I still remember the feelings when I first started learning about routing in client side applications. At the time I was just a wee lad still getting my feet wet with this whole "Single Page Application" thing and I'd be lying if I said it didn't take a poop all over my brain. From the beginning it's as if my brain treated my application code and my router code as two unique and distinct ideas. They were like step brothers who didn't like each other but were forced to live together anyway.
Over the last few years I have, probably at this point against your approval, been fortunate enough to be able to teach this idea of routing to other developers. Unfortunately, it turns out that most of our brains seem to be wired similarly to mine. I think there's a few reasons for this. First, routing in general is pretty complex. This makes, for those library authors, finding the right abstraction over routing even more complex. Second, because of this complexity, consumers of routing libraries tend to blindly trust the abstraction without really understanding what's going on under the hood. In this tutorial, we'll dive into solving both problems. First the later by recreating our own simplified version of React Router v4 which will then shed some light on the former, that is, whether or not RRv4 is a reasonable abstraction.
Here's our app code we'll be using to test our ~React Router implementation once we've built it. You can play around with the final example here
const Home = () => (
<h2>Home</h2>
)
const About = () => (
<h2>About</h2>
)
const Topic = ({ topicId }) => (
<h3>{topicId}</h3>
)
const Topics = ({ match }) => {
const items = [
{ name: 'Rendering with React', slug: 'rendering' },
{ name: 'Components', slug: 'components' },
{ name: 'Props v. State', slug: 'props-v-state' },
]
return (
<div>
<h2>Topics</h2>
<ul>
{items.map(({ name, slug }) => (
<li key={name}>
<Link to={`${match.url}/${slug}`}>{name}</Link>
</li>
))}
</ul>
{items.map(({ name, slug }) => (
<Route key={name} path={`${match.path}/${slug}`} render={() => (
<Topic topicId={name} />
)} />
))}
<Route exact path={match.url} render={() => (
<h3>Please select a topic.</h3>
)}/>
</div>
)
}
const App = () => (
<div>
<ul>
<li><Link to="/">Home</Link></li>
<li><Link to="/about">About</Link></li>
<li><Link to="/topics">Topics</Link></li>
</ul>
<hr/>
<Route exact path="/" component={Home}/>
<Route path="/about" component={About}/>
<Route path="/topics" component={Topics} />
</div>
)
If you're unfamiliar with React Router v4, here's the basic premise. Route
s render some UI when a URL matches a location you specify in the Route's path
prop. Link
s provide a declarative, accessible way to navigate around your app. In other words, the Link
component allows you to update the URL, and the Route
component changes your UI based on that new URL. The focus of this tutorial isn't actually on teaching the basics of RRV4, so if the code above is still confusing, head over to the official docs, play around with the examples, and once you're more comfortable, come back.
The first thing you should notice is that we've introduced two components that were given to us by the router into our app, Link
and Route
. My favorite aspect of React Router v4 is that the API is "Just Components™". What this means is that if you're already familiar with React, the same intuition you have about components and how to compose them, will continue to hold true in regards to your routing code. And even more convenient for our use case here, because we're already familiar with how to create components, creating our own React Router will be nothing more than doing what we're already familiar with, creating more components.
We're going to start off by creating our Route
component. Before we dive into the code, let's go ahead and check out the API (which conveniently is just which props it takes).
In our example above, you'll notice that <Route>
can take in three props. exact
, path
, and component
. This means the propTypes
for our Route
component currently look like this,
static propTypes = {
exact: PropTypes.bool,
path: PropTypes.string,
component: PropTypes.func,
}
There's a few subtleties here. First, the reason path
isn't required is because if a Route
isn't given a path, it will automatically be rendered. Second, the reason component
isn't marked as required either is because there are actually a few different ways to tell React Router the UI you want to render if the path matches. One way that isn't in our example above is with the render
prop. It looks like this,
<Route path='/settings' render={({ match }) => {
return <Settings authed={isAuthed} match={match} />
}} />
render
allows you to conveniently inline a function which returns some UI rather than creating a separate component. So we'll add that to our propTypes as well,
static propTypes = {
exact: PropTypes.bool,
path: PropTypes.string,
component: PropTypes.func,
render: PropTypes.func,
}
Now we know what props Route
receives, let's talk again about what it actually does. Route "renders some UI when the URL matches a location you specify in the Route's path
prop". Based off of that definition, we know that <Route>
is going to need some functionality which checks if the current URL matches the component's path
prop. If it does, we'll render some UI. If it doesn't, we'll do nothing by returning null.
