Ember model proxying deprecation confusion - ember.js

I'm trying to work through deprecations messages in Ember 1.11 about object proxying in templates, and have run into confusion.
On an edit route, for example, I was loading a single record in the model hook. Then, using the setupController hoook, I was setting other collection variables to enable select menus. This provoked deprecation notices about proxying; I attempted to resolve those by setting all data in the model hook using an Ember.RSVP.hash as the model.
I now have a problem where, if I load the page directly, the model is correct, but if I visit from a link-to, the model is incorrect. I found information on similar problems that led me to try using setupController to set the model, based on this.get('model'), hoping that would grab the model from the route's model funtion, but this also did not work. The setupController action is being invoked, but it doesn't seem to be affecting the template for some reason, which is where I've had a roadblock.
show.hbs:
{{link-to model}}
edit.hbs –– data present if direct page load, empty if visiting from show link-to
{{model.foo.name}}
edit route:
renderTemplate: function (controller,model) {
this.render('foo/form');
},
model: function (project) {
return Ember.RSVP.hash({
foo: this.store.find("foo", foo.foo_id),
bar: this.store.find('bar')
});
}
also tried this for edit route:
renderTemplate: function (controller,model) {
this.render('foo/form');
},
model: function (project) {
return Ember.RSVP.hash({
foo: this.store.find("foo", foo.foo_id),
bar: this.store.find('bar')
});
},
setupController: function (controller, model) {
controller.set('model', this.get('model');
}

I've found one simple way around this, not sure if there is a better "Ember way" –– I see that because link-to is passing the model, it is assumed that there is no model hook required; passing model.id instead of model as link-to argument does cause the model hook to be invoked.
{{link-to model.id}}
instead of:
{{link-to model}}

Related

setupController no longer working for Ember route

I can not longer setup a route that has a hash of multiple models. What I use to do is this:
import Ember from 'ember';
export default Ember.Route.extend({
model() {
return Ember.RSVP.hash({
products: this.store.findAll('product'),
suppliers: this.store.findAll('supplier')
});
},
setupController(controller, model) {
controller.set('products', model.products);
controller.set('suppliers', model.suppliers);
}
});
When I do this I can't loop through each {{#each products}} but I still can with {{#each model.products}}
Object proxying is deprecated, so you should refer to your models with a model. prefix.
The correct way to do this would be to leave out your setupController and use the model as an object with .products and .suppliers keys on it.
You'll still have to {{#each model.products as |product|}} unless you want to alias them in your controller like this:
products: Ember.computed.alias("model.products")
I'd recommend sticking with model.products in your template though to alleviate any confusion as to where that particular data came from. Removing it would imply it's controller-related vs. model data fetched in the route's model hook.
EDIT: The above solution was done in Ember 1.13.5 and should also work in any 2.x version.

Ember Unbound & Belongsto

I have no problem doing {{unbound title}} or
{{#each file}}
{{unbound filename}}
{{/each}}
on a model.
BUT,
all belongsto object in ember is really problematic for me.
None of ways below work
{{unbound location.address}}
and
{{with location}}
{{unbound address}}
{{/with}}
both of these two result in empty output
At the time your model is being processed, any belongsTo relationships are not resolved yet. Since you're not binding, it can't retroactively update once that data is available either. I found this workaround yesterday, helping my solve my (similar) issues with belongsTo: https://github.com/emberjs/data/issues/1405
For existing records, you need to do something like the following in
your route:
// your-route beforeModel: function() { var self = this; return
Em.RSVP.hash({
firstBelongsTo: this.store.find('first-belongs-to'),
secondBelongsTo: this.store.find('second-belongs-to')
}).then(function (models) {
self.controllerFor('this-route').setProperties(models); });
And in your controller, be sure to declare the properties before
setting them as Ember tries to throw then into content when they don't
exist:
// your-controller App.MyController = Ember.Controller.extend({
firstBelongsTo: null, secondBelongsTo: null });
By returning a promise in the beforeModel hook, you are telling the
route to resolve the promise BEFORE loading the model, which also mean
before any rendering occurs. This gives your application time to load
the data up front before binding it to the select boxes.

Ember.js - How do I insert a controller?

