Part 2 — Tools that talk to systems · wrap-up
Cheat sheet & self-check
Every command from this section on one page.
Building blocks
flag.String("cluster", "", "cluster name") | Standard-library flag |
go get github.com/spf13/cobra@latest | Add cobra for subcommands |
slog.New(slog.NewJSONHandler(os.Stderr, nil)) | Structured JSON logs to stderr |
signal.NotifyContext(ctx, os.Interrupt, syscall.SIGTERM) | A context cancelled on Ctrl-C or SIGTERM |
os.Exit(2) | Exit with a specific code (deferred calls don't run!) |
Patterns
client := &http.Client{Timeout: 10 * time.Second} | Never use a client without a timeout |
req, _ := http.NewRequestWithContext(ctx, http.MethodGet, url, nil) | Request that stops when ctx is cancelled |
defer resp.Body.Close() | Always close bodies (connection reuse) |
json.NewDecoder(resp.Body).Decode(&out) | Decode JSON straight from the response |
if resp.StatusCode >= 300 { } | Errors from the server are not Go errors: check the status |
Patterns
var wg sync.WaitGroup; wg.Add(1); go func() { defer wg.Done() }() | Start work and wait for it |
g, ctx := errgroup.WithContext(ctx); g.SetLimit(10) | Run tasks, at most 10 at once, cancel all on first error |
select { case v := <-ch: case <-ctx.Done(): } | Wait for whichever happens first |
var mu sync.Mutex; mu.Lock(); defer mu.Unlock() | Protect shared state |
go test -race ./... / go run -race . | Detect data races |
