@bcoe any chance of a quick merge? It's fully downwards-compatible…

@wmertens @Qard this feature feels a little bit magical to me. It's not really stopping on an unknown argument, it's essentially stopping parsing as soon as a positional argument is hit rather than a flag argument right?

In the cases where I've needed to split command line arguments (usually because one bin is instrumenting another bin) I've used a helper like this:

https://github.com/bcoe/c8/blob/master/lib/parse-args.js#L60

I wonder if instead of adding this additional parsing option, we could find a way to formalize this helper in yargs? I think this is the use-case we're all trying to solve.

I'm swamped with open-source projects these days, so don't get to all my GitHub issues, perhaps I could convince you both to join me in slack here:

http://devtoolscommunity.herokuapp.com/

☝️ we've been using this slack to discuss JavaScript tooling related issues.

FYI I use the following workaround to stop parsing at the first positional argument:

function parseToPositional(yargs: yargs.Argv, cmd: string[]) {
  let newCmd = cmd.map((c, i) => (c.startsWith('-') ? c : c + ':' + i.toString()))
  let positional = yargs.parse(newCmd)
  if (!positional._.length) return yargs.parse(cmd)

  let position = Number(positional._[0].substr(positional._[0].lastIndexOf(':') + 1))

  let result = yargs.parse(cmd.slice(0, position))
  result._ = result._.concat(cmd.slice(position))
  return result
}

Appending ':index' to each of the non-switch segments ensures that the boolean switches are correctly interpreted and that we're not finding an argument to a switch instead of a positional argument - e.g. cmd --arg positional --other other positional should not find the first positional.