Dependency injection in React

Written by Simohamed Marhraoui ✏️

Dependency injection (DI) is a pattern where components necessary for your code to run are hot-swappable. This means that your dependencies are not hard-coded in your implementation and can change as your environment changes.

Dependency injection in JavaScript

To illustrate the principles of DI, imagine an npm module that exposes the following ping function:

export const ping = (url) => {
  return new Promise((res) => {
    fetch(url)
      .then(() => res(true))
      .catch(() => res(false))
  })
}

Using the ping function in a modern browser would work just fine.

import { ping } from "./ping"

ping("https://logrocket.com").then((status) => {
  console.log(status ? "site is up" : "site is down")
})

But running this code inside Node.js would throw an error because fetch is not implemented in Node.js. However, there are many fetch implementations and polyfills for Node.js that we can use.

DI allows us to turn fetch into an injectable dependency of ping, like so:

export const ping = (url, fetch = window.fetch) => {
  return new Promise((res) => {
    fetch(url)
      .then(() => res(true))
      .catch(() => res(false))
  })
}

We are not required to give fetch a default value of window.fetch, but not requiring us to include it every time we use ping makes for a better development experience.

Now, in a Node environment, we can use node-fetch in conjunction with our ping function, like so:

import fetch from "node-fetch"
import { ping } from "./ping"

ping("https://logrocket.com", fetch).then((status) => {
  console.log(status ? "site is up" : "site is down")
})

Working with multiple dependencies

If we have multiple dependencies, it wouldn’t be feasible to keep adding them as parameters: func(param, dep1, dep2, dep3,…). Instead, a better option is to have an object for dependencies:

const ping = (url, deps) => {
  const { fetch, log } = { fetch: window.fetch, log: console.log, ...deps }

  log("ping")

  return new Promise((res) => {
    fetch(url)
      .then(() => res(true))
      .catch(() => res(false))
  })
}

ping("https://logrocket.com", {
  log(str) {
    console.log("logging: " + str)
  }
})

Our parameter deps will be spread into an implementation object and will override the functions that it provides. By destructuring from this modified object, the surviving properties will be used as dependencies.

Using this pattern, we can choose to override one dependency but not the others.

Dependency injection in React

While working with React, we make heavy use of custom hooks to fetch data, track user behavior, and perform complex calculations. Needless to say that we do not wish to (nor can we) run these hooks on all environments.

Tracking a page visit during testing will corrupt our analytics data, and fetching data from a real backend would translate to slow-running tests.

Testing is not the only such environment. Platforms like Storybook streamline documentation and can do without using many of our hooks and business logic.

Dependency injection via props

Take the following component, for example:

import { useTrack } from '~/hooks'

function Save() {
  const { track } = useTrack()

  const handleClick = () => {
    console.log("saving...")
    track("saved")
  }

  return <button onClick={handleClick}>Save</button>
}

As mentioned before, running useTrack (and by extension, track) is something to avoid. Therefore, we will convert useTrack into a dependency of the Save component via props:

import { useTracker as _useTrack } from '~/hooks'

function Save({ useTrack = _useTrack }) {
  const { track } = useTrack()

  /* ... */
}

By aliasing our useTracker to avoid name collision and using it as a default value of a prop, we preserve the hook in our app and have the ability to override it whenever the need arises.

The name _useTracker is one naming convention out of many: useTrackImpl, useTrackImplementation, and useTrackDI are all widely used conventions when trying to avoid collision.

Inside Storybook, we can override the hook as such, using a mocked implementation.

import Save from "./Save"

export default {
  component: Save,
  title: "Save"
}

const Template = (args) => <Save {...args} />
export const Default = Template.bind({})

Default.args = {
  useTrack() {
    return { track() {} }
  }
}

Using TypeScript

When working with TypeScript, it is useful to let other developers know that a dependency injection prop is just that, and use the exact typeof implementation to retain type safety:

function App({ useTrack = _useTrack }: Props) {
  /* ... */
}

interface Props {
  /**
   * For testing and storybook only.
   */
  useTrack?: typeof _useTrack
}

Dependency injection via the Context API

Working with the Context API makes dependency injection feel like a first-class citizen of React. Having the ability to redefine the context in which our hooks are run at any level of the component comes in handy when switching environments.

Many well-known libraries provide mocked implementations of their providers for testing purposes. React Router v5 has MemoryRouter, while Apollo Client provides a MockedProvider. But, if we employ a DI-powered approach, such mocked providers are not necessary.

In development, we can use a bare queryClient with all the default options intact.

import { QueryClient, QueryClientProvider } from "react-query"
import { useUserQuery } from "~/api"

const queryClient = new QueryClient()

function App() {
  return (
    <QueryClientProvider client={queryClient}>
      <User />
    </QueryClientProvider>
  )
}

function User() {
  const { data } = useUserQuery()
  return <p>{JSON.stringify(data)}</p>
}

But when testing our code, features like retries, re-fetch on window focus, and cache time can all be adjusted accordingly.

// storybook/preview.js
import { QueryClient, QueryClientProvider } from "react-query"

const queryClient = new QueryClient({
  queries: {
    retry: false,
    cacheTime: Number.POSITIVE_INFINITY
  }
})

/** @type import('@storybook/addons').DecoratorFunction[] */
export const decorators = [
  (Story) => {
    return (
      <QueryClientProvider client={queryClient}>
        <Story />
      </QueryClientProvider>
    )
  },
]

Dependency injection in React is not exclusive to hooks, but also JSX, JSON, and anything that we wish to abstract away or change under different circumstances.

Alternatives to dependency injection

Depending on the context, dependency injection might not be the right tool for the job. Data-fetching hooks, for instance, are better mocked using an interceptor (like MSW) instead of injecting hooks all over your test code, and outright mocking functions remains an advanced and cumbersome tool for bigger problems.

Why should you use dependency injection?

Reasons to use DI:

No overhead in development, testing, or production

Extremely easy to implement

Does not require a mocking/stubbing library because it’s native to JavaScript

Works for all your stubbing needs, such as components, classes, and regular functions

Reasons to not use DI:

Clutters your imports and components’ props/API

Might be confusing to other developers

Conclusion

In this article, we took a look at a library-free guide to dependency injection in JavaScript and make the case for its use in React for testing and documentation. We used Storybook to illustrate our use of DI, and finally, reflected on reasons why you should and shouldn’t use DI in your code.

Full visibility into production React apps

Debugging React applications can be difficult, especially when users experience issues that are hard to reproduce. If you’re interested in monitoring and tracking Redux state, automatically surfacing JavaScript errors, and tracking slow network requests and component load time, try LogRocket.

The LogRocket Redux middleware package adds an extra layer of visibility into your user sessions. LogRocket logs all actions and state from your Redux stores.

Modernize how you debug your React apps — start monitoring for free