Xe's Go Style Guide

Write Go code following the conventions and patterns used by Xe Iaso.

Project Structure

├── cmd/            # Main applications (each subdirectory is a binary)
│   ├── x/          # CLI with subcommands
│   │   ├── main.go
│   │   └── cmd/    # Subcommand packages
│   └── sakurajima/ # Service binaries
│       ├── main.go
│       └── internal/  # Command-specific internal packages
├── internal/       # Private packages shared across commands
├── web/            # Web-related services and API clients
└── gen/            # Generated code (protobuf)

Package naming: lowercase, single words preferred (slog, flagenv, kahless).

CLI Pattern (CRITICAL)

All command-line tools MUST call flagenv.Parse() and then flag.Parse().

package main

import (
    "flag"

    "github.com/facebookgo/flagenv"
)

var (
    bind  = flag.String("bind", ":8080", "HTTP bind address")
    dbLoc = flag.String("db-loc", "", "Database location")
)

func main() {
    flagenv.Parse()
    flag.Parse()
}

Configuration loading order: flags → env vars (flagenv) → flags (final override).

Subcommands

Use github.com/google/subcommands for subcommand-based CLIs:

package main

import (
    "context"
    "flag"
    "os"

    "github.com/facebookgo/flagenv"
    "github.com/google/subcommands"
)

func main() {
    flagenv.Parse()

    subcommands.Register(subcommands.HelpCommand(), "")
    subcommands.Register(subcommands.FlagsCommand(), "")
    subcommands.Register(subcommands.CommandsCommand(), "")
    subcommands.Register(&serveCmd{}, "")
    subcommands.Register(&versionCmd{}, "")

    os.Exit(int(subcommands.Execute(context.Background())))
}

type serveCmd struct {
    dbLoc string
}

func (c *serveCmd) Name() string     { return "serve" }
func (c *serveCmd) Synopsis() string { return "Start the server" }
func (c *serveCmd) Usage() string {
    return "serve [flags]\nStart the HTTP server.\n"
}

func (c *serveCmd) SetFlags(f *flag.FlagSet) {
    f.StringVar(&c.dbLoc, "db-loc", "", "Database location")
}

func (c *serveCmd) Execute(ctx context.Context, f *flag.FlagSet, args ...interface{}) subcommands.ExitStatus {
    // Implementation
    return subcommands.ExitSuccess
}

Flag names use kebab-case: --db-loc, --grpc-bind, --slog-level.

Error Handling

Define package-level sentinel errors:

var ErrNotFound = errors.New("store: key not found")
var ErrCantDecode = errors.New("store: can't decode value")

Wrap errors with context using %w:

if err != nil {
    return nil, fmt.Errorf("failed to connect to database: %w", err)
}
return nil, fmt.Errorf("ollama: error encoding request: %w", err)

Combine multiple validation errors with errors.Join:

func (t *Toplevel) Valid() error {
    var errs []error
    if err := t.Bind.Valid(); err != nil {
        errs = append(errs, fmt.Errorf("invalid bind block:\n%w", err))
    }
    if len(errs) != 0 {
        return fmt.Errorf("invalid configuration file:\n%w", errors.Join(errs...))
    }
    return nil
}

Check errors with errors.Is() and errors.As():

if errors.Is(err, ErrNotFound) {
    // handle not found
}

Custom error types implement Error() and slog.LogValue():

type Error struct {
    WantStatus, GotStatus int
    URL                   *url.URL
    Method                string
    ResponseBody          string
}

func (e Error) Error() string {
    return fmt.Sprintf("%s %s: wanted status code %d, got: %d: %v",
        e.Method, e.URL, e.WantStatus, e.GotStatus, e.ResponseBody)
}

func (e Error) LogValue() slog.Value {
    return slog.GroupValue(
        slog.Int("want_status", e.WantStatus),
        slog.Int("got_status", e.GotStatus),
        slog.String("url", e.URL.String()),
        slog.String("method", e.Method),
        slog.String("body", e.ResponseBody),
    )
}

Logging

Use log/slog (never the plain log package). Output JSON to stderr. Use -slog-level flag for runtime control.

slog.Info("starting up",
    "bind", *bind,
    "db-loc", *dbLoc,
)
slog.Error("failed to create dao", "err", err)
slog.Debug("processing", "count", len(items))

Always use "err" as the key for errors. Implement LogValue() for complex types:

func (s *Show) LogValue() slog.Value {
    return slog.GroupValue(
        slog.String("title", s.GetTitle()),
        slog.String("disk_path", s.GetDiskPath()),
    )
}

HTTP Patterns

Middleware pattern:

func PasswordMiddleware(username, password string, next http.Handler) http.Handler {
    return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
        user, pass, ok := r.BasicAuth()
        if !ok || user != username || pass != password {
            http.Error(w, "Unauthorized", http.StatusUnauthorized)
            return
        }
        next.ServeHTTP(w, r)
    })
}

Chaining middleware:

var h http.Handler = topLevel
h = xffMW.Handler(h)
h = cors.Default().Handler(h)
h = FlyRegionAnnotation(h)

HTTP client with context and error handling:

func (c *Client) doRequest(ctx context.Context, method, path string, wantCode int, body io.Reader) (*http.Response, error) {
    req, err := http.NewRequestWithContext(ctx, method, u.String(), body)
    if err != nil {
        return nil, fmt.Errorf("failed to create request: %w", err)
    }

    resp, err := c.http.Do(req)
    if err != nil {
        return nil, fmt.Errorf("request failed: %w", err)
    }

    if resp.StatusCode != wantCode {
        return nil, web.NewError(wantCode, resp)
    }
    return resp, nil
}

Server lifecycle with errgroup:

g, gCtx := errgroup.WithContext(ctx)

g.Go(func() error {
    ln, err := net.Listen("tcp", cfg.Bind.HTTP)
    if err != nil {
        return fmt.Errorf("failed to listen: %w", err)
    }
    return rtr.HandleHTTP(gCtx, ln)
})

return g.Wait()

Testing

Tests co-located with source (*_test.go). Use table-driven tests:

func TestDomainValid(t *testing.T) {
    t.Parallel()

    for _, tt := range []struct {
        name        string
        input       Domain
        err         error
        errContains string
    }{
        {name: "simple happy path", input: Domain{Name: "example.com"}},
        {name: "invalid domain", input: Domain{Name: "\uFFFD.com"}, err: ErrInvalidDomainName},
    } {
        t.Run(tt.name, func(t *testing.T) {
            err := tt.input.Valid()
            if tt.err != nil {
                if !errors.Is(err, tt.err) {
                    t.Logf("want: %v", tt.err)
                    t.Logf("got:  %v", err)
                    t.Error("got wrong error")
                }
            }
        })
    }
}

Testing best practices:

• Use tt as the loop variable in table-driven tests
• Always include name field for subtests
• Use t.Helper() for helper functions
• Use t.Parallel() for independent tests
• Use t.TempDir() for temporary directories
• Use httptest.NewServer() for HTTP mocking
• Use errors.Is() for error comparison
• Include t.Logf("want: %v", x) and t.Logf("got: %v", y) for debugging

func loadConfig(t *testing.T, fname string) config.Toplevel {
    t.Helper()
    // ...
}

Code Style

• Use go fmt/goimports for formatting
• Tabs for indentation
• camelCase for variables
• PascalCase for exported identifiers
• Files use snake_case
• Packages use lower-case module names