How to Speed Up Your Tests via :build_stubbed

Rspec is an awesome thing that was created for ruby community. Most of us write tests. However, sometimes in large projects our test becomes very slow. So, each launch of the test really hurts and it does not meter whether you launch your test before commit/push or on CI. When it takes over 30 minutes to pass your test suit,  something definitely went wrong.

There are many things that you can improve: database clearing, perform caching, stub external requests and so on.

In our project, we use FactoryGirl (FactoryBot). Here I will try to describe how :build_stubbed can help us to improve the test speed.



In this case we will create comment object and all association for it.


Warning: Here is a misconception, we did not save object but we still will have requests to our database if factory has an associations!

:build does not create object -  but it creates all association to that object.

factory :comment do
  association :post
(0.1ms)  begin transaction
Post Create (0.5ms)  INSERT INTO "posts" DEFAULT VALUES
(0.6ms)  commit transaction



:build_stubbed does not call database at all. It just creates and assigns attributes to an object to make it behave like instantiated object. It have an assigned id. Thats why we have such a speed-up.

What about association?

Association also works. build_stubbed will create the associations via build_stubbed, while build will create real association.

comment = FactoryBot.build_stubbed(:comment)
#<Comment:0x007f94d2b92df0 id: 1002, post_id: 1001, body: "text">
#<Post:0x007f94d5883440 id: 1001, name: nil>

Also, do you actually need to reference the post in every single test?

Note: remember that build_stubbed stubs next methods:

  • persisted?
  • new_record?
  • created_at

Also remember that next methods will cause an exception:

  • update_attribute
  • update_column
  • save
  • destroy
  • connection
  • reload
RuntimeError: stubbed models are not allowed to access the database - Comment#save()


Using build_stubbed is not a perfect solution for every situation. However, it can be a great approach in some cases. For example, when you test models or services.