I'm using Devise (v2.1.2) with Omniauth for user verification. I'm working on a functional test for a controller that takes a JSON object as the POST body and thus using the technique from this question to set the raw POST body. This works fine for development, but when I run tests I get an exception on a method that's completely unauthenticated:
NoMethodError: undefined method `user' for nil:NilClass
Example test:
test "should be able to create an item" do
m = FactoryGirl.attributes_for(:item)
raw_post :create, {}, m.to_json
assert_response :success
end
None of my models have a user method, and nothing in this controller uses authentication, so I was pretty confused. A full stack trace shows that the error comes from the first line of this function in Devise:
def sign_out_all_scopes(lock=true)
users = Devise.mappings.keys.map { |s| warden.user(:scope => s, :run_callbacks => false) }
warden.raw_session.inspect
warden.logout
expire_devise_cached_variables!
warden.clear_strategies_cache!
warden.lock! if lock
users.any?
end
So it looks like in my functional tests (and only in my functional tests) the warden object is nil.
Why is this function being called on an unauthenticated request?
Why doesn't the warden object exist here?
What can I do to fix it?
No idea, Devise is doing its own thing.
See 1.
Include Devise::TestHelpers.
The Devise documentation says that you need to include the helpers in order to use them. It does not say that if you don't include the helpers your functional tests will fail, including those that don't use any authentication, but that's what happens.
(Note the JSON handling here, which I originally thought was the problem, ended up being just a red herring. Even with standard post or get you will have this problem.)
Related
I'm updating an application that was working with a very old version of RSpec (2.9.0). It works fine with 3.8, but I get a deprecation error about the following code:
response = Net::HTTPOK.new(1.0, "200", "OK")
response.stub(:content_type => 'text/json', :body => contents_raw)
Now, this code is a mock callout to an external API, and it is deep inside my code's libraries. The application I'm testing is a Sinatra app, so I'm using "get" from Rack::Test::Methods to test my app, but then deep inside the app itself is this response.stub
Apparently I should be using "double()" and "allow(object).to receive(...)", but all the examples I've seen are for using double directly in your test's "it" block, which this code is nowhere near. If I actually try to use double I just get a no method error.
WebMocks seems like a very large hammer to just replace this single call.
Maybe the right thing to do is to make a superclass of Net::HTTPOK and pass the response data in wit that superclass's new?
Huh. OK, so the wrapper class method turned out to be far easier than I expected. In my spec file I added:
class HttpWrapper < Net::HTTPOK
def initialize(data, *args)
#data = data
super(*args)
end
def content_type
'text/json'
end
def body
#data
end
end
(Bizarre detail: if I replace every instance of "data" above with "body", it explodes horribly.)
And at the block of code in question I now have:
response = HttpWrapper.new(contents_raw, 1.0, "200", "OK")
And that seems to have done the trick just fine.
If this is a terrible idea for some reason, hopefully someone will let me know. :)
the best way to mock http, use gem which called iswebmock
I am trying to do some rspec unit tests on a method in my model. The method returns a promise, and when resolved, the name of the person. The method is not the problem as I know that it works correctly. Here is my test code:
it 'should return correct name' do
report = Report.new(first_name: 'Testy', last_name: 'Testerson')
report.save!
expect(report.name).to eql('Testy Testerson')
end
When I test it, I get the following error:
Failure/Error: expect(report.name).to eql('Testy Testerson')
TypeError:
can't convert Promise to Array (Promise#to_ary gives Promise)
While debugging, I used the following line to inspect the returned value of the method:
puts report.name.inspect
And I got the following response:
#<Promise(70319926955580): "Testy Testerson">
The error seems to be happening because it tests the promise against the expected value. Why am I getting this error?
Using report.name.value fixes this issue
When running code on the server, calls to store return a promise that is already resolved. But on the client the promise won't be resolved yet. Someone (forget the name atm) is working on adding support to promise directly into opal-rspec, but at the moment a returned promise won't wait for opal-rspec. The plan is once thats ready we'll add more tools to volt to make it easier for developers to test in both MRI and opal (like we do with Volt itself).
You can call .value on a promise to get back its value, but only if the promise has resolved. The safer way to do it is to use a .then block:
report.name.then do |name|
expect(name).to eq('Bob')
end
Hopefully that helps.
I am running a padrino application and have started with the included mailers. I want to test that a mail is sent and had previously had no trouble accessing the Mail::TestMailer object to look at the mails delivered during the test.
That is the background about what I am doing but not precisely the question. I want to know how can a module become available to the runtime environment.
I have this test in two versions
first
def test_mailer
Mail::TestMailer.deliveries.clear
get '/owners/test'
e = Mail::TestMailer.deliveries.pop
puts e.to.to_s
end
second
def test_mailer
get '/owners/test'
Mail::TestMailer.deliveries.clear
e = Mail::TestMailer.deliveries.pop
puts e.to.to_s
end
In the second version this test fails with the error message NoMethodError: undefined method to' for nil:NilClass This makes sense to me. I clear the messages then ask for the last one which should be nil. However when I run the test on the first version the error is NameError: uninitialized constant OwnersControllerTest::Mail
So somehow the get method is causing the Mail object/module to be made available. I don't understand how it can do this. I don't know if this is a rack-test or padrino thing so am unsure what extra information to copy in here.