So my understanding from the Ember docs is that the pattern for views/controllers/models is as follows:
[view] <- [controller] <- [model]
(with views consuming controllers consuming models)
In my previous experience using Ember, I'd set up a view to consume a model, like so:
{{#with blogpost}}
{{#view MyApp.BlogPostView contentBinding="this"}}
<h1>{{title}}</h1>
<p>{{content}}</p>
{{/view}}
{{/with}}
Now say I create a controller:
MyApp.BlogPostController = Ember.BlogPostController.extend()
Where do I initialize this controller?
Looking at the Ember docs, it seems like this happens automatically if the controller is associated with a route, but what if I just want an ad-hoc controller which ties together a view and a model? This could be for an arbitrary component on my page.
Am I responsible for instanciating the controller? Should I use some kind of controllerBinding attribute? Will it be instantiated automatically with my model, or with my view?
Any advice appreciated; I'm comfortable with the model/view pattern in Ember, but I'm having some difficulty working out where controllers fit in.
Looking at the Ember docs, it seems like this happens automatically if the controller is associated with a route
This is correct, a controller associated with a route will be automatically instantiated by ember when needed.
but what if I just want an ad-hoc controller which ties together a view and a model? This could be for an arbitrary component on my page. Am I responsible for instanciating the controller? Should I use some kind of controllerBinding attribute? Will it be instantiated automatically with my model, or with my view?
There are different way's to get your arbitrary controller instantiated automatically by ember without the needs of doing it yourself.
For the examples, let's assume you have a controller which is not associated with any routes called LonelyController,
App.LonelyController = Ember.ArrayController.extend({
content: ['foo', 'bar', 'baz']
});
Approach 1
Let's assume you have a route and you hook into setupController, if you try here to request you LonelyController with this.controllerFor('lonely'); this will make ember instantiate it for you:
App.IndexRoute = Ember.Route.extend({
setupController: function(controller, model) {
this.controllerFor('lonely').get('content');
// the above line will retrive successfully
// your `LonelyController`'s `content` property
}
});
Approach 2
Another possible way to get your LonelyController automatically instantiated by ember would be by defining a dependence with the needs API in another controller:
App.IndexController = Ember.ObjectController.extend({
needs: 'lonely',
someAction: function() {
this.get('controllers.lonely').get('content');
// the above line will retrive successfully
// your `LonelyController`'s `content` property
}
});
Using the needs API you could also doing something like this:
App.IndexController = Ember.ObjectController.extend({
needs: 'lonely',
lonelysContentBinding: 'controllers.lonely.content',
someAction: function() {
this.get('lonelysContent');
// the above line will retrive successfully
// your `LonelyController`'s `content` property
}
});
There are also some other combinations of the mentioned methods to get your LonelyController automatically instantiated, but I guess this should be more clear by now.
One last tip: to get a clue of what ember creates automatically under the hood you could also enable the generation logging to observe this in your console, which is very helpful, by doing:
var App = Ember.Application.create({
LOG_ACTIVE_GENERATION: true
});
Hope it helps.

Ember, Ember-data, and jquery-ui.dialog, "Oh my!"

