Go to file
Pavel Borzenkov a6af419fff README: update TextUnmarshaler example
Values are much more convenient to use in argument structs, so update
README to use them instead of pointers in the example as we now support
this.

Signed-off-by: Pavel Borzenkov <pavel.borzenkov@gmail.com>
2018-11-20 12:32:32 +03:00
Godeps vendor: update go-scalar to the latest version 2018-11-20 12:18:45 +03:00
vendor/github.com vendor: update go-scalar to the latest version 2018-11-20 12:18:45 +03:00
.gitignore Initial commit 2015-10-31 18:30:06 -07:00
.travis.yml add coveralls to .travis.yml 2015-11-04 09:12:41 -08:00
LICENSE Initial commit 2015-10-31 18:30:06 -07:00
README.md README: update TextUnmarshaler example 2018-11-20 12:32:32 +03:00
doc.go Altered help tag parsing to reduce the constraints on help text content; old behaviour is retained for backward compatibility 2017-10-02 14:18:41 +01:00
example_test.go Adding separate tag option 2017-03-04 09:13:12 +07:00
parse.go Fix the problem with errors 2018-05-14 22:18:05 +03:00
parse_test.go parse_test: add tests covering new TextUnamarshaler value support 2018-11-20 12:29:36 +03:00
usage.go Custom parsers implementing encoding.TextMarshaler() can have default values printed via --help 2018-04-13 14:46:24 +10:00
usage_test.go fixed example comment, test coverage issue 2018-04-16 11:07:48 +10:00

README.md

GoDoc Build Status Coverage Status Report Card

Structured argument parsing for Go

go get github.com/alexflint/go-arg

Declare the command line arguments your program accepts by defining a struct.

var args struct {
	Foo string
	Bar bool
}
arg.MustParse(&args)
fmt.Println(args.Foo, args.Bar)
$ ./example --foo=hello --bar
hello true

Required arguments

var args struct {
	ID      int `arg:"required"`
	Timeout time.Duration
}
arg.MustParse(&args)
$ ./example
Usage: example --id ID [--timeout TIMEOUT]
error: --id is required

Positional arguments

var args struct {
	Input   string   `arg:"positional"`
	Output  []string `arg:"positional"`
}
arg.MustParse(&args)
fmt.Println("Input:", args.Input)
fmt.Println("Output:", args.Output)
$ ./example src.txt x.out y.out z.out
Input: src.txt
Output: [x.out y.out z.out]

Environment variables

var args struct {
	Workers int `arg:"env"`
}
arg.MustParse(&args)
fmt.Println("Workers:", args.Workers)
$ WORKERS=4 ./example
Workers: 4
$ WORKERS=4 ./example --workers=6
Workers: 6

You can also override the name of the environment variable:

var args struct {
	Workers int `arg:"env:NUM_WORKERS"`
}
arg.MustParse(&args)
fmt.Println("Workers:", args.Workers)
$ NUM_WORKERS=4 ./example
Workers: 4

You can provide multiple values using the CSV (RFC 4180) format:

var args struct {
    Workers []int `arg:"env"`
}
arg.MustParse(&args)
fmt.Println("Workers:", args.Workers)
$ WORKERS='1,99' ./example
Workers: [1 99]

Usage strings

var args struct {
	Input    string   `arg:"positional"`
	Output   []string `arg:"positional"`
	Verbose  bool     `arg:"-v" help:"verbosity level"`
	Dataset  string   `help:"dataset to use"`
	Optimize int      `arg:"-O" help:"optimization level"`
}
arg.MustParse(&args)
$ ./example -h
Usage: [--verbose] [--dataset DATASET] [--optimize OPTIMIZE] [--help] INPUT [OUTPUT [OUTPUT ...]] 

Positional arguments:
  INPUT 
  OUTPUT

Options:
  --verbose, -v            verbosity level
  --dataset DATASET        dataset to use
  --optimize OPTIMIZE, -O OPTIMIZE
                           optimization level
  --help, -h               print this help message

As the example above shows, the help tag can be used in conjunction with arg, or instead. When used together, they can appear in either order.

