Integration test: `assert_select 'a[href=?]'` fails for paginated pages - ruby-on-rails

I have many integration tests that look something like:
first_page_of_users = User.page(1) # Using kaminari for pagination.
first_page_of_users.each do |user|
assert_select 'a[href=?]', user_path(user)
end
These tests fail with an error message such as:
Expected at least 1 element matching "a[href="/users/1"]", found 0..
I added puts #response.body to the tests to see the code that it responded with.
For all the users included in the response body it has correct hrefs, such as href="/users/75938613". However, the numbers behind /users/ are much higher, and "users1" is indeed not present (not just the url not present but the entire record isn't on that page). If I ensure that there are only a few fixtures so that no pagination applies, the tests pass. This makes me feel as if with pagination the order that first_page_of_users = User.page(1) applies differs from the order in records that first_page_of_users.each do |user| assert_select 'a[href=?]', user_path(user) expects...
What could be causing this error? It applies to different pages and models of my app.
Update: For another model even when I try
Organization.all.each do |org|
puts #response.body
assert_select 'a[href=?]', organization_path(org)
end
I get the error:
Expected at least 1 element matching "a[href="/organizations/fix0"]",
The organization with the name "fix0" is not present at all in the body that it reponds with. My (shortened) fixtures are below and they confirm there should not be any organization with that name. Any idea what's going on?
one:
name: abcd
activated: true
activated_at: 2015-04-16 11:38:53
two:
name: efgh
activated: true
activated_at: 2015-04-16 11:38:53
<% 35.times do |n| %>
organization_<%= n %>:
name: <%="fix#{n}" %>
activated: true
activated_at: <%= Time.zone.now %>
<% end %>

When integration testing, it's important to keep track of what should show up where and test just that. Because we have too many records for the first page in this case, anything that would land on later pages will cause the failing test. As you suspect, something like this should verify that the first page of results is present:
Organization.order('organizations.name','desc').page(1).each do |organization|
assert_select 'a[href=?]', organization_path(organization)
end
I recommend keeping a smaller data set for testing, because in this case you should also 'get' page two in your test and do the assertions above for 'page(2)' as well. This means this integration test has to issue two get requests (plus any setup you may need) and check the results, and making a habit of that can slow down your suite.

Related

Failing Rails test based on lack of set session variable - how to set in Rails 5.2?

I have the following very simple test which has always passed.
test "should get index" do
sign_in #user
get companies_url
assert_response :success
end
However, I'm now receiving the following error.
Error:
CompaniesControllerTest#test_should_get_index:
ActionView::Template::Error: No route matches {:action=>"show", :controller=>"companies", :id=>nil}, missing required keys: [:id]
app/views/layouts/application.html.erb:53:in `_app_views_layouts_application_html_erb__2582383126246718868_70101260952060'
test/controllers/companies_controller_test.rb:16:in `block in <class:CompaniesControllerTest>'
What I've changed in my app is that I've built a sidebar (which I load via application.html.erb so it's loaded on all views (new, show, edit), which lets a user switch between various companies they "own" - which changes a session variable which we use to alter the content of the sidebar.
If we dig into the line that seems to be failing app/views/layouts/application.html.erb:53 this is how it looks:
<div class="list-group-item">
<%= link_to 'Company', company_path(session[:current_company]) unless current_user.companies.empty? %>
</div>
If I remove this link_to line then the tests pass.
My guess is that the show view is trying to load, including the sidebar which doesn't have a session[:current_company] set so the view crashes. However, in Rails 5.2 you cannot set/test session variables as far as I understand, so I'm wondering what the best way for me to set the testing up to make this pass? I do set a value for this session within my application controller a user signs in though:
def after_sign_in_path_for(resource_or_scope)
# Set a default current company scope for a use after signing in
session[:current_company] = current_user.companies.first.id unless current_user.companies.empty?
companies_path
end
Perhaps in the link_to from from within the sidebar I could add a default value to make sure we're always sending in a company, regardless of whether it's the session[:current_company] or not?
Altering the line to skip creating the link if there is no session variable present
<%= link_to 'Company', company_path(session[:current_company]) unless current_user.companies.empty? || !session[:current_company] %>
Seemed to do the trick and make all the tests pass. Would love some feedback on whether this is a good solution or not though! :)

What is the quickest way to determine why an integration test failed?

