Skip to content

Back to React Recipes

The `useEffect` Hook

The useEffect hook lets your function component run side effects.

What are side effects (or, for short, effects)? They are "things that happen" inside the component, which may affect other components or alter the DOM. They can't be done during the render phase. Things like:

This hook serves a purpose similar to some lifecycle methods inside class components:

While in a class component a particular effect was spread among many lifecycle methods, useHook lets you group the feature in a single function that gets executed after each render. (Optionally, you can skip an effect on some renders).

The signature of the useEffect hook:

useEffect(effect [, dependencies]);


An effect may look like this:

const Component = props => {
const updateDocTitle = () => {
document.title = `Current component: ${props.title}`;

return <div>My component</div>;

This effect is a simple function that does something. And it will do it after each render. But what's the point of setting the same title over and over again? The dependencies array lets us control how often the effect is called. If we rewrite it as:

const updateDocTitle = () => {
document.title = `Current component: ${props.title}`;

useEffect(updateDocTitle, [props.title]);

...this particular effect will only be executed when props.title changes. (The comparison is made as with You can have more than one dependency in the array, and the effect will be executed if any item in the array has changed.

Additionally, passing an empty array [] as the dependencies will make the effect run only on the initial render, much like you'd have in componentDidMount for a class component.

Note that in general you'd write effects as anonymous functions:

useEffect(() => {
document.title = `Current component: ${props.title}`;
}, [props.title]);

but I had extracted it to a named function to make a point: it's a different function for each render cycle, and when referencing variables from the component's scope in the effect, it will capture their values at that particular moment in time. This is important when you specify arrays of dependencies, because useEffect will ignore a particular updateDocTitle function until a dependency changes. When a dependency does change, the hook will use the freshest instance of updateDocTitle.

Effects with cleanup

When you update document.title, it's something you don't have to clean up after. But other things, such as subscribing to an event emitter or adding a DOM listener, needs to be undone — actions that need reactions. Otherwise we'd be piling up the same event listener over and over again.

To cover this, you can return a cleanup function from your effect:

const mouseMove() {
let listener = e => { console.log(e) };
document.addEventListener('mousemove', listener);

// cleanup function
return () => {
document.removeEventListener('mousemove', listener);


📖 For lack of a better name, I call mouseMove a symmetrical function (TODO link to javascript-patterns repo), or a function that encapsulate both an action and its opposite.

After each render, React will call a fresh instance of mouseMove, adding the event listener but, crucially — it will call the previous instance's cleanup function beforehand, thus removing the previous listener.

Conditional execution of useEffect

We know that one rule of Hooks is you need to call all of them on each render, so no if statements or loops. So to implement an effect hook that gets called only in certain situations, we can make it depend on... something. For example, if you want an effect to run whenever the component becomes active, and stop the moment the compoent goes idle, you can:

import { useState, useEffect } from 'react';

const Component = (props) => {
let [ active, setActive ] = useState(false);

let mouseMove = () => {
document.addEventListener('mousemove', ...);
return () => {
document.removeEventListener('mousemove', ...);
useEffect(mouseMove, [ active ]);

So we can manipulate the state variable active by calls to setActive(true) and setActive(false) in order to control when an effect takes place.

Bonus: What if we want to have the effect not run while active is false, but run after each render while active is true? We can write the dependencies as [active ? {} : null]. That's because {} !== {} but null === null. (Technically, it uses which is slightly different than simple equality, but this particular example stands nonetheless).

Open questions

See also

Futher reading