イベントとエフェクトを切り離す
イベントハンドラは同じインタラクションを再度実行した場合のみ再実行されます。イベントハンドラとは異なり、エフェクトは、プロパティや state 変数のような読み取った値が、前回のレンダー時の値と異なる場合に再同期を行います。また、ある値には反応して再実行するが、他の値には反応しないエフェクトなど、両方の動作をミックスさせたい場合もあります。このページでは、その方法を説明します。
このページで学ぶこと
- イベントハンドラとエフェクトの選択方法
- エフェクトがリアクティブで、イベントハンドラがリアクティブでない理由
- エフェクトのコードの一部をリアクティブにしない場合の対処法
- エフェクトイベントとは何か、そしてエフェクトイベントからエフェクトを抽出する方法
- エフェクトイベントを使用してエフェクトから最新の props と state を読み取る方法
イベントハンドラとエフェクトのどちらを選ぶか
まず、イベントハンドラとエフェクトの違いについておさらいしましょう。
チャットルームのコンポーネントを実装している場合を想像してください。要件は次のようなものです:
- コンポーネントは選択されたチャットルームに自動的に接続する
- 「Send」ボタンをクリックすると、チャットにメッセージが送信される
あなたはそのためのコードはすでに実装されているが、それをどこに置くか迷っているとしましょう。イベントハンドラを使うべきか、エフェクトを使うべきか。この質問に答える必要があるたびに、なぜそのコードが実行される必要があるのかを考えてみてください。
特定のインタラクションに反応して実行されるイベントハンドラ
ユーザの立場からすると、メッセージの送信は、特定の「送信」ボタンがクリックされたから起こるはずです。それ以外の時間や理由でメッセージを送信すると、ユーザはむしろ怒るでしょう。そのため、メッセージの送信はイベントハンドラで行う必要があります。イベントハンドラを使えば、特定のインタラクションを処理することができます:
function ChatRoom({ roomId }) {
const [message, setMessage] = useState('');
// ...
function handleSendClick() {
sendMessage(message);
}
// ...
return (
<>
<input value={message} onChange={e => setMessage(e.target.value)} />
<button onClick={handleSendClick}>Send</button>;
</>
);
}
イベントハンドラを使えば、ユーザがボタンを押したときだけ sendMessage(message)
が実行されるようにすることができます。
同期が必要なときに実行されるエフェクト
また、コンポーネントをチャットルームに接続しておく必要があることを思い出してください。そのコードはどこに記述されるのでしょうか?
このコードを実行する理由は、何か特定のインタラクションではありません。ユーザがなぜ、どのようにチャットルームの画面に移動したかは問題ではありません。ユーザがチャットルームの画面を見て、対話できるようになった今、このコンポーネントは、選択されたチャットサーバに接続されたままである必要があります。チャットルーム・コンポーネントがアプリの初期画面であり、ユーザが何のインタラクションも行っていない場合でも、接続する必要があります。これがエフェクトである理由です:
function ChatRoom({ roomId }) {
// ...
useEffect(() => {
const connection = createConnection(serverUrl, roomId);
connection.connect();
return () => {
connection.disconnect();
};
}, [roomId]);
// ...
}
このコードを使用すると、ユーザが行った特定のインタラクションに関係なく、現在選択されているチャットサーバへの接続が常にアクティブであることを確認することができます。ユーザがアプリを開いただけであろうと、別の部屋を選んだだけであろうと、別の画面に移動して戻ってきただけであろうと、このエフェクトはコンポーネントが現在選択されている部屋と同期していることを保証し、必要なときはいつでも再接続するようにします。
import { useState, useEffect } from 'react'; import { createConnection, sendMessage } from './chat.js'; const serverUrl = 'https://localhost:1234'; function ChatRoom({ roomId }) { const [message, setMessage] = useState(''); useEffect(() => { const connection = createConnection(serverUrl, roomId); connection.connect(); return () => connection.disconnect(); }, [roomId]); function handleSendClick() { sendMessage(message); } return ( <> <h1>Welcome to the {roomId} room!</h1> <input value={message} onChange={e => setMessage(e.target.value)} /> <button onClick={handleSendClick}>Send</button> </> ); } export default function App() { const [roomId, setRoomId] = useState('general'); const [show, setShow] = useState(false); return ( <> <label> Choose the chat room:{' '} <select value={roomId} onChange={e => setRoomId(e.target.value)} > <option value="general">general</option> <option value="travel">travel</option> <option value="music">music</option> </select> </label> <button onClick={() => setShow(!show)}> {show ? 'Close chat' : 'Open chat'} </button> {show && <hr />} {show && <ChatRoom roomId={roomId} />} </> ); }
リアクティブな値とリアクティブなロジック
直感的に言うと、イベントハンドラは、例えばボタンをクリックするなど、常に「手動」でトリガされます。一方、エフェクトは「自動」であり、同期を保つために必要な回数だけ実行され、再実行されます。
もっと正確な考え方があります。
コンポーネントの body 内で宣言された props 、state 、変数をリアクティブ値と呼びます。この例では、serverUrl
はリアクティブ値ではありませんが、roomId
と message
はリアクティブ値です。これらは、レンダーのデータフローに参加しています:
const serverUrl = 'https://localhost:1234';
function ChatRoom({ roomId }) {
const [message, setMessage] = useState('');
// ...
