I've inherited a codebase that follows the format of: a router sets a controller, the controller fetches the collection/model needed, then the controller set/passes the view the collection/model.
The current View I'm working on loads a collection, and now I need to build in a feature where I fetch a single model after the view has rendered, based on an id clicked (note the model is from a different collection).
I only want to load the model when/if they click a button. So my question is, can I setup the model/fetch in the View, or should I be doing that in the controller? Is there a backbone best practice when adopting a controller/view setup like this?
I primarily ask because it seems easier for me to add this new feature right in the View. But is that a bad practice? I thought so, so I started down the path of triggering an event in the View for the controller to the fetch the model, and then somehow pass that back to the View (but I'm not sure really how to even do that)...it seems like a lot of unnecessary hoop jumping?
Its OK to fetch collection via views. As 'plain' backbone does not Controller, View in charge of it responsibilities.
But imho fetch collections via controller is better practice, its easier to scale and support and test.
The only difficulty is to set communication between Controller and View context event. One of the approach is trigger Message Bus event on context event like
events: {
'click .some': function() {
this.trigger('someHappend', { some: data })
}
}
and listen to this event in Controller
this.on(someViewInstance, 'someHappend', function() {
// do something in controller
})
If you already inherited code with structure you described you'd better follow it. Also you might be interested in MarionetteJS as significant improvement. Also highly recommend you to checkout BackboneRails, screencasts not free but very usefull, especially in large scale app maintain
Related
I would like to implement a page like this
I have the following folder structure:
-root/
--root.controller.js
--root.module.js
--root.routes.js
--root.html
--filters/
----filters.controller.js
----filters.module.js
----filters.routes.js
----filters.html
--tableData/
----tableData.controller.js
----tableData.module.js
----tableData.routes.js
----tableData.html
--graph/
----graph.controller.js
----graph.module.js
----graph.routes.js
----graph.html
I would like to know the best approach in order to manage call from each controller in the best way.
I mean, from filters, I can change selection on dropdown or update other fields values, and then I would update data on tabledata. And if I change selection on tabledata I would update graph.
Root url is '/root/{itemId}' and I cannot add some other value on querystring.
How can I manage internal variables and methods?
thanks a lot
From my understanding, it's better to go for events $emit , $broadcast & on to achieve the same.
As I can see, you are having independent controllers with no relationship with most of the the other controllers.
Few suggestions to implement it:
use $rootScope.$emit and $rootScope.on if they controllers have no relationship. Make sure you are removing it manually though (can be ignored if the app is not too heavy but should be kept in mind). This link will be surely helpful . Eg:
If the filter is changed, an event filter-changed will be triggered. The graph.controller and table.controller will be informed and will render the UI accordingly.
Create a messaging service from where the controllers will subscribe and unsubscribe the events. That will keep numbers of events in check. You'll be aware of how many events have actually been created in the application.
I'm reading up in choosing the correct client-side framework to segment/modularize my frontend code in Widgets.
Basically what I have/want is:
a complex website with multiple pagetypes, so no single-page application.
all pages are able to render a complete page WITHOUT the use of javascript. IOW: javascript is used as enrichment only.
Lots of pages have a very dynamic way in which widgets can be shown on screen. To overcome complexity at the server-side I've modularized my code into widgets (composite pattern), where each widget is responsible for it's own:
server-side controller code
server-side templating (using hogan/mustache)
routing endpoints, should it need to be called from the client
structural css (css converning the structure of the widget as opposed to the look&feel)
a server-side RegionManager ultimately decides which widgets are rendered and where they are rendered on screen. Endresults is that the RegionManager spits out the entire html (server-generated) as the composite of the rendering of all of it's widgets.
Now, some of these widgets DO have client-side logic and need rerendering on the client. Take a searchpage for instance, which needs to be able to update through ajax. (I've described this process, which uses DRY templating on client and server, here)
What I ultimately want is that, given I already use the composite pattern on the server, to extend this to the client somehow so that a Widget (1 particular logic block on the screen) contains all mentioned server-side code, plus all needed client-side code.
I hope this makes sense.
Would Marionette be suited to be used as a client side framework in this scenario? I'm asking since I'm not 100% sure if the concept of a Marionette Module is what I describe as being a Widget in above scenario. (I'm mentioning Twitter Flight in my question, since I believe this would be a fit, but it currently is so new that I'm hesitant to go with it at the moment_
I think basically what I'm asking is if anybody has some experience doing something along these lines.
