Building pages with Storybook
Storybook helps you build any component, from small “atomic” components to composed pages. But as you move up the component hierarchy toward the level of pages, you end up dealing with more complexity.
There are many ways to build pages in Storybook. Here are common patterns and solutions.
- Pure presentational pages.
- Connected components (e.g., network requests, context, browser environment).
Teams at the BBC, The Guardian, and the Storybook maintainers themselves build pure presentational pages. If you take this approach, you don't need to do anything special to render your pages in Storybook.
It's straightforward to write components to be fully presentational up to the screen level. That makes it easy to show in Storybook. The idea is that you do all the messy “connected” logic in a single wrapper component in your app outside of Storybook. You can see an example of this approach in the Data chapter of the Intro to Storybook tutorial.
- Easy to write stories once components are in this form.
- All the data for the story is encoded in the args of the story, which works well with other parts of Storybook's tooling (e.g. controls).
Your existing app may not be structured in this way, and it may be difficult to change it.
Fetching data in one place means that you need to drill it down to the components that use it. This can be natural in a page that composes one big GraphQL query (for instance), but other data fetching approaches may make this less appropriate.
It's less flexible if you want to load data incrementally in different places on the screen.
When you are building screens in this way, it is typical that the inputs of a composite component are a combination of the inputs of the various sub-components it renders. For instance, if your screen renders a page layout (containing details of the current user), a header (describing the document you are looking at), and a list (of the subdocuments), the inputs of the screen may consist of the user, document and subdocuments.
In such cases it is natural to use args composition to build the stories for the page based on the stories of the sub-components:
This approach is beneficial when the various subcomponents export a complex list of different stories. You can pick and choose to build realistic scenarios for your screen-level stories without repeating yourself. Your story maintenance burden is minimal by reusing the data and taking a Don't-Repeat-Yourself(DRY) philosophy.
If you need to render a connected component in Storybook, you can mock the network requests to fetch its data. There are various layers in which you can do that.
If you are using a provider that supplies data via the context, you can wrap your story in a decorator that provides a mocked version of that provider. For example, in the Screens chapter of the Intro to Storybook tutorial, we mock a Redux provider with mock data.
Connected applications such as Twitter, Instagram, amongst others, are everywhere, consuming data either from REST or GraphQL endpoints. If you're working in an application that relies on either of these data providers, you can add Mock Service Worker (MSW) via Storybook's MSW addon to mock data alongside your app and stories.
Mock Service Worker is an API mocking library. It relies on service workers to capture network requests and provides mocked data in response. The MSW addon adds this functionality into Storybook, allowing you to mock API requests in your stories.
If you're working with pure presentational screens, adding stories through args composition is recommended. You can easily encode all the data via args, removing the need for handling it with "wrapper components". However, this approach loses its flexibility if the screen's data is retrieved from a RESTful endpoint within the screen itself. For instance, if your screen had a similar implementation to retrieve a list of documents:
To test your screen with the mocked data, you could write a similar set of stories:
The mocked data (i.e.,
TestData) will be injected via parameters, enabling you to configure it per-story basis.
In addition to mocking RESTful requests, the other noteworthy feature of the MSW addon is the ability to mock incoming data from any of the mainstream GraphQL clients (e.g., Apollo Client, URQL or React Query). For instance, if your screen retrieves the user's information and a list of documents based on a query result, you could have a similar implementation:
To test your screen with the GraphQL mocked data, you could write the following stories:
It is also possible to mock imports directly, as you might in a unit test, using Webpack’s aliasing. It's advantageous if your component makes network requests directly with third-party libraries.
We're going to use isomorphic-fetch as an example.
Inside a directory called
__mocks__, create a new file called
isomorphic-fetch.js with the following code:
The code above creates a decorator which reads story-specific data off the story's parameters, enabling you to configure the mock on a per-story basis.
To use the mock in place of the real import, we use webpack aliasing:
Add the decorator you've just implemented to your storybook/preview.js:
Finally, we can set the mock values in a specific story. Let's borrow an example from this blog post:
Another mocking approach is to use libraries that intercept calls at a lower level. For instance, you can use
fetch-mock to mock fetch requests specifically.
It's possible to avoid mocking the dependencies of connected "container" components entirely by passing them around via props or React context. However, it requires a strict split of the container and presentational component logic. For example, if you have a component responsible for data fetching logic and rendering DOM, it will need to be mocked as previously described.
It’s common to import and embed container components amongst presentational components. However, as we discovered earlier, to render them within Storybook, we’ll likely have to mock their dependencies or the imports themselves.
Not only can this quickly grow to become a tedious task, but it’s also challenging to mock container components that use local states. So, instead of importing containers directly, a solution to this problem is to create a React context that provides the container components. It allows you to freely embed container components as usual, at any level in the component hierarchy without worrying about subsequently mocking their dependencies; since we can swap out the containers themselves with their mocked presentational counterpart.
We recommend dividing context containers up over specific pages or views in your app. For example, if you had a
ProfilePage component, you might set up a file structure as follows:
It’s also often helpful to set up a “global” container context (perhaps named
GlobalContainerContext) for container components that may be rendered on every page of your app and adding them to the top level of your application. While it’s possible to place every container within this global context, it should only provide globally required containers.
Let’s look at an example implementation of this approach.
First, create a React context, and name it
ProfilePageContext. It does nothing more than export a React context:
ProfilePage is our presentational component. It will use the
useContext hook to retrieve the container components from
In the context of Storybook, instead of providing container components through context, we’ll instead provide their mocked counterparts. In most cases, the mocked versions of these components can often be borrowed directly from their associated stories.
If the same context applies to all
ProfilePage stories, we can also use a decorator.
Now, in the context of your application, you’ll need to provide
ProfilePage with all of the container components it requires by wrapping it with
For example, in Next.js, this would be your
If you’ve set up
GlobalContainerContext, you’ll need to set up a decorator within Storybook’s
preview.js to provide context to all stories. For example: