The Ember project is excited to announce the release of Ember v7.3. This is a standard minor release as part of the Ember Release Train process.
This release brings a new way to create reactive state that doesn't need a class, makes a serious dent in the size of the JavaScript bundle for apps that are on the modern build system, and fixes a handful of long-standing bugs in the router and elsewhere! 🎉😃
Ember.js 7.3
Ember.js 7.3 introduces one new feature per RFC #1071: tracked can now be used outside of classes, and, arguably more importantly, both forms of tracked let you configure equality so that setting a value to what it already was no longer triggers a re-render. The release also includes some internal restructuring that lets bundlers drop a lot more unused code from ember-source, and ships eight bugfixes. There are no new deprecations.
tracked() outside of classes
Since Ember Octane, the way you create reactive state in Ember has been to put a @tracked property on a class:
import Component from '@glimmer/component';
import { tracked } from '@glimmer/tracking';
export default class Counter extends Component {
@tracked count = 0;
increment = () => this.count++;
}
This works great and is still what we recommend for the vast majority of app code, but it does mean that if you want a single reactive value you first need a class to put it on. That gets in the way in a few places: helpers and modifiers that are written as plain functions, tests that want a bit of state to poke at, and demos where every extra line of boilerplate is a distraction from the thing you are actually trying to show.
Ember 7.3 implements RFC #1071, which does two things to the existing tracked import. First, when you call it as a function with an initial value it returns a standalone reactive value usable outside of classes:
import { tracked } from '@glimmer/tracking';
const count = tracked(0);
const increment = () => count.value++;
<template>
Count is: {{count.value}}
<button {{on "click" increment}}>add one</button>
</template>
Reading count.value in a template (or in a getter that a template uses) entangles with the render similarly to reading a @tracked property would, and writing to it causes a re-render. Alongside .value there are four function short-hands:
count.get(); // same as reading count.value
count.set(2); // same as assigning count.value
count.update((n) => n + 1); // write based on the current value, without consuming it
count.freeze(); // prevent any further writes
One nice pattern that this unlocks is keeping mutable state private to a class while exposing a read-only view of it:
import { tracked } from '@glimmer/tracking';
export class Session {
#user = tracked(null);
get user() {
return this.#user.value;
}
async login() {
this.#user.value = await fetchCurrentUser();
}
}
If any of this looks familiar it's because the idea has been floating around the ecosystem for a while. It was prototyped as Cell in Starbeam and has been available to Ember developers as cell from ember-resources. Now it's built in, with no extra import. Rather than @tracked being "magic, we can utilize tracked() as a storytelling tool to demystify how @tracked works.
Configurable equality
Until now, setting a @tracked property always notified consumers, even when you set it to the exact value it already held, so this.count = this.count re-rendered everything that read count. That was a historical choice and there was no way to change it. Both forms of tracked now accept an options object with an equals function that decides whether a write counts as a change.
The non-decorator form defaults to Object.is, so count.value = count.value does not re-render, and you can pass { equals: () => false } to get the old always-notify behaviour. The @tracked decorator keeps its always-notify default for backwards compatibility, and you can now opt a property into equality-based notification:
class Counter {
@tracked({ equals: (a, b) => a === b }) count = 0;
// this no longer causes a re-render
noop = () => (this.count = this.count);
}
You can read more, including a couple of edge cases around passing plain objects as the initial value, in the API docs for tracked.
Introduced in emberjs/ember.js PR #21471
Smaller bundles for Vite apps
In the Ember 7.2 release blog we talked about setting type: "module" on ember-source and said that it wasn't enough on its own to shrink your bundle. This release is where that starts to pay off.
ember-source now declares "sideEffects": false in its package.json. That is a hint to bundlers like Vite and Rollup that importing one of Ember's internal modules never has side effects on any other module, which means the bundler is free to drop any module that your app never actually uses. Along with that, some of Ember's internals have been restructured so that a small app no longer pulls in the old rendering pipeline and some classic-era pieces.
The effect is significant. The compressed JavaScript for the hello-world app in Ember's own smoke tests went from around 64.5 kB before this work started to about 37 kB with these changes, or roughly 42% smaller. If you are curious about the details, the numbers are tracked in PR #21456.
