For truly one-off things, I'd agree. But how many of these do you want to reinvent: sorting by multiple columns, grouping, aggregate functions, file locking, deleting, inserting, updating, selecting by multiple criteria, etc. etc. A six line DBD::CSV or DBD::AnyData script can accomplish all of those. And when it comes to output or integrating the handling of that table into a larger context, do you want to reinvent all of that or leverage the many DBI extension modules for those purposes?
I think it's foolhardy to say that either the database way or the non-database way is always the right way, it depends entirely on context. And as far as howitzers go, it is a matter of trading the howitzer of including modules (which may make very little difference in the long run) against the howitzer of coding all the details from file opening to parsing to sorting.