The task:
Open a form in a lightbox to create a new "event"; the opened form should be bookmarkable.
The road blocks:
There are examples of opening a lightbox using {{action}} tags, but could not find one that opened in its own route.
There are many examples using older versions of ember.js.
There is not a lot of documentation related to ember-data and REST (I know, I know...it isn't "production ready").
The problem:
The fields in the form were not being tied to a backing model so "null" was being posted to my servlet (a Spring controller).
My very first iteration was not too far off from the final outcome (jsfiddle). The thing that finally made it works swapping this:
EP.EventsNewRoute = Ember.Route.extend({
...
setupController : function(controller, model) {
controller.set("model", model);
},
...
});
...for this:
EP.EventsNewRoute = Ember.Route.extend({
...
setupController : function(controller, model) {
this.controllerFor("events-new").set("model", model);
},
...
});
The question:
Why does the setupController function need to call controllerFor in order to properly set up the model?
And finally, since I struggled to find a fully-functional example, I wanted to make this accessible (and hopefully discover improvements).
Here's the fiddle: http://jsfiddle.net/6thJ4/1/
Here are a few snippets.
HTML:
<script type="text/x-handlebars">
<div>
<ul>
{{#linkTo "events.new" tagName="li"}}
Add Event
{{/linkTo}}
</ul>
</div>
{{outlet events-new}}
</script>
<script type="text/x-handlebars" data-template-name="events-new">
<form>
<div>
<label>Event Name:</label>
{{view Ember.TextField valueBinding="name"}}
</div>
<div>
<label>Host Name:</label>
{{view Ember.TextField valueBinding="hostName"}}
</div>
</form>
</script>
JavaScript:
...
EP.Router.map(function() {
this.resource("events", function() {
this.route("new");
});
});
EP.EventsNewRoute = Ember.Route.extend({
model : function() {
return EP.Event.createRecord();
},
setupController : function(controller, model) {
//controller.set("model", model); // Doesn't work? Why not?
this.controllerFor("events-new").set("model", model); // What does this do differently?
},
...
});
EP.EventsNewController = Ember.ObjectController.extend({
save : function() {
this.get("content.transaction").commit(); // "content.store" would commit _everything modified_, we only have one element changed, so only "content.transaction" is necessary.
}
});
EP.EventsNewView = Ember.View.extend({
...
});
EP.Event = DS.Model.extend({
name : DS.attr("string"),
hostName : DS.attr("string")
});
Resources:
http://emberjs.com/guides/routing/setting-up-a-controller/
http://emberjs.com/guides/getting-started/toggle-all-todos/ (trying to mimic what I learned, but morph the add-new to a new route)
Writing a LightboxView causes problems / Integrating DOM Manipulating jQuery-Plugins makes actions unusable (lightbox "example")
Dependable views in Ember (another lightbox "example" but doesn't have routes for the lightbox opening)
Why does the setupController function need to call controllerFor in order to properly set up the model?
Ember makes URLs a very integral part of its conventions. This means that the state of your application is represented by the route it is on. You've grokked most of this correctly. But there are couple of subtle nuances, that I will clarify below.
First consider an app with the following URLs,
/posts - shows a list of blog posts.
/posts/1 - shows a single blog post.
And say clicking on a post in the list at /posts takes you to /posts/1.
Given this scenario, there 2 ways a user will get to see the post at /posts/1.
By going to /posts and clicking on the 1st post.
By typing in /posts/1, via bookmarks etc.
In both these cases, the PostRoute for /posts/1 will need the model corresponding to Post id 1.
Consider the direct typing scenario first. To provide a way to lookup the id=1 post model, you would use,
model: function(params) {
return App.Post.find(params.post_id);
}
Your template for post will get the model and it can render using it's properties.
Now consider the second scenario. Clicking on post with id=1 takes you to /posts/1. To do this your template would use linkTo like this.
{{#linkTo 'post' post}} {{post.title}} {{/linkTo}}
Here you are passing in the post model to the linkTo helper. It then serializes the data for the post into a URL, ie:- '/posts/1'. When you click on this link Ember realizes that it needs to render the PostRoute but it already has the post model. So it skips the model hook and directly calls setupController.
The default setupController is setup to simply assign the model on the controller. It's implemented to do something like,
setupController: function(controller, model) {
controller.set('model', model);
}
If you do not need to set custom properties on your controller, you don't need to override it. Note: if you are augmenting it with additional properties you still need to call _super to ensure that the default setupController behaviour executes.
setupController: function(controller, model) {
this._super.apply(this, arguments);
controller.set('customProp', 'foo');
}
One final caveat, If you are using linkTo and the route does not have dynamic segments, then the model hook is still called. This exception makes sense if you consider that you were linking to the /posts route. Then the model hook has to fire else Ember has no data to display the route.
Which brings us to the crux of your question. Nearly there, I promise!
In your example you are using linkTo to get to the EventsNewRoute. Further your EventsNewRoute does not have dynamic segments so Ember does call the model hook. And controller.set("model", model); does work in so much as setting the model on the controller.
The issue is to do with your use of renderTemplate. When you use render or {{render}} helper inside a template, you are effectively getting a different controller to the one you are using. This controller is different from the one you set the model on, hence the bug.
A workaround is to pass the controller in the options, which is why renderTemplate gets this controller as an argument.
renderTemplate: function(controller) {
this.render("events-new", {
outlet : "events-new", controller: controller
});
}
Here's an updated jsfiddle.
Final Note: Unrelated to this question, you are getting the warning,
WARNING: The immediate parent route ('application') did not render into the main outlet and the default 'into' option ('events') may not be expected
For that you need to read this answer. Warning, it's another wall of text! :)

How to use multiple models with a single route in EmberJS / Ember Data?