This only affects apps that consume the ESM sources of ember-source directly, which is every app on the Embroider and Vite build system that has been the default since Ember 6.8. If you are still on the classic ember-cli build you won't see any change, which is one more reason to look at the Vite codemod if you haven't yet.
Introduced in emberjs/ember.js PR #21456 and PR #21462
Bug Fixes
Ember.js 7.3 introduces 8 bugfixes:
- #21203 Fix
@model becomes undefined or changes to the wrong route's model during Glimmer component willDestroy
- #21591 Destroy dynamic modifiers that were set after the initial render, fixing a memory leak
- #21409 Fix query params trigger model refresh unnecessarily
- #21410 Fix query param redirects during active transitions
- #21521 Treat nullish
<LinkTo> @query as an empty query object
- #21524 Allow
CoreObject#init to be called with no arguments
- #21406 Add a helpful assertion when
{{component}} is given an unsupported argument
- #21407 Improve
{{debugger}} message for template-only components
A couple of these deserve a special mention. The first one fixes a bug that was reported in 2020 and, from the git history, has probably existed since @model was introduced in Ember 3.14. If a component in a route template read @model in its willDestroy hook while you were transitioning to a different route, it could see undefined or, worse, the other route's model. The fix adds a check on the controller's identity so the outlet cannot be redirected mid-teardown, and the new smoke test covers transitions to sibling, parent, cousin, and unrelated routes.
The second one is a memory leak that has been with us since Ember 3.25. If a dynamic modifier like {{this.mod}} started out as undefined and was set to a real modifier after the first render (or was swapped for a different modifier later), its destructor never ran when the element went away, so anything the modifier had set up, such as the floating-ui observers in ember-primitives, leaked. The fix registers the updating opcode with its block so teardown reaches it. The PR is tagged for backport to an LTS.
The query param fixes are part of a larger effort to improve the router's test coverage, and each of them closes an issue that people have been hitting for years: parent routes no longer re-run their model hook when you transition to a child route with unchanged query params, and redirecting from beforeModel back to the same route with different query params no longer loses those params or crashes on a direct visit. And if you have ever had a <LinkTo> blow up because you passed @query={{this.maybeParams}} and it happened to be null, that now works as expected.
Finally, a small security hardening that is not in the list above: the guard that stops set() from @ember/object walking through __proto__ and constructor in a path now also blocks prototype, closing a prototype pollution edge case. See PR #21451.
Documentation
The API docs for {{each}} and {{each-in}} now document that they support native Set and Map respectively, which they have done since the iterable refactor but never said so. See PR #21523. The Ember.Templates.helpers docs also link to @ember/helper correctly again (PR #21573).
Ember CLI 7.3
Ember CLI 7.3 is a maintenance release for ember-cli itself: no new features, no deprecations, and no bugfixes. The only changes there are the routine dependency updates that happen as part of the release train: ember-cli itself picked up small updates to things like testem and morgan, and the default app blueprint now generates apps on ember-source 7.3 with the current @embroider/macros, @glint/template, and eslint-plugin-warp-drive versions. None of those updates crossed a major version boundary, so there is nothing to do when you upgrade. As with 7.2, newly generated apps still get WarpDrive 5.8.
Tests run through testem directly
The @ember/app-blueprint that ember new uses did get one change worth knowing about. Since Ember 6.8, pnpm test in a new app has built the app with Vite and then handed the built output to ember test --path dist. In 7.3 that second step calls testem directly:
"test": "vite build --mode development && testem ci --port 0"
with a cwd: 'dist' line added to testem.cjs so testem serves the built app. Nothing changes about how your tests are written or which browser runs them. ember test was a thin wrapper around testem for Vite apps, and now it calls testem directly, removing a layer that could make it unclear which tool owned the flags you were passing. If you already have an app you do not need to change anything, but if you want the same setup you can copy the script and the one config line.
Introduced in ember-cli/ember-app-blueprint PR #307
Thank You!
A final note: this post was drafted with the help of artificial intelligence, and a human reviewed it in its entirety.
As a community-driven open-source project with an ambitious scope, each of these releases serves as a reminder that the Ember project would not have been possible without your continued support. We are extremely grateful to our contributors for their efforts.
Log in
