I've created a BookService that is used for storing (add), getting (getAll) and removing (remove) books. Now I want to use this BookService inside my controller to publish some methods that I can later use in my view. My controller looks like the following:
app.controller('BookController', ['BookService', '$scope', function(BookService, $scope) {
$scope.addBook = function() { BookService.add(); }
$scope.getAllBooks = function() { BookService.getAll(); }
$scope.removeBook = function() { BookService.remove(); }
}]);
As you can see, the controller is just a proxy that forwards the method calls to the service. Is this in general bad practice? And is there a better way to solve that using Angular.js? Calling the service directly from the view without a controller seems, to me like a bigger problem.
Best Practice
From what I can see this is perfectly good practice, I'm assuming that you're using $scope here to bind these services to a button or something similar. The Idea of using a factory or a service is to create some well organised reusable code that you can then instantiate in your controller exactly like you have done.
Controllers can get quite complex so moving basic functionality to your service is the best thing to do. In your case you have a relatively simple controller so you may not see the benefit of coding in this way. But I assure you that as your code base grows you will be required to code like this just to keep things readable.
Note
Just as an extra note I've added a data process that you might take with an example of a database request
database > API > service/factory > controller > view
This would be the correct path to take when requesting something from a database.
What you have shown is not bad practice; it is in fact recommended that you abstract resource management logic (like adding, removing and finding resources of the same type) into a separate service, for organization and maintainability and reusability purposes.
I do understand from your example how the controller seems to be just a proxy to the service though, and why you would think it's repetitive and possibly redundant. Let me expand on your example to see if I can explain the reasons for abstracting the logic. In doing so, let's consider only the adding operation, even though the logic would apply to others as well.
Reason 1: adding a book (or any resource for that matter) in a web app is usually more complicated than just a bookService.add() call. It requires having some sort of form to enter the details, a submit button to handle the actual request, possible massaging of the entered data, adding new fields like time stamps and author ids, and persisting to some data storage. The first two or three of those steps would be applicable to the screen or view from which you are adding, so it makes sense to put that in a controller (the view model), whereas the rest of the steps should go into the generic service.
Now say you have another screen where you see a friend's list of books and you can add it to your list by selecting a book. In that case, the controller logic for that view's controller will differ from the previous example, but the service steps will still be identical, and thus do not need to be duplicated since the service can be injected into both controllers.
That's a big win.
Example 2: let's say you've already split the code between controller and service as describe in example 1 above. Then you decide that your data storage needs have changed and now you want to update your logic to save to a different database system. Now instead of having to change all the controllers, you just change your services and the everything works.
This is also a really big win.
Yes you'ved done this correctly. Just keep in mind that as you scaled up your app in complexity, you'll want to have isolated scopes, (and) decomposition into component directives. Each directive will have its own controller and template fragment.
Related
I am working on a project where in there are almost 90+ modules.
All modules has a set of input fields and on submit the data should be saved on the server.
At any given point in time only one module is active. But there can be open modules in the background.
Submit button is common to all modules, meaning there is only one Submit button throughout the application.
Below picture explains it more.
The prime motto is to keep the individual module changes to minimum and a way to handle certain things(validation, reload etc) in the module from a central place.
The current approach I am planning is,
Use a 'moduleInit' directive that all module should include in its
partial.
The directive takes the $scope of the module and pass it to a
common service/factory (pushConfigService)
The pushConfigService stores and keep this scope as long as the
module is open. Once the scope is destroyed the reference of the
same will be removed from the pushConfigService.
The footer panel is another directive with Submit button in it and
calls a save function in the pushConfigService which in turn calls
a $scope function in the module to get the form data.
pushConfigService talks to a bunch of other services like
dirtyChecker, apiGenerator and finally post data to the server.
Each module will have a set of scope methods defined with some standard names. Eg: _submit, _onSubmit, _cancel, _reload etc.
Another way to handle this, broadcast the submit event and each module listens to the same. There is possibility more actions will be added to the footer panel.
So I am little bit hesitant to use the broadcast approach.
My question, Is it a good idea to pass controller scope to a service? Any alternate suggestions?
Thanks in advance.
I believe your core concept is a nice way to handle this setup. Yet I'd suggest to split business logic from UI. I don't have a sample of your code so it is a little hard to build an exact example. Yet since you're using the $scope variable I'm going to assume you're not using a styleguide like or similar to John Papa's. His ways encourage you to not use the $scope and to stay close to actual JavaScript "classes".
How does this make a difference?
Instead of passing the whole scope, you'd be able to just pass the instance of your specific module. For one it is less confusing to you and colleagues to have a concrete interface to operate on instead of having to figure out the composition of given scope. In addition it prevents services from being able to alter the $scope.
The latter could be considered a good practice. Having just the controllers alter the scope make it easy to find the code which alters and manages the UI. From there on the controller could access services to do the actual logic.
Taking it one step further
So passing the class instance instead of scope should be an easy adjustment to the already proposed setup. But please consider the following setup as well.
