Monday, December 4, 2023
HomeJavaScriptNo New Options – React Weblog

No New Options – React Weblog

At this time, we’re publishing the primary Launch Candidate for React 17. It has been two and a half years since the earlier main launch of React, which is a very long time even by our requirements! On this weblog publish, we’ll describe the position of this main launch, what adjustments you may count on in it, and how one can do that launch.

No New Options

The React 17 launch is uncommon as a result of it doesn’t add any new developer-facing options. As a substitute, this launch is primarily targeted on making it simpler to improve React itself.

We’re actively engaged on the brand new React options, however they’re not part of this launch. The React 17 launch is a key a part of our technique to roll them out with out leaving anybody behind.

Particularly, React 17 is a “stepping stone” launch that makes it safer to embed a tree managed by one model of React inside a tree managed by a distinct model of React.

Gradual Upgrades

For the previous seven years, React upgrades have been “all-or-nothing”. Both you keep on an previous model, otherwise you improve your entire app to a brand new model. There was no in-between.

This has labored out to this point, however we’re operating into the bounds of the “all-or-nothing” improve technique. Some API adjustments, for instance, deprecating the legacy context API, are not possible to do in an automatic manner. Though most apps written at this time don’t ever use them, we nonetheless assist them in React. We have now to decide on between supporting them in React indefinitely or leaving some apps behind on an previous model of React. Each of those choices aren’t nice.

So we wished to supply another choice.

React 17 permits gradual React upgrades. Whenever you improve from React 15 to 16 (or, quickly, from React 16 to 17), you’ll normally improve your entire app without delay. This works nicely for a lot of apps. However it could actually get more and more difficult if the codebase was written various years in the past and isn’t actively maintained. And whereas it’s doable to make use of two variations of React on the web page, till React 17 this has been fragile and brought on issues with occasions.

We’re fixing a lot of these issues with React 17. Because of this when React 18 and the following future variations come out, you’ll now have extra choices. The primary choice might be to improve your entire app without delay, such as you might need achieved earlier than. However additionally, you will have an choice to improve your app piece by piece. For instance, you would possibly resolve emigrate most of your app to React 18, however hold some lazy-loaded dialog or a subroute on React 17.

This doesn’t imply you need to do gradual upgrades. For many apps, upgrading unexpectedly remains to be one of the best answer. Loading two variations of React — even when certainly one of them is loaded lazily on demand — remains to be not splendid. Nevertheless, for bigger apps that aren’t actively maintained, this feature could make sense to contemplate, and React 17 permits these apps to not get left behind.

To allow gradual updates, we’ve wanted to make some adjustments to the React occasion system. React 17 is a significant launch as a result of these adjustments are doubtlessly breaking. In follow, we’ve solely needed to change fewer than twenty elements out of 100,000+ so we count on that the majority apps can improve to React 17 with out an excessive amount of bother. Inform us if you happen to run into issues.

Demo of Gradual Upgrades

We’ve ready an instance repository demonstrating the right way to lazy-load an older model of React if obligatory. This demo makes use of Create React App, however it needs to be doable to comply with an analogous method with some other software. We welcome demos utilizing different tooling as pull requests.


We’ve postponed different adjustments till after React 17. The objective of this launch is to allow gradual upgrades. If upgrading to React 17 had been too troublesome, it could defeat its objective.

Modifications to Occasion Delegation

Technically, it has all the time been doable to nest apps developed with totally different variations of React. Nevertheless, it was moderately fragile due to how the React occasion system labored.

In React elements, you normally write occasion handlers inline:

<button onClick={handleClick}>

The vanilla DOM equal to this code is one thing like:

myButton.addEventListener('click on', handleClick);

Nevertheless, for many occasions, React doesn’t really connect them to the DOM nodes on which you declare them. As a substitute, React attaches one handler per occasion kind straight on the doc node. That is known as occasion delegation. Along with its efficiency advantages on giant utility timber, it additionally makes it simpler so as to add new options like replaying occasions.

React has been doing occasion delegation robotically since its first launch. When a DOM occasion fires on the doc, React figures out which part to name, after which the React occasion “bubbles” upwards by your elements. However behind the scenes, the native occasion has already bubbled as much as the doc degree, the place React installs its occasion handlers.

Nevertheless, this can be a downside for gradual upgrades.

In case you have a number of React variations on the web page, all of them register occasion handlers on the high. This breaks e.stopPropagation(): if a nested tree has stopped propagation of an occasion, the outer tree would nonetheless obtain it. This made it troublesome to nest totally different variations of React. This concern just isn’t hypothetical — for instance, the Atom editor bumped into this 4 years in the past.

For this reason we’re altering how React attaches occasions to the DOM beneath the hood.

