To improve performance, consider using `export LC_ALL=C` (credit: https://benhoyt.com/writings/count-words/). But note that it sorts uppercase and lowercase separately, and doesn't handle Unicode well.