inst/examples/calls/README.md

In muttest: Mutation Testing

Aggregate — Symmetric Inputs Hide Function Swaps

What this demonstrates

When test inputs are chosen so that two semantically different functions (any/all, max/min, sum/prod) return the same result, swapping one for the other produces no observable difference. The mutant survives. Choosing discriminating inputs — where the two functions return different values — kills the mutant.

The function

has_positive uses any, all_positive uses all, range_width uses max and min, and total uses sum. Each of these function calls is a target for call-name mutation.

The weak test

The tests use inputs where any and all agree (all-positive vectors), check only the sign of range_width (non-negative), and verify only the type of total (numeric). None of these assertions can detect the corresponding function swap.

The surviving mutant

• any → all survives with all-positive input: any(c(1,2,3) > 0) = TRUE, all(c(1,2,3) > 0) = TRUE — same result.
• max → min survives: min(c(1,5,3)) - min(c(1,5,3)) = 0, which is still ≥ 0.
• sum → prod survives: prod(c(1,2,3)) = 6 = sum(c(1,2,3)) — coincidentally equal.

The fix

Use mixed inputs that produce different results for each function pair. For any/all, pass a vector with some positives and some non-positives. For range_width, assert the exact value. For total, use inputs where sum ≠ prod.

Key rule

When a call-name mutant survives, choose inputs that discriminate between the original function and its swap — the two functions must return different values for the test to be useful.

muttest documentation built on May 14, 2026, 5:10 p.m.