}
これらのようなリアクティブな値は、再レンダーによって変更される可能性があります。例えば、ユーザが message
を編集したり、ドロップダウンで別の roomId
を選択することがあります。イベントハンドラとエフェクトは、それぞれ異なる方法で変化に対応します:
- **イベントハンドラ内のロジックはリアクティブではない。**ユーザが同じ操作(クリックなど)を再度行わない限り、再度実行されることはありません。イベントハンドラは、その変更に「反応」することなく、リアクティブ値を読み取ることができます。
- **エフェクト内のロジックはリアクティブである。**エフェクトがリアクティブ値を読み取る場合、依存関係としてそれを指定する必要があります。そして、再レンダーによってその値が変更された場合、React は新しい値でエフェクトのロジックを再実行します。
この違いを説明するために、先ほどの例をもう一度見てみましょう。
イベントハンドラ内のロジックはリアクティブではない
このコードの行を見てみてください。このロジックはリアクティブであるべきでしょうか、そうではないでしょうか?
// ...
sendMessage(message);
// ...
ユーザから見れば、**message
の変更は、メッセージを送りたいということではありません。**あくまでも、ユーザが入力していることを意味します。つまり、メッセージを送るロジックはリアクティブであってはならないのです。リアクティブ値が変わったからと言って、再び実行されるべきではないのです。だから、イベントハンドラの中にあるのです:
function handleSendClick() {
sendMessage(message);
}
イベントハンドラはリアクティブではないので、sendMessage(message)
はユーザが送信ボタンをクリックしたときのみ実行されます。
エフェクト内のロジックはリアクティブである
では、この行に戻りましょう:
// ...
const connection = createConnection(serverUrl, roomId);
connection.connect();
// ...
ユーザからすると、**roomId
の変更は、別の部屋に接続したいことを意味します。**つまり、ルームに接続するためのロジックはリアクティブであるべきなのです。これらのコードは、リアクティブ値に「ついていける」ようにし、その値が異なる場合は再度実行するようにします。だから、エフェクトの中にあるのです:
useEffect(() => {
const connection = createConnection(serverUrl, roomId);
connection.connect();
return () => {
connection.disconnect()
};
}, [roomId]);
エフェクトはリアクティブなので、createConnection(serverUrl, roomId)
と connection.connect()
は、roomId
の異なる値ごとに実行されます。エフェクトは、現在選択されているルームに同期したチャット接続を維持します。
Extracting non-reactive logic out of Effects
Things get more tricky when you want to mix reactive logic with non-reactive logic.
For example, imagine that you want to show a notification when the user connects to the chat. You read the current theme (dark or light) from the props so that you can show the notification in the correct color:
function ChatRoom({ roomId, theme }) {
useEffect(() => {
const connection = createConnection(serverUrl, roomId);
connection.on('connected', () => {
showNotification('Connected!', theme);
});
connection.connect();
// ...
However, theme
is a reactive value (it can change as a result of re-rendering), and every reactive value read by an Effect must be declared as its dependency. Now you have to specify theme
as a dependency of your Effect:
function ChatRoom({ roomId, theme }) {
useEffect(() => {
const connection = createConnection(serverUrl, roomId);
connection.on('connected', () => {
showNotification('Connected!', theme);
});
connection.connect();
return () => {
connection.disconnect()
};
}, [roomId, theme]); // ✅ All dependencies declared
// ...
