Finding the Bug in a Loop

Until now every loop arrived with the same question attached: what does this print. This set asks a harder one. You are told what a piece of code was meant to do, you are shown code that does not do it, and you find the difference yourself.

Nothing here tells you which line to look at. In real work nothing does, so the method has to.

Reading a loop and nodding along is the most reliable way to miss a bug in it, because you end up reading the intention already in your head rather than the code on the page. Tracing means writing down what every variable holds after every pass, in order, skipping nothing.

So it prints 3. Written out like that there is nowhere for a mistake to hide, and the pass where the numbers stop matching what you expected is the pass holding the bug.

A loop has three zones, and a fault lives in exactly one of them.

Before the loop. What each variable starts at, and whether that line is outside the loop. Write those starting values down before you trace a single pass.

Inside the loop. Which lines are in the body at all, which of them are inside the if, and what order they run in.

After the loop. Which variable gets printed or returned, and whether that line is outside.

A counter should go up once for each item that qualifies, and once only. Hold a counter against that contract line by line: which passes raise it, and by how much.

So compare the answer against the number of items before accepting it. A count larger than the list is not a close call, it is proof.

The first pass and the last pass are where loops fail; the middle usually looks after itself. Before the first pass, write down what every variable holds and what the first item does to it. At the other end, write down which item the loop was looking at when it stopped.

Check those two passes by hand, every time.

Summing keeps a running total and counting keeps a running tally. The same skeleton keeps a running largest: hold one variable, and replace what it holds whenever the item you are looking at beats it. A running smallest is the same thing with the comparison turned round.