Use case:
I'm writing system tests using Geb/Selenium (so outside of angular).
I want to decorate $http to log all requests/responses at run time.
and here's the catch: without touching the source code.
Before you rush to answer "use $provide#decorator", for example,
http://blog.xebia.com/2014/08/08/extending-angularjs-services-with-the-decorate-method/
That solution for this use case means adding a test hook into production code... that's normally a bad thing I want to avoid if possible.
Update: Geb allows you to run Javascript in the browser window. So just for the heck of it I ran the tutorial code to decorate $http. Unfortunately, it didn't work because apparently you can't re-config the app after it's been loaded. But even if it did work, this brings up another interesting point---I need to override $http before any modules have had a chance to use it.
Since decorating $http service would be the cleanest way of doing this, you can avoid polluting production code by using something like ng-constants and gulp/grunt to only add decoration code for a 'test' environment.
See related Q/A here: How do I configure different environments in Angular.js?
If you are inclined on changing this at runtime(where runtime takes place in a test environment), you may need to go 'closer to the metal' and deal with XMLHttpRequests: Add a "hook" to all AJAX requests on a page
Related
I am taking over a big (50+ modules) AngularJS project, from another programmer, which did not do it right, so I have several migration questions:
There are lots of usages of JS functions like setTimeout and setInterval. It will be very easy to change to $timeout and $interval (because they use the same syntax, so it is just find and replace), but should I bother?
The entire project is without services, and all data requests run in the controllerss. Should I make time to create services for all controllers, the most important ones, or none? (I know that "if it works dont fix it", but from your experience does this make your life easier?)
The ENTIRE project uses $.ajax, with over a thousand requests. I do not have the time to migrate all requests to use $http, but I will try over time. In the meanwhile, should I create a service like $http_o and replace all "$.ajax(" strings in all files to $http_o, so the service will pretty much get a normal ajax request syntax, and send it using the $http service.
Every controller's services are written in a variable name, and not with a string at the start (function($scope) instead of ['$scope', function($scope)). Is there a fast way to change all of them to use the normal syntax, so I can use a minifier, or do I have to do it by hand? Should I do it?
I will obviously try my best to rework modules to use correct MVC and angular rules in my spare time at work, but this will happen way ahead in the future.
It's not as trivial as you might think i.e. $interval applies a $rootScope.$apply() at end so it will trigger $digest cycle, with lots of intervals it might slow down app
It's all about readability and maintainability. I'd say yes, do separate services and DRY the application otherwise when you will have to debug it in 3 months time it's going to be hell
A smart regex could work for you although the advantage of using $http over $.ajax is that you don't have to use tricks to trigger change in view - i.e. manually run $digest or $apply
You can do it easily with https://github.com/olov/ng-annotate
I am reading the source code for an app which reads as follows:
angular.module('graffio', [
'graffio.signupController',
'graffio.loginController',
'graffio.mainController',
'ui.router'
])
I have quite a few questions! So much confusion...
How is code loading these controllers which are defined later in the
code?
Why do I even need to state which controllers I want in my
app? Why can't I just have all the ones I declare?
Why does the angular documentation use the word 'injectable' so much? Isn't
injectable just the same as require?
With Angular, you can group as much or as little code into a module as you like. And you can compose modules together, like the author of the app you are looking at has done. The final module will have all of the services, config blocks, routes, controllers, directives, filters and so on that are in all of the modules it depends on as well as its own module.
This author has chosen to put each controller into its own module. Which is why the main module needs to depend on each of those modules. In my opinion, this seems like overkill, but it is what has been done, and all you need to do is understand it, not agree with it.
In answer to your other questions:
How is code loading these controllers which are defined later in the code?
When your code first runs, all the modules will be declared, and populated with directives, routes, controllers, services and so on. Nothing gets used yet. So long as when the code you have above is run, the other modules have already been declared then everything is fine (this may be done by a build process such as a Grunt task).
Then, when the document.ready event occurs, Angular looks through your HTML for an ng-app directive that says which module to load as your application. It then does what it calls the "bootstrap" process for that module.
Why do I even need to state which controllers I want in my app? Why can't I just have all the ones I declare?
Because this author has put each controller in their own module. If you put all the controllers you want into one module, then you don't need to declare them like that.
Why does the angular documentation use the word 'injectable' so much? Isn't injectable just the same as require?
Sort of. It is similar in that you are asking for a dependency by name and then using it.
It is different in that with require you can't typically modify what value is retrieved for a given dependency at runtime. With dependency injection, you can swap or modify dependencies at run time if you so choose (before your app starts properly and the dependencies are injected into the code using them).
This is perfect for testing, so that you can use a mock (fake) version of a dependency for testing purposes, instead of using the normal version. Think of a service that makes a call to get some data from the server and returns a promise. You can just create a mock version of that service that doesn't call to the server, just returns some test data immediately. This makes your unit tests fast, and means you don't need to be running your web server for them to work.
I have an application made with angularJS, whole application is composed of IIFEs(Immediately Invoked Function Expression). Every module, directive, controller is itself an IIFE and they are around 100s.
I want to know what is the performance pitfall when an app has so many IIFEs.
It is OK to use IIFEs with AngularJS?
