Clearly the fastest way to parse this is to load it all into RAM, and work backwards from the end. That has the advantage of putting the digits in ascending ord

Clearly the fastest way to parse this is to load it all into RAM, and work backwards from the end. That has the advantage of putting the digits in ascending order, and then the delimiter, then the string... until EOF or NL is encountered.