Let's see what this looks like in code, trusting that we'll build out the matching function, which we'll call matchPath
, later.
class Route extends Component {
static propTypes = {
exact: PropTypes.bool,
path: PropTypes.string,
component: PropTypes.func,
render: PropTypes.func,
}
render () {
const {
path,
exact,
component,
render,
} = this.props
const match = matchPath(
location.pathname, // global DOM variable
{ path, exact }
)
if (!match) {
// Do nothing because the current
// location doesn't match the path prop.
return null
}
if (component) {
// The component prop takes precedent over the
// render method. If the current location matches
// the path prop, create a new element passing in
// match as the prop.
return React.createElement(component, { match })
}
if (render) {
// If there's a match but component
// was undefined, invoke the render
// prop passing in match as an argument.
return render({ match })
}
return null
}
}
Now Route
is looking pretty solid. If the current location matches the path
prop that was passed in, we render some UI, if not, we don't do anything.
Let's take a step back for a moment and talk about routing in general. In a client side application, there are really only two ways for the user to update the URL. The first way is clicking on an anchor tag and the second is by clicking the back/forward button. Foundationally, our router needs to be aware of the current URL and render UI based on it. What this also means is that our router needs to be aware of when the URL changes, so that it can figure out which new UI to display based on that new URL. If we know that the only way to update a URL is through an anchor tag or the forward/back button, we can plan for and react to those changes. We'll get into anchor tags a little later when we build out our <Link>
component, but for now, I want to focus on the back/forward buttons. React Router uses History's .listen
method to listen to changes for the current URL, but to avoid bringing in another library, we'll use HTML5's popstate
event. popstate
, which will be fired whenever the user clicks on the forward or back button, is exactly what we need. Because it's the Route
s that are rendering the UI based off the current URL, it makes sense to also give Route
s the ability to listen for and re-render whenever a popstate
event occurs. By re-rendering, each Route
will re-check to see if they match with the new URL. If they do, they'll render UI, if not, they'll do nothing. Let's see what this looks like now,
class Route extends Component {
static propTypes: {
path: PropTypes.string,
exact: PropTypes.bool,
component: PropTypes.func,
render: PropTypes.func,
}
componentWillMount() {
addEventListener("popstate", this.handlePop)
}
componentWillUnmount() {
removeEventListener("popstate", this.handlePop)
}
handlePop = () => {
this.forceUpdate()
}
render() {
const {
path,
exact,
component,
render,
} = this.props
const match = matchPath(location.pathname, { path, exact })
if (!match)
return null
if (component)
return React.createElement(component, { match })
if (render)
return render({ match })
return null
}
}
You should notice that all we've done is add a popstate
listener when the component mounts, and when the popstate
event is fired, we call forceUpdate
which will kick off a re-render.
Now, no matter how many <Route>
s we're rendering, each of them will listen for, re-match, and re-render based on the forward/back buttons.
One thing we've been "hand waving" over up until this point has been our matchPath
function. This function is pivotal to our router because it's the function which is going to decide if a current URL matches the path of a <Route>
component as we talked about above. One nuance to matchPath
is we need to make sure that we take into account <Route>
s exact
prop. If you're not familiar with what exact
does, here's an explanation straight from the docs
When
true
, will only match if the path matches thelocation.pathname
exactly.
| path | location.pathname | exact | matches? |
| /one
| /one/two
| true
| no |
| /one
| /one/two
| false
| yes |
Now, let's dive into the implementation of our matchPath
function. If you look back at our Route
component, you'll see that the signature for matchPatch
looks like this,
const match = matchPath(location.pathname, { path, exact })
Where match
is either an object or null depending on if there was a match. Based on that signature, we can build out the first part of matchPath
like this,
const matchPatch = (pathname, options) => {
const { exact = false, path } = options
}
Here we're using some ES6 magic. We're saying "create a variable called exact which equates to options.exact, unless that's undefined, then set it to false. Also create a variable called path which eqautes to options.path".