It seems there are quite some different ways to handle and process the data provided by the module/end user. This logic is now implemented in the controller. One might think some of these modules share similar handling methods (big assumption there). You could move this logic to, so to speak, saving strategies, in services. On activation of a module, this module will set its preferred saving strategy in the service which handles the submit button click. Or more precisely, the save data method which should be called from the onClick handler in the controller.
Now these services/strategies might be shared among controllers, potentially setting up for a better workflow and less duplicated code.
i'm very confuse about how ember controller works.
I'm starting with ember and ember-cli now, and i would like to understand more about how controller works.
If i have an nested route called new, inside a events resource, i should i have:
models/event
routes/events/new
templates/events/new
What about controllers?? I just work one simple controller, or should i use controllers/events/new too?
There isn't any generator command that will create every resource for me? I need call one by one?
Thanks.
What about controllers?? I just work one simple controller, or should i use controllers/events/new too?
This mainly depends on what is your controller needs to do. If it's only the essential stuff the controller does anyways, Ember will create that controller under the hood for you and automatically bubble actions up to its parent controller.
No better place than Ember guides to read what a controller is used for:
The simplest definition is:
Controllers allow you to decorate your models with display logic.
This means that you basically use them as the main communication layer between your route and your template. Essentially, you model comes from your route, through your controller and into your template. Actions happening in the template go up to the controller and then to the route. Therefore, controller is essentially the middle layer where you user your model (and other data) to control what is shown to the user, control what a user can do, control where can they navigate etc.
However, be aware of the plan for the future:
Controllers are very much like components, so much so that in future versions of Ember, controllers will be replaced entirely with components. At the moment, components cannot be routed to, but when this changes, it will be recommended to replace all controllers with components.
This means, that right now, controller responsibility is limited to two things:
Maintaining application state based on the current route
Handling or bubbling user actions that pass through the controller layer when moving from a component to a route.
All actions triggered on a template are first looked up on the controller, if it is not handled or bubbled (by return true) in the controller, they are looked up on the route.
Therefore, controllers for your /events or events/new routes aren't necessary at all, only if you want to handle things happening on those routes right away (in a smaller scope) instead of allowing everything to bubble up to the ApplicationController.
There isn't any generator command that will create every resource for me? I need call one by one?
Yes. Since, if you don't specifically create a controller, Ember just generates one for you behind the scenes. You need to specify where you want to handle things yourself.
You should visit the link I gave above (or here it is again) to the Ember guides that have many more examples in much more detail.
Note: Sorry for the length of the post, but the reason I decided not to break it down in separate questions was because I find these issues hard to address without a complex problem like this. I am dazzled, and a bit afraid, that I'm trying to force Angular to do stuff for me which is not the 'Angular way'. Any advice would be highly appreciated, and would probably get me on the right track with Angular.
My problem is as follows:
I have a dynamically generated form, controlled by myFormCtrl. I want to go extremely modular: I want to use it whenever and wherever. This means, that sometimes I'll need to put it somewhere as-is, sometimes I need to nest forms dynamically (like when I change a form value, and other sub-form appears), or control two separate forms as one in a view by a parent controller, with one 'Save' button for both. The myFormCtrl uses $scope.type_id and $scope.rowid to know, which record should it display from the database. The records are then Ajax-fetched by a service, and saved under the myFromCtrl's $scope.formItems. When saved, the form sends back data to the server (via service) with the type_id and scope credentials, so the restful api knows where to put the record.
In theory that would be really easy to do in Angular.js.
It definitely would be in every object-orientated language: The parent class could call a public method of getFormValues() of myFormCtrl. Now that can't be done in Angular: parent can't read children's scope.
For me, it seems, that this is not a simple 'how to communicate between controllers' issue. I know that the answer to that question is services, events, scope inheritance.
Also, a number of other problems seem to emerge from each solution I found sofar.
So I have a myFormCtrlBase class, which does basic stuff, and other more advanced classes extend this one. I also have a formControls.html, and a formLayout.html partial. The first contains an ng-switch, and gives the appropriate input element based on $scope.formItem.controltype, the second has the html layout of a common form, ng-including formControls.html at the right places. It utilizes ng-repeat="formItem in formItems", so that's where formControls.html's $scope.formItem comes from.
When I want the form to have a different layout for example, I create a customFormLayout.html partial ng-controlled by the myFormCtrl class.
First question: what if my form layout can't be put in an ng-repeat?
Like when form elements need to be placed scattered across the page, or form layout is not something which could be fit in an ng-repeat loop. my formControls.html still expects a $scope.formItem to work with. Easy OO solution: parent puts formItem in child's scope.
My solution: I created a <formItemScopeChanger formItemScope="formItems[1]"> directive which gets formItems[1] as an attribute, and turns it to $scope.formItem variable. This solutions feels messy: directives are not meant to be used like this. Doesn't seem very Angulary. Is this really the best solution?
Second question: Is ng-init really that evil?
