Hacker News new | past | comments | ask | show | jobs | submit login
Avoiding Common Backbone.js Pitfalls (ozkatz.github.com)
154 points by liordegani on March 27, 2013 | hide | past | favorite | 13 comments



Great article highlighting some common areas for attention -- especially when first moving from vanilla jQuery to something like Backbone ... but important for any basic Model/View layer separation.

For what it's worth, all 5 of these points are addressed by the Backbone docs, and there are some further nuances worth mentioning:

1. listenTo() is great when you're connecting from an object (frequently a View) that will be removed long before the object that its observing is removed. But when the two (the observer, and the observed) tend to live and die together in the same cycles, no special care is needed, as garbage collection will simply remove both together.

2. An easier way to avoid multiple DOM reflows when rendering collections, is to simply use a less granular template instead of a document fragment. You can even render the "item" template directly within the optimized "collection" template (remember, a template is just a function that takes data and returns HTML). For example:

    <div class="books-collection">
      <% books.each(function(book) { %>
        <%= templates.bookItem({model: book}) %>
      <% }); %>
    </div>
3. Absolutely. Addressed in the FAQ: http://backbonejs.org/#FAQ-bootstrap. While bootstrapping models, be careful that you don't inject arbitrary user JSON into the page without properly escaping it first -- if there's any data shared publicly between different users of your app. Otherwise they can throw a "</script>" into their post, and ruin your day.

4. All of Backbone's "save" calls are optimistic by default. If you'd rather a specific call be pessimistic instead, just pass "{wait: true}", in order to wait for the server to respond with a 200 OK before emitting the change on the client.

5. Yep. Instead of putting further model changes in a "success" callback, simply have a listener waiting for the "change" event. Then Ajax is out of the equation, and whenever the model state is considered to be "change"'d on the client-side, that change propagates appropriately through the system.


Thanks for the comments, valid points. regarding #2, it gets tricky when the model subviews do more than just render the template (listen on model changes and/or DOM events from their element).

Regarding #4, I totally agree. This is why I'm talking explicitly about the "success" callback - it is often misused.


For those who have not yet tried it, I would highly recommend TIm Branyen's Backbone.LayoutManager [1]. It solves items 1 and 2 very well.

[1] https://github.com/tbranyen/backbone.layoutmanager


I've used Marionette up until now but this is very interesting. It looks like this solves one problem really well, but marionette solves a bunch of other problems with views/app management.


Respectfully disagree with point 3 - "Doing unnecessary XHR requests on page load".

It's better to have your main app page completely static so it can be cached. That said, don't use ajax either - load the initial data via its own script tag.


Doesn't that also involve an extra HTTP call? I discussed the same issue as OP a while ago: http://branchandbound.net/blog/web/2012/11/unify-server-side...

Someone in the comments there mentioned a nice alternative to embeddding the data in a <script> tag to avoid the problem jashkenas raised (</script> in the JSON). Namely, setting it in a data- attribute.


Still, optimizing the initial load is a good idea for applications like this. As an alternative, you could create an API service that returns all relevant information for the first page view in one block, instead of half a dozen smaller AJAX requests.


Nitpick: I'm not sure that append() will trigger a reflow for every call. That will only happen if you interleave calls to a reflow-required property such as clientWidth. Anyone have a definitive answer?


I often find people get hung up on the part where they already have an app that isn't RESTful and doesn't necessarily return a clean JSON object on page load.

Front loading data is an easy way around this and if you're savvy enough you can create a fairly clean, agnostic shim to programmatically handle this for you (so long as you can make your backbone info align with your backend model structure).


Does anyone know if marionette takes care of the DOM reflow problem?

I'm looking at the source and it looks like it does not.

https://github.com/marionettejs/backbone.marionette/blob/mas...


Is backbone the best Javascript framework to go with these days? I'm keen to switch to this style of web development but I'm waiting for the community to pick a winner.


I strongly recommend AngularJS. Huge reduction in application code complexity. There are other good frameworks, like Ember, but the only one I wouldn't recommend is backbone. The only things that it has going for it are that it is easier to get started (but harder to continue), and that if you really love to hand roll stuff you get to: both because it doesn't really do that much.


It could be, but it also depends highly on your JavaScripts strengths. With them you could take backbone.js, which is quite lightweight and easy to learn, and create an extended SPA framework that screams.




Join us for AI Startup School this June 16-17 in San Francisco!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: