An immutable React state management library with a simple mutable API, memoized selectors, and structural sharing. Powered by Immer.
The benefits of immutable state are clear, but maintaining that immutable state can sometimes be burdensome and verbose: updating a value more than one or two levels deep in your state tree can require lots of object/array spreading, and it's relatively easy to accidently mutate something.
react-copy-write lets you use straightforward mutations to update an immutable state tree, thanks to Immer. Since Immer uses the copy-on-write technique to update immutable values, we get the benefits of structural sharing and memoization. This means react-copy-write not only lets you use simple mutations to update state, but it's also very efficient about re-rendering.
react-copy-write requires React 16.3 or later, as it depends on the new React.createContext API.
yarn add react-copy-writereact-copy-write exports a function which takes your base state and returns an object with Provider and Consumer components, along with a few other utility methods. You can use a single state instance for your entire app, or create multiple different independant state instances that interleave throughout your app.
import createState from "react-copy-write";
// You can namespace the components by accessing them as properties,
// e.g., State.Provider / State.Consumer
const State = createState({
user: null,
loggedIn: false
});
// Or destructure if you'd like
const { Provider, Consumer } = createState({
user: null,
loggedIn: false
});The Provider component is what provides the state (crazy, right?). When the Provider component mounts the initial state will be whatever you passed to createState. The Provider component takes no props, and expects that all the associated Consumer components will be rendered as descendants.
class App extends React.Component {
render() {
return (
<State.Provider>
<AppBody />
</State.Provider>
);
}
}You can only ever render a single instance of a given Provider.
Consumer components let you consume or observe some portion of the state. By default that portion will be all of the state.
Lets look at this through a series of examples. Here we have a UserAvatar component, that wants to render an avatar for some given user.
const UserAvatar = ({ id }) => (
<State.Consumer>
{state => (
<div className="avatar">
<img src={state.users[id].avatar.src} />
</div>
)}
</State.Consumer>
);The State.Consumer component expects a render callback as a child, just like the React context consumer it wraps. That render callback will be called with the current state. The problem with this is that whenever any value in state changes, UserAvatar will be re-rendered, even though it's only using a single property from a single, nested object.
To avoid the problem of observing too much state, Consumer components let you pass in a selector function. A selector takes the current state and returns only the subset of that state that the Consumer cares about. Refactoring UserAvatar, we get:
const UserAvatar = ({ id }) => (
<State.Consumer selector={state => state.users[id].avatar}>
{avatar => (
<div>
<img src={avatar.src} />
</div>
)}
</State.Consumer>
);Using the selector prop, we pick out the avatar object from this user's entry in state. Now, UserAvatar will only ever re-render if that avatar object changes. This is possible because react-copy-write is powered by Immer and Immer uses structural sharing, which means that it will re-use unchanged portions of the state tree whenever it's updated.
You may be tempted to use selectors to derive some new kind of state. This is a common and useful pattern as your state shape rarely maps 1:1 to your view. The problem with doing it in selectors is that react-copy-write needs selectors to main referential equailty on every render.
For example, maybe you have a list of blog posts and you want to filter them based on their author.
const UserPosts = ({ userId }) => (
<State.Consumer
selector={state => state.posts.filter(post => post.id === userId)}
>
{userPosts => userPosts.map(post => <Post {...post} />)}
</State.Consumer>
);This works, but now everytime the selector function is called, a new filtered array is returned. That means that UserPosts will re-render needlessly. If you can, move that filtering into the render callback.
const UserPosts = ({ userId }) => (
<State.Consumer selector={state => state.posts}>
{posts => {
const filteredPosts = posts.filter(post => post.id === userId)
return userPosts.map(post => <Post {...post} />)
}
</State.Consumer>
)Now those Post components will only re-render if UserPosts is re-rendered (a new userID) or if state.posts gets updated somewhere else. selector relies on referrential equality checks between renders, so avoid returning any new objects or arrays. Since it's relying on === feel free to return primitive values like strings or numbers which maintain that strict equality between instances.
