Loop Coverage

software-engineering testing

Definition

Loop Coverage

Loop coverage is a structural testing metric designed to test the validity of loop structures. It ensures that every loop is executed 0 times, 1 time, and >1 times.

Concept

Simple metrics like statement or branch coverage often test loops only once (the “standard” pass). Loop coverage specifically targets boundary conditions related to iteration logic.

Required Test Cases

For a simple loop while (C) { ... }:

1. Zero Iterations: The loop condition C is initially false. The body is skipped entirely.

• Goal: Verify code behaviour when the loop is bypassed (e.g., handling empty input arrays).

2. One Iteration: The loop body runs exactly once, then C becomes false.

• Goal: Verify basic “execute once” logic.

3. Multiple Iterations: The loop body runs more than once (e.g., 2 or “typical” usage).

• Goal: Verify that state updates correctly between iterations (e.g., incrementing counters, moving array pointers).

Lecture Insight: Why >1 matters

In the lecture example (reversing an array), a bug existed where the index pointer j was never decremented inside the loop.

• 0 Iterations: Passed (empty array returns empty).
• 1 Iteration: Passed (array of size 1 returns itself; j didn’t need to move).
• >1 Iterations: Failed. The loop ran twice, but j pointed to the same spot, overwriting data incorrectly. Only the >1 case revealed the bug.

Example

Consider a simple summation loop:

public int sum(int n) {
    int s = 0;
    while (n > 0) { // Condition C
        s += n;
        n--;
    }
    return s;
}
 

Coverage Analysis

To achieve 100% Loop Coverage, we need three distinct test inputs:

1. 0 Iterations: n = 0

   • n > 0 is false immediately. Body skipped. Returns 0.

2. 1 Iteration: n = 1

   • Pass 1: 1 > 0 (True), s=1, n=0.
   • Pass 2: 0 > 0 (False). Returns 1.

3. Multiple Iterations: n = 2

   • Pass 1: 2 > 0 (True), s=2, n=1.
   • Pass 2: 1 > 0 (True), s=3, n=0.
   • Pass 3: 0 > 0 (False). Returns 3.