Well the question is whether punch_card_don needs to do that. I'm not clear on that. It seems to me he's just comparing things on known positions, not searching for them, but I may easily be wrong. I have no idea whether pack()ing the data (at whatever stage of the processing or generation) can be used or whether it would help or just complicate things. I was just trying to suggest something that punch_card_don might have omited, hoping it will be of some use.

Besides ... even if he does need to search for the substrings it might still be beneficial to pack the data for storage. He might save up to 7/8 of the disk access. Which might help even if he has to unpack the data to process them.