Running it over a test file of about 1G in size results in 15G RSS.

Problems start with slurping the entire thing upfront, I don't know how it manages to amass 15 x memory overhead on top of it.

more from gnu coreutils sits at around 2MB.

How to repro:

1. generate a file of about 1G: perl -e 'print "meh\n" x (256 * 1024 * 1024)' > testfile
2. cargo run -r testfile