Path should format when 'base' missing and 'name' and 'ext' exist

[reggi]

This test shows that name and ext are not taken into account when using path.format without a base.

The common usage this is preventing is this:

var src = path.parse(srcpath)
src.base = null
if(src.ext === '') src.ext = '.js'
src = path.format(src)

However this doesn't do anything.

var path = require('path')
var assert = require('assert')

// pulled from https://nodejs.org/api/path.html#path_path_format_pathobject
/* global describe, it */

describe('node path', function () {

  it('should work as documented', function () {
    assert.equal(path.format({
      root: '/',
      dir: '/home/user/dir',
      base: 'file.txt',
      ext: '.txt',
      name: 'file'
    }), '/home/user/dir/file.txt')
  })

  it('should not work if missing base', function () {
    assert.notEqual(path.format({
      root: '/',
      dir: '/home/user/dir',
      ext: '.txt',
      name: 'file'
    }), '/home/user/dir/file.txt')
  })

  it('should show ext and name are irrelevant', function () {
    assert.equal(path.format({
      root: '/',
      dir: '/home/user/dir',
      ext: '.txt',
      name: 'file'
    }), path.format({
      root: '/',
      dir: '/home/user/dir'
    }))
  })

})

https://github.com/nodejs/io.js/blob/master/lib/path.js#L584

I can write a pull request, if anyone else thinks this is a good idea.

[sam-github]

Your example code above is really confusing, its not clear whether you are posting a unit test that should pass, but does not pass right now, or that passes right now, but that should not.

For future readers, the problem is that the unit test does pass, and that's lame:

> path.format({ root: '/', dir: '/home/user/dir', base: 'file.txt', ext: '.js', name: 'other' })
'/home/user/dir/file.txt'

This is somewhat as I would expect, because it parallels how url.format() works when it uses host if it exists, and only uses the component hostname and port properties if host does not exist.

> path.format({ root: '/', dir: '/home/user/dir', ext: '.js', name: 'other' })
'/home/user/dir/'

This is bizarre, I would expect name and .ext to be concatenated into the file base, as would @reggi I assume.

Also, note that path.format() is undocumented: https://iojs.org/api/path.html#path_path_format_pathobject

One example of how it behaves with a single set of inputs is not documentation :-(. Its also a particularly bad example, because it implies that the root, name, and ext properties are input... they are not, they are completely ignored.

@reggi I'd +1 a fix for this, particularly one that actually added documentation in detail, in a form paralleling url.format(), that is sufficiently detailed to be able to predict the output from any set of inputs, without experimenting at the repl. For example:

> path.format({ root: '/root', base: 'helo.you', ext: '.js', name: 'other' })
'helo.you'

What happened to root? Is root even used? Seems not...

[reggi]

@sam-github Thanks for looking into the topic and sorry for the ambiguous test, should have mentioned that it passes. I completely agree on paralleling url.format().

[Trott]

Except for the documentation issues cited by @sam-github, this all appears to be fixed. If no one beats me to it (and please, feel free), I'll try to get around to documenting this soon so we can close this out.

[Trott]

Documentation landed in a97dfa0.