I have the following:
test "should update deal" do
patch user_deal_url(#user, #deal), params: { deal: { name: #deal.name+'xxx' } }
# p #response.body
# p #deal.errors.full_messages
assert_select ".alert", false # should not have alert
assert_redirected_to dashboard_url
follow_redirect!
assert_select ".alert", "Deal was successfully updated."
end
It only says:
Failure:
DealsControllerTest#test_should_update_deal [C:/Users/Chloe/workspace/test/controllers/deals_controller_test.rb:73]:
Expected response to be a <3XX: redirect>, but was a <200: OK>
I thought I could print #deal.errors.full_messages but it's not the same #deal in the controller. I'm able to print the entire response body and copy it into Notepad++ and search for the error messages, but it's so tedious. I thought I can assert that .alert doesn't exist and it would tell me what it actually was. I added
assert_select "#error_explanation", false # should not have error
But that only told me that it exists, not why it failed.
Expected exactly 0 elements matching "#error_explanation", found 1..
Expected: 0
Actual: 1
So how do you quickly determine why a model is failing to save or update in a controller during testing?
Rails 5.0.2
Reference: http://guides.rubyonrails.org/testing.html
My solution causes subsequent assert_select statements to fail! Just having puts css_select('#error_explanation') makes the last assert_select fail, but commenting it out makes it succeed. Why?
I entered a Rails bug for that in the mean time: https://github.com/rails/rails/issues/29367
I was able to put
puts css_select('#error_explanation')
right after the put/patch and it prints
# Running:
<div id="error_explanation">
<h2>
2 errors
prohibited this deal from being saved:
</h2>
<ul>
<li>Size can't be blank</li>
<li>Size is not a number</li>
</ul>
</div>
F
Failure:
I'm still looking for something better as leaving this line in causes subsequent assert_selects to fail!

Rspec2 + Rails4: Testing for displayed model fields in form partial

Question
I'd like to write tests that check the model fields that are displayed in my "show" and "form" partials. I succeeded for "show", not for "form".
Main constrain: The solution must be able to loop through an Array that contains each names of the model fields.
I believe this case can be interesting for anyone that is trying to shorten his test script files, while having many fields, and having a complete control over what's displayed and what's not, so I'll put some efforts trying to find a solution, with your help if you please :)
Form view
Nothing fancy
= form_for #user do |f|
= f.select :field_1, options_from_collection_for_select ...
= f.text_field :field_2
...
Actual situation
I found an easy way for the "show" partial, here is how my spec file looks like:
def user_fields_in_show_view
[:field_1, :field_2, ..., :field_n]
end
it 'display fields' do
user_fields_in_show_view.each do |field|
User.any_instance.should_receive(field).at_least(1).and_call_original
end
render
end
This works well.
-
But the exact same technique does not work in the "form" partial, using the same code
def user_fields_in_form_view # List of fields need to be different, because of the associations
[:field_1_id, :field_2, ..., :field_n]
end
it 'display fields' do
user_fields_in_form_view.each do |field|
User.any_instance.should_receive(field).at_least(1).and_call_original
end
render
end
It whines like this:
Failure/Error: Unable to find matching line from backtrace
Exactly one instance should have received the following message(s) but didn't: field1_id, field_2, ..., field_n
# Backtrace is long and shows only rspec/mock, rspec/core, rspec/rails/adapters, and spork files
What I tried so far
1- I commented out the stub part of my tests and output rendered to the console, to manually check what's generated by my view, and yes the fields are correctly generated.
2- I replaced User.any_instance by the model I assign to the view, error is slightly different but it still not working
it 'display fields' do
user = create :user
assign :user, user
user_fields_in_form_view.each do |field|
user.should_receive(field).at_least(1).and_call_original
end
render
end
Gives:
Failure/Error: user.should_receive(field).at_least(1).and_call_original
(#<User:0x0000000506e3e8>).field_1_id(any args)
expected: at least 1 time with any arguments
received: 0 times with any arguments
3- I change the code so the it is inside the loop, like this:
user_fields_in_form_view.each do |field|
it 'display fields' do
user = create :user
assign :user, user
user.should_receive(field).at_least(1).and_call_original
render
end
end
Same result as above
And I run out of options. I suspect the internals of FormBuilder to play a bad trick on me but I can't figure it out, I'm not very knowledgeable with those yet. Thanks for reading
I usually try to write unit test as simple as possible. Loops in unit tests don't add much readability and are not very good practice in general. I'd rewrite the test like this:
it 'should display user name and email' do
# note: `build` is used here instead of `create`
assign :user, build(:user, first_name: 'John', last_name: 'Doe', email: 'jd#example.com')
render
rendered.should have_content 'John'
rendered.should have_content 'Doe'
rendered.should have_content 'jd#example.com'
end
Thus, we're not limiting the view in how it should render the first and the last name. For example, if our view uses the following (bad) code in order to render user's full name, then your test will fail, but my test will work just fine, because it tests the behaviour of the view, not its internals:
<%= user.attributes.values_at('first_name', 'middle_name').compact.join(' ') %>
Moreover, multiple assertions in one test is a bad smell too. Going one step further, I'd replace this test with three smaller ones:
it "should display user's first name" do
assign :user, build(:user, first_name: 'John')
render
expect(rendered).to include 'John'
end
it "should display user's last name" do
assign :user, build(:user, last_name: 'Doe')
render
expect(rendered).to include 'Doe'
end
it "should display user's email" do
assign :user, build(:user, email: 'jd#example.com')
render
expect(rendered).to include 'jd#example.com'
end
========
UPD: Let's make it more dynamic in order to avoid tons of repetition. Tis doesn't answers why your spec fails, but hopefully represents working tests:
%i(first_name last_name email).each do |field|
it "should display user's #{field}" do
user = build(:user)
assign :user, user
render
expect(rendered).to include user.public_send(field)
end
end
In order to make these tests more reliable, make sure that user factory doesn't contain repetitive data.
I am not quite sure how you build your form, but if you use the form_for, simple_form_for or formtastic_form_for helpers, actually you are using a different object. You write something like (assume the basic form_for)
= form_for #user do |f|
and all methods are relayed to object f. Now f.object will point to #user, but is it a copy of #user or #user itself, I don't know. So I would expect that User.any_instance should work.
Regardless, when doing a view test, it is not important how the contents of a field are set, it is important that the contents of a field are set correctly. Suppose you build your forms yourself, you switch to another library to build your forms, and all your tests break, because it retrieves the data differently. And that should not matter.
So I am with #DNNX, and in your view tests you should test the content of the rendered HTML and not how the data is retrieved.

