Commander.js: How to stop parsing for options after first argument?

[karfau]

I have the following usage pattern:

runex [options] <module> [args...]
where args should be everything that is listed after module, even if it matches any option I have configured using .option().

Is there a way to tell commander to stop on the first positional argument it found and pass the rest to args?
Or a way to hook into commander event listeners to achieve this?
(A related question of understanding: are arguments only supported on commands or also on the root level?)

I tried the following code:

const runnable$ = '<runnable> [args...]';
new Command('runex').usage(`[options] ${runnable$}`)
  .option(
    '-r, --require <module>', '0..n modules for node to require', collectDistinct, []
  )
  .arguments(runnable$)
  .parse(process.argv]

but the following case is not working as expected:

runex -r ts-node/register path/to/script -r

since it exits with

error: option '-r, --require <module>' argument missing

I'm not sure but #193 might be related.

I'm happy to support adding this as a new option.

[shadowspawn]

Is there a way to tell commander to stop on the first positional argument it found and pass the rest to args?

No, but you can use -- to tell Commander to stop looking for options and treat the remaining arguments as operands. e.g.

runex -r ts-node/register path/to/script -- -r

https://pubs.opengroup.org/onlinepubs/9699919799/basedefs/V1_chap12.html#tag_12_02

Guideline 10:
The first -- argument that is not an option-argument should be accepted as a delimiter indicating the end of options. Any following arguments should be treated as operands, even if they begin with the '-' character.

[shadowspawn]

(A related question of understanding: are arguments only supported on commands or also on the root level?)

Both. You are actually adding arguments to the root level command in your example!

See https://github.com/tj/commander.js#specify-the-argument-syntax

You use .arguments to specify the arguments for the top-level command, and for subcommands they are included in the .command call.

[shadowspawn]

I'm happy to support adding this as a new option.

Positional option parsing does come up sometimes in issues, and I am interested, if you want to look into it. I suspect it may be tricky to add!

[karfau]

In the mean time I switched to yargs-parser that offers that option. And since my lib is still small regarding options, I'm handwriting the usage part for now.
Feel free to close this issue.

[shadowspawn]

Related previous issue: #1033

Closing as details provided and OP no longer interested.

[karfau]

Actually I'm still not convinced by my move away from commander. I like it's "fail early" approach more than the flexibility others offer.

Where/How would you suggest to add this feature?
Do you think it should be implemented as some (opt-in) option, so it is not a breaking change?
I will start to have a look at the code base, to get some ideas, but any hint is welcome.

[shadowspawn]

Where/How would you suggest to add this feature?

parseOptions removes the options it recognises and splits the remaining arguments into "known" and "unknown". This is where you could stop processing options when find an "operand" (i.e. something which is not an option or an option value).

(The approach of splitting the arguments is flawed because it can change the argument order, and I am hoping to eventually have a go at rewriting the logic... Edit: working on it now...)

Note there is special processing for -- and everything afterwards gets thrown into "known". I think this new mode would be somewhat the opposite, all the options after the first operand would get thrown into "unknown" for later processing.

Do you think it should be implemented as some (opt-in) option, so it is not a breaking change?

Yes. Backwards compatibility is important.

My first thought is most people will either want (all) global options, or (all) positional options. I'm thinking there is a new call to switch the default behaviour (rather than control per option).

e.g. .makeOptionsPositional

I will start to have a look at the code base, to get some ideas, but any hint is welcome.

Good luck. I'm busy elsewhere in the code so won't be doing a deep dive at the moment, but ask if you have questions I might be able to help with in the meantime.

[shadowspawn]

Opened PR for rewrite of .parseOptions to fix argument ordering issues: #1138

[karfau]

If I understand your last message correctly, it will be much easier to add the feature I requested in v5 / when that PR landed, right?

(Sorry for the above redundant "noise", was just a rebase of a branch. I removed the link from the commit message now.)

[shadowspawn]

A little easier, and a lot less likely to get a merge conflict. :-)

[shadowspawn]

Second PR open now too: #1145
I am not working on any other havoc to .parseOptions()!

[shadowspawn]

Found another related issue for positional options: #797

[shadowspawn]

No activity in a month. I did some research and rethinking the higher level goals. Closing in favour of #1229.

[shadowspawn]

Pull Request opened to add .enablePositionalOptions() and .passThroughOptions(): #1427

[karfau]

Sorry, currently I'm having a hard time following all the threads and descriptions to solve my requirement.
Are you able to answer my initial question with relation to the features available now?

Thx

[shadowspawn]

Just add .passThroughOptions():

const runnable$ = '<runnable> [args...]';
new Command('runex').usage(`[options] ${runnable$}`)
  .passThroughOptions() // <------
  .option(
    '-r, --require <module>', '0..n modules for node to require', collectDistinct, []
  )
  .arguments(runnable$)
  .action((runnable, args, options, command) => {
    console.log(`runex: ${runnable} ${args}`);
  })
  .parse(process.argv)

% node index.js -r ts-node/register path/to/script -r 
runex: path/to/script -r