38 Go interview questions on performance, profiling & testing, with detailed answers and code.
Importing net/http/pprof for its side effect registers handlers under /debug/pprof/ on http.DefaultServeMux . The gotcha: if your public API server also uses…
The CPU profiler is a sampling profiler: the runtime asks the OS for a timer signal (SIGPROF on Unix) at 100 Hz by default, and on each tick records the stack of the…
The heap profile records allocation sites, sampled on average once per runtime.MemProfileRate bytes (default 512 KiB). It offers four sample types: inuse_space /…
Symptom: runtime.NumGoroutine() (export it as a metric) grows monotonically along with memory. Diagnose with the goroutine profile: curl -s…
Both are disabled by default because they add overhead. Block profile : where goroutines waited on synchronization primitives — channel send/recv, select , sync.Mutex ,…
pprof aggregates samples into "where is time spent"; the execution tracer records a timeline of events : goroutine create/start/block/unblock, syscalls, network poller…
b.Loop() returns true while the benchmark should keep iterating. It fixes three classic b.N mistakes: Setup cost : the timer is reset on the first call, so expensive…
func square(x int) int { return x * x } func BenchmarkSquare(b *testing.B) { for i := 0; i < b.N; i++ { square(42) } } It typically reports something like 0.25 ns/op —…
b.ReportAllocs() (or the -benchmem flag for all benchmarks) adds B/op and allocs/op columns. Allocations are often the best predictor of real-world cost because they…
func TestParseDuration(t *testing.T) { tests := []struct { name string in string want time.Duration wantErr bool }{ {name: "seconds", in: "5s", want: 5 * time.Second},…
func TestOrder(t *testing.T) { defer fmt.Println("defer") t.Cleanup(func() { fmt.Println("cleanup") }) for i := range 2 { t.Run(fmt.Sprint(i), func(t *testing.T) {…
Parallel tests in a package run concurrently up to -parallel (default GOMAXPROCS ). Packages themselves run in parallel as separate processes ( -p ). Top-level parallel…
defer : fine in a test without parallel subtests; runs when the function returns. t.Cleanup(fn) : runs after the test and all its subtests complete, in LIFO order.…
Since Go 1.18, func FuzzXxx(f *testing.F) defines a coverage-guided fuzz target. f.Add supplies seed inputs; f.Fuzz takes a function whose first parameter is *testing.T…
Testing code with timeouts, tickers, and background goroutines usually means real sleeps (slow and flaky) or injecting a fake clock everywhere. testing/synctest…
Go's implicit interface satisfaction lets the consumer define a small interface describing only what it needs; production code passes the real type, tests pass a…
Handlers : call the handler directly with httptest.NewRequest and httptest.NewRecorder — no network, fast. Clients : spin up httptest.NewServer (real loopback listener)…
The toolchain rewrites source to increment counters per basic block. go test -coverprofile=cover.out ./... go tool cover -func=cover.out # per-function percentages go…
PGO (GA in Go 1.21) feeds a CPU profile from production into the compiler so it optimizes hot paths more aggressively. The Go team reports typical gains of 2–14% CPU.…
The compiler inlines functions whose body cost is under a budget (~ 80 AST nodes ). Inlining removes call overhead, but the bigger win is that it enables further…
type Point struct{ X, Y int } func a() Point { p := Point{1, 2}; return p } // (1) func b() *Point { p := Point{1, 2}; return &p } // (2) func c() int { p := &Point{1,…
Go is memory safe: every s[i] gets a runtime check that panics if i is out of range. The SSA backend's prove pass removes checks it can prove redundant. In tight loops…
sync.Pool is a per-P cache of reusable temporary objects that reduces allocation rate and GC pressure for hot, short-lived objects (buffers, encoders). Semantics you…
Strings are immutable, so s += x in a loop allocates a new string and copies everything each iteration: O(n²) bytes copied and n allocations. func concatPlus(parts…
package main import ( "fmt" "strings" ) func main() { var a strings.Builder a.WriteString("hello") b := a b.WriteString(" world") fmt.Println(b.String()) } It panics :…
Go's GC is a concurrent, non-generational, non-moving mark-sweep collector. Pacing: GOGC (default 100): the next GC triggers when the heap grows by GOGC% over the live…
func main() { s := make([]int, 0, 2) a := append(s, 1) b := append(s, 2) fmt.Println(a[0], b[0], len(s), cap(s)) var grow []int prev := -1 for i := range 2000 { grow =…
CPUs cache memory in 64-byte lines (128 on some ARM). If two cores write to different variables on the same cache line, the line ping-pongs between cores via the…
Each field is aligned to its type's alignment (int64/pointers: 8 on 64-bit), and the struct's size is rounded up to its largest alignment. Poor ordering inserts padding.…
A normal string(b) or []byte(s) copies because strings must be immutable. The compiler already avoids the copy in several cases: Map lookups: m[string(b)] (lookup only,…
go test -race / go build -race instruments every memory access and uses ThreadSanitizer's happens-before (vector clock) algorithm. It reports when two goroutines access…
Options for gating slow tests: testing.Short() : skip under go test -short . Build constraints: //go:build integration at the top of the file; run with go test…
Functions named ExampleXxx in _test.go files serve as both documentation (rendered on pkg.go.dev next to the identifier) and tests. If the function ends with an //…
A structured answer matters more than any one tool: Correlate metrics : CPU, RSS, goroutine count, GC cycles/sec and GC CPU fraction ( runtime/metrics ), request rate,…
Benchmarks show allocs/op, but nobody reads benchmark output on every PR. For code that must stay allocation-free (encoders, loggers, request routing), lock it in with a…
b.RunParallel(body) starts GOMAXPROCS × SetParallelism(p) goroutines (p defaults to 1) and splits the b.N iterations among them; each goroutine loops on pb.Next() . The…
The old map was an array of buckets holding 8 entries each, with overflow buckets chained on collisions and incremental evacuation into a doubled array on growth. Go…
The mark phase must scan every live object that contains pointers . Objects whose type has no pointers are allocated in "noscan" spans and are never scanned. So GC CPU…