Today the Ember project is releasing version 3.1.0 of Ember.js, Ember Data, and Ember CLI.
This release kicks off the 3.2 beta cycle for all sub-projects. We encourage our
community (especially addon authors) to help test these beta builds and report
any bugs before they are published as a final release in six weeks' time. The
ember-try addon is a great way to
continuously test your projects against the latest Ember releases.
You can read more about our general release process here:
Ember.js
Ember.js is the core framework for building ambitious web applications.
Changes in Ember.js 3.1
Ember 3.1 is a minor release containing several new features and bug fixes. It includes a bump of Glimmer VM, Ember's rendering implementation, to version 0.30.5.
Named Arguments (1 of 4)
Named arguments are here! This allows you to reference component arguments as {{@name}}
when passed in as {{user-profile name="Zoey"}}
.
From RFC 276: Until now, the way to access named arguments passed in from the caller was to reference {{name}}
. The (first) problem with this approach is that the {{name}}
syntax is highly ambigious, as it could be referring to a local variable (block param), a helper or a named argument from the caller (which actually works by accessing auto-reflected {{this.name}}
) or a property on the component class (such as a computed property).
Since the {{foo}}
syntax still works on Component
(which is the only kind of components available today) via the auto-reflection mechanism, we are not really in a rush to migrate the community (and the guides, etc) to using the new syntax. In the meantime, this could be viewed as a tool to improve clarity in templates.
While we think writing {{@foo}}
would be a best practice for new code going forward, the community can migrate at its own pace one component at a time.
ES5 Getters for Computed Properties (2 of 4)
Ember's object system has long used set
and get
to access properties. These APIs came from the codebase's origins in SproutCore, and predated ES5's defineProperty
. In recent years, native JavaScript setter and getter implementations have become fast and mature.
Starting in Ember 3.1 (and described in RFC281) you are now able to read the value of a computed property using a native ES5 getter. For example, this component which uses computed properties:
import Component from '@ember/component';
import { computed } from '@ember/object';
export default Component.extend({
name: computed('firstName', 'lastName', function() {
return `${this.get('firstName')} ${this.get('lastName')}`;
}),
message: computed('name', function() {
return `Hello ${this.get('name')}!`;
});
});
Can be re-written using ES5 getters:
import Component from '@ember/component';
import { computed } from '@ember/object';
export default Component.extend({
name: computed('firstName', 'lastName', function() {
return `${this.firstName} ${this.lastName}`;
}),
message: computed('name', function() {
return `Hello ${this.name}!`;
})
});
Legacy get
features are not deprecated or removed in 3.1. In fact there are
several cases where you must still use get
:
- If you are calling
get
with a chained path. For example inthis.get('a.b.c')
ifb
isundefined
the return value isundefined
. Converting this
tothis.a.b.c
whenb
isundefined
would instead raise an exception. - If your object is using
unknownProperty
you must continue to useget
. Using an ES5 getter on an object withunknownProperty
will cause an assertion failure in development. - Ember Data returns promise proxy objects when you read an async relationship and from other API. Ember proxy objects, including promise proxies, still require that you call
get
to read values.
With these caveats in mind, how should you know if you can convert a get
call to a native getter? If you have code where get
is called on this
, you likely can convert it. If you have a get
on another object, anything.get('foo')
, you should exercise caution when converting to a native getter.
The community-provided es5-getter-ember-codemod is a great way to convert your existing codebase to ES5 getters. It follows the conservative guidelines and only converts this.get
. Note that it cannot make all possible conversions to the new API, nor can it ensure 100% of the conversions it makes are correct. If your app has poor test coverage or you lack any confidence in your ability to make regression checks, a manual and gradual conversion process may be more appropriate.
Thanks to Chris Garrett for pushing forward work on ES5 getters with support from Godfrey Chan, Robert Jackson, and Kris Selden). Thanks to Jonathan Jackson for his work on the codemod.
Introducing Optional Features (3 of 4)
Because major releases of Ember are not supposed to make breaking changes without prior deprecation, the project has been extremely conservative about changing behaviors that don't have a clear deprecation path. As a result, we've had several quirks of the framework linger into the 3.x series.
To give the project a path forward when a breaking change is mandatory, we've released the @ember/optional-features
addon. Today this addon is opt-in via installation as an NPM dependency. In a future release of Ember it will become part of the default application blueprint.
