Misleading "Usage:" when slice option and positional arguments?

[marco-m]

Hello, thanks a lot for go-arg, I love it!

consider the following program (using go-arg v1.1.0)

package main

import (
	"fmt"
	arg "github.com/alexflint/go-arg"
)

func main() {
	var args struct {
		Foo []string
		A   string `arg:"positional,required"`
	}
	arg.MustParse(&args)
	fmt.Printf("%+v\n", args)
}

Let's ask for its help:

$ ./bin/positional -h
Usage: positional [--foo FOO] A

Positional arguments:
  A

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

If I try to follow the "Usage" example, I get:

$ ./bin/positional --foo 1 x
Usage: positional [--foo FOO] A
error: a is required

which is very confusing for the end user (and at first for the programmer too, then the programmer understands the reason :-D)

After thinking about it, I tried the classic -- to mean end of flag processing, but it didn't work:

 $ ./bin/positional --foo 1 -- x
Usage: positional [--foo FOO] A
error: a is required

After a while I saw the light and inverted the order, and it worked as expected:

$ ./bin/positional x --foo 1
{Foo:[1] A:x}

So I guess I have two questions:

1. Would it be possible/suitable to make the -- trick work? (Is this maybe a regression of Positional arguments with dashes #55 ?)
2. Would it make sense, in presence of a required positional argument AND a slice of options, to invert the usage, something like:

Usage: positional A [--foo FOO]

or any other way to give an hint to the user?

[alexflint]

Woah this is definitely a bug. Both of the examples you gave are supposed to work:

./bin/positional --foo 1 x
./bin/positional --foo 1 -- x

If these are broken then go-arg has a regression in a recent release. I'll investigate.

[alexflint]

Oh on first read I didn't see that foo is a []string. Given that, I still expect the second example you gave to work (but not the first):

./bin/positional --foo 1 -- x   // is supposed to work given the program you gave
./bin/positional --foo 1 x     // not expected to work

[marco-m]

Would it make sense then to change the usage text to mention the -- if the positional is required? Maybe:

Usage: positional [--foo FOO --] A