What's Coming in Ember in 2014

– By Tom Dale

Every few months, the Ember core team likes to get together to discuss issues face-to-face and set our priorities for the following quarter.

This time, we all converged on Portland, Oregon for the weekend to figure out where to focus our energy in 2014.

Build Tools & Modules

We spent the majority of our time discussing a unified plan that combines ES6 modules with fast, robust, full-featured build tools for Ember developers.

One thing I love about the Ember community is that it values quality over raw implementation speed. For us, it's more important to solve problems than to solve scenarios, because scenario solving leads to fragmented solutions that don't scale or compose as your app grows.

So far, we've held off on endorsing any particular build tool because, while we've investigated all of them thoroughly, the currently-available options all have fatal flaws that make them unpleasant to use as your app gets larger.

Many people have also been asking us about modules. When we started Ember, there were no popular module formats. Since then, both CommonJS modules on the server and AMD modules in the browser have exploded in popularity. While both are great options, their ecosystems are disconnected and the two communities frequently engage in emacs/vi-style holy wars.

Fortunately, TC39, the committee in charge of maintaining JavaScript, recognized the need for a unified module system. ES6 modules offer the best of both AMD and CommonJS modules while being integrated directly into the language.

Just as the Ember community was one of the first to fully embrace JavaScript promises, we have spent the last few months working on making ES6 modules a reality and bringing them to production.

Two notable projects have come out of this work:

  • ES6 Module Transpiler, which can transpile JavaScript written in the ES6 module syntax down into browser globals, AMD or CommonJS.
  • Ember App Kit, a foundation for building Ember apps that uses ES6 modules (via the transpiler) instead of globals.

It's important to note that we have been planning for module support for some time. In fact, one of the last features that was blocking our initial 1.0 release was to have the entire framework go through an object called the resolver.

The resolver is the part of our dependency injection system that is responsible for determining naming conventions. For example, imagine a user visits your Ember application at /posts. By default, Ember will look for a template called posts, ask the App.PostsRoute for a model, and hook it all up to the App.PostsController. But these globals are not hardcoded into the framework—instead, there's a default resolver that encapsulates all of these naming conventions.

To write an Ember app that uses modules, swap out the default resolver for one that requires named modules instead of looking for globals.

In order to make it really nice, however, we believe we need command-line tools to smooth the process of creating, developing, and deploying apps.

Our focus for the first quarter of 2014 will be to develop first-class build tools that leverage existing tools where they fit our needs, and write new ones where the current options fall down. We'll make modules a first-class citizen, and soon every developer who starts using Ember will also be using ES6 modules. We're excited about a world where browser and server JavaScript packages live together in harmony.

To give you an idea of what we're thinking, here are some example CLI commands and their effects. Note that this is all under heavy development and is subject to change.

Initializing an App

ember new my-app

This initializes a new Ember.js project and generates stub directories, an app, a router, etc.

Starting the Development Server

ember server

Running Tests

ember test

Runs the QUnit tests using Karma (PhantomJS plus available browsers), although we are planning migrate to testem as we've found its output easier to work with.

You can also host the QUnit test harness for manual testing by running:

ember test:server

Adding Packages

The Ember CLI supports bower natively. Just use the commands you're used to:

bower install moment

The lingua franca module format in Ember is ES6 modules. We will do the work to ensure that bower modules are usable by Ember apps automatically, by shimming AMD, CommonJS and browser global packages:

  1. Provide a shim.json and shim directory for AMD shims
  2. Name anonymous module based on filesystem location (including versions)
  3. Concatenate modules into app.js (with source map)
  4. Inherit source maps from bower (if provided)

Even if a library was shipped as AMD, you would be able to use it as though it was an ES6 module:

import md from "markdown";

Every time a file changes, the final concatenated file loaded by the browser (app.js, app.css, etc.) is locked and any HTTP requests for it block until the new version is compiled.

Enhanced Inspector Features

If you're developing your app using the Ember CLI tools, the Chrome and Firefox inspectors get even better. The development server will open a socket that the inspector can connect to that will provide additional environmental information to the browser extension.

There are lots of exciting things you can do once your browser and development tools are talking to each other, but some of the ideas we've tossed around are:

  • Seeing a list of installed bower packages
  • Seeing the results of test runs
  • Linting and auditing your codebase
  • Configuring package.json using visual editor

Filesystem Layout

Here is a rough example of the kind of layout you might see when initializing a new Ember app via the CLI:

  underscore.js    // bower installed
  ember-histogram/ // incubator for packages
modules/           // non-MVC stuff

"Pod" Directory Structure

Right now, many Ember projects have adopted a Rails-style directory layout where everything is grouped by type:


We discussed moving this to a layout where related features are grouped together in "pods" of functionality:


This proposed directory structure is still up for lots of discussion. We've been looking at many different real world apps to see if this makes managing the source code for Ember apps any easier.