This addon does nothing by default, but provides a command-line interface to enable and disable breaking changes in Ember. Two optional features are being introduced in Ember 3.1.
To install ember-optional-features:
ember install @ember/optional-features
Thanks to Godfrey Chan and Robert Jackson for their work on the optional features system.
New Optional Feature: Application Template Wrapper
Ember applications have long created a wrapping div
around their rendered content: . With ember-optional-features, this functionality can now be turned off:
ember feature:disable application-template-wrapper
Disabling this feature will stop Ember from creating a div
around the application. This change may require alterations to your application's CSS, or to any other code that depends upon the presence of the div
.
Additionally, enabling this feature will prompt you to optionally run a codemod to add the application div
to the application.hbs
of your application.
Although disabling this feature will eventually be the default for Ember, leaving the feature enabled is not deprecated in this release. You can read more details about this optional feature and the motivations for introducing it in RFC #280.
New Optional Feature: Template-only Glimmer Components
Ember components implicitly create an element in the DOM where they are invoked, and the contents of their templates are then treated as "innerHTML" inside that DOM element. For example, this component template:
{{!-- app/templates/components/hello-world.hbs --}}
Hello World!
When invoked as:
{{!-- app/templates/index.hbs --}}
{{hello-world}}
Would render with an implicit div
:
Hello World!
The treatment of templates as "innerHTML" in Ember makes several parts of the framework's API harder to learn. For example, setting a class on an element in a template is straight forward, and any developer comfortable with HTML should be comfortable doing so. However adding a class to the implicit component div
is more difficult, requiring the developer to create a JavaScript file for the component and use the classNames
property.
To resolve this tension, Glimmer components shift templates to be treated as "outerHTML". There is no implicit div
. All the DOM elements created by the renderer are in a template.
The "Template-only Glimmer Component" feature provides a first practical step in this direction. You can enable this feature by running:
ember feature:enable template-only-glimmer-components
Once enabled, any component template file without a corresponding JavaScript file will behave like "outerHTML". For example the component file:
{{!-- app/templates/components/hello-world.hbs --}}
Hello World!
Without any corresponding JavaScript file, and invoked as:
{{!-- app/templates/index.hbs --}}
{{hello-world}}
Would render without an implicit div, as follows:
Hello World!
Enabling this feature may require changes to your application's CSS, or to any other code dependent upon the presence of div
s for JavaScript-free components. In practice, most applications and nearly all addons use the Ember CLI generators for new components, which include a JavaScript file. If your application has template-only components which rely on a backing EmberComponent
class, for example, if they have an injected-by-type service, note that they would also lose access to that backing class.
However, enabling this feature will prompt you to optionally run a codemod which creates backing classes for all template-only components, meaning both the implicit div
and backing class are retained.
Although enabling this feature will eventually be the default for Ember, leaving the feature disabled is not deprecated in this release. You can read more details about this optional feature and the motivations for introducing it in RFC #278.
Positional Params Bug Fix (4 of 4)
Ember introduced contextual components in Ember 2.3. Contextual components close over arguments and are intended to do so in a manner consistent with closures in JavaScript.
As the implementation of contextual components has been refined in the Glimmer VM, a notable discrepancy has been noticed in how they handle positional params. Given the following template:
{{#with (component 'x-foo' 1 2 3) as |comp|}}
{{component comp 4 5 6}}
{{/with}}
The params of 4, 5, 6
would override those of 1, 2, 3
. Normal closure implementations would instead have appended the arguments to result in a positional argument list of 1, 2, 3, 4, 5, 6
.
In Ember 3.1 we've corrected the implementation to act like a proper closure. In researching the impact of this breaking bug fix we found no known public addons or applications which would be impacted.
For more information about this change see
emberjs/ember.js#15287.
Deprecations in Ember 3.1
Deprecations are added to Ember.js when an API will be removed at a later date.
Each deprecation has an entry in the deprecation guide describing the migration
path to more stable API. Deprecated public APIs are not removed until a major
release of the framework.
Consider using the
ember-cli-deprecation-workflow
addon if you would like to upgrade your application without immediately addressing
deprecations.
Two new deprecations are introduced in Ember.js 3.1:
- Calling
array.get('@each')
is deprecated.@each
may only be used as dependency key. - The private APIs
propertyWillChange
andpropertyDidChange
will be removed after the first LTS of the 3.x cycle. You should remove any calls topropertyWillChange
and replace any calls topropertyDidChange
withnotifyPropertyChange
. This applies to both the Ember global version and the EmberObject method version.
