Resolve components asynchronously, with support for code splitting and advanced server side rendering use cases.
const AsyncProduct = asyncComponent({
resolve: () => System.import('./Product'),
LoadingComponent: ({ productId }) => <div>Loading {productId}</div>, // Optional
ErrorComponent: ({ error }) => <div>{error.message}</div> // Optional
});
<AsyncProduct productId={1} /> // 🚀
This library does not require that you use either Webpack or Babel. Instead it provides you a generic and "pure" Javascript/React API which allows for the expression of lazy-loaded Components. It's Promise
-based API naturally allows you to take advantage of modern code splitting APIs (e.g import()
, System.import
, require.ensure
).
- Supports any major code splitting API.
- Show a
LoadingComponent
until your component is resolved. - Show an
ErrorComponent
if your component resolution fails. - Prevents flash-of-content by tracking already resolved Components.
- Full server side rendering support, allowing client side state rehydration, avoiding React checksum errors.
Imagine you had the following Product
component:
export default function Product({ id }) {
return <div>Product {id}</div>
}
To make this asynchronous create a new file that wraps it with asyncComponent
, like so:
import { asyncComponent } from 'react-async-component';
export default asyncComponent({
resolve: () => System.import('./Product')
});
I recommend that you use the following folder/file structure:
|- components
|- AsyncProduct
|- index.js // contains asyncComponent
|- Product.js // The component you want resolved asynchronously
Now, you can simply import AsyncProduct
anywhere in your application and use it exactly as you would any other component.
For example:
import AsyncProduct from './components/AsyncProduct'
export default function MyApp() {
return (
<div>
<h1>Welcome to My App</h1>
<AsyncProduct id={1337} />
</div>
)
}
The asynchronous component factory. Config goes in, an asynchronous component comes out.
config
(Object) : The configuration object for the async Component. It has the following properties available:resolve
(() => Promise) : A function that should return aPromise
that will resolve the Component you wish to be async.LoadingComponent
(Component, Optional, default:null
) : A Component that will be displayed until your async Component is resolved. All props will be passed to it.ErrorComponent
(Component, Optional, default:null
) : A Component that will be displayed if any error occurred whilst trying to resolve your component. All props will be passed to it as well as anerror
prop containing theError
.name
(String, Optional, default:'AsyncComponent'
) : Use this if you would like to name the created async Component, which helps when firing up the React Dev Tools for example.autoResolveES2015Default
(Boolean, Optional, default:true
) : Especially useful if you are resolving ES2015 modules. The resolved module will be checked to see if it has a.default
and if so then the value attached to.default
will be used. So easy to forget to do that. 😀env
(String, Optional) : Provide either'node'
or'browser'
so you can write your own environment detection. Especially useful when using PhantomJS or ElectronJS to prerender the React App.serverMode
(Boolean, Optional, default:'resolve'
) : Only applies for server side rendering applications. Please see the documentation on server side rendering. The following values are allowed.'resolve'
- Your asynchronous component will be resolved and rendered on the server. It's children will be checked to see if there are any nested asynchronous component instances, which will then be processed based on theserverMode
value that was associated with them.'defer'
- Your asynchronous component will not be rendered on the server, instead deferring rendering to the client/browser.'boundary'
- Your asynchronous component will be resolved and rendered on the server. However, if it has a nested asynchronous component instance within it's children that component will be ignored and treated as being deferred for rendering in the client/browser instead (it's serverMode will be ignored). We highly recommend that you consider usingdefer
as much as you can.
A React Component.
export default asyncComponent({
resolve: () => import('./Product'),
LoadingComponent: ({ id }) => <div>Loading product {id}</div>
})
export default asyncComponent({
resolve: () => import('./Product'),
ErrorComponent: ({ error }) => <div>{error.message}</div>
})
export default asyncComponent({
resolve: () => new Promise(resolve =>
// Webpack's code splitting API w/naming
require.ensure(
[],
(require) => {
resolve(require('./Product'));
},
'ChunkName'
)
)
})
Currently only useful when building server side rendering applications. Wraps your application allowing for efficient and effective use of asynchronous components.
asyncContext
(Object) : Used so that you can gain hooks into the context for server side rendering render tracking and rehydration. See thecreateAsyncContext
helper for creating an instance.rehydrateState
(Object, Optional) : Used on the "browser-side" of a server side rendering application (see the docs). This allows you to provide the state returned by the server to be used to rehydrate the client appropriately.
Creates an asynchronous context for use by the <AsyncComponentProvider />
. The context is an object that exposes the following properties to you:
getState()
(() => Object) : A function that when executed gets the current state of the<AsyncComponentProvider />
. i.e. which async components resolved / failed to resolve etc. This is especially useful for server sider rendering applications where you need to provide the server rendered state to the client instance in order to ensure the required asynchronous component instances are resolved prior to render.
