I've been meaning to learn Google's Angular.js framework for a while now. I of course have experience with straight up jQuery, but that inevitably becomes a mishmash of code. I do have some experience with Backbone.js, which has been battle tested on many of websites such as Pandora, but something about angular seems a bit cleaner.
At the time of writing I am working on a web project in Go. I intend to make the Go backend serve an API which an Angular app will consume.
Application bootstrapping involves creating an injector used for dependency injection, the injector then creates a root scope to become the context of the application, and then angular compiles the DOM rooted at the
ngApp root element, which processes directives and bindings.
$scope object allows explicit control over what models or parts of models are exposed to the view. The object is initialized in a controller, which is responsible for providing initial model values and defining functions. Models themselves can be any object.
Scopes inherit prototypically from each other. A variable defined in a higher-level scope is accessible to child-scopes. Writing to a variable in a child-scope doesn't propagate up the inheritance chain. Parent scopes can be directly referenced using the
$parent property, but it's use is generally discouraged because it embeds an assumption about the overall DOM structure.
Because of the asymmetric behavior in reading/writing scope properties at different scopes, it's generally encouraged to bind to object properties, not scope properties. This has the effect of always consulting the prototype chain, since it first looks up the object, which it finds in its ancestor up the prototype chain, then it changes/sets the value on that object. Whereas setting a single primitive (i.e. non-object) does not consult the prototype chain and simply creates a new value in the child scope that shadows the parent value.
Scopes form a tree rooted at the
$rootScope, and events can be propagated through the hierarchy. Events are sent up the tree with
$emit and down the tree with
$broadcast. Scope event handlers are registered with
$on, and as with DOM events, we can
stopPropagation. It's advised to use scope events sparingly, instead opting for two-way data binding.
Scopes pose memory considerations, and are generally destroyed when they're no longer needed. This can be controlled explicitly with
It's possible to register object-creating recipes that are interpreted by the
$injector to provide wired-up instances. Objects created by the
$injector are called services, and these are singletons. Only one instance of any given service is created.
There are ways to control how objects are created. The first and simplest way is to register a pre-defined instance with
value. Objects registered with
value can't express dependencies.
The simplest way to register object recipes that can express dependencies is by registering the constructor function with
service. This is generally useful for registering pre-existing constructor functions.
factory method is more flexible than the previous two, since it can be used to register any arbitrary object-creating function. As a result of this flexibility, it's the most common way to register objects into the dependency injection system.
Constants can be registered in order to be used by services. The problem with constants is that they have to be supplied as soon as a service expresses a dependency on it:
Providers is the generic way of registering services. A provider must return an object containing a
$get property which is a factory function returning a instance of the service. Provider functions can have additional methods and properties, which is useful for configuration of the service.
Providers can be configured before they produce any object instances. To facilitate this, module lifecycles consist of a configuration phase where recipes are collected and configured, and a run phase where it's possible to execute any post-instantiation logic.
The configuration phase allows providers to be configured as follows:
The run phase allows work to be done upon the application bootstrap, similar to the
main entry function in programs, although Angular supports multiple run blocks.