Slimming Down Ember.js (Project Svelte)

We believe there are several areas in the Ember codebase that have the potential to be slimmed down substantially. We'd like to pick a target filesize and have a sprint to get the total built framework size under that target.

We haven't picked a target size yet, but we tossed around the 50k figure. We'll have more information about this once we've made a list of good candidates for slimming action.

Distributing Components

There are a number of open issues around distributing reusable, third-party components and libraries of components. Typically we defer to the Web Components spec, but there are areas where the spec does not currently provide any guidance. As we deal with them, we are keen to provide feedback to the Web Components spec authors with how we have chosen to solve the problems.

Other limitations are due to certain features not being available in older browsers, and we are working on polyfilling and working around those issues so that you can enjoy the power of Web Components today in browsers all the way back to IE8.


Right now, Ember components share a global namespace. If I have a component called area-graph and you have a component called area-graph and I want to use your component in my app, we will have a collision.

Soon, components in packages will be addressable by a fully-qualified path.

Here's a notional area graph component:

{{area-graph@d3 xValue=responseTimes}}

If you find yourself typing the fully-qualified path often, you will be able to alias the helper in the lexical scope of the Handlebars template:

{{import "area-graph" from "d3"}}

We may also add a syntax for making all helpers in a package available:

{{import "d3"}}

Template Versioning/Compilation

Handlebars compiles templates down into an intermediate AST. This AST can change between Handlebars versions. Additionally, Handlebars syntax may be extended, or change in version 2.0.

Given that components are distributed on bower, should they be shipped with their templates pre-compiled, or should we leave it up to ember-cli to compile them on-demand?

Currently, we are leaning towards shipping templates in their raw source form, but we need to do some more investigation.

Scoped CSS

<style scoped> is not available today in many browsers. Distributed components will get a uuid associated with them, and all of their CSS rules will be wrapped in a selector that limits them to elements with that uuid as a class name.

Item Controller

Instead of doing {{#each itemController="postItem"}}, you can just define App.PostsItemController (or the app/controllers/posts-item module) and it will be wired up automatically.


HTMLBars is on track and will hopefully be available behind a feature flag sometime in the new year. Yehuda spent some time going over the implementation with Kris Selden and Alex Matchneer. They are going to be taking over the effort of integrating the HTMLBars compiler into Ember proper.

If you're not familiar with HTMLBars, it is a compiler that knows about both HTML5 and Handlebars syntax. This is important for two reasons.

First, it will allow us to deprecate the {{bind-attr}} Handlebars helper. Instead of this:

<img {{bind-attr src=imageUrl}}>

You will be able to say:

<img src="{{imageUrl}}">

Second, it will also allow us to eliminate the ugly metamorph.js script tags that we currently use to track bound values in the DOM.

Here's what the DOM looks like before HTMLBars:

And after:

Eliminate jQuery Dependency

Once we move over to HTMLBars, the interaction between Ember.js and the DOM gets smaller. We can probably treat jQuery as an optional dependency and only use it if it is available via globals or as an AMD module. We just want to make sure that we remove components/views using jQuery if its available so that jQuery UI widgets that have stored data in the DOM via jQuery.data() get cleaned up appropriately.


We're still working on having animation support be built in to the framework, but we don't have any specific API proposals to share at this time. As always, we prioritize getting the API correct the first time rather than having to re-do it in a later release.

IE8 Support

Despite the imminent End of Life status of Windows XP, we will continue supporting Internet Explorer 8. We know many Ember.js users still need to target enterprise and education customers, who will be on IE8 for some time.


EmberConf 2014: It's happening. We'll have more details available very soon, but I will say that you should keep March as free as you can.


Just kidding.

How You Can Help

As we've said before, Ember is a completely community-driven project. I'd like to thank all of the core team members who pay out-of-pocket for travel and accommodation so they can be at these meetings.

It has been extremely rewarding to watch the Ember community grow, and also to see it assume a position of leadership in the JavaScript community.

We need your help to keep pushing the boundaries of what's possible in the browser. If you'd like to get started helping out, please join us in #emberjs-dev on Freenode, where our helpful community can point you in the right direction. (If you just need help with developing your Ember app, please keep those questions in #emberjs.)

If there's anything we've forgotten to mention here, please feel free to let us know what you'd like to see in the comments below!

Ember Data

We've got lots of exciting stuff coming in Ember Data. Expect an update on our roadmap there in a separate blog post soon.