Your initial solution works for 10,000 rows. How would you change your approach if the dataset was 100 gigabytes?