In React 17, React will now not connect occasion handlers on the doc degree. As a substitute, it’s going to connect them to the basis DOM container into which your React tree is rendered:

const rootNode = doc.getElementById('root');
ReactDOM.render(<App />, rootNode);

In React 16 and earlier, React would do doc.addEventListener() for many occasions. React 17 will name rootNode.addEventListener() beneath the hood as an alternative.

A diagram showing how React 17 attaches events to the roots rather than to the document

Because of this transformation, it’s now safer to embed a React tree managed by one model inside a tree managed by a distinct React model. Notice that for this to work, each of the variations would must be 17 or larger, which is why upgrading to React 17 is essential. In a manner, React 17 is a “stepping stone” launch that makes subsequent gradual upgrades possible.

This transformation additionally makes it simpler to embed React into apps constructed with different applied sciences. For instance, if the outer “shell” of your app is written in jQuery, however the newer code inside it’s written with React, e.stopPropagation() contained in the React code would now forestall it from reaching the jQuery code — as you’ll count on. This additionally works within the different course. When you now not like React and need to rewrite your app — for instance, in jQuery — you can begin changing the outer shell from React to jQuery with out breaking the occasion propagation.

We’ve confirmed that quite a few issues reported over the years on our difficulty tracker associated to integrating React with non-React code have been fastened by the brand new conduct.


You may be questioning whether or not this breaks Portals outdoors of the basis container. The reply is that React additionally listens to occasions on portal containers, so this isn’t a problem.

Fixing Potential Points

As with every breaking change, it’s doubtless some code would must be adjusted. At Fb, we needed to modify about 10 modules in complete (out of many hundreds) to work with this transformation.

For instance, if you happen to add guide DOM listeners with doc.addEventListener(...), you would possibly count on them to catch all React occasions. In React 16 and earlier, even if you happen to name e.stopPropagation() in a React occasion handler, your customized doc listeners would nonetheless obtain them as a result of the native occasion is already on the doc degree. With React 17, the propagation would cease (as requested!), so your doc handlers wouldn’t hearth:

doc.addEventListener('click on', perform() {

You possibly can repair code like this by changing your listener to make use of the seize section. To do that, you may go { seize: true } because the third argument to doc.addEventListener:

doc.addEventListener('click on', perform() {
}, { seize: true });

Notice how this technique is extra resilient total — for instance, it’s going to most likely repair current bugs in your code that occur when e.stopPropagation() known as outdoors of a React occasion handler. In different phrases, occasion propagation in React 17 works nearer to the common DOM.

Different Breaking Modifications

We’ve saved the breaking adjustments in React 17 to the minimal. For instance, it doesn’t take away any of the strategies which have been deprecated within the earlier releases. Nevertheless, it does embrace a couple of different breaking adjustments which have been comparatively protected in our expertise. In complete, we’ve needed to modify fewer than 20 out of 100,000+ our elements due to them.

Aligning with Browsers

We’ve made a few smaller adjustments associated to the occasion system:

  • The onScroll occasion now not bubbles to forestall frequent confusion.
  • React onFocus and onBlur occasions have switched to utilizing the native focusin and focusout occasions beneath the hood, which extra carefully match React’s current conduct and generally present additional data.
  • Seize section occasions (e.g. onClickCapture) now use actual browser seize section listeners.

These adjustments align React nearer with the browser conduct and enhance interoperability.


Though React 17 switched from focus to focusin beneath the hood for the onFocus occasion, be aware that this has not affected the effervescent conduct. In React, onFocus occasion has all the time bubbled, and it continues to take action in React 17 as a result of typically it’s a extra helpful default. See this sandbox for the totally different checks you may add for various explicit use instances.

No Occasion Pooling

React 17 removes the “occasion pooling” optimization from React. It doesn’t enhance efficiency in trendy browsers and confuses even skilled React customers:

perform handleChange(e) {
  setData(information => ({
    textual content: e.goal.worth

It is because React reused the occasion objects between totally different occasions for efficiency in previous browsers, and set all occasion fields to null in between them. With React 16 and earlier, you need to name e.persist() to correctly use the occasion, or learn the property you want earlier.

In React 17, this code works as you’ll count on. The previous occasion pooling optimization has been absolutely eliminated, so you may learn the occasion fields everytime you want them.

This can be a conduct change, which is why we’re marking it as breaking, however in follow we haven’t seen it break something at Fb. (Perhaps it even fastened a couple of bugs!) Notice that e.persist() remains to be obtainable on the React occasion object, however now it doesn’t do something.

