Best practice for saving an entire collection?

Question

Say that I have a Collection and I've made changes to many of its Models. What's the best way to save all of the changes using a single HTTP request?

Answer 1

Usually REST backends handle single instance creation/update. You would need to change that to accept an array of objects.

That said, on the client side, you would need to go directly to the Backbone.sync function

Backbone.sync = function(method, model, options)

In this case your model should be an array of model. The method should be "create" or "save" and the options take the same type of options as a jQuery ajax call (error, success, etc.)

Answer 2

This code adds a new method to the collection prototype just to call the save method of those models that had changed. It worked for me:

Backbone.Collection.prototype.saveAll = function(options) {
 return $.when.apply($, _.map(this.models, function(m) {
   return m.hasChanged() ? m.save(null, options).then(_.identity) : m;
 }));
};

Gist link: https://gist.github.com/julianitor/701c677279bac1529b88

Answer 3

You should extend Backbone.Collection, giving it a save() method that would check each of its models hasChanged().

Then it should call Backbone.sync, which you'll probably have to extend a little into a custom sync function. If you do use a custom Backbone.sync function, then be sure to set it on your collection.

var CollectionSync = function(method, model, [options]) {
    // do similar things to Backbone.sync
}

var MyCollection = Backbone.Collection.extend({
    sync: CollectionSync,
    model: MyModel,
    getChanged: function() {
        // return a list of models that have changed by checking hasChanged()
    },
    save: function(attributes, options) {
        // do similar things as Model.save
    }
});

A different approach (using a model to represent the collection) is here: "How" to save an entire collection in Backbone.js - Backbone.sync or jQuery.ajax?

I also like https://stackoverflow.com/a/7986982/137067

Answer 4

I'm going to do the Wrong Thing here and quote Wikipedia regarding proper RESTful practices: a PUT to example.com/resources should replace the entire collection with another collection. Based on this, when we had to support editing multiple items simultaneously, we wrote up this contract.

1. The client sends {"resources": [{resource1},{resource2}]}
2. The server replaces the entire collection with the new information from the client, and returns the information after it's been persisted: {"resources": [{"id":1,...},{"id":2,...}]}

We wrote the server half of the contract in Rails, but here's the client half (in CoffeeScript, sorry!):

class ChildElementCollection extends Backbone.Collection
  initialize: ->
    @bind 'add', (model) -> model.set('parent_id', @parent.id)

  url: -> "#{@parent.url()}/resources" # let's say that @parent.url() == '/parent/1'
  save: ->
    response = Backbone.sync('update', @, url: @url(), contentType: 'application/json', data: JSON.stringify(children: @toJSON()))
    response.done (models) => @reset models.resources

I thought this was a lot easier to implement then overriding Backbone.sync. One comment on the code, our collections were always child objects, which should explain why the code sets a "parent_id" whenever an object is added to the collection, and how the root of the URL is the parent's URL. If you have root-level collections that you want to modify, then just remove the @parent business.