I don't disagree stylistically, but these changes do add about 5% to the benchmark times on my machine, which isn't ideal.

That is strange, but looks like it's second commit causes slowdown, first actually gives ~5% speedup
Try again with only first of two commits

Yup, first commit on its own seems to give a (marginal) speed-up. I think that makes sense; len(foo) looks like a function call but I believe it reduces to just a variable access, so storing it in a variable explicitly just makes more work for the compiler.

Push the branch back to just the first commit and I'll merge this.

@eapache done