Impact Cleanup Timing

We’re making the timing of the useEffect cleanup perform extra constant.

useEffect(() => {
  return () => {      };});

Most results don’t have to delay display updates, so React runs them asynchronously quickly after the replace has been mirrored on the display. (For the uncommon instances the place you want an impact to dam paint, e.g. to measure and place a tooltip, choose useLayoutEffect.)

Nevertheless, when a part is unmounting, impact cleanup capabilities used to run synchronously (much like componentWillUnmount being synchronous in courses). We’ve discovered that this isn’t splendid for bigger apps as a result of it slows down giant display transitions (e.g. switching tabs).

In React 17, the impact cleanup perform all the time runs asynchronously — for instance, if the part is unmounting, the cleanup runs after the display has been up to date.

This mirrors how the results themselves run extra carefully. Within the uncommon instances the place you would possibly need to depend on the synchronous execution, you may change to useLayoutEffect as an alternative.


You may be questioning whether or not which means you’ll now be unable to repair warnings about setState on unmounted elements. Don’t fear — React particularly checks for this case, and does not hearth setState warnings within the quick hole between unmounting and the cleanup. So code cancelling requests or intervals can virtually all the time keep the identical.

Moreover, React 17 will all the time execute all impact cleanup capabilities (for all elements) earlier than it runs any new results. React 16 solely assured this ordering for results inside a part.

Potential Points

We’ve solely seen a few elements break with this transformation, though reusable libraries might have to check it extra completely. One instance of problematic code could seem like this:

useEffect(() => {
  return () => {

The issue is that someRef.present is mutable, so by the point the cleanup perform runs, it could have been set to null. The answer is to seize any mutable values inside the impact:

useEffect(() => {
  const occasion = someRef.present;
  return () => {

We don’t count on this to be a standard downside as a result of our eslint-plugin-react-hooks/exhaustive-deps lint rule (be sure to use it!) has all the time warned about this.

Constant Errors for Returning Undefined

In React 16 and earlier, returning undefined has all the time been an error:

perform Button() {

That is partly as a result of it’s simple to return undefined unintentionally:

perform Button() {
  <button />;

Beforehand, React solely did this for sophistication and performance elements, however didn’t verify the return values of forwardRef and memo elements. This was on account of a coding mistake.

In React 17, the conduct for forwardRef and memo elements is in line with common perform and sophistication elements. Returning undefined from them is an error.

let Button = forwardRef(() => {
  <button />;

let Button = memo(() => {
  <button />;

For the instances the place you need to render nothing deliberately, return null as an alternative.

Native Part Stacks

Whenever you throw an error within the browser, the browser provides you a stack hint with JavaScript perform names and their areas. Nevertheless, JavaScript stacks are sometimes not sufficient to diagnose an issue as a result of the React tree hierarchy might be simply as essential. You need to know not simply {that a} Button threw an error, however the place within the React tree that Button is.

To resolve this, React 16 began printing “part stacks” when you’ve gotten an error. Nonetheless, they was once inferior to the native JavaScript stacks. Particularly, they weren’t clickable within the console as a result of React didn’t know the place the perform was declared within the supply code. Moreover, they had been largely ineffective in manufacturing. Not like common minified JavaScript stacks which might be robotically restored to the unique perform names with a sourcemap, with React part stacks you had to decide on between manufacturing stacks and bundle dimension.

In React 17, the part stacks are generated utilizing a distinct mechanism that stitches them collectively from the common native JavaScript stacks. This allows you to get the absolutely symbolicated React part stack traces in a manufacturing surroundings.

The way in which React implements that is considerably unorthodox. At the moment, the browsers don’t present a method to get a perform’s stack body (supply file and site). So when React catches an error, it’s going to now reconstruct its part stack by throwing (and catching) a brief error from inside every of the elements above, when it’s doable. This provides a small efficiency penalty for crashes, however it solely occurs as soon as per part kind.

When you’re curious, you may learn extra particulars in this pull request, however for probably the most half this precise mechanism shouldn’t have an effect on your code. Out of your perspective, the brand new characteristic is that part stacks at the moment are clickable (as a result of they depend on the native browser stack frames), and that you could decode them in manufacturing such as you would with common JavaScript errors.

The half that constitutes a breaking change is that for this to work, React re-executes components of a few of the React capabilities and React class constructors above within the stack after an error is captured. Since render capabilities and sophistication constructors shouldn’t have uncomfortable side effects (which can also be essential for server rendering), this could not pose any sensible issues.

Eradicating Personal Exports

Lastly, the final notable breaking change is that we’ve eliminated some React internals that had been beforehand uncovered to different tasks. Particularly, React Native for Internet used to rely on some internals of the occasion system, however that dependency was fragile and used to interrupt.

In React 17, these personal exports have been eliminated. So far as we’re conscious, React Native for Internet was the one undertaking utilizing them, and so they have already accomplished a migration to a distinct method that doesn’t rely on these personal exports.

