Multiple positionals with multiple values

[superstator]

I want to build a CLI in Rust, using clap, which can accept some number of positional arguments with a set number of values each, and a final argument accepting whatever positional values were left over. For example, ./app a b c with a positional argument accepting 2 values, and a positional argument accepting leftovers, would assign a b to the first arg, and c to the second. I'm able to do this with a positional argument that accepts a single value:

        let cmd = Command::new("main")
            .arg(Arg::new("ONE").action(ArgAction::Set).num_args(1))
            .arg(Arg::new("TWO").action(ArgAction::Append));

        let matches = cmd.get_matches_from(&["bin","a","b","c"]);

        assert_eq!(matches.get_many::<String>("ONE").unwrap().collect::<Vec<_>>(), &["a"]);
        assert_eq!(matches.get_many::<String>("TWO").unwrap().collect::<Vec<_>>(), &["b","c"]);

But if I try to have the first arg accept two values like this:

        let cmd = Command::new("main")
            .arg(Arg::new("ONE").action(ArgAction::Set).num_args(2))
            .arg(Arg::new("TWO").action(ArgAction::Append));

        let matches = cmd.get_matches_from(&["bin","a","b","c"]);

        assert_eq!(matches.get_many::<String>("ONE").unwrap().collect::<Vec<_>>(), &["a","b"]);
        assert_eq!(matches.get_many::<String>("TWO").unwrap().collect::<Vec<_>>(), &["c"]);

I get the error Positional argument [TWO]... *must* have required(true) or last(true) set because a prior positional argument ([ONE] [ONE]) has num_args(1..). If I make argument TWO required, I just get a different error, and if I make them both required the last argument is always attached to TWO, making it impossible to accept something like ./app a b or ./app a b c d and still have ONE capture the first two values.

The only thing I've found that gets me the functionality I want is to have a single generic positional argument with .action(ArgAction::Append) and then carve up those matches myself, but then I lose out on the helpful help messages clap would otherwise generate. Is there another way?

[epage]

The asserts and parser have likely not been hardened against accepting a fixed num_args and instead assumes num_args is variable.

Feel free to open an issue about this.

[superstator]

Ok. If I dig into it a little would you be open to a PR?