Makes JSON resource enumerable, despite method_missing magic

[TheLonelyGhost]

Per Slack discussion:

obj = json(content: '[{"Name": "abc"}, {"Name": "def"}]')

obj.each do |node|
  describe node do
    it { should eq nil }
  end
end

Above does not respond as expected due to method_missing magic. One expects 2 tests to execute, but each does not iterate over its data. This PR corrects that and includes all of the same enumerable abilities on the json (and by extension, yaml) resource that its data contains.

Yes, I am aware this could also be accomplished by doing obj.params.each do instead of obj.each do, but I don't feel this is very intuitive. We're exposing the :[] method via method_missing, why not make it react like a hash or array too?

[jquick]

Thanks @TheLonelyGhost ! This looks pretty slick. You mind adding a unit test for this?

[TheLonelyGhost]

@jquick I believe I did under test/unit/resources/json_test.rb. Or is there something more you’d like me to do with regards to test support?

[jquick]

Thanks @TheLonelyGhost !

[TrevorBramble]

This is great, just have a couple of questions. =^)

[TrevorBramble]

Reading above, and parse(), it looks like there should always be a valid hash for @params, what's the scenario you have in mind where this condition would be false?

[TheLonelyGhost]

I originally considered json(content: 'null') or something similar that is neither an array nor a JSON object. Upon actually testing this just now, I see that JSON.parse() actually fails with anything other than those two. That was news to me.

From the look of things "valid JSON" was once defined as either an array or object, but ECMAScript 5 expanded it to include any data type such as "fizz buzz", null, 3.14, and others. It doesn't appear ruby's JSON parser has caught up yet.

At this point, this line doesn't really hurt anything, but builds in support for when and if the JSON parser is updated to the ES5 standard definition of "valid" JSON.

[TrevorBramble]

This is purpose built for params instead of delegation to some injected receiver, so I wonder if Params should be part of the name? Maybe EnumerableParamsDelegator?

[TheLonelyGhost]

I didn't like the dependency on the params attribute either, but minimum viable solution and all. I agree that your suggestion is a better name, but I'm definitely more interested in if you have a solution that delegates to an injected receiver. Ideas?

[clintoncwolfe]

Seems sensible, and the test makes sense and passes. Thanks, @TheLonelyGhost