Play with this example and see if you can spot the problem with this user experience:
import { useState, useEffect } from 'react'; import { createConnection, sendMessage } from './chat.js'; import { showNotification } from './notifications.js'; const serverUrl = 'https://localhost:1234'; function ChatRoom({ roomId, theme }) { useEffect(() => { const connection = createConnection(serverUrl, roomId); connection.on('connected', () => { showNotification('Connected!', theme); }); connection.connect(); return () => connection.disconnect(); }, [roomId, theme]); return <h1>Welcome to the {roomId} room!</h1> } export default function App() { const [roomId, setRoomId] = useState('general'); const [isDark, setIsDark] = useState(false); return ( <> <label> Choose the chat room:{' '} <select value={roomId} onChange={e => setRoomId(e.target.value)} > <option value="general">general</option> <option value="travel">travel</option> <option value="music">music</option> </select> </label> <label> <input type="checkbox" checked={isDark} onChange={e => setIsDark(e.target.checked)} /> Use dark theme </label> <hr /> <ChatRoom roomId={roomId} theme={isDark ? 'dark' : 'light'} /> </> ); }
When the roomId
changes, the chat re-connects as you would expect. But since theme
is also a dependency, the chat also re-connects every time you switch between the dark and the light theme. That’s not great!
In other words, you don’t want this line to be reactive, even though it is inside an Effect (which is reactive):
// ...
showNotification('Connected!', theme);
// ...
You need a way to separate this non-reactive logic from the reactive Effect around it.
Declaring an Effect Event
Use a special Hook called useEffectEvent
to extract this non-reactive logic out of your Effect:
import { useEffect, useEffectEvent } from 'react';
function ChatRoom({ roomId, theme }) {
const onConnected = useEffectEvent(() => {
showNotification('Connected!', theme);
});
// ...
Here, onConnected
is called an Effect Event. It’s a part of your Effect logic, but it behaves a lot more like an event handler. The logic inside it is not reactive, and it always “sees” the latest values of your props and state.
Now you can call the onConnected
Effect Event from inside your Effect:
function ChatRoom({ roomId, theme }) {
const onConnected = useEffectEvent(() => {
showNotification('Connected!', theme);
});
useEffect(() => {
const connection = createConnection(serverUrl, roomId);
connection.on('connected', () => {
onConnected();
});
connection.connect();
return () => connection.disconnect();
}, [roomId]); // ✅ All dependencies declared
// ...
This solves the problem. Note that you had to remove onConnected
from the list of your Effect’s dependencies. Effect Events are not reactive and must be omitted from dependencies.
Verify that the new behavior works as you would expect:
import { useState, useEffect } from 'react'; import { experimental_useEffectEvent as useEffectEvent } from 'react'; import { createConnection, sendMessage } from './chat.js'; import { showNotification } from './notifications.js'; const serverUrl = 'https://localhost:1234'; function ChatRoom({ roomId, theme }) { const onConnected = useEffectEvent(() => { showNotification('Connected!', theme); }); useEffect(() => { const connection = createConnection(serverUrl, roomId); connection.on('connected', () => { onConnected(); }); connection.connect(); return () => connection.disconnect(); }, [roomId]); return <h1>Welcome to the {roomId} room!</h1> } export default function App() { const [roomId, setRoomId] = useState('general'); const [isDark, setIsDark] = useState(false); return ( <> <label> Choose the chat room:{' '} <select value={roomId} onChange={e => setRoomId(e.target.value)} > <option value="general">general</option> <option value="travel">travel</option> <option value="music">music</option> </select> </label> <label> <input type="checkbox" checked={isDark} onChange={e => setIsDark(e.target.checked)} /> Use dark theme </label> <hr /> <ChatRoom roomId={roomId} theme={isDark ? 'dark' : 'light'} /> </> ); }
You can think of Effect Events as being very similar to event handlers. The main difference is that event handlers run in response to a user interactions, whereas Effect Events are triggered by you from Effects. Effect Events let you “break the chain” between the reactivity of Effects and code that should not be reactive.
Reading latest props and state with Effect Events
Effect Events let you fix many patterns where you might be tempted to suppress the dependency linter.
For example, say you have an Effect to log the page visits:
function Page() {
useEffect(() => {
logVisit();
}, []);
// ...
}
Later, you add multiple routes to your site. Now your Page
component receives a url
prop with the current path. You want to pass the url
as a part of your logVisit
call, but the dependency linter complains:
function Page({ url }) {
useEffect(() => {
logVisit(url);
}, []); // 🔴 React Hook useEffect has a missing dependency: 'url'
// ...
