I've added 3 columns of 3 dictionaries. I've put some random chars at the end to get lines length the same as his files. And, again, short regex (he uses short regex in this part of the code) was as fast as index (and even slightly faster). I don't know why, but the best he can do, it's to try with his own data and his own perl (I've used perl 5.20.0)