytes part is encoding agnostic. you just feed the reversed bytes to the csv parser then re-reverse both the yielded rows and the cells bytes and get the original order of the bytes themselves.