For example, the following:
Ember.propertyWillChange(object, 'someProperty');
doStuff(object);
Ember.propertyDidChange(object, 'someProperty');
object.propertyWillChange('someProperty');
doStuff(object);
object.propertyDidChange('someProperty');
Should be changed to:
doStuff(object);
Ember.notifyPropertyChange(object, 'someProperty');
doStuff(object);
object.notifyPropertyChange('someProperty');
If you are an addon author and need to support both Ember applications greater than 3.1 and less than 3.1 you can use the polyfill ember-notify-property-change-polyfill.
For more details on changes in Ember.js 3.1, please review the
Ember.js 3.1.0 release page.
Upcoming Changes in Ember.js 3.2
New Features (1)
let template helper
Ember.js 3.2 introduces a new feature, the block let
template helper.
This helper enables you to introduce bindings in your templates, without having to make them properties of the respective controller or component.
For example, imagine you want to pass the same hash of options to two different different components, you can now do the following:
{{#let (hash theme="fairyfloss" indentation=4) as |options|}}
{{code-editor options=options}}
{{code-preview options=options}}
{{/let}}
You can read more about it in RFC #286 - Block let
template helper.
Deprecations (3)
Three new deprecations are introduced in Ember.js 3.2:
Ember.Logger
Use of Ember.Logger
is deprecated. You should replace any calls to Ember.Logger
with calls to console
. Read more about this deprecation on the deprecation page.
Router#route
The Router#route
private API has been renamed to Router#_route
to avoid collisions with user-defined properties or methods. Read more about this deprecation on the deprecation page.
defineProperty
This next one typically won't affect most apps, but it might affect some addons. You'll need to replace directly assigned computed properties, and use defineProperty
to define computed properties instead. Read more about this deprecation on the deprecation page.
For more details on the upcoming changes in Ember.js 3.2, please review the
Ember.js 3.2.0-beta.1 release page.
Ember Data
Ember Data is the official data persistence library for Ember.js applications.
Ember Data 3.1 contains bug fixes and build improvements for Ember Data.
Changes in Ember Data 3.1
New Features (1)
#5273 client-side-delete semantics unloadRecord
Deprecations (2)
Two new deprecations are introduced in Ember Data 3.1.
Ember.Map
was a private API provided by Ember (for quite some time). Unfortunately, Ember Data madeEmber.Map
part of its public API surface via documentation blocks.Ember.Map
is [scheduled for deprecation](scheduled for deprecation), after we make sure that Ember Data will continue working after this feature is deprecated and removed.Ember.Map
differs from nativeMap
in a few ways:Ember.Map
has customcopy
andisEmpty
methods which are not present in nativeMap
Ember.Map
adds a staticcreate
method (which instantiates itself withnew Ember.Map()
)Ember.Map
does not accept constructor argumentsEmber.Map
does not have:@@species
@@iterator
entries
values
This implementation adds a deprecated backwards compatibility for:copy
isEmpty
This is needed because Map
requires instantiation with new
, and by default Babel transpilation will do superConstructor.apply(this, arguments)
which throws an error with native Map
. The desired code (if we lived in an "only native class" world) would be:
export default class MapWithDeprecations extends Map {
constructor(options) {
super();
this.defaultValue = options.defaultValue;
}
get(key) {
let hasValue = this.has(key);
if (hasValue) {
return super.get(key);
} else {
let defaultValue = this.defaultValue(key);
this.set(key, defaultValue);
return defaultValue;
}
}
}
For more details on changes in Ember Data 3.1, please review the
Ember Data 3.1.0 release page.
Upcoming changes in Ember Data 3.2
Lazy Relationship Payloads (1 of 4)
Due to implementation details in the parsing of lazy relationships, polymorphic relationships were not supported, causing newly encountered polymorphic types to dereference previous payloads.
This issue is now addressed.
Ember Data Feature Flag Removal (2 of 4)
Ember Data 3.2 removes all current feature flags for Ember Data. These feature flags have gone stale and Ember Data is going to
attempt to go a different direction with some of the planned changes for 2018. Many of these feature flags have been around for a long time. If your app depends on enabling these feature flag to run, please reach out to the Ember Data team by opening a github issue on the Ember Data repo and the Ember Data team will try to assist with the transition.
