Suggestions needed on analysing large number of csv files

Congrats on coming up with an efficient solution! Would that approach work though when the data doesn’t all fit in memory? Does it stream or swap/map to disk?