From reading the docs, it looks like you have to (or should) assign a model to a route like so:
App.PostRoute = Ember.Route.extend({
model: function() {
return App.Post.find();
}
});
What if I need to use several objects in a certain route? i.e. Posts, Comments and Users? How do I tell the route to load those?
Last update forever: I can't keep updating this. So this is deprecated and will likely be this way. here's a better, and more up-to-date thread EmberJS: How to load multiple models on the same route?
Update: In my original answer I said to use embedded: true in the model definition. That's incorrect. In revision 12, Ember-Data expects foreign keys to be defined with a suffix (link) _id for single record or _ids for collection. Something similar to the following:
{
id: 1,
title: 'string',
body: 'string string string string...',
author_id: 1,
comment_ids: [1, 2, 3, 6],
tag_ids: [3,4]
}
I have updated the fiddle and will do so again if anything changes or if I find more issues with the code provided in this answer.
Answer with related models:
For the scenario you are describing, I would rely on associations between models (setting embedded: true) and only load the Post model in that route, considering I can define a DS.hasMany association for the Comment model and DS.belongsTo association for the User in both the Comment and Post models. Something like this:
App.User = DS.Model.extend({
firstName: DS.attr('string'),
lastName: DS.attr('string'),
email: DS.attr('string'),
posts: DS.hasMany('App.Post'),
comments: DS.hasMany('App.Comment')
});
App.Post = DS.Model.extend({
title: DS.attr('string'),
body: DS.attr('string'),
author: DS.belongsTo('App.User'),
comments: DS.hasMany('App.Comment')
});
App.Comment = DS.Model.extend({
body: DS.attr('string'),
post: DS.belongsTo('App.Post'),
author: DS.belongsTo('App.User')
});
This definition would produce something like the following:
With this definition, whenever I find a Post, I will have access to a collection of comments associated with that post, and the comment's author as well, and the user which is the author of the post, since they are all embedded. The route stays simple:
App.PostsPostRoute = Em.Route.extend({
model: function(params) {
return App.Post.find(params.post_id);
}
});
So in the PostRoute (or PostsPostRoute if you're using resource), my templates will have access to the controller's content, which is the Post model, so I can refer to the author, simply as author
<script type="text/x-handlebars" data-template-name="posts/post">
<h3>{{title}}</h3>
<div>by {{author.fullName}}</div><hr />
<div>
{{body}}
</div>
{{partial comments}}
</script>
<script type="text/x-handlebars" data-template-name="_comments">
<h5>Comments</h5>
{{#each content.comments}}
<hr />
<span>
{{this.body}}<br />
<small>by {{this.author.fullName}}</small>
</span>
{{/each}}
</script>
(see fiddle)
Answer with non-related models:
However, if your scenario is a little more complex than what you described, and/or have to use (or query) different models for a particular route, I would recommend to do it in Route#setupController. For example:
App.PostsPostRoute = Em.Route.extend({
model: function(params) {
return App.Post.find(params.post_id);
},
// in this sample, "model" is an instance of "Post"
// coming from the model hook above
setupController: function(controller, model) {
controller.set('content', model);
// the "user_id" parameter can come from a global variable for example
// or you can implement in another way. This is generally where you
// setup your controller properties and models, or even other models
// that can be used in your route's template
controller.set('user', App.User.find(window.user_id));
}
});
And now when I'm in the Post route, my templates will have access to the user property in the controller as it was set up in setupController hook:
<script type="text/x-handlebars" data-template-name="posts/post">
<h3>{{title}}</h3>
<div>by {{controller.user.fullName}}</div><hr />
<div>
{{body}}
</div>
{{partial comments}}
</script>
<script type="text/x-handlebars" data-template-name="_comments">
<h5>Comments</h5>
{{#each content.comments}}
<hr />
<span>
{{this.body}}<br />
<small>by {{this.author.fullName}}</small>
</span>
{{/each}}
</script>
(see fiddle)
Using Em.Object to encapsulate multiple models is a good way to get all data in model hook. But it can't ensure all data is prepared after view rendering.
Another choice is to use Em.RSVP.hash. It combines several promises together and return a new promise. The new promise if resolved after all the promises are resolved. And setupController is not called until the promise is resolved or rejected.
App.PostRoute = Em.Route.extend({
model: function(params) {
return Em.RSVP.hash({
post: // promise to get post
comments: // promise to get comments,
user: // promise to get user
});
},
setupController: function(controller, model) {
// You can use model.post to get post, etc
// Since the model is a plain object you can just use setProperties
controller.setProperties(model);
}
});
