Go Mastery 5: Testing, Profiling and Internals

1. Testing

• *_test.go, TestXxx(t *testing.T)
• Table-driven tests with t.Run subtests
• t.Helper(), t.Cleanup(), t.TempDir(), t.Setenv()
• t.Parallel()
• TestMain
• Golden files
• httptest.NewServer / httptest.NewRecorder
• Mocking by interfaces (hand-written fakes beat generated mocks when possible)
• testify (optional), go-cmp for diffs
• Coverage: go test -coverprofile=c.out && go tool cover -html=c.out
• Integration tests with testcontainers-go
• testing/synctest for concurrent code (newer Go)

Fuzzing

• func FuzzXxx(f *testing.F), go test -fuzz=.

Benchmarks

• func BenchmarkXxx(b *testing.B), b.Loop() / for i := 0; i < b.N; i++
• -benchmem, -count, benchstat for comparing runs
• Avoid compiler optimizing away results

2. Profiling and performance

• pprof: CPU, heap, goroutine, block, mutex profiles
• net/http/pprof in running services
• go tool pprof -http=:8080 cpu.prof
• go tool trace: execution tracer
• Escape analysis: go build -gcflags="-m"
• Reduce allocations: preallocate slices (make([]T, 0, n)), reuse buffers, sync.Pool, strings.Builder
• Struct field ordering and padding
• Avoid premature optimization: measure first
• PGO (profile-guided optimization)

Visual: the performance workflow

flowchart TD
  A["Write a benchmark"] --> B["Measure: go test -bench -benchmem"]
  B --> C["Profile: pprof CPU and heap"]
  C --> D{"Hotspot found?"}
  D -- "yes" --> E["Change ONE thing"]
  E --> F["Re-run benchmark, compare with benchstat"]
  F --> D
  D -- "no" --> G["Stop. Good enough."]

Rule: measure first, optimize second.

3. Runtime internals (for true mastery)

• Garbage collector: concurrent tri-color mark-and-sweep, GOGC, GOMEMLIMIT
• Goroutine stacks (growable)
• The scheduler: G, M, P, work stealing, preemption
• How maps, slices, strings, interfaces, and channels are laid out in memory
• runtime package: NumGoroutine, ReadMemStats, GC
• Read: A Guide to the Go Garbage Collector (go.dev/doc/gc-guide)

Practice

1. Write benchmarks comparing += vs strings.Builder vs bytes.Buffer
2. Take a slow function, profile it with pprof, and cut allocations by half
3. Fuzz a parser and fix what it finds
4. Reach 80%+ meaningful coverage on one of your projects

---

Previous: Go Mastery 4: Standard Library and Tooling
Next: Go Mastery 6: Backend and Production Go

Back to the roadmap