Also flatten-duplicate-arrays when used with dot-notation

[JeroenVdb]

Background: yargs/yargs#777

When using dot-notations I also want arrays to be flattened. This was the behaviour in Yargs 5 but not anymore in Yargs 6.

# GOOD
$ node good-yargs-parser.js --metric.list AA --metric.list BB --metricList CC --metricList DD
{ _: [],
  metric: { list: [ 'AA', 'BB' ] },
  metricList: [ 'CC', 'DD' ],
  '$0': 'yargstest.js' }

# BAD
$ node bad-yargs-parser.js --metric.list AA --metric.list BB --metricList CC --metricList DD
{ _: [],
  metric: { list: [ 'AA', ['BB'] ] },
  metricList: [ 'CC', 'DD' ],
  '$0': 'yargstest.js' }

[bcoe]

@JeroenVdb at this time we don't yet fully support setting types for inner keys on objects, e.g.,

foo.bar: array, I think that adding this support could potentially lead to a combinatoric explosion of little tweaks we need to make in the parser -- I think I'd prefer that, if you want to enforce types on an object, you instead apply a coerce function on the top-level key:

something like this:

var argv = require('./')('--foo.a 99 --foo.b 33')
  .coerce('foo', function (obj) {
    obj.a = obj.a * 2
    obj.b = obj.b * 2
    return obj
  })
  .argv

console.log(argv)

Having said this!, I think you're correct in your analysis of yargs/yargs#777:_

It feels like --metric.list AA --metric.list BB should create an array on the list key, not two arrays.

Would you be willing to modify your patch to bring us back to the 5.x behavior?

[JeroenVdb]

@bcoe I don't think I understand correct. Doesn't my patch restore the previous 5.X behaviour?

[bcoe]

@JeroenVdb I just didn't like that we were starting to check foo.bar for object variables ... perhaps this is the minimal change to get behavior back to a reasonable spot though 👍

I'm going to be at Hack Illinois this coming weekend, working with students on yargs, expect a lot of things to land (also, this is a good time to get additional issues open).

[bcoe]

@JeroenVdb closing in favor of: #86 (which I believe solves the more general case).