Cannot use Capybara finders

[PikachuEXE]

I am trying to write a test for a link for its attributes

Currently I am using

it {should have_css("a.btn[data-method='post'][href='#{some_path)}']")}

But it's too long and hard to read

So I have this:

subject do
  find_link(name)
end

it {should exist}

specify {subject[:href].should eq some_path}
specify {subject[:class].should eq '.btn'}
specify {subject['data-method'].should eq 'post'}

specify {subject[:rel].should eq :tooltip}
specify {subject[:title].should eq "Some Action"}

But find_link method is not defined

Actually I want to use have_link but the option only supports href

[PikachuEXE]

Now I am using some custom matcher:

# Source:
# https://groups.google.com/forum/?fromgroups=#!topic/ruby-capybara/v1h4npzN6OI

# Check for any <css> tag with _all_ of the specified attributes.
# Does _not_ check the text after the tag.
# Still returns true even if the tag has more attributes than those specified.
# Example:
#   it { should have_css_tag_and_attributes('div',
#        id: "left_sidebar", class: "span2") }
def has_selector_with_attributes?(selector, attributes = {})

  raise ArgumentError.new "attributes must be a hash with at least one element" if
      !attributes.is_a?(Hash) || attributes.empty?
  attr_string =  attributes.map{|k,v| "@#{k}='#{v}'"}.join(' and ')
  xpath = "//#{selector}[#{attr_string}]"
  has_xpath?(xpath, nil)

end

def has_link_with_attributes?(attributes = {})
  has_selector_with_attributes?(:a, attributes)
end

But I cannot use it with locator (id or link text)

[alexandru-calinoiu]

I ma having the same issues also

[apotonick]

Sorry guys, I don't get it - can you please reduce your issue to the minimum and give me that example? Thahanks!

[PikachuEXE]

Cannot use capybara finders such as find_link in rspec-cells I want to use it since I want to test some attributes of a link, not just href

[alexandru-calinoiu]

Or the following code sample fails with "no method defined have_link ..."

subject { cell }
visit user_path(user)
it { should have_link('Settings', href: edit_user_path(user)) }

[alexandru-calinoiu]

I got it, I had capybara only in the cucumber test group in the gemfile, once I've put it in test group also all was good, clue was in cell_example_group#15

[PikachuEXE]

I have capybara in test group all the time... Let me check it later :/

[therealjasonkenney]

If you do this first:

subject { Capybara::Node::Simple.new(render_cell(:cell_name, 'stuff')) }

You can use matchers with the subject

I noticed after updating rspec-cells, I had to do that for my tests. My suggestion is... a method that wraps either one in a capybara::node::simple if capybara is loaded.

[therealjasonkenney]

Found it!

You have to order the Gemfile so capybara is first! This is because ruby evaluates the module before Capybara is loaded at load time!

[therealjasonkenney]

I was going to try and do a patch and PR, however loading capybara causes everything to 'splode due to Rails.root being undefined. :( Am I doing something wrong? How do people test for WebRat and Capybara in gems like these?

[seuros]

the subject should be a capybara session, not a simplenode.

[apotonick]

Should we offer an alternative ::subject?

[therealjasonkenney]

subject is NOT something that should be specified in the lib. After all sometimes a subject is a cell, a node, a session, it's whatever is defined in the specific test as subject.

The current problem was render_cell was returning ActiveSupport::SafeBuffer rather than a Capybara::Simple::Node. (You can't return a session in a unit test without invoking a full browser emulation, which I thought is the domain of feature and request specs).

And this is mostly tied to the fact that the decision to include Capybara code in the gem is based on capybara being loaded first.

I think that route should be, just put the capybara code in rspec/cells/capybara and webrat respectively, then have people require it in the rspec helper.

I could make a PR, but again for the life of me I have no idea how to get capybara working in the test without stubbing out Rails.root, which seems wrong.

[apotonick]

Absolutely not, as long as it's only Rails.root this is totally fine.

Alternatively, you can use a dummy Rails app, I do this in all my gems that provide support for Rails. Here's an example of how that would work, it's really easy once you got the code: https://github.com/apotonick/kaminari-cells/blob/master/test/fake_app/rails_app.rb

Please PR - I personally do not use Rspec and am happy to merge your stuff. What do you think, @seuros ?

[seuros]

I will look into this today.

[apotonick]

Rspec-cells 0.3 always returns a Capybara::Node::Simple instance from ViewModel#call when Capybara is loaded, this should be fixed.