In some cases, deriving state involves reading from other parts of your state. Maybe we want the UserPosts component to read the user's ID from state too. If you didn't heed the above advice, you might try something like:
const UserPosts = () => (
<State.Consumer selector={state => ({ posts: state.posts, userId: state.user.id }}>
{({posts, userId}) => {
const filteredPosts = posts.filter(post => post.id === userId)
return posts.map(post => <Post id={post.id} />)
}
</State.Consumer>
)This is just as bad as filtering in the selector, since a new object is returned each time. A naive solution (AKA, what I tried to do first) would be to nest Consumers.
const UserPosts = () => (
<State.Consumer selector={state => state.posts}>
{posts => (
<State.Consumer selector={state => state.user.id}>
{userId => {
const filteredPosts = posts.filter(post => post.id === userId);
return posts.map(post => <Post id={post.id} />);
}}
</State.Consumer>
)}
</State.Consumer>
);This is a good thought, but the problem is that Consumers are very protective about re-renders. They'll only render if the state they're observing changes. So if state.posts changes but state.user.id doesn't, it won't update. You could wrap the inner Consumer in it's own component and pass in posts as a prop to trigger a render. This isn't a terrible solution, but it means you're creating a bunch of wrapper components where you normally wouldn't.
To solve this, Consumers can accept an array of selectors.
const UserPosts = () => (
<State.Consumer selector={[state => state.posts, state => state.userId]}>
{[posts, userId] =>
const filteredPosts = posts.filter(post => post.id === userId)
return posts.map(post => <Post id={post.id} />)
)}
</State.Consumer>
)Now the Consumer will re-render if any of the selectors return a new value.
The render callback you pass as a child to Consumer components take a second argument; a mutate function that lets you mutate a draft of the current state, processed by Immer as an immutable state update. If you're wondering how you can get immutable state by mutating state, go check out the Immer repo's README.
Let's start implementing that Post component we've been using:
const Post = ({ id }) => (
<div className="post">
<State.Consumer selector={state => state.posts[id]}>
{post => (
<>
<h1>{post.title}</h1>
<img src={post.image} />
<p>{post.body}</p>
<button>Praise</button>
</>
)}
</State.Consumer>
</div>
);Post just renders a div with a title, an image, some text, and a button to "Praise". We want post.praiseCount to be incremented everytime that button is clicked.
const Post = ({ id }) => (
<div className="post">
<State.Consumer selector={state => state.posts[id]}>
{(post, mutate) => (
<>
<h1>{post.title}</h1>
<img src={post.image} />
<p>{post.body}</p>
<button
onClick={() =>
// Here's the magic:
mutate(draft => {
draft.posts[id].praiseCount += 1;
})
}
>
Praise
</button>
</>
)}
</State.Consumer>
</div>
);Mutate the value you want to change, and an immutable state update will be processed. Only those Consumer components that were observing the praiseCount state will be re-rendered. Here's another example; a simple search bar.
const SearchBar = () => (
<div className="search-bar">
{/* Use a selector to only observe state.search */}
<State.Consumer selector={state => state.search}>
{(search, mutate) => (
<input
value={state}
onChange={event =>
mutate(draft => {
// Update draft.search (which will end up being state.search) via mutation
draft.search = event.currentTarget.value;
})
}
/>
)}
</State.Consumer>
</div>
);One issue with mutate being provided via a render callback is that you now have to either inline the functions calling it in render, or pass it as a prop to another component to use it in another lifecycle.
The State object returned from createState also provides a method called createMutator. Since it's also bound to the same state instance as the returned Provider and Consumer, you can use it to make state updates outside of the render callback.
const createMutator = State.createMutator;
// Statically define your mutation method. If this were a class component, you
// could define it as an instance property.
const setSearch = createMutator((draft, event) => {
const {value} = event.currentTarget;
draft.search = value;
});
const SearchBar = () => (
<div className="search-bar">
<State.Consumer selector={state => state.search}>
{(search) => (
<input value={state} onChange={setSearch} />
)}
</State.Consumer>
</div>
)