In this way you get all models before view rendering. And using Em.Object doesn't have this advantage.
Another advantage is you can combine promise and non-promise. Like this:
Em.RSVP.hash({
post: // non-promise object
user: // promise object
});
Check this to learn more about Em.RSVP: https://github.com/tildeio/rsvp.js
But don't use Em.Object or Em.RSVP solution if your route has dynamic segments
The main problem is link-to. If you change url by click link generated by link-to with models, the model is passed directly to that route.
In this case the model hook is not called and in setupController you get the model link-to give you.
An example is like this:
The route code:
App.Router.map(function() {
this.route('/post/:post_id');
});
App.PostRoute = Em.Route.extend({
model: function(params) {
return Em.RSVP.hash({
post: App.Post.find(params.post_id),
user: // use whatever to get user object
});
},
setupController: function(controller, model) {
// Guess what the model is in this case?
console.log(model);
}
});
And link-to code, the post is a model:
{{#link-to "post" post}}Some post{{/link-to}}
Things become interesting here. When you use url /post/1 to visit the page, the model hook is called, and setupController gets the plain object when promise resolved.
But if you visit the page by click link-to link, it passes post model to PostRoute and the route will ignore model hook. In this case setupController will get the post model, of course you can not get user.
So make sure you don't use them in routes with dynamic segments.
For a while I was using Em.RSVP.hash, however the problem I ran into was that I didn't want my view to wait until all models were loaded before rendering. However, I found a great (but relatively unknown) solution thanks to the folks at Novelys that involves making use of the Ember.PromiseProxyMixin:
Let's say you have a view that has three distinct visual sections. Each of these sections should be backed by its own model. The model backing the "splash" content at the top of the view is small and will load quickly, so you can load that one normally:
Create a route main-page.js:
import Ember from 'ember';
export default Ember.Route.extend({
model: function() {
return this.store.find('main-stuff');
}
});
Then you can create a corresponding Handlebars template main-page.hbs:
<h1>My awesome page!</h1>
<ul>
{{#each thing in model}}
<li>{{thing.name}} is really cool.</li>
{{/each}}
</ul>
<section>
<h1>Reasons I Love Cheese</h1>
</section>
<section>
<h1>Reasons I Hate Cheese</h1>
</section>
So let's say in your template you want to have separate sections about your love/hate relationship with cheese, each (for some reason) backed by its own model. You have many records in each model with extensive details relating to each reason, however you'd like the content on top to render quickly. This is where the {{render}} helper comes in. You can update your template as so:
<h1>My awesome page!</h1>
<ul>
{{#each thing in model}}
<li>{{thing.name}} is really cool.</li>
{{/each}}
</ul>
<section>
<h1>Reasons I Love Cheese</h1>
{{render 'love-cheese'}}
</section>
<section>
<h1>Reasons I Hate Cheese</h1>
{{render 'hate-cheese'}}
</section>
Now you'll need to create controllers and templates for each. Since they're effectively identical for this example, I'll just use one.
Create a controller called love-cheese.js:
import Ember from 'ember';
export default Ember.ObjectController.extend(Ember.PromiseProxyMixin, {
init: function() {
this._super();
var promise = this.store.find('love-cheese');
if (promise) {
return this.set('promise', promise);
}
}
});
You'll notice that we are using the PromiseProxyMixin here, which makes the controller promise-aware. When the controller is initialized, we indicate that the promise should be loading the love-cheese model via Ember Data. You'll need to set this property on the controller's promise property.
Now, create a template called love-cheese.hbs:
{{#if isPending}}
<p>Loading...</p>
{{else}}
{{#each item in promise._result }}
<p>{{item.reason}}</p>
{{/each}}
{{/if}}
In your template, you'll be able to render different content depending on the state of promise. When your page initially loads, your "Reasons I Love Cheese" section will display Loading.... When the promise is loaded, it will render all the reasons associated for each record of your model.
Each section will load independently and not block the main content from rendering immediately.
This is a simplistic example, but I hope everyone else finds it as useful as I did.
If you're looking to do something similar for many rows of content, you may find the Novelys example above even more relevant. If not, the above should work fine for you.
This might not be best practice and a naïve approach, but it shows conceptually how you would go about having on multiple models available on one central route:
App.PostRoute = Ember.Route.extend({
model: function() {
var multimodel = Ember.Object.create(
{
posts: App.Post.find(),
comments: App.Comments.find(),
whatever: App.WhatEver.find()
});
return multiModel;
},
setupController: function(controller, model) {