Add require 'mail' to your test helper.
The issue is explained here: https://github.com/padrino/padrino-framework/issues/1797
Sorry if this is plain simple. i am new to ruby as well as rspec and it seems rspec is a very 'obscure' world (esp when coming from a .net background).
In my 'spec', i have:
before(:each) do
expect(File).to receive(:exist?).with("dummy.yaml").and_return (true)
end
This works fine for all my 'examples', except one where i want it to return false.
expect(File).to receive(:exist?).with("non_existent.yaml").and_return (false)
This obviously fails my test because although "non_existent.yaml" expectation was met, the "dummy.yaml" was not:
(<File (class)>).exist?("dummy.yaml")
expected: 1 time with arguments: ("dummy.yaml")
received: 0 times
So how can i do a 'Reset' on 'File.exist?' (a class method mock) before i setup the new expectation for it? (... "non_existent.yaml"..)
i googled and it yielded:
RSpec::Mocks.proxy_for(your_object).reset
but this gives me:
NoMethodError:
undefined method `proxy_for' for RSpec::Mocks:Module
I could not find anywhere in the documentation that this is how you should do it, and past behaviors goes to show that this solution might also change in the future, but apparently this is how you can currently do it:
RSpec::Mocks.space.proxy_for(your_object).reset
I would follow #BroiSatse's remark, though, and think about re-designing the tests, aiming to move the expectation from the before block. The before block is meant for setup, as you say, and the setup is a very weird place to put expectations.
I'm not sure how you came to this design, but I can suggest two possible alternatives:
If the test is trivial, and will work anyway, you should create one test with this explicit expectation, while stubbing it for the other tests:
before(:each) do
allow(File).to receive(:exist?).with("dummy.yaml").and_return (true)
end
it "asks if file exists" do
expect(File).to receive(:exist?).with("dummy.yaml").and_return (true)
# do the test...
end
If the expectation should run for every test, since what changes in each scenario is the context, you should consider using shared examples:
shared_examples "looking for dummy.yaml" do
it "asks if file exists" do
expect(File).to receive(:exist?).with("dummy.yaml").and_return (true)
# do the test...
end
end
it_behaves_like "looking for dummy.yaml" do
let(:scenario) { "something which sets the context"}
end
You might also want to ask myron if there is a more recommended/documented solution to reset mocked objects...
This worked for me to unmock a specific method from a class:
mock = RSpec::Mocks.space.proxy_for(MyClass)
mock.instance_variable_get(:#method_doubles)[:my_method].reset
Note: Same logic of
RSpec::Mocks.space.proxy_for(MyClass).reset which resets all methods
Expanding on #Uri Agassi's answer and as I answered on another similar question, I found that I could use RSpec::Mocks.space.registered? to check if a method was a mock, and RSpec::Mocks.space.proxy_for(my_mocked_var).reset to reset it's value.
Here is the example I included in my other answer:
Example: Resetting a mocked value
For example, if we wanted to reset this mock back to it's unmocked
default value, we can use the RSpec::Mocks.space.proxy_for helper to
find our mock, then reset it:
# when
# Rails.configuration.action_controller.allow_forgery_protection == false
# and
# allow(Rails.configuration.action_controller).to receive(:allow_forgery_protection).and_return(true)
RSpec::Mocks.space.registered?(Rails.configuration.action_controller)
# => true
Rails.configuration.action_controller.allow_forgery_protection
# => true
RSpec::Mocks.space.proxy_for(Rails.configuration.action_controller).reset
Rails.configuration.action_controller.allow_forgery_protection
# => false
Notice however that the even though the mock value has been reset, the
mock remains registered?:
RSpec::Mocks.space.registered?(Rails.configuration.action_controller)
# => true
When using "expect_any_instance" I had success using the following method to change the mock (e.g. our example: Putting out a Twitter post and returning a different tweet id)
expect_any_instance_of(Twitter::REST::Client).to receive(:update).and_return(Hashie::Mash.new(id: "12"))
# post tweet
RSpec::Mocks.space.verify_all
RSpec::Mocks.space.reset_all
expect_any_instance_of(Twitter::REST::Client).to receive(:update).and_return(Hashie::Mash.new(id: "12346"))
# post another tweet
In Sinatra tests, env['SERVER_NAME'] defaults to www.example.com. How can I set this to some arbitrary domain?
Capybara has .default_host method, but not using Capybara.
Or, is it possible to change the env[DEFAULT_HOST]?
Using RSpec, Sinatra, WebMock.
EDIT: Adding env['SERVER_NAME'] = 'www.foo.com' to RSpec test raises exception:
NameError: undefined local variable or method 'env' for #<RSpec::Core::ExampleGroup::Nested_1::Nested_1:0x007fe6ce3b5ff8>
The env helper is only accessible within a Sinatra app.
One way to change it is when making a request:
get "/blah", {}, {'HTTP_SERVER_NAME' => 'www.foo.com' }
The 3rd argument of a rack/test get or post is the headers hash.