golang-testing

Define a slice of structs with name, inputs, and expected outputs, then loop over cases calling t.Run for each subtest. This pattern covers many scenarios including error cases with minimal code and is the idiomatic Go testing standard.

Write functions starting with Benchmark that loop b.N times, then run go test -bench=. -benchmem. Use b.ResetTimer after setup and b.Run for sub-benchmarks comparing different input sizes or implementations.

Yes, Go 1.18+ includes built-in fuzzing via testing.F. Add seed corpus inputs with f.Add, define property assertions in f.Fuzz, and run with go test -fuzz=FuzzName -fuzztime=30s.

Define interfaces for dependencies, then create mock structs with function fields matching the interface methods. Inject the mock into your service under test to control behavior per test case without external libraries.

Flakiness often comes from time.Sleep calls, shared state, or missing tt := tt capture in parallel subtests. Detect flaky tests by running go test -count=10 and use channels or conditions instead of sleeps.

Run go test -race -coverprofile=coverage.out ./... then use go tool cover -func to check totals. You can fail the CI build when coverage drops below a threshold like 80 percent using a shell check.