I was more viewing this in context of the article, which goes on to micro-optimize a version with an IMO unreasonable set of assumptions.

If it's instead explicitly given that this is a quick litmus test, the choices of the first pass implementation don't really matter, and you'll talk about complexities after, then sure.