How good is using libs like browserify and requirejs with AngularJS for managing the dependencies?
Can you please throw some light on this?
The question you need to ask first is whether you've got some internal parts within the IIFE that you don't want to expose to the global scope.
The whole point of creating a closure in this manner is to mimic encapsulation and avoid polluting the global scope.
The performance pitfall is not so easy to measure; I think the performance issue is negligible when you create the IIFE (don't forget you're just creating a function). One performance issue that I might think of is that when you reference a function variable form an inner function, you need to travel through the scope chain and get it from the closure object.
I would recommend you to look at patterns like module pattern, revealing module pattern etc. I personally use the revealing module pattern.
Browserify and requireJS are two libraries that implement two different specs; the commonJS and AMD respectively. These two specifications try to accommodate a functionality that is not supported by ES3 or ES5; That is a way of defining module and then loading them in specified locations.
If you want to define modules and load them synchronously in a similar manner to hose nodeJS works, you can use Browserify. In addition, Browserify allows you to use the same modules both for client-side and server-side (as long as you're using nodeJS).
On the other hand if you want to asynchronously load your modules you can go with AMD and requireJS, but you won't be able to reuse them on the back-end.
Finally, bare in mind that everything you've mentioned is not directly connected to angularJS; those are some good JavaScript practises to overcome some issues of the language itself. It can be well used whether you're working with angular or not.
I would recommend that you use either browserify or requireJS; it will benefit you in the long run. Just imagine having 100 JS files; you would need to out the manually into your html in the correct order based on the dependency graph. You can easily come across issues like race conditions where one file should have been inserted before another one.
As for the performance overhead of IIFEs, you shouldn't have any serious issues.
As another answer said, IIFEs are a good general practice, independent of AngularJS.
If you're concerned about the performance of IIFEs, go ahead and check out these performance tests on JSPerf (or write your own):
http://jsperf.com/iife-antecedents-in-js
http://jsperf.com/immediate-anonymous-function-invocation
While some ways of IIFE creation are clearly much slower than others, I wouldn't worry about the performance of IIFEs unless you're in a large loop.
I'm attempting to write some UI tests for a RequireJS-based Backbone application, utilizing FluentAutomation.SeleniumWebDriver and NUnit. The HTML page in question contains a typical data-main attribute for loading the RequireJS module for the application. My struggle is in properly detecting when the application is fully loaded with these tools; the only thing I've gotten to work consistently so far is using an explicit wait in seconds, like so:
I.Open("http://myapp")
.Wait(5)
.Enter("foo").In("input[name=username]")
.Enter("bar").In("input[name=password]")
.Click("button")
.Wait(5)
.Expect.Text("Welcome").In("#welcome");
This is less than ideal -- my test as written above will always take at least 10 seconds to run, when in reality the app might be "ready" much faster than that. What I'd like to be able to do is something like this:
I.Open("http://myapp")
.WaitUntil(() => I.Assert.Exists("input[name=username]"))
.Enter("foo").In("input[name=username]")
.Enter("bar").In("input[name=password]")
.Click("button")
.WaitUntil(() => I.Assert.Exists("#welcome"))
.Expect.Text("Welcome").In("#welcome");
However, this doesn't work -- using WaitUntil here actually seems to prevent the app from loading, for reasons unclear to me, as I simply receive timeout exceptions after the default wait period (30 seconds), stating that it was unable to locate the element in question within that timeframe.
I see that Selenium 2 provides a WebDriverWait for this kind of scenario, and possibly that would work here, but am unsure how I would use this within FluentAutomation (and a quick search of the FluentAutomation code on GitHub doesn't seem to indicate it's in use within the library).
What can I use in FluentAutomation to properly wait for a RequireJS module (or DOM loaded by it) to be ready?
Additional details:
This might not be a RequireJS compatibility problem at all. I've looked further into the app and found that what's happening after the Click("button") is actually a window.location.replace -- not a RequireJS async module load. It's the one place in the app that this is occurring, apparently. So, is a window.location redirect a known scenario that would cause problems with WaitUntil, and is there an alternate approach (aside from a simple Wait(5)) that would properly handle this?
All,
We are developing our app with AngularJS and time and again we keep running into the issue of Angular running or loading controllers twice. This becomes a little more intrusive when testing our controllers and more specifically when working with Testacular with jasmine's SpyOn's (since they get triggered before our code runs). So, our question is, is there such a thing as a constructor or init method that Angular is guaranteed to call when instantiating the controllers w/o having to hack work-arounds in the test code? TIA.
If you are specifying the controller in your router, then your template doesn't need to specify the controller via an ng-controller tag. Doing so will double load your controller.
Your controllers shouldn't be loaded twice, unless you are doing something wrong. You shouldn't have to hack any work-arounds.
You might want to provide a concrete example of how you are loading the controllers (and/or the partials which are associated with the controllers). It sounds like you probably have a routing issue which is causing your view to be loaded twice for each request.
Assuming that is a routing-related issue, unless/until you can provide more information to help people to help you, you may want to read this document. Pay particular mention to sections that mention 'redirect' and 'HTML 5' mode:
http://docs.angularjs.org/guide/dev_guide.services.$location
With so little information to go on, I can't answer your question, but perhaps that link will help you to help yourself. :)