// now you have here model.posts, model.comments, etc.
// as promises, so you can do stuff like
controller.set('contentA', model.posts);
controller.set('contentB', model.comments);
// or ...
this.controllerFor('whatEver').set('content', model.whatever);
}
});
hope it helps
Thanks to all the other excellent answers, I created a mixin that combines the best solutions here into a simple and reusable interface. It executes an Ember.RSVP.hash in afterModel for the models you specify, then injects the properties into the controller in setupController. It does not interfere with the standard model hook, so you would still define that as normal.
Example use:
App.PostRoute = Ember.Route.extend(App.AdditionalRouteModelsMixin, {
// define your model hook normally
model: function(params) {
return this.store.find('post', params.post_id);
},
// now define your other models as a hash of property names to inject onto the controller
additionalModels: function() {
return {
users: this.store.find('user'),
comments: this.store.find('comment')
}
}
});
Here is the mixin:
App.AdditionalRouteModelsMixin = Ember.Mixin.create({
// the main hook: override to return a hash of models to set on the controller
additionalModels: function(model, transition, queryParams) {},
// returns a promise that will resolve once all additional models have resolved
initializeAdditionalModels: function(model, transition, queryParams) {
var models, promise;
models = this.additionalModels(model, transition, queryParams);
if (models) {
promise = Ember.RSVP.hash(models);
this.set('_additionalModelsPromise', promise);
return promise;
}
},
// copies the resolved properties onto the controller
setupControllerAdditionalModels: function(controller) {
var modelsPromise;
modelsPromise = this.get('_additionalModelsPromise');
if (modelsPromise) {
modelsPromise.then(function(hash) {
controller.setProperties(hash);
});
}
},
// hook to resolve the additional models -- blocks until resolved
afterModel: function(model, transition, queryParams) {
return this.initializeAdditionalModels(model, transition, queryParams);
},
// hook to copy the models onto the controller
setupController: function(controller, model) {
this._super(controller, model);
this.setupControllerAdditionalModels(controller);
}
});
https://stackoverflow.com/a/16466427/2637573 is fine for related models. However, with recent version of Ember CLI and Ember Data, there is a simpler approach for unrelated models:
import Ember from 'ember';
import DS from 'ember-data';
export default Ember.Route.extend({
setupController: function(controller, model) {
this._super(controller,model);
var model2 = DS.PromiseArray.create({
promise: this.store.find('model2')
});
model2.then(function() {
controller.set('model2', model2)
});
}
});
If you only want to retrieve an object's property for model2, use DS.PromiseObject instead of DS.PromiseArray:
import Ember from 'ember';
import DS from 'ember-data';
export default Ember.Route.extend({
setupController: function(controller, model) {
this._super(controller,model);
var model2 = DS.PromiseObject.create({
promise: this.store.find('model2')
});
model2.then(function() {
controller.set('model2', model2.get('value'))
});
}
});
Adding to MilkyWayJoe's answer, thanks btw:
this.store.find('post',1)
returns
{
id: 1,
title: 'string',
body: 'string string string string...',
author_id: 1,
comment_ids: [1, 2, 3, 6],
tag_ids: [3,4]
};
author would be
{
id: 1,
firstName: 'Joe',
lastName: 'Way',
email: 'MilkyWayJoe#example.com',
points: 6181,
post_ids: [1,2,3,...,n],
comment_ids: [1,2,3,...,n],
}
comments
{
id:1,
author_id:1,
body:'some words and stuff...',
post_id:1,
}
...
I believe the link backs are important so that the full relationship is established. Hope that helps someone.
You could use the beforeModel or afterModel hooks as these are always called, even if model is not called because you're using dynamic segments.
As per the asynchronous routing docs:
The model hook covers many use cases for pause-on-promise transitions, but sometimes you'll need the help of the related hooks beforeModel and afterModel. The most common reason for this is that if you're transitioning into a route with a dynamic URL segment via {{link-to}} or transitionTo (as opposed to a transition caused by a URL change), the model for the route you're transitioning into will have already been specified (e.g. {{#link-to 'article' article}} or this.transitionTo('article', article)), in which case the model hook won't get called. In these cases, you'll need to make use of either the beforeModel or afterModel hook to house any logic while the router is still gathering all of the route's models to perform a transition.
So say you have a themes property on your SiteController, you could have something like this:
themes: null,
afterModel: function(site, transition) {
return this.store.find('themes').then(function(result) {
this.set('themes', result.content);
}.bind(this));
},
setupController: function(controller, model) {
controller.set('model', model);
controller.set('themes', this.get('themes'));
}