I think just using Backbone.js is perfect for this type of application you are describing. You have probably already read this, but most of the backbone literature is focused around your views having associated server generated JSON models and collections, then using the View's render function to generate (on the client) the HTML UI that represents the model/collection.
However it doesn't have to be used this way. In fact there is nothing stopping you attaching views to existing elements that contain content already, which gives you all of the benefits of Backbone's modularity, events system and so on. I often use views that have no model or collection, purely because I like the conformity of style. I have also used an approach like I describe below in the cases where I have had to work with older, existing applications that have not yet got, or never will have a nice REST API, but they do provide content in HTML.
Firstly, lets assume the following HTML represents one of your widgets:
<div id="widget">
<div class="widget-title"></div>
<div class="widget-body">
<!-- assume lots more html is in here -->
Do something!
</div>
</div>
In this case, you could use backbone with a single Widget Model. This would be a very simple model, like this:
App.WidgetModel = Backbone.Model.extend({
intialize: function () {
this.url = this.options.url;
}
});
Take note of the fact the Widget receives a URL as a parameter to its constructor/initialize function. This widget model would represent many of your widgets (and of course you could adopt this general approach with more complicated models and pluck different data from the rendered HTML). So next for your views. As you probably know, normally you pass most views a model or collection when you instantiate them. However in this case, you could create the Widget model in your View's initialize Function and pass it a URL from the pre-rendered HTML as follows:
App.WidgetView = App.View.ComboboxView = Backbone.View.extend({
initialize: function () {
this.model = new App.WidgetModel({}, { url: this.$("a").attr("href") });
}
// rest of the view code
});
So instantiating the view would be something like:
new App.WidgetView({el: $("#widget")})'
By doing all of the above you can do pretty much everything else that backbone offers you and its modular and encapsulated nicely, which is what you are after.
The end result of this whole approach is:
You have rendered the Widget UI as pure HTML which (I assume) is functional without JavaScript.
You attach a View to the existing HTML.
You pass into the View as options, content by extracted (such as a URL) from the rendered HTML with jQuery.
The View is responsible for instantiating the Model passing on the relevant options the model needs (such as a URL).
This means all dynamic server side content is intially contained in the rendered HTML and your View is a modular JavaScript component that can do stuff to it, which I think is the end result you're after.
So you mentioned that you would like to have AJAX functionality for your widgets and that fine with this approach too. Using this approach, you can now use the standard Backbone fetch and save functions on the Widget model to get new content. In this example it is from the URL retrieved from the rendered HTML. When you get the response, you can use the view's, render function, or other finer grained functions to update the HTML on the page as required.
A few points:
The only thing to look out for is that you'll need to change the content type of the fetch and save functions to "text/html" if that's what the server is providing. For example:
this.model.fetch({
type: "POST",
contentType: "text/html"
});
Lastly, the model I have proposed is instantiated with no content. However if your ajax calls are a content type of "text/html", you may need to play around with you model so it can store this content in its attributes collection properly. See this answer for more information.
I am trying to test drive a view event using Jasmine and the problem is probably best explained via code.
The view looks like:
App.testView = Backbone.View.extend({
events: { 'click .overlay': 'myEvent' },
myEvent: function(e) {
console.log('hello world')
}
The test looks something like:
describe('myEvent', function() {
it('should do something', function() {
var view = new App.testView();
view.myEvent();
// assertion will follow
});
});
The problem is that the view.myEvent method is never called (nothing logs to the console). I was trying to avoid triggering from the DOM. Has anyone had similar problems?
(Like I commented in the question, your code looks fine and should work. Your problem is not in the code you posted. If you can expand your code samples and give more info, we can take another look at it. What follows is more general advice on testing Backbone views.)
Calling the event handler function like you do is a legitimate testing strategy, but it has a couple of shortcomings.
It doesn't test that the events are wired up correctly. What you're testing is that the callback does what it's supposed to, but it doesn't test that the action is actually triggered when your user interacts with the page.
If your event handler needs to reference the event argument or the test will not work.
I prefer to test my views all the way from the event:
var view = new View().render();
view.$('.overlay').click();
expect(...).toEqual(...);
Like you said, it's generally not advisable to manipulate DOM in your tests, so this way of testing views requires that view.render does not attach anything to the DOM.
The best way to achieve this is leave the DOM manipulation to the code that's responsible for initializing the view. If you don't set an el property to the view (either in the View.extend definition or in the view constructor), Backbone will create a new, detached DOM node as view.el. This element works just like an attached node - you can manipulate its contents and trigger events on it.
So instead of...
