Ok, thank you for your answer.

How can sorting on each step be a best practice?
But it is clever without doubt.

This is not a mistake. The first translations (Ruby, Python, etc... 8 years ago) didn't test for duplicates. Duplicates were introduced by the Haskell translator (@bkaes) and gave many fiery discourses (see at the very bottom of this page). I never had enough time to modify that and don't think I will have. All I can do is to modify the description by replacing "Beware: r must be without duplicates." by "Beware: in some languages r must be without duplicates.". Hope this will satisvy you though not quite satisfying...

Your solution fails on this test:
in_array(["foo bar", "foo", "bar"], ["foo", "bar"])
Sorry.

This solution allows duplicates in the answer and thus is wrong.

Actually this approach is a huge loser in benchmark tests.

Lol! The author should add test cases.

Ok. What if n = 1 and m = 0?

What if m == 0 ?

Brilliant!