Checking for multiple copies of a selector in RSpec

I'm having a problem with one of my tests in RSpec. The test in question is supposed to test whether there is more than one post on the index page, by checking for the existence of more than one <article class="post"> tag on the page. My current implementation looks like this:
it "should have more than one post" do
get :index
response.should have_selector("article", :class => "post", :count => 2)
end
For some reason this test fails, despite showing multiple instances of the <article class="post"> tag in the debug output. The specific error it give is
Failures:
1) PostsController GET 'index' should have more than one posts
Failure/Error: response.should have_selector("article", :class => "post", :count => 2)
expected following output to contain a <article class='post'/> tag:
Note that if I remove the :count => 2 option the test passes without incident.
Does anyone know why this test is failing?
I'm guessing you are using Capybara under RSpec here or Webrat? With the count option your test is asserting that the response contains exactly 2 posts, rather than at least 2. Does the minimum option do what you want?
See https://github.com/jnicklas/capybara/blob/master/lib/capybara/node/matchers.rb and http://rubydoc.info/github/jnicklas/capybara/master/Capybara/Node/Matchers#has_selector%3F-instance_method

Rails 2.3.x - How to stub a helper method (that gets called from a view) in a functional test (no RSpec)?

Please don't tell me "search more" or other stuff cause all solutions for similar question fail.
Simple:
I have a functional tests. I want to make a simple get and see if proper content gets rendered
test "displays headline if user should see it" do
get :index
assert_match /headline/, response.body
end
test "doesn't display headline if user shouldn't see it" do
get :index
assert_no_match /headline/, response.body
end
and a simple view
<% if show_headline?(arg) %>
headline
<% end %>
and a helper:
module TheHelper
def show_headline?(arg)
arg ? hard_code_logic : even_harder_logic
end
end
so what I need is to do in test something like:
test "displays headline if user should see it" do
Something.stubs(:show_headline?).returns(true)
get :index
assert_match /headline/, response.body
end
test "doesn't display headline if user shouldn't see it" do
Something.stubs(:show_headline?).returns(false)
get :index
assert_no_match /headline/, response.body
end
The question is what is Something? I want to stub it cause I have helpers tested in unit/helpers.
After the get helper module gets remixed into the controller class. Please don't give me links to other answers, I read them (but of course I could have read the wrong ones) and they don't work for me. I use Rails 2.3.10 with mocha 0.9.8.
Things that don't work:
TheController.any_instance.stubs(:show_headline?)
ActionView::Base.any_instance...
#controller.stubs...
UPDATE:
the only mock that worked was:
<% self.stubs(:show_headline?).returns(true) >%
<% if show_headline?(arg) %>
headline
<% end %>
but of course I will not use that... maybe it is a clue

Resources