7 days of WordPress plugins, themes & templates - for free!* Unlimited asset downloads! Start 7-Day Free Trial
  1. Code
  2. Ruby

Ruby Page Objects for Capybara Connoisseurs

Scroll to top
Read Time: 13 mins
Final product imageFinal product imageFinal product image
What You'll Be Creating

What Are Page Objects?

I’ll give you the short pitch first. It’s a design pattern to encapsulate markup and page interactions—specifically to refactor your feature specs. It’s a combination of two very common refactoring techniques: Extract class and Extract method—which don’t have to happen at the same time because you can gradually build up to extracting a whole class via a new Page Object.

This technique allows you to write high-level feature specs that are very expressive and DRY. In a way, they are acceptance tests with application language. You might ask, aren’t specs written with Capybara already high-level and expressive? Sure, for developers who write code on a daily basis, Capybara specs read just fine. Are they DRY out of the box? Not really—actually, certainly not!

When you look at these examples of feature specs, where do you see opportunities to make this read better, and how could you extract information to avoid duplication? Also, is this high-level enough for easy modeling of user stories and for non-technical stakeholders to understand?

In my mind, there are a couple of ways to improve this and to make everybody happy—developers who can avoid fiddling with the details of interacting with the DOM while applying OOP, and other non-coding team members having no trouble jumping between user stories and these tests. The last point is nice to have, for sure, but the most important benefits come mostly from making your DOM-interacting specs more robust.

Encapsulation is the key concept with Page Objects. When you write your feature specs, you’ll benefit from a strategy to extract the behaviour that is driving through a test flow. For quality code, you want to capture the interactions with particular sets of elements on your pages—especially if you stumble upon repeating patterns. As your application grows, you want / need an approach that avoids spreading that logic all over your specs.

”Well isn’t that overkill? Capybara reads just fine!” you say?

Ask yourself: Why wouldn’t you have all the HTML implementation details in one place while having more stable tests? Why shouldn’t UI interaction tests have the same quality as tests for application code? Do you really want to stop there?

Due to everyday changes, your Capybara code is vulnerable when spread all over—it introduces possible breakpoints. Let’s say a designer wants to change the text on a button. No biggie, right? But do you want to adapt to that change in one central wrapper for that element in your specs, or do you prefer to do that all over the place? I thought so!

There are many refactorings possible for your feature specs, but Page Objects offer the cleanest abstractions for encapsulating user-facing behaviour for pages or more complex flows. You don’t have to simulate the whole page(s) though—focus on the essential bits that are necessary for user flows. No need to overdo it!

Acceptance Tests / Feature Specs

Before we move on to the heart of the matter, I’d like to take a step back for people new to the whole testing business and clear up some of the lingo that is important in this context. People more familiar with TDD won’t miss much if they skip ahead.

What are we talking about here? Acceptance testing usually comes in at a later stage of projects to evaluate if you have been building something of value for your users, product owner, or other stakeholder. These tests are usually run by customers or your users. It’s sort of a check if the requirements are being met or not. There is something like a pyramid for all sorts of testing layers, and acceptance tests are near the very top. Because this process often includes non-technical folks, a high-level language for writing these tests is a valuable asset for communicating back and forth.

Feature specs, on the other hand, are a bit lower in the testing food chain. A lot more high-level than unit tests, which focus on the technical details and business logic of your models, feature specs describe flows on and between your pages.

Tools like Capybara help you to avoid doing this manually, meaning that you rarely have to open your browser to test stuff manually. With these kinds of tests, we like to automate these tasks as much as possible and test-drive the interaction through the browser while writing assertions against pages. By the way, you don’t use get, put, post or delete as you do with request specs.

Feature specs are very similar to acceptance tests—sometimes I feel the differences are too blurry to really care about the terminology. You write tests that exercise your whole application, which often involves a multi-step flow of user actions. These interaction tests show if your components work in harmony when they are brought together.

In Ruby land, they are the main protagonist when we’re dealing with Page Objects. Feature specs themselves are already very expressive, but they can be optimized and cleaned up by extracting their data, behaviour and markup into a separate class or classes.

I hope that clearing up this blurry terminology will help you see that having Page Objects is a bit like doing acceptance-level testing while writing feature specs.


Maybe we should go over this really quickly as well. This library describes itself as an “Acceptance test framework for web applications”. You can simulate user interactions with your pages via a very powerful and convenient domain-specific language. In my personal opinion, RSpec paired with Capybara offers the best way to write your feature specs at the moment. It lets you visit pages, fill in forms, click on links and buttons, and look for markup on your pages, and you are able to easily combine all kinds of these commands to interact with your pages through your tests.

You basically can avoid opening the browser yourself to test this stuff manually most of the time—which is not only less elegant but also a lot more time-consuming and error-prone. Without this tool, the process of “outside-in-testing”–you drive your code from high-level tests down into your unit-level tests—would be a lot more painful and possibly therefore more neglected.

In other words, you start writing these feature tests which are based on your user stories, and from there you go down the rabbit hole until your unit tests provide the coverage your feature specs needed. After that, when your tests are green of course, the game starts anew and you go back up to continue with a fresh feature test.


Let’s look at two simple examples of feature specs that let M create classified missions which can then be completed.

In the markup, you have a list of missions, and a successful completion creates an additional class completed on the li of that particular mission. Straightforward stuff, right? As a first approach, I started with small, very common refactorings that extract common behaviour into methods.