Because of this the older variations of React Native for Internet received’t be suitable with React 17, however the newer variations will work with it. In follow, this doesn’t change a lot as a result of React Native for Internet needed to launch new variations to adapt to inside React adjustments anyway.

Moreover, we’ve eliminated the ReactTestUtils.SimulateNative helper strategies. They’ve by no means been documented, didn’t do fairly what their names implied, and didn’t work with the adjustments we’ve made to the occasion system. If you would like a handy method to hearth native browser occasions in assessments, take a look at the React Testing Library as an alternative.

Set up

We encourage you to strive React 17.0 Launch Candidate quickly and elevate any points for the issues you would possibly encounter within the migration. Take into account that a launch candidate is extra more likely to comprise bugs than a secure launch, so don’t deploy it to manufacturing but.

To put in React 17 RC with npm, run:

npm set up react@17.0.0-rc.3 react-dom@17.0.0-rc.3

To put in React 17 RC with Yarn, run:

yarn add react@17.0.0-rc.3 react-dom@17.0.0-rc.3

We additionally present UMD builds of React by way of a CDN:

<script crossorigin src=""></script>
<script crossorigin src=""></script>

Consult with the documentation for detailed set up directions.



React DOM

  • Delegate occasions to roots as an alternative of doc. (@trueadm in #18195 and others)
  • Clear up all results earlier than operating any subsequent results. (@bvaughn in #17947)
  • Run useEffect cleanup capabilities asynchronously. (@bvaughn in #17925)
  • Use browser focusin and focusout for onFocus and onBlur. (@trueadm in #19186)
  • Make all Seize occasions use the browser seize section. (@trueadm in #19221)
  • Don’t emulate effervescent of the onScroll occasion. (@gaearon in #19464)
  • Throw if forwardRef or memo part returns undefined. (@gaearon in #19550)
  • Take away occasion pooling. (@trueadm in #18969)
  • Cease exposing internals that received’t be wanted by React Native Internet. (@necolas in #18483)
  • Connect all identified occasion listeners when the basis mounts. (@gaearon in #19659)
  • Disable console within the second render go of DEV mode double render. (@sebmarkbage in #18547)
  • Deprecate the undocumented and deceptive ReactTestUtils.SimulateNative API. (@gaearon in #13407)
  • Rename personal discipline names used within the internals. (@gaearon in #18377)
  • Don’t name Person Timing API in improvement. (@gaearon in #18417)
  • Disable console in the course of the repeated render in Strict Mode. (@sebmarkbage in #18547)
  • In Strict Mode, double-render elements with out Hooks too. (@eps1lon in #18430)
  • Permit calling ReactDOM.flushSync throughout lifecycle strategies (however warn). (@sebmarkbage in #18759)
  • Add the code property to the keyboard occasion objects. (@bl00mber in #18287)
  • Add the disableRemotePlayback property for video parts. (@tombrowndev in #18619)
  • Add the enterKeyHint property for enter parts. (@eps1lon in #18634)
  • Warn when no worth is offered to <Context.Supplier>. (@charlie1404 in #19054)
  • Warn when memo or forwardRef elements return undefined. (@bvaughn in #19550)
  • Enhance the error message for invalid updates. (@JoviDeCroock in #18316)
  • Exclude forwardRef and memo from stack frames. (@sebmarkbage in #18559)
  • Enhance the error message when switching between managed and uncontrolled inputs. (@vcarl in #17070)
  • Maintain onTouchStart, onTouchMove, and onWheel passive. (@gaearon in #19654)
  • Repair setState hanging in improvement inside a closed iframe. (@gaearon in #19220)
  • Repair rendering bailout for lazy elements with defaultProps. (@jddxf in #18539)
  • Repair a false optimistic warning when dangerouslySetInnerHTML is undefined. (@eps1lon in #18676)
  • Repair Take a look at Utils with non-standard require implementation. (@just-boris in #18632)
  • Repair onBeforeInput reporting an incorrect occasion.kind. (@eps1lon in #19561)
  • Repair occasion.relatedTarget reported as undefined in Firefox. (@claytercek in #19607)
  • Repair “unspecified error” in IE11. (@hemakshis in #19664)
  • Repair rendering right into a shadow root. (@Jack-Works in #15894)
  • Repair movementX/Y polyfill with seize occasions. (@gaearon in #19672)
  • Use delegation for onSubmit and onReset occasions. (@gaearon in #19333)
  • Enhance reminiscence utilization. (@trueadm in #18970)

React DOM Server

React Take a look at Renderer

Concurrent Mode (Experimental)



Please enter your comment!
Please enter your name here

Most Popular

Recent Comments