Feature Flag ds-improved-ajax
(3 of 4)
During the Ember Data 3.2 beta cycle, the Ember Data team is planning on releasing an addon that will support the ds-improved-ajax
API.
Feature Flag ds-pushpayload-return
(4 of 4)
If you rely on the ds-pushpayload-return
feature flag, you can use the following pattern to manually serialize the API response and push the record into the store.
export function pushPayload(store, modelName, rawPayload) {
let ModelClass = store.modelFor(modelName);
let serializer = store.serializerFor(modelName);
let jsonApiPayload = serializer.normalizeResponse(store, ModelClass, rawPayload, null, 'query');
return store.push(jsonApiPayload);
}
// Before
this.get('store').pushPayload(modelName, rawPayload);
// After
import { pushPayload } from '/utils/push-payload';
pushPayload(this.get('store'), modelName, rawPayload);
Deprecations in Ember Data 3.2
There are no new deprecations planned for Ember Data 3.2.
For more details on the upcoming changes in Ember Data 3.2, please review the Ember Data 3.2.0-beta.1 release page.
Ember CLI
Ember CLI is the command line interface for managing and packaging Ember.js applications.
Upgrading Ember CLI
You may upgrade Ember CLI separately from Ember.js and Ember Data! You can do this either with yarn or npm.
To upgrade your projects using yarn
, run:
yarn upgrade ember-cli
To upgrade your projects using npm
, run:
npm install --save-dev ember-cli
After running the upgrade command, run ember init
inside of the project directory to apply the blueprint changes.
You can preview those changes for applications and addons.
Changes in Ember CLI 3.1
New Features (2)
ember test:all
Previously, npm test
would run all configured scenarios of ember-try
. This was confusing due to npm test
and ember test
having different behaviors, as well as npm test
doing different things depending on whether it was being run by an app or an addon. The fact that the command would also run several hard to cancel processes for the ember-try
scenarios also worsened the developer experience.
To address this, npm test
was changed to run ember test
, and a new npm test:all
was introduced with the old behavior of running ember-try
scenarios.
Yarn lock file detection
Ember CLI will now correctly detect if the project is part of a Yarn workspace root, and adequately use Yarn instead of npm.
Deprecations (0)
No new deprecations are introduced in Ember CLI 3.1.
For more details on the changes in Ember CLI 3.1 and detailed upgrade instructions, please review the Ember CLI 3.1.0 release page.
Upcoming Changes in Ember CLI 3.2
New Features (3)
Qunit Dom - In order to make DOM assertions more readable, the qunit-dom
dependency will be added by default to all apps and addons. Opt out by removing it from your package.json file. See https://github.com/simplabs/qunit-dom-codemod to ease migration (#7605).
This is, in our opinion, totally awesome. It means that this code:
assert.equal(this.element.querySelector('.title').textContent.trim(), 'Hello World!');
becomes this:
assert.dom('.title').hasText('Hello World!');
See what I mean? Totally awesome. <3
Experiments with more efficient transpilation - Until now, addons were responsible for compiling their own JS/HBS/CSS and returning AMD/CSS. Now they return the raw code, and the app uses its own processors (babel, htmlbars) to compile it. This is required to do proper tree-shaking and code-splitting. Delayed transpilation (#7501) and all-at-once addon optimization after compilation (#7650) have been added. Additionally, more comprehensive methods to detect if ember-cli is being run within CI or not have also been added (#7637) - see https://github.com/watson/ci-info/.
Module Unification (new file layout) Continues - You can now generate an addon using the Module Unification layout (#7490)! Use the command MODULE_UNIFICATION=true ember addon my-addon
to try it out (#7658). We also improved the logic to support addons that use Module Unification (#7660), added the blueprint for a dummy app to addons that use Module Unification (#7667), and updated the version of Ember used in Module Unification (#7678).
Deprecations (1)
ember-cli-babel 5
This release of Ember CLI deprecates ember-cli-babel
5.x. Babel 6 support has been out for a long time now and works quite well. Babel 5 support is deprecated and is expected to be dropped soon.
For more details on the changes in Ember CLI 3.2.0-beta.1 and detailed upgrade
instructions, please review the Ember CLI 3.2.0-beta.1 release page.
Thank you to @GavinJoyce, @Turbo87, @cibernox, @iezer, @kellyselden, @raytiley, @t-sauer, and @thetimothyp
for your incredible work on ember-cli!
Thank You!
As a community-driven open-source project with an ambitious scope, each of these releases serve 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.