The Dutch Flag Problem

This problem was first proposed by E.W Dijkstra in 1976 (in the chapter 14 of his book "A Discipline of Programming"). It is a very classical variation of the sorting algorithms since then. It is sometimes useful in real life, but it became famous for its pedagogical properties: its complexity is not trivial, but not complex either. It can easily be formally proven, but it's not absolutely trivial either.

What can I do to improve this PLM universe?

As usual, there are several things that could be done in the code of this universe to improve it: