How to test model's callback method independen

问题:

I had a method in a model:

class Article < ActiveRecord::Base
  def do_something
  end
end

I also had a unit test for this method:

# spec/models/article_spec.rb
describe "#do_something" do
  @article = FactoryGirl.create(:article)
  it "should work as expected" do
    @article.do_something
    expect(@article).to have_something
  end
  # ...several other examples for different cases
end

Everything was fine until I found it's better to move this method into a after_save callback:

class Article < ActiveRecord::Base
  after_save :do_something

  def do_something
  end
end

Now all my tests about this method broken. I have to fix it by:

• No more specific call to do_something because create or save will trigger this method as well, or I'll meet duplicate db actions.
• Change create to build
• Test respond_to

• Use general model.save instead of individual method call model.do_something

  describe "#do_something" do
    @article = FactoryGirl.build(:article)
    it "should work as expected" do
      expect{@article.save}.not_to raise_error
      expect(@article).to have_something
      expect(@article).to respond_to(:do_something)
    end
  end
  

The test passed but my concern is it's no longer about the specific method. The effect will be mixed with other callbacks if more added.

My question is, is there any beautiful way to test model's instance methods independently that becoming a callback?

回答1:

Callback and Callback behavior are independent tests. If you want to check an after_save callback, you need to think of it as two things:

1. Is the callback being fired for the right events?
2. Is the called function doing the right thing?

Assume you have the Article class with many callbacks, this is how you would test:

class Article < ActiveRecord::Base
  after_save    :do_something
  after_destroy :do_something_else
  ...
end

it "triggers do_something on save" do
  expect(@article).to receive(:do_something)
  @article.save
end

it "triggers do_something_else on destroy" do
  expect(@article).to receive(:do_something_else)
  @article.destroy
end

it "#do_something should work as expected" do
  # Actual tests for do_something method
end

This decouples your callbacks from behavior. For example, you could trigger the same callback method article.do_something when some other related object is updated, say like user.before_save { user.article.do_something }. This will accomodate all those.

So, keep testing your methods as usual. Worry about the callbacks separately.

Edit: typos and potential misconceptions Edit: change "do something" to "trigger something"

回答2:

You can use shoulda-callback-matchers to test existence of your callbacks without calling them.

describe Article do
  it { should callback(:do_something).after(:save) }
end

If you also want to test the behaviour of the callback:

describe Article do
  ...

  describe "#do_something" do
    it "gives the article something" do
      @article.save
      expect(@article).to have_something
    end
  end
end

回答3:

This is more of a comment than an answer, but I put it here for the syntax highlighting...

I wanted a way to skip the callbacks in my tests, this is what I did. (This might help with the tests that broke.)

class Article < ActiveRecord::Base
  attr_accessor :save_without_callbacks
  after_save :do_something

  def do_something_in_db
    unless self.save_without_callbacks
      # do something here
    end
  end
end

# spec/models/article_spec.rb
describe Article do
  context "after_save callback" do
    [true,false].each do |save_without_callbacks|
      context "with#{save_without_callbacks ? 'out' : nil} callbacks" do
        let(:article) do
          a = FactoryGirl.build(:article)
          a.save_without_callbacks = save_without_callbacks
        end
        it do
          if save_without_callbacks
            # do something in db
          else
            # don't do something in db
          end
        end
      end
    end
  end
end

回答4:

In the spirit of Sandi Metz and minimalist testing, the suggestion in https://stackoverflow.com/a/16678194/2001785 to confirm the call to a possibly private method does not seem right to me.

Testing a publicly-observable side-effect or confirming an outgoing command message makes more sense to me. Christian Rolle provided an example at http://www.chrisrolle.com/en/blog/activerecord-callback-tests-with-rspec.

回答5:

describe "#do_something" do

 it "gives the article something" do

  @article = FactoryGirl.build(:article)

   expect(@article).to have_something

 @article.save
end

end