I second the suggestion to check out DBD::CSV.

The idea of being "SQL like" and having the upgrade path to a "real" SQL is certainly a plus.

As to performance, I've found that for reasonably small tables, selecting is no slower than for a real db since there's no overhead of talking to the database. Insert/Update/Delete are murder for large files since I think it reads/writes the entire file for each operation.

But since the data is in a plain text file, for large update operations (like building the search index), you could probably do a Perl-only choice that would be very fast to write the file, then SQL to read.

And finally, when you don't want to do SQL, you can always use vi to "update the table".