View.extend({el: '#container'});
...or...
new View({el:'#container'});
...you should initialize your views as follows:
var view = new View();
$("#container").html(view.render().el);
Defining your views like this has multiple benefits:
Enables testing views fully without attaching them to DOM.
The views become reusable, you can create multiple instances and render them to different elements.
If your render method does some complicated DOM manipulation, it's faster to perform it on an detached node.
From a responsibility point of view you could argue that a view shouldn't know where it's placed, in the same way a model should not know what collection it should be added to. This enforces better design of view composition.
IMHO, this view rendering pattern is a general best practice, not just a testing-related special case.
In a backbone application what is the best practice regarding when a model is fetched? I can see the following possibilities.
View calls the model fetch method
Some other JavaScript code calls the fetch model? if so when and what structure would this code have? is this the missing controller concept in Backbone?
A few best practives:
1 Collections and models that are necessary from the very first milliseconds of the app's life should be 'bootstrapped' in place (so there shouldn't be need to fetch them to gain access to vital data)
So when the user is served the correct pages from the server, the models and collections should be already in place (nice example from backbone.js docs)
var ExampleCollection = new Backbone.Collection();
ExampleCollection.reset(<%= #your_collection_data.to_json() %>); // Or whatever your server-side language requires you to do, this is a ruby example
2 The rest can be fetched just in time
The models and collections that aren't needed at the moment your app is initialized can be fetched whenever you feel like it, but I think that the logical time to do that is when the user expresses intent to use those models. E.g. user presses a button to open a view that needs some model/collection -> fetch that collection, user wants to clear unsaved changes from a model -> fetch that model from the server to get the last saved status of the model, and so forth. Usually the place where the fetching is bound to happen is the view that 'owns' the model/collection, because it relays the users actions to the model and displays the model's state to the user.
But like it was said, Backbone.js isn't strict about when a model or collection should be fetched. It can be done anywhere in the app, just make sure you do it only when it's necessary.
Hope this helps!
If you want to be standard, your view must render one time when initialize and listen for the change event of the Model and re render the view every time that model changes, that is all. (regarding what does View needs to do when fetch is completed)
And for call the model.fetch() if you follow the standard that I said, no matters where the fetch is called your view will be updated.
Some people could have a module named load in the view where do something like this:
load : function(){
this.model.fetch();
}
Others could do external fetch call, like this:
var myModel = new YourModel();
var myView = new SomeView( {model : model} );
//Probably you could render with the default data in the while model is fetched
myView.render();
model.fetch();
What is the best way to bind events to a Backbone boilerplate application? I've been trying to bind my events directly to the models associated with my views, in my views, but it doesn't seem to be working. I see within 'namespace.js', that there is an app key that extends Backbone.Events like so:
// Keep active application instances namespaced under an app object.
app: _.extend({}, Backbone.Events)
I don't fully understand how to use it...
I was able to get things working without the boilerplate, but it does provide some very cool functionality, so I'd love to be able to use it. Thanks!
ADDED
the code I was using was with the underscore bind method like so:
this.module.bind('change', this.render);
But then, I realized that 'this.model' is returning undefined, and so this doesn't work. I really am not sure how the boilerplate wants me to reference my model from the view.
I'm not sure if it is a typo that you copied from your code or a typo you only entered here, but I believe this.module (which IS undefined) should be this.model, which you also must be sure to pass in when you instantiate your view, of course, as so:
myView = new BBView({model: myModel});
then you can say this.model.bind('change', this.render); or this.model.on('change', this.render); for the most recent version of Backbone
I frequently bind my views to change events on my models in this way.
As for extending Backbone.Events, the way I have used it is to create an independent "event aggregator" that will help connect events between different views on your page. Let's say for example you have some action on one view that needs to set off an action on another view. In this case, you can pass your event aggregator object as an option to each of your views when you instantiate them, so that they can both trigger events or bind to events on a common object (i.e. your event aggregator).
whatsUp = _.extend({}, Backbone.Events) // the aggregator
myFirstView = new FirstBBView ({whatsUp: whatsUp});
(the aggregator shows up as this.options.whatsUp inside the view)
mySecondView = new SecondBBView2 ({whatsUp: whatsUp});
inside FirstBBView:
this.options.whatsUp.bind('specialEvent', function(arg1,arg2) {
// do stuff
});
inside SecondBBView, when something important happens:
this.options.whatsUp.trigger('specialEvent', {arg1: 'some data', arg2: 'more data'});
For a great explanation, see this great article by Derick Bailey