Earlier I mentioned "the reason path
isn't required is because if a Route
isn't given a path, it will automatically be rendered". Well since it's indirectly our matchPath
function which decides if something is rendered or not (by whether there's a match), let's add that functionality now.
const matchPatch = (pathname, options) => {
const { exact = false, path } = options
if (!path) {
return {
path: null,
url: pathname,
isExact: true,
}
}
}
Now comes the matching part. React Router uses pathToRegex for this, we'll simplify things and just use a simple Regex.
const matchPatch = (pathname, options) => {
const { exact = false, path } = options
if (!path) {
return {
path: null,
url: pathname,
isExact: true,
}
}
const match = new RegExp(`^${path}`).exec(pathname)
}
If you're not familiar with .exec
, it's going to return an array containing the matched text if it finds a match, otherwise it returns null.
Here is every match
when our example app routes to `/topics/components
| path | location.pathname | return value |
| /
| /topics/components
| ['/']
|
| /about
| " " | null
|
| /topics
| " " | ['/topics']
|
| /topics/rendering
| " " | null
|
| /topics/components
| " " | ['/topics/components']
|
| /topics/props-v-state
| " " | null
|
| /topics
| " " | ['/topics']
|
Notice that we're getting a
match
for every<Route>
that's in our app. That's because, well, each<Route>
callsmatchPath
in its render method.
Now that we know what the match
that .exec
is returning, all we need to do now is figure out if there's a match.
`javascript
const matchPatch = (pathname, options) => {
const { exact = false, path } = options
if (!path) {
return {
path: null,
url: pathname,
isExact: true,
}
}
const match = new RegExp(^${path}
).exec(pathname)
if (!match) {
// There wasn't a match.
return null
}
const url = match[0]
const isExact = pathname === url
if (exact && !isExact) {
// There was a match, but it wasn't
// an exact match as specified by
// the exact prop.
return null
}
return {
path,
url,
isExact,
}
}
`
Earlier I mentioned how there's really just two ways to update the URL if you're the user, via the back/forward buttons, or clicking on an achor tag. We've taken care of re-rendering on back/forward clicks via the popstate
event listener in our Route
, now let's take care of the anchor tag by building our our Link
component.
The API for Link
looks like this,
<Link to='/some-path' replace={false} />
Where to
is a string and is the location to link to and replace
is a boolean which when true, clicking the link will replace the current entry in the history stack instead of adding a new one.
Adding those propTypes to our Link component, we get this,
javascript
class Link extends Component {
static propTypes = {
to: PropTypes.string.isRequired,
replace: PropTypes.bool,
}
}
Now we know that the render method in our Link
component needs to return an anchor tag, but we obviously don't want to cause a full page refresh every time we switch routes, so we'll hijack the anchor tag by adding a onClick
handler to it
`javascript
class Link extends Component {
static propTypes = {
to: PropTypes.string.isRequired,
replace: PropTypes.bool,
}
handleClick = (event) => {
const { replace, to } = this.props
event.preventDefault()
// route here.
}
render() {
const { to, children} = this.props
return (
<a href={to} onClick={this.handleClick}>
{children}
</a>
)
}
}
`
Now all that's lacking is actually changing the current location. To do this React Router uses History's push
and replace
methods, but we'll use HTML5's pushState and replaceState methods to avoid adding in a dependency.
We're hand waving over the History library in this post as a way to avoid external dependencies but it's crucial for the real React Router code as it normalizes the differences in managing session history in various browser environments.
Both pushState
and replaceState
take in three arguments. The first is an object which is associated with the new history entry - we don't need this functionality so we'll just pass in an empty object. The second is a title, which we also don't need so we'll pass in null. The third, and the one we'll actually use, is a relative URL.
`javascript
const historyPush = (path) => {
history.pushState({}, null, path)
}
const historyReplace = (path) => {
history.replaceState({}, null, path)
}
`
Now inside of our Link
component, we'll invoke historyPush
or historyReplace
depending on the replace
prop,
`javascript
class Link extends Component {
static propTypes = {
to: PropTypes.string.isRequired,
replace: PropTypes.bool,
}
handleClick = (event) => {
const { replace, to } = this.props
event.preventDefault()
replace ? historyReplace(to) : historyPush(to)
}
render() {
const { to, children} = this.props
return (
<a href={to} onClick={this.handleClick}>
{children}
</a>
)
}
}
`
Now there's just one more, albeit crucial addition we need to make. If you were to play around with our example app with our current router code, you'd notice a pretty big problem. When you navigate around, the URL would update, but the UI would stay exactly the same. This is because even though we're changing the location with our historyReplace
or historyPush
functions, our <Route>
s are unaware of that change and don't know that they should re-render and re-match. To solve this problem, we need to keep track of which <Route>
s have been rendered and call forceUpdate
on them whenever a route changes.
