testing-guide

Derive test cases from requirements and acceptance criteria rather than implementation code. Use AAA structure (Arrange, Act, Assert), name tests as '[method] should [expected behavior] when [condition]', and include at least one concrete assertion per test.

Weak assertions like toBeDefined, toBeTruthy, and not.toBeNull verify almost nothing about actual behavior. Tests using them as the sole assertion pass even when logic is broken, so they are treated as blockers; use specific value or structure assertions instead.

Mutation testing automatically modifies code to check whether tests detect the changes, producing a mutation score. Supported tools include Stryker for JavaScript/TypeScript, mutmut for Python, PIT for Java, Stryker.NET for C#, Mull for C/C++, Muter for Swift, go-mutesting for Go, and cargo-mutants for Rust.

Run branch coverage analysis after requirement-driven tests are complete: enumerate conditionals, exception paths, guard clauses, and loop boundaries, map them against existing tests, then generate supplementary tests annotated with @covers-branch for defensive logic gaps.

The standard target is 80% for both line and branch coverage, with 60% as the minimum acceptable floor. Coverage is a necessary but not sufficient condition, so pair it with mutation score targets of 80% for core business logic.

Mock only external dependencies you do not own, such as databases, network requests, file systems, and third-party APIs. Never mock internal implementation details or pure functions, and always verify mock call arguments and invocation counts.