Default values

var args struct {
	Foo string
	Bar bool
}
args.Foo = "default value"
arg.MustParse(&args)

Arguments with multiple values

var args struct {
	Database string
	IDs      []int64
}
arg.MustParse(&args)
fmt.Printf("Fetching the following IDs from %s: %q", args.Database, args.IDs)
./example -database foo -ids 1 2 3
Fetching the following IDs from foo: [1 2 3]

Arguments that can be specified multiple times, mixed with positionals

var args struct {
    Commands  []string `arg:"-c,separate"`
    Files     []string `arg:"-f,separate"`
    Databases []string `arg:"positional"`
}
./example -c cmd1 db1 -f file1 db2 -c cmd2 -f file2 -f file3 db3 -c cmd3
Commands: [cmd1 cmd2 cmd3]
Files [file1 file2 file3]
Databases [db1 db2 db3]

Custom validation

var args struct {
	Foo string
	Bar string
}
p := arg.MustParse(&args)
if args.Foo == "" && args.Bar == "" {
	p.Fail("you must provide one of --foo and --bar")
}
./example
Usage: samples [--foo FOO] [--bar BAR]
error: you must provide one of --foo and --bar

Version strings

type args struct {
	...
}

func (args) Version() string {
	return "someprogram 4.3.0"
}

func main() {
	var args args
	arg.MustParse(&args)
}
$ ./example --version
someprogram 4.3.0

Embedded structs

The fields of embedded structs are treated just like regular fields:


type DatabaseOptions struct {
	Host     string
	Username string
	Password string
}

type LogOptions struct {
	LogFile string
	Verbose bool
}

func main() {
	var args struct {
		DatabaseOptions
		LogOptions
	}
	arg.MustParse(&args)
}

As usual, any field tagged with arg:"-" is ignored.

Custom parsing

You can implement your own argument parser by implementing encoding.TextUnmarshaler:

package main

import (
	"fmt"
	"strings"

	"github.com/alexflint/go-arg"
)

// Accepts command line arguments of the form "head.tail"
type NameDotName struct {
	Head, Tail string
}

func (n *NameDotName) UnmarshalText(b []byte) error {
	s := string(b)
	pos := strings.Index(s, ".")
	if pos == -1 {
		return fmt.Errorf("missing period in %s", s)
	}
	n.Head = s[:pos]
	n.Tail = s[pos+1:]
	return nil
}

// optional: implement in case you want to display a default value in the usage string
func (n *NameDotName) MarshalText() (text []byte, err error) {
	text = []byte(fmt.Sprintf("%s.%s", n.Head, n.Tail))
	return
}

func main() {
	var args struct {
		Name NameDotName
	}
	// set default
	args.Name = NameDotName{"file", "txt"}
	arg.MustParse(&args)
	fmt.Printf("%#v\n", args.Name)
}
$ ./example --help
Usage: test [--name NAME]

Options:
  --name NAME [default: file.txt]
  --help, -h             display this help and exit

$ ./example
main.NameDotName{Head:"file", Tail:"txt"}

$ ./example --name=foo.bar
main.NameDotName{Head:"foo", Tail:"bar"}

$ ./example --name=oops
Usage: example [--name NAME]
error: error processing --name: missing period in "oops"

Description strings

type args struct {
	Foo string
}

func (args) Description() string {
	return "this program does this and that"
}

func main() {
	var args args
	arg.MustParse(&args)
}
$ ./example -h
this program does this and that
Usage: example [--foo FOO]

Options:
  --foo FOO
  --help, -h             display this help and exit

API Documentation

https://godoc.org/github.com/alexflint/go-arg

Rationale

There are many command line argument parsing libraries for Go, including one in the standard library, so why build another?

The shortcomings of the flag library that ships in the standard library are well known. Positional arguments must preceed options, so ./prog x --foo=1 does what you expect but ./prog --foo=1 x does not. Arguments cannot have both long (--foo) and short (-f) forms.

Many third-party argument parsing libraries are geared for writing sophisticated command line interfaces. The excellent codegangsta/cli is perfect for working with multiple sub-commands and nested flags, but is probably overkill for a simple script with a handful of flags.

The main idea behind go-arg is that Go already has an excellent way to describe data structures using Go structs, so there is no need to develop more levels of abstraction on top of this. Instead of one API to specify which arguments your program accepts, and then another API to get the values of those arguments, why not replace both with a single struct?

Backward Compatibility Notes

The tags have changed recently. Earlier versions required the help text to be part of the arg tag. This is still supported but is now deprecated. Instead, you should use a separate help tag, described above, which removes most of the limits on the text you can write. In particular, you will need to use the new help tag if your help text includes any commas.