React Router gets around this problem by having using a combination of setState, context, and history.listen inside of a Router component you wrap your code with.
To keep our router simple, we'll keep track of which <Route>
s have been rendered by pushing their instances to an array, then whenever a location change occurs, we can loop through that array and call forceUpdate on all the instances.
`javascript
let instances = []
const register = (comp) => instances.push(comp)
const unregister = (comp) => instances.splice(instances.indexOf(comp), 1)
`
Notice we've created two functions. We'll call register
whenever a <Route>
is mounted and call unregister
whenever it unmounts. Then, whenever we call historyPush
or historyReplace
(which we will every time a user clicks on a <Link>
), we can loop through those instances and forceUpdate
.
Let's update our <Route>
component first,
`javascript
class Route extends Component {
static propTypes: {
path: PropTypes.string,
exact: PropTypes.bool,
component: PropTypes.func,
render: PropTypes.func,
}
componentWillMount() {
addEventListener("popstate", this.handlePop)
register(this)
}
componentWillUnmount() {
unregister(this)
removeEventListener("popstate", this.handlePop)
}
...
}
`
Now, let's update historyPush
and historyReplace
`javascript
const historyPush = (path) => {
history.pushState({}, null, path)
instances.forEach(instance => instance.forceUpdate())
}
const historyReplace = (path) => {
history.replaceState({}, null, path)
instances.forEach(instance => instance.forceUpdate())
}
`
🎉 now whenever a <Link>
is clicked and the location changes, each <Route>
will be aware of that and re-match and re-render.
Now, our full router code looks like this code below, and our example app above works perfectly with it.
`javascript
import React, { PropTypes, Component } from 'react'
let instances = []
const register = (comp) => instances.push(comp)
const unregister = (comp) => instances.splice(instances.indexOf(comp), 1)
const historyPush = (path) => {
history.pushState({}, null, path)
instances.forEach(instance => instance.forceUpdate())
}
const historyReplace = (path) => {
history.replaceState({}, null, path)
instances.forEach(instance => instance.forceUpdate())
}
const matchPath = (pathname, options) => {
const { exact = false, path } = options
if (!path) {
return {
path: null,
url: pathname,
isExact: true
}
}
const match = new RegExp(^${path}
).exec(pathname)
if (!match)
return null
const url = match[0]
const isExact = pathname === url
if (exact && !isExact)
return null
return {
path,
url,
isExact,
}
}
class Route extends Component {
static propTypes: {
path: PropTypes.string,
exact: PropTypes.bool,
component: PropTypes.func,
render: PropTypes.func,
}
componentWillMount() {
addEventListener("popstate", this.handlePop)
register(this)
}
componentWillUnmount() {
unregister(this)
removeEventListener("popstate", this.handlePop)
}
handlePop = () => {
this.forceUpdate()
}
render() {
const {
path,
exact,
component,
render,
} = this.props
const match = matchPath(location.pathname, { path, exact })
if (!match)
return null
if (component)
return React.createElement(component, { match })
if (render)
return render({ match })
return null
}
}
class Link extends Component {
static propTypes = {
to: PropTypes.string.isRequired,
replace: PropTypes.bool,
}
handleClick = (event) => {
const { replace, to } = this.props
event.preventDefault()
replace ? historyReplace(to) : historyPush(to)
}
render() {
const { to, children} = this.props
return (
<a href={to} onClick={this.handleClick}>
{children}
</a>
)
}
}
`
Bonus: The React Router API also comes with a <Redirect>
component. Using the code we've previously written, creating this component is pretty straight forward
`javascript
class Redirect extends Component {
static defaultProps = {
push: false
}
static propTypes = {
to: PropTypes.string.isRequired,
push: PropTypes.bool.isRequired,
}
componentDidMount() {
const { to, push } = this.props
push ? historyPush(to) : historyReplace(to)
}
render() {
return null
}
}
`
Notice this component isn't actually rendering any UI, instead, it's acting purely as a route director, hence the name.
I hope this has helped you create a better mental model of what's happening in React Router while also helping you to gain an appreciation for React Router's elegance and "Just Components" API. I've always said that React will make you a better JavaScript developer. I now also believe that React Router will make you a better React developer. Because everything is just components, if you know React, you know React Router.
ps. shoutsout to Ryan who created the first implementation of the mini router and helped me with this post.