blob: ec3a6bb06b78b889491fff5978f56f4cd00b824c [file] [log] [blame]
// Copyright 2018 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
// Package tool is a harness for writing Go tools.
package tool
import (
// This file is a harness for writing your main function.
// The original version of the file is in
// It adds a method to the Application type
// Main(name, usage string, args []string)
// which should normally be invoked from a true main as follows:
// func main() {
// (&Application{}).Main("myapp", "non-flag-command-line-arg-help", os.Args[1:])
// }
// It recursively scans the application object for fields with a tag containing
// `flag:"flagnames" help:"short help text"``
// uses all those fields to build command line flags. It will split flagnames on
// commas and add a flag per name.
// It expects the Application type to have a method
// Run(context.Context, args...string) error
// which it invokes only after all command line flag processing has been finished.
// If Run returns an error, the error will be printed to stderr and the
// application will quit with a non zero exit status.
// Profile can be embedded in your application struct to automatically
// add command line arguments and handling for the common profiling methods.
type Profile struct {
CPU string `flag:"profile.cpu" help:"write CPU profile to this file"`
Memory string `flag:"profile.mem" help:"write memory profile to this file"`
Trace string `flag:"profile.trace" help:"write trace log to this file"`
// Application is the interface that must be satisfied by an object passed to Main.
type Application interface {
// Name returns the application's name. It is used in help and error messages.
Name() string
// Most of the help usage is automatically generated, this string should only
// describe the contents of non flag arguments.
Usage() string
// ShortHelp returns the one line overview of the command.
ShortHelp() string
// DetailedHelp should print a detailed help message. It will only ever be shown
// when the ShortHelp is also printed, so there is no need to duplicate
// anything from there.
// It is passed the flag set so it can print the default values of the flags.
// It should use the flag sets configured Output to write the help to.
// Run is invoked after all flag processing, and inside the profiling and
// error handling harness.
Run(ctx context.Context, args ...string) error
type SubCommand interface {
Parent() string
// This is the type returned by CommandLineErrorf, which causes the outer main
// to trigger printing of the command line help.
type commandLineError string
func (e commandLineError) Error() string { return string(e) }
// CommandLineErrorf is like fmt.Errorf except that it returns a value that
// triggers printing of the command line help.
// In general you should use this when generating command line validation errors.
func CommandLineErrorf(message string, args ...interface{}) error {
return commandLineError(fmt.Sprintf(message, args...))
// Main should be invoked directly by main function.
// It will only return if there was no error. If an error
// was encountered it is printed to standard error and the
// application exits with an exit code of 2.
func Main(ctx context.Context, app Application, args []string) {
s := flag.NewFlagSet(app.Name(), flag.ExitOnError)
if err := Run(ctx, s, app, args); err != nil {
fmt.Fprintf(s.Output(), "%s: %v\n", app.Name(), err)
if _, printHelp := err.(commandLineError); printHelp {
// TODO(adonovan): refine this. It causes
// any command-line error to result in the full
// usage message, which typically obscures
// the actual error.
// Run is the inner loop for Main; invoked by Main, recursively by
// Run, and by various tests. It runs the application and returns an
// error.
func Run(ctx context.Context, s *flag.FlagSet, app Application, args []string) error {
s.Usage = func() {
if app.ShortHelp() != "" {
fmt.Fprintf(s.Output(), "%s\n\nUsage:\n ", app.ShortHelp())
if sub, ok := app.(SubCommand); ok && sub.Parent() != "" {
fmt.Fprintf(s.Output(), "%s [flags] %s", sub.Parent(), app.Name())
} else {
fmt.Fprintf(s.Output(), "%s [flags]", app.Name())
if usage := app.Usage(); usage != "" {
fmt.Fprintf(s.Output(), " %s", usage)
fmt.Fprint(s.Output(), "\n")
p := addFlags(s, reflect.StructField{}, reflect.ValueOf(app))
if err := s.Parse(args); err != nil {
return err
if p != nil && p.CPU != "" {
f, err := os.Create(p.CPU)
if err != nil {
return err
if err := pprof.StartCPUProfile(f); err != nil {
return err
defer pprof.StopCPUProfile()
if p != nil && p.Trace != "" {
f, err := os.Create(p.Trace)
if err != nil {
return err
if err := trace.Start(f); err != nil {
return err
defer func() {
log.Printf("To view the trace, run:\n$ go tool trace view %s", p.Trace)
if p != nil && p.Memory != "" {
f, err := os.Create(p.Memory)
if err != nil {
return err
defer func() {
runtime.GC() // get up-to-date statistics
if err := pprof.WriteHeapProfile(f); err != nil {
log.Printf("Writing memory profile: %v", err)
return app.Run(ctx, s.Args()...)
// addFlags scans fields of structs recursively to find things with flag tags
// and add them to the flag set.
func addFlags(f *flag.FlagSet, field reflect.StructField, value reflect.Value) *Profile {
// is it a field we are allowed to reflect on?
if field.PkgPath != "" {
return nil
// now see if is actually a flag
flagNames, isFlag := field.Tag.Lookup("flag")
help := field.Tag.Get("help")
if isFlag {
nameList := strings.Split(flagNames, ",")
// add the main flag
addFlag(f, value, nameList[0], help)
if len(nameList) > 1 {
// and now add any aliases using the same flag value
fv := f.Lookup(nameList[0]).Value
for _, flagName := range nameList[1:] {
f.Var(fv, flagName, help)
return nil
// not a flag, but it might be a struct with flags in it
value = resolve(value.Elem())
if value.Kind() != reflect.Struct {
return nil
p, _ := value.Addr().Interface().(*Profile)
// go through all the fields of the struct
for i := 0; i < value.Type().NumField(); i++ {
child := value.Type().Field(i)
v := value.Field(i)
// make sure we have a pointer
if v.Kind() != reflect.Ptr {
v = v.Addr()
// check if that field is a flag or contains flags
if fp := addFlags(f, child, v); fp != nil {
p = fp
return p
func addFlag(f *flag.FlagSet, value reflect.Value, flagName string, help string) {
switch v := value.Interface().(type) {
case flag.Value:
f.Var(v, flagName, help)
case *bool:
f.BoolVar(v, flagName, *v, help)
case *time.Duration:
f.DurationVar(v, flagName, *v, help)
case *float64:
f.Float64Var(v, flagName, *v, help)
case *int64:
f.Int64Var(v, flagName, *v, help)
case *int:
f.IntVar(v, flagName, *v, help)
case *string:
f.StringVar(v, flagName, *v, help)
case *uint:
f.UintVar(v, flagName, *v, help)
case *uint64:
f.Uint64Var(v, flagName, *v, help)
log.Fatalf("Cannot understand flag of type %T", v)
func resolve(v reflect.Value) reflect.Value {
for {
switch v.Kind() {
case reflect.Interface, reflect.Ptr:
v = v.Elem()
return v