When you say "match the keys", do you mean that you look for occurrences of identical keys? Or are you looking for substrings?

If it is the former, you should read one line at a time, split it, and store it in a hash. (No need to the store the whole file in memory).

And the you read the second file line by line, consulting your hash for each of them.

An entirely different solution is to view them as database tables, and use DBI together with DBD::CSV and do a join on the tables.