Oauth Model Concern Test Coverage Stubs

Question

I am trying to figure out the best way to reach 100% test coverage on this class. I have outlined my full spec and I am hoping someone can point me in the right direction. My assumption is stubbing the Oauth2 request would do this, but I can not seem to make that work.

I'm using Rails 4.

Spec

RSpec.describe 'AppOmniAuthentication', type: :concern do
  let(:klass) { User }
  let(:user) { create(:user) }
  let(:user_oauth_json_response) do
    unfiltered_oauth_packet = load_json_fixture('app_omni_authentication_spec')
    unfiltered_oauth_packet['provider'] = unfiltered_oauth_packet['provider'].to_sym
    unfiltered_oauth_packet['uid'] = unfiltered_oauth_packet['uid'].to_i
    unfiltered_oauth_packet
  end

  before do
    OmniAuth.config.test_mode = true
    OmniAuth.config.mock_auth[:app] = OmniAuth::AuthHash.new(
      user_oauth_json_response,
      credentials: { token: ENV['APP_CLIENT_ID'], secret: ENV['APP_CLIENT_SECRET'] }
    )
  end

  describe '#from_omniauth' do
    let(:app_oauth) { OmniAuth.config.mock_auth[:app] }

    it 'returns varying oauth related data for Bigcartel OAuth response' do
      data = klass.from_omniauth(app_oauth)
      expect(data[:provider]).to eq(user_oauth_json_response['provider'].to_s)
      expect(data[:uid]).to eq(user_oauth_json_response['uid'].to_s)
      expect(data[:email]).to eq(user_oauth_json_response['info']['email'])
      expect(data[:customer_ids]).to eq(user_oauth_json_response['extra']['raw_info']['customer_ids'])
    end
  end

  describe '#refresh_access_token!' do
    it 'false if OAuth2 Fails' do
      allow(user).to receive(:access_token) { true }
      allow(user).to receive(:refresh_access_token!) { false }
      allow(user).to receive(:result).and_raise(OAuth2::Error)
      expect(user.refresh_access_token!).to be_falsey
    end

    it 'false if refresh fails' do
      allow(user).to receive(:access_token) { true }
      allow(user).to receive(:refresh_access_token!) { false }
      expect(user.refresh_token!).to be_falsey
    end

    it 'true if new token' do
      allow(user).to receive(:access_token) { true }
      allow(user).to receive(:refresh_access_token!) { true }
      expect(user.refresh_token!).to be_truthy
    end

    it 'true when refreshed' do
      auth_token = OpenStruct.new(token: FFaker::Lorem.characters(50),
                                  refresh_token: FFaker::Lorem.characters(50),
                                  expires_at: 5.days.from_now)
      allow(user).to receive_message_chain('access_token.refresh!') { auth_token }
      expect(user.refresh_access_token!).to be_truthy
    end
  end

  describe '#refresh_token!' do
    it 'false if no access token' do
      allow(user).to receive(:access_token) { false }
      expect(user.refresh_token!).to be_falsey
    end

    it 'false if refresh fails' do
      allow(user).to receive(:access_token) { true }
      allow(user).to receive(:refresh_access_token!) { false }
      expect(user.refresh_token!).to be_falsey
    end

    it 'true if new token is saved' do
      allow(user).to receive(:access_token) { true }
      allow(user).to receive(:refresh_access_token!) { true }
      expect(user.refresh_token!).to be_truthy
    end
  end

  describe '#token expired?' do
    it 'true if valid' do
      expect(user.token_expired?).to be_falsey
    end

    it 'false if expired' do
      user.token_expires_at = 10.days.ago
      expect(user.token_expired?).to be_truthy
    end
  end
end

UPDATE

I altered the current spec to:

it 'false if OAuth2 Fails' do
  allow(OAuth2::AccessToken).to receive(:access_token) { Class.new }
  allow(OAuth2::AccessToken).to receive_message_chain('access_token.refresh!') { raise OAuth2::Error.new('ERROR') }
  binding.pry
end

However, now I am getting the following error:

WebMock::NetConnectNotAllowedError: Real HTTP connections are disabled. Unregistered request..

Answer 1

It looks like the key is the second line here:

def refresh_access_token!
  result = access_token.refresh!
  store_token(result)
  save
rescue OAuth2::Error
  false
end

All your other tests stub access_token. If you had a test that called the real method, it would cover the missing lines in access_token, client, strategy, and settings.

Now strategy,settings, and client are all pretty boring: the first two are essentially just constants. And client doesn't do anything just by initializing it. So calling those should be no big deal. That leaves just access_token. You can see that that initializes an OAuth2::AccessToken, whose code is on Github. The initializer is also pretty boring. It just saves the inputs to use later.

So I would stub its refresh! method: once to return a valid-looking refresh token, and once to raise an OAuth2::Error. You can use expect_any_instance_of to do that. If that makes you uneasy you could also stub :new itself, and have it return your own fake object:

o = expect(OAuth2::AccessToken).to receive(:new) { Object.new }
expect(o).to receive(:refresh!) { raise OAuth2::Error.new("something here") }

It looks like it might be a little bit of a nuisance to construct an OAuth2::Error since it takes a request object, but I don't see anything too complicated there, just messy.

That should give you 100% coverage!

Answer 2

(Posted solution on behalf of the question author).

This is now solved.

  describe '#refresh_access_token!' do
    it 'false if OAuth2 Fails' do
      response = OpenStruct.new(error: 'ERROR')
      allow(OAuth2::AccessToken).to receive_message_chain(:new, :refresh!) { raise OAuth2::Error.new(response) }
      expect(user.refresh_token!).to be_falsey
    end

    it 'true if new token' do
      allow(OAuth2::AccessToken).to receive_message_chain(:new, :refresh!) { @auth_token }
      expect(user.refresh_token!).to be_truthy
    end
  end

The stubbed token looks like:

@auth_token = OpenStruct.new(token: FFaker::Lorem.characters(50),
                           refresh_token: FFaker::Lorem.characters(50),
                           expires_at: 5.days.from_now)

After researching the concept that Paul Jungwirth posted, stubbing the methods in the gem prevented the Webmock alert from firing (as intended) and I could trigger the raise correctly. The raise had to match the gem's response setup as well.