info: add requires_sudo? and --machine-readable

[phinze]

• requires_sudo reports whether we can detect that the cask
  install will need sudo privileges
• the --machine-readable option for brew cask info changes the
  output format to JSON, which is more suitable for parsing from
  external tools

I didn't attempt to include artifact details in the JSON output just
yet, since the first goal of this feature is just to tell
puppet-brewcask (in boxen) whether it need sudo for a given Cask.

refs #8817

[vitorgalvao]

A note on --machine-readable: @rolandwalker commented on us not having any pseudo-verbs starting with -- as being a good thing. This would break that streak.

[phinze]

Oh hey - thanks for the pointer to that conversation; I had missed it.

Unlike the linked example, to me this does feel like more of a true modifier to the info verb than new pseudo-verb. It's the same data, just formatted differently. Like I could see the flag just as easily being called --format={text,json} (an option defaulting to text).

Does it strike you more as something needing a fresh top-level verb?

[radeksimko]

Like I could see the flag just as easily being called --format={text,json} (an option defaulting to text).

👍 I was thinking that json may not be the best format if/when people start parsing it with standard shell tools like grep, awk etc. Nonetheless it will obviously do the job for puppet-brewcask as we can parse JSON in Ruby there.

[vitorgalvao]

Does it strike you more as something needing a fresh top-level verb?

Not necessarily. I linked to it because it seemed relevant to the conversation, but I have no strong opinion on the subject either way.

I was thinking that json may not be the best format if/when people start parsing it with standard shell tools like grep, awk etc.

Agreed.

[phinze]

I was thinking that json may not be the best format if/when people start parsing it with standard shell tools like grep, awk etc.

I hear you there, and I did give it some thought before deciding on JSON. The trouble is that there's not really a standard "awk-friendly" data format, whereas JSON is becoming more and more widely used for data transport between tools. Personally I've found in the last year or so jq has become indispensable for day-to-day unix piping needs.

All that being said, I see there could be a good case for us to add an additional awk/grep friendly output format - so how about I make room for that possibility by switching to the --format=json arg style?

[radeksimko]

so how about I make room for that possibility by switching to the --format=json arg style?

👍

[radeksimko]

Regarding the "awk-friendly" data format, I think we could be inspired by vagrant:
https://github.com/mitchellh/vagrant/blob/5e6b2bc55b873e2fb17ea115911c7717f2efa968/lib/vagrant/ui.rb#L72-106

[rolandwalker]

+1. JSON support is a great thing, even if we support other formats later.

The default text format of info could use some love, too.

[blardo]

👍 Working with Boxen to install Vagrant and VirtualBox runs into this same issue. This work is foundational in making that integration smoother. Excited for when this hits.

[tamasflamich]

Thank you for this. Adding --machine-readable would be a welcome change for me as well. May I suggest to include the version of the formula in the JSON too?

[sebroeder]

Is there anything we can do to get this PR completed/merged? Asking for a Boxen user 😆

[phinze]

Anyone else is welcome to pick this up. For now - closing it to reduce noise on the repo.