Although there are other ways, of course, to deal with stuff like sign_in_as, create_classified_mission_named and so on, it’s easy to see how fast these things can start to suck and to pile up.

UI-related specs often don’t get the OO treatment they need / deserve, I think. They have the reputation of providing too little bang for the buck, and of course developers are not very fond of times when they have to touch markup stuff a lot. In my mind, that makes it even more important to DRY these specs up and make it fun to deal with them by throwing in a couple of Ruby classes.

Let’s do a little magic trick where I hide the Page Object’s implementation for now and only show you the end result applied to the feature specs above:



Doesn’t read too bad, huh? You basically create expressive wrapper methods on your Page Objects that let you deal with high-level concepts—instead of fiddling everywhere with the intestines of your markup all the time. Your extracted methods do this kind of dirty work now, and that way shotgun surgery is not your problem anymore.

Put differently, you encapsulate most of the noisy, down-in-the-weeds DOM interacting code. I’ve got to say, though, that sometimes intelligently extracted methods in your feature specs are sufficient and read a bit better since you can avoid dealing with Page Object instances. Anyway, let’s take a look at the implementation:


What you see is a plain old Ruby object—Page Objects are in essence very simple classes. Normally you don’t instantiate Page Objects with data (when the need occurs, of course, you can) and you create mostly a language via the API that a user or a non-technical stakeholder on a team might use. When you think about naming your methods, I think it’s good advice to ask yourself the question: How would a user describe the flow or the action taken?

I should maybe add that without including Capybara, the music stops pretty fast.

You probably wonder how these custom matchers work:

RSpec generates these custom matchers based on predicate methods on your Page Objects. RSpec converts them by removing the ? and changes has to have. Boom, matchers from scratch without much fuzz! A bit magic, I’ll give you that, but the good kind of wizardry, I’d say.

Since we parked our Page Object at specs/support/features/pages/missions.rb, you also need to make sure that the following is not commented out in spec/rails_helper.rb.

If you run into a NameError with an uninitialized constant Pages, you’ll know what to do.

If you’re curious what happened to the sign_in_as method, I extracted it into a module at spec/support/sign_in_helper.rb and told RSpec to include that module. This has nothing to do with Page Objects directly—it just makes more sense to store test functionality like sign in in a more globally accessible way than via a Page Object.


And you need to let RSpec know that you want access to this helper module:


Overall, it’s easy to see that we succeeded in hiding away the Capybara specifics—like finding elements, clicking links, etc. We can now focus on the functionality and less on the actual structure of the markup, which is now encapsulated in a Page Object—the DOM structure should be the least of your concerns when you test something as high-level as feature specs.


Setup stuff like factory data belongs in the specs and not in Page Objects. Also, assertions are probably better placed outside your Page Objects to achieve a separation of concerns.

There are two different perspectives on the topic. Advocates for putting assertions into Page Objects say it helps with avoiding duplication of assertions. You can provide better error messages and achieve a better “Tell, Don’t Ask” style. On the other hand, advocates for assertion-free Page Objects argue that it’s better not to mix responsibilities. Providing access to page data and assertion logic are two separate concerns and lead to bloated Page Objects when mixed. Page Object’s responsibility is access to the state of pages, and assertion logic belongs to specs.

Page Objects Types

Components represent the smallest units and are more focused—like a form object, for example.

Pages combine more of these components and are abstractions of a full page.

Experiences, as you guessed by now, span the whole flow across potentially many different pages. They are more high-level. They focus on the flow the user experiences while they interact with various pages. A checkout flow which has a couple of steps is a good example to think about this.

When & Why?

It’s a good idea to apply this design pattern a little bit later in a project’s life cycle—when you have amassed a little bit of complexity in your feature specs and when you can identify repeating patterns like DOM structures, extracted methods or other commonalities that are consistent on your pages.

So you probably shouldn’t start writing Page Objects right away. You approach these refactorings gradually when the complexity and size of your application / tests grow. Duplications and refactorings that need a better home through Page Objects will be easier to spot over time.

My recommendation is to start with extracting methods in your feature specs locally. Once they hit critical mass, they’ll look like obvious candidates for further extraction, and most of them will probably fit the profile for Page Objects. Start small, because premature optimization leaves nasty bite marks!

Final Thoughts

Page Objects provide you the opportunity to write clearer specs that read better and are overall a lot more expressive because they are more high-level. Besides that, they offer a nice abstraction for everybody who likes to write OO code. They hide the specifics of the DOM and also enable you to have private methods that do the dirty work while being unexposed to the public API. Extracted methods in your feature specs don’t offer quite the same luxury. The API of Page Objects doesn’t need to share the nitty-gritty Capybara details.

For all the scenarios when design implementations change, your descriptions of how your app should work don’t need to change when you use Page Objects—your feature specs are more focused on user-level interactions and don’t care so much about the specifics of the DOM implementations. Since change is inevitable, Page Objects become critical when applications grow and also aid understanding when the sheer size of the application means drastically increased complexity.

Did you find this post useful?
Want a weekly email summary?
Subscribe below and we’ll send you a weekly email summary of all new Code tutorials. Never miss out on learning about the next big thing.
Looking for something to help kick start your next project?
Envato Market has a range of items for sale to help get you started.