NOTE: This section only really applies if you would like to have control over the behaviour of how your asyncComponent instances are rendered on the server. If you don't mind your asyncComponents always being resolved on the client only then you need not do any of the below. In my opinion there is great value in just server rendering your app shell and having everything else resolve on the client, however, you may have very strict SEO needs - in which case, we have your back.
This library has been designed for interoperability with react-async-bootstrapper
.
react-async-bootstrapper
allows us to do a "pre-render parse" of our React Element tree and execute an asyncBootstrap
function that are attached to a components within the tree. In our case the "bootstrapping" process involves the resolution of asynchronous components so that they can be rendered "synchronously" by the server. We use this 3rd party library as it allows interoperability with other libraries which also require a "bootstrapping" process (e.g. data preloading as supported by react-jobs
).
Firstly, install react-async-bootstrapper
:
npm install react-async-bootstrapper
Now, let's configure the "server" side. You could use a similar express
(or other HTTP server) middleware configuration:
import React from 'react'
import { renderToString } from 'react-dom/server'
import { AsyncComponentProvider, createAsyncContext } from 'react-async-component' // 👈
import asyncBootstrapper from 'react-async-bootstrapper' // 👈
import serialize from 'serialize-javascript'
import MyApp from './shared/components/MyApp'
export default function expressMiddleware(req, res, next) {
// Create the async context for our provider, this grants
// 👇 us the ability to tap into the state to send back to the client.
const asyncContext = createAsyncContext()
// 👇 Ensure you wrap your application with the provider.
const app = (
<AsyncComponentProvider asyncContext={asyncContext}>
<MyApp />
</AsyncComponentProvider>
)
// 👇 This makes sure we "bootstrap" resolve any async components prior to rendering
asyncBootstrapper(app).then(() => {
// We can now render our app 👇
const appString = renderToString(app)
// Get the async component state. 👇
const asyncState = asyncContext.getState()
const html = `
<html>
<head>
<title>Example</title>
</head>
<body>
<div id="app">${appString}</div>
<script type="text/javascript">
// Serialise the state into the HTML response 👇
window.ASYNC_COMPONENTS_STATE = ${serialize(asyncState)}
</script>
</body>
</html>`
res.send(html)
})
}
Then on the "client" side you would do the following:
import React from 'react'
import { render } from 'react-dom'
import { AsyncComponentProvider, createAsyncContext } from 'react-async-component' // 👈
import asyncBootstrapper from 'react-async-bootstrapper' // 👈
import MyApp from './components/MyApp'
// 👇 Get any "rehydrate" state sent back by the server
const rehydrateState = window.ASYNC_COMPONENTS_STATE
// Ensure you wrap your application with the provider,
// 👇 and pass in the rehydrateState.
const app = (
<AsyncComponentProvider rehydrateState={rehydrateState}>
<MyApp />
</AsyncComponentProvider>
)
// We run the bootstrapper again, which in this context will
// ensure that all components specified by the rehydrateState
// 👇 will be resolved prior to render.
asyncBootstrapper(app).then(() => {
// 👇 Render the app
render(app, document.getElementById('app'))
});
It is worth us highlighting exactly how we go about resolving and rendering your asyncComponent
instances on the server. This knowledge will help you become aware of potential issues with your component implementations as well as how to effectively use our provided configuration properties to create more efficient implementations.
When running react-async-bootstrapper
on the server the helper has to walk through your react element tree (depth first i.e. top down) in order to discover all the asyncComponent
instances and resolve them in preparation for when you call the ReactDOM.renderToString
. As it walks through the tree it has to call the componentWillMount
method on your Components and then the render
methods so that it can get back the child react elements for each Component and continue walking down the element tree. When it discovers an asyncComponent
instance it will first resolve the Component that it refers to and then it will continue walking down it's child elements (unless you set the configuration for your asyncComponent
to not allow this) in order to try and discover any nested asyncComponent
instances. It continues doing this until it exhausts your element tree.
Although this operation isn't as expensive as an actual render as we don't generate the DOM it can still be quite wasteful if you have a deep tree. Therefore we have provided a set of configuration values that allow you to massively optimise this process. See the next section below.
As discussed in the "SSR AsyncComponent Resolution Process" section above it is possible to have an inefficient implementation of your asyncComponent
instances. Therefore we introduced a new configuration object property for the asyncComponent
factory, called serverMode
, which provides you with a mechanism to optimise the configuration of your async Component instances. Please see the API documentation for more information.
Understand your own applications needs and use the options appropriately . I personally recommend using mostly "defer" and a bit of "boundary". Try to see code splitting as allowing you to server side render an application shell to give the user perceived performance. Of course there will be requirements otherwise (SEO), but try to isolate these components and use a "boundary" as soon as you feel you can.
You can see a "live" version here. This is a deployment of the "React, Universally" starter kit that makes use of this library. Open the network tab and then click the menu items to see the asynchronous component resolving in action.
Let me know if you have any...