node package manager
Orgs are free. Discover, share, and reuse code in your team. Create a free org ยป

backout

Introduction

There already is a much fuller featured KnockoutJS solution for Backbone Models/Collections found at Knockback, that does pretty much everything this project does and more. The goal of this project is slightly different. Instead of being concerned with how to map all your Backbone data it focuses solely on making sure events are propogated and reduce the amount of mapping needed.

When one first encounters Knockout the first hurdle is to conceptualize what a ViewModel is in the MVVM pattern used. Since Knockout doesn't really focus what the Models or the ViewModels look like it's definitely the 'View' in the MV* since it doesn't push you along any particular convention. This makes it very flexible and minimal but it also leads to a lot of confusion. How to represent state has been a struggle in Web Applications since day one. What generally happens is you sprint off wrapping everything in observables before even really thinking whether you should be. Then the problem becomes: How can I best map all my data from the server. There are plugins for that or you can use one of many packages to integrate with a model framework. Ultimately the problem still ends up falling on the lose concept of what a ViewModel is. In practice people tend to write 2 types of ViewModels:

  • Application Logic View Models: these view models contain specific logic require to handle UI manipulation on your pages, sections, panels, controls, etc..
  • Data Model View Models: these view models wrap model/collection data with observables that can be interacted with.

The second category of View Models are the more problematic of the 2 since there is a sort of imperative to keep code DRY to have these shared. However, in practice they can't completely live outside of the context they are being used in and tend to bloat over time. More critically if the instances are shared, perhaps as some sort of centralized store you end up that more likely to create undesired dependencies. Especially when you have to consider mapping nested model structures. Ultimately in the same way Models are their own tree parallel to the Views, the View Models also follow this pattern and cross connecting mess is almost unavoidable.

This was the primary solution React was trying to solve. By componentizing the Application Logic View Models to own their own state representation and essentially removing Data Model View Models from the equation by forcing directionality in store interactions it's not nearly as easy to create this tangled web. The same patterns can be applied to Knockout. Most Knockout bindings save (checked, and value) are only one way. If anything the 2 way binding is just a convenience to reduce unnecessary boilerplate.

Knockout is built for fine grained control. The advantage of this becomes immediately obvious if you've ever tried to integrate your own store logic or deal with nested data structures. Given Knockout is already very efficient only redrawing changes in the DOM it's ok to loosen how many things get recalculated in certain cases to promote React-like simplicity. That is the goal of this project. Instead of attempting to map all the model attributes in every situation we just set tell Knockout to recalculate values when certain fields on the model or collections change. It's less efficient but not by much since it will only render the DOM if the value changed and the update cycles are still localized. Since most binding is one way it lets you deal with the models and collections directly in the view in many cases instead of mapped observables.

Documentation

Backout provides two Knockout Extenders that are wrapped in simple to use specialized Observables.

  • ko.observableModel(model, ...keys) - creates an observable that notifies subscribers on setting the model, and/or on key change events
  • ko.observableCollection(collection) - creates an observable that notifies subscribers on setting the collection and/or on add, remove, reset, and sort
  • ko.observableAttribute(model, key) - creates an ko.observable that is bound to the model attribute. Useful for 2 way binding.
  • ko.observableAttributes(model, ...keys) - returns an object with an observableAttribute for each key

So to set up a View Model with Backout one would use methods like the following:

var user_model = new Model({
  first_name: 'John',
  last_name: 'Smith'
});

this.user = ko.observableModel(user_model);  // create an observable model that notifies it's observers on any change

this.user().get('first_name') // 'John'

this.ignorant_user = ko.observableModel(user_model, 'first_name');  // create an observable model that only notifies when the first_name has changed

Object.assign(this, ko.observableAttributes(user_model, 'first_name', 'last_name')); // create a ko.observable for each attribute specified on the model

this.first_name('Jack');

user_model.get('first_name'); // 'Jack'

this.list = ko.observableCollection(new Collection());

this.list().reset([user_model]);

From here you could bind to the DOM like:

<span data-bind="text: user().get('first_name')"></span>

<input type="text" data-bind="value: first_name" />

// iterate a list and map on the fly
<ul data-bind="foreach: list().models">
  <li data-bind="with: ko.observableModel($data)">
    <span data-bind="text: get('first_name')"></span>
  </li>
</ul>

Test

TODO: Write tests