}
Think about what you want the code to do. You want to log a separate visit for different URLs since each URL represents a different page. In other words, this logVisit
call should be reactive with respect to the url
. This is why, in this case, it makes sense to follow the dependency linter, and add url
as a dependency:
function Page({ url }) {
useEffect(() => {
logVisit(url);
}, [url]); // ✅ All dependencies declared
// ...
}
Now let’s say you want to include the number of items in the shopping cart together with every page visit:
function Page({ url }) {
const { items } = useContext(ShoppingCartContext);
const numberOfItems = items.length;
useEffect(() => {
logVisit(url, numberOfItems);
}, [url]); // 🔴 React Hook useEffect has a missing dependency: 'numberOfItems'
// ...
}
You used numberOfItems
inside the Effect, so the linter asks you to add it as a dependency. However, you don’t want the logVisit
call to be reactive with respect to numberOfItems
. If the user puts something into the shopping cart, and the numberOfItems
changes, this does not mean that the user visited the page again. In other words, visiting the page is, in some sense, an “event”. It happens at a precise moment in time.
Split the code in two parts:
function Page({ url }) {
const { items } = useContext(ShoppingCartContext);
const numberOfItems = items.length;
const onVisit = useEffectEvent(visitedUrl => {
logVisit(visitedUrl, numberOfItems);
});
useEffect(() => {
onVisit(url);
}, [url]); // ✅ All dependencies declared
// ...
}
Here, onVisit
is an Effect Event. The code inside it isn’t reactive. This is why you can use numberOfItems
(or any other reactive value!) without worrying that it will cause the surrounding code to re-execute on changes.
On the other hand, the Effect itself remains reactive. Code inside the Effect uses the url
prop, so the Effect will re-run after every re-render with a different url
. This, in turn, will call the onVisit
Effect Event.
As a result, you will call logVisit
for every change to the url
, and always read the latest numberOfItems
. However, if numberOfItems
changes on its own, this will not cause any of the code to re-run.
さらに深く知る
In the existing codebases, you may sometimes see the lint rule suppressed like this:
function Page({ url }) {
const { items } = useContext(ShoppingCartContext);
const numberOfItems = items.length;
useEffect(() => {
logVisit(url, numberOfItems);
// 🔴 Avoid suppressing the linter like this:
// eslint-disable-next-line react-hooks/exhaustive-deps
}, [url]);
// ...
}
After useEffectEvent
becomes a stable part of React, we recommend never suppressing the linter.
The first downside of suppressing the rule is that React will no longer warn you when your Effect needs to “react” to a new reactive dependency you’ve introduced to your code. In the earlier example, you added url
to the dependencies because React reminded you to do it. You will no longer get such reminders for any future edits to that Effect if you disable the linter. This leads to bugs.
Here is an example of a confusing bug caused by suppressing the linter. In this example, the handleMove
function is supposed to read the current canMove
state variable value in order to decide whether the dot should follow the cursor. However, canMove
is always true
inside handleMove
.
Can you see why?
import { useState, useEffect } from 'react'; export default function App() { const [position, setPosition] = useState({ x: 0, y: 0 }); const [canMove, setCanMove] = useState(true); function handleMove(e) { if (canMove) { setPosition({ x: e.clientX, y: e.clientY }); } } useEffect(() => { window.addEventListener('pointermove', handleMove); return () => window.removeEventListener('pointermove', handleMove); // eslint-disable-next-line react-hooks/exhaustive-deps }, []); return ( <> <label> <input type="checkbox" checked={canMove} onChange={e => setCanMove(e.target.checked)} /> The dot is allowed to move </label> <hr /> <div style={{ position: 'absolute', backgroundColor: 'pink', borderRadius: '50%', opacity: 0.6, transform: `translate(${position.x}px, ${position.y}px)`, pointerEvents: 'none', left: -20, top: -20, width: 40, height: 40, }} /> </> ); }
The problem with this code is in suppressing the dependency linter. If you remove the suppression, you’ll see that this Effect should depend on the handleMove
function. This makes sense: handleMove
is declared inside the component body, which makes it a reactive value. Every reactive value must be specified as a dependency, or it can potentially get stale over time!
