loops hurt so bad

:: Programming Languages, Python

By: John Clements

Why, professor Clements, why? Why are you taking all of our loops away?

I feel like the Grinch, sometimes, taking all of the pretty little candy loops away from the wide-eyed first-year students.

But the fact is… that I really don’t like loops.

  • Loops in every popular language simply execute a block of code repeatedly. This means that your code has to perform mutation in order to allow a value to escape from the code. This requires before-and-after reasoning, and introduces a huge new source of bugs into your program. By contrast, there’s almost no chance to tangle the action of the caller and the callee in a recursive call.

  • Corollary of this: you can easily update loop variables in the wrong order: e.g.: i += 1, sum += arr[i]. Oops, bug.

  • Loops can’t easily be debugged; in a functional style, each loop iteration is a recursive call for which the student can write an explicit test. Not possible using loops.

  • In functional style, you can’t forget to update a variable; each recursive call must contain values for all loop variables, or you get a sensible (and generally statically detectable) error.

  • You can’t write a loop in a non-tail-calling fashion. You have to do all of the work on the way down. Traversing a tree is basically impossible (to do it, you’re going to wind up building a model of the stack, turing tar pit etc.).

  • Finally, writing in a functional style is about 80% of the way to proving your code correct; you have to choose a name for the action of the recursive call, and you can usually state an invariant on the relation between the inputs and the outputs without difficulty.

Nothing here is new; it’s just a way of blowing off steam while grading my students’ terrible, terrible code. Sigh.

(EDIT: I should add… none of this applies to list comprehension forms such as for/list; those are dandy. Not sure? Run down the list of bullet points and check for yourself.)