The author of the original code has “lied” to React by saying that the Effect does not depend ([]
) on any reactive values. This is why React did not re-synchronize the Effect after canMove
has changed (and handleMove
with it). Because React did not re-synchronize the Effect, the handleMove
attached as a listener is the handleMove
function created during the initial render. During the initial render, canMove
was true
, which is why handleMove
from the initial render will forever see that value.
If you never suppress the linter, you will never see problems with stale values.
With useEffectEvent
, there is no need to “lie” to the linter, and the code works as you would expect:
import { useState, useEffect } from 'react'; import { experimental_useEffectEvent as useEffectEvent } from 'react'; export default function App() { const [position, setPosition] = useState({ x: 0, y: 0 }); const [canMove, setCanMove] = useState(true); const onMove = useEffectEvent(e => { if (canMove) { setPosition({ x: e.clientX, y: e.clientY }); } }); useEffect(() => { window.addEventListener('pointermove', onMove); return () => window.removeEventListener('pointermove', onMove); }, []); return ( <> <label> <input type="checkbox" checked={canMove} onChange={e => setCanMove(e.target.checked)} /> The dot is allowed to move </label> <hr /> <div style={{ position: 'absolute', backgroundColor: 'pink', borderRadius: '50%', opacity: 0.6, transform: `translate(${position.x}px, ${position.y}px)`, pointerEvents: 'none', left: -20, top: -20, width: 40, height: 40, }} /> </> ); }
This doesn’t mean that useEffectEvent
is always the correct solution. You should only apply it to the lines of code that you don’t want to be reactive. In the above sandbox, you didn’t want the Effect’s code to be reactive with regards to canMove
. That’s why it made sense to extract an Effect Event.
Read Removing Effect Dependencies for other correct alternatives to suppressing the linter.
Limitations of Effect Events
Effect Events are very limited in how you can use them:
- Only call them from inside Effects.
- Never pass them to other components or Hooks.
For example, don’t declare and pass an Effect Event like this:
function Timer() {
const [count, setCount] = useState(0);
const onTick = useEffectEvent(() => {
setCount(count + 1);
});
useTimer(onTick, 1000); // 🔴 Avoid: Passing Effect Events
return <h1>{count}</h1>
}
function useTimer(callback, delay) {
useEffect(() => {
const id = setInterval(() => {
callback();
}, delay);
return () => {
clearInterval(id);
};
}, [delay, callback]); // Need to specify "callback" in dependencies
}
Instead, always declare Effect Events directly next to the Effects that use them:
function Timer() {
const [count, setCount] = useState(0);
useTimer(() => {
setCount(count + 1);
}, 1000);
return <h1>{count}</h1>
}
function useTimer(callback, delay) {
const onTick = useEffectEvent(() => {
callback();
});
useEffect(() => {
const id = setInterval(() => {
onTick(); // ✅ Good: Only called locally inside an Effect
}, delay);
return () => {
clearInterval(id);
};
}, [delay]); // No need to specify "onTick" (an Effect Event) as a dependency
}
Effect Events are non-reactive “pieces” of your Effect code. They should be next to the Effect using them.
まとめ
- Event handlers run in response to specific interactions.
- Effects run whenever synchronization is needed.
- Logic inside event handlers is not reactive.
- Logic inside Effects is reactive.
- You can move non-reactive logic from Effects into Effect Events.
- Only call Effect Events from inside Effects.
- Don’t pass Effect Events to other components or Hooks.
チャレンジ 1/4: Fix a variable that doesn’t update
This Timer
component keeps a count
state variable which increases every second. The value by which it’s increasing is stored in the increment
state variable. You can control the increment
variable with the plus and minus buttons.
However, no matter how many times you click the plus button, the counter is still incremented by one every second. What’s wrong with this code? Why is increment
always equal to 1
inside the Effect’s code? Find the mistake and fix it.
import { useState, useEffect } from 'react'; export default function Timer() { const [count, setCount] = useState(0); const [increment, setIncrement] = useState(1); useEffect(() => { const id = setInterval(() => { setCount(c => c + increment); }, 1000); return () => { clearInterval(id); }; // eslint-disable-next-line react-hooks/exhaustive-deps }, []); return ( <> <h1> Counter: {count} <button onClick={() => setCount(0)}>Reset</button> </h1> <hr /> <p> Every second, increment by: <button disabled={increment === 0} onClick={() => { setIncrement(i => i - 1); }}>–</button> <b>{increment}</b> <button onClick={() => { setIncrement(i => i + 1); }}>+</button> </p> </> ); }