Say, form is not put in the view by $routeProvider, but in a custom partial: rent-a-car.html. Here, I want to have a form where the user can select a car, and an other form, where I get his contacts. The two forms work with different $scope.type_id's, so there need to be two different forms:
<h1>Rent a car!</h1>
<div ng-controller="myFormCtrl" ng-init="type_id='rentOrder'">
<div ng-include="'formLayout.html'"></div>
</div>
<h2>Your contact information</h2>
<div ng-controller="myFormCtrl" ng-init="type_id='User';rowid='{{userData.rowid}}'">
<div ng-include="'formLayout.html'"></div>
</div>
Angular docs sais, that the only appropriate use of ng-init is when aliasing ng-repeat values. I don't see what the problem is with the example above - it is still the cleanest solution, isn't it?
I use the same technique with nested forms - I put a controller in with a template, initialized from the html by ng-init, and shown/hidden with an ng-if condition.
BTW, this is the only real initialization technique I found beside writing a new controllers (extending myFormCtrlBase). In an OO language, parent would write into the child's scope and then initialize it.
Perhaps my approach is influenced by my previously used languages and programming techniques, and is absolutely wrong.
Some would say, 'get init values from parent scopes!', but I can't seem to understand how that would be safe and efficient. I'd need to do $scope.type_id=($scope.type_id || $routeParams.type_id) with every scope property, which is first: really not nice to look at, second: is risky. Maybe this is a single form in a simple template, but somewhere in the scope hierarchy, there is a possibility, that it will find a completely different type_id. Perhaps it will be a completely different controller's type_id.
I don't see how using '.'-s in my scope variables would help. I has the same risk as I see it.
Third question: how to handle rentACar.html submission?
When hitting a Save button on my rentACar.html page, the rentACarCtrl (the controller in charge of the model of the view) should somehow retrieve the values of the two forms, and handle the validation and submission. I can't seem to understand how the common mantra 'controllers communicate through services' would be applicable here. A service for only to these two forms?
Am I on the right track? Every one of these solutions seem quirky. I feel lost :)
+ 1 question: Even after all this hassle, I can't seem to find a good reason why Angular wouldn't let parents call children's public stuff. Is there a good reason? Most of the above problems would have an easy answer in every true OO js framework.
You need to think about how you would test the logic of each of these components. Ask yourself how each of these 'features' work in isolation.
A few tips to help get you back on track:
Try and say away from a 'base' controller, I have hit many dead ends with scope inheritance, the logic gets muddled and hard to follow. Also this affects testing, because you find yourself having to stand up more objects than should be necessary for a test
Favor a singleton (angular service) for shared state over scope inheritance (a parent controller)
Create a directive and bind to the shared services state before using ng-include (prefer interacting with a service over scope inheritance)
Use an event pattern when another service or controller needs to now about events triggered from directives. A shared service (state) can listen for those events
What your asking is quite complex and I would like to help, Try to focus on one feature at a time and provide some code, I can show you how to use a shared service and the event pattern once you provide some examples
Also, taking a test first approach will often reveal the best 'Angular Way' of doing things.
Thanks to Mark-Sullivan, and a lot of work, trial-and-error attempts, the whole thing has boiled down to this. I'd like to get feedback from Mark, and other Angular gurus about this. What do you think?
You don't do class/prototypical inheritance in Angular.js. It is hard to test, and thats a big problem. For those, who are looking for 'inheritance' in Angular, I recommend this:
Your base class is the controller. The controller is an abstract model anyways, so it is perfect for that purpose. Use a $scope.init() function in your controller, but don't call it from there!
If you want to 'extend' your controller's functionality, use directives. In you directive link() function, call the controller's $scope.init(). (when compiling, angular runs controllers first, and directive link functions after). If scope had a $scope.name='base', in the directive link you will be able to redefine $scope.name=child, and after that, run $scope.init().
But wait! But this only allows a single-level inheritance. - Yes, thats true. But if you are looking for multilevel inheritance, you should use Services.
Multilevel inheritance is nothing else, but sharing the same code in a hierarchical class structure. For this purpose, use Services, and throw in these services with the dependency injector into your directives. Soo easy. This should be easy to accomplish, easy to understand, and tests run smooth.
Directives are very powerful tools, because you can dynamically combine partials with controllers.
I'm writing a large scale marionette application, which is ran initially from a router / controller.
Here's my question -- is it good practice to include other functions in your controller that aren't meant for routes?
So say I have the following method:
index : function() {
alert('test!');
}
is it consistent with best practices to be able to declare other functions in the controller not called when routes are initialized? Like so:
noRouteAssociated: function() {
alert('test!');
}
index: function() {
this.noRouteAssociated();
}
Obviously this is a simplified example, but I am trying to avoid putting large amounts of information or function declarations inside of methods only because they're associated with routers.
The roles and responsibilities of controllers are best illustrated by #davidsulc in this SO post and better yet his new book.
Generally speaking, it's okay to include methods that aren't meant for routes, if they're controlling the workflow of your app. Event triggering is a good example, but if you want to change the appearance of something or retrieve data from a database, you should move these methods to a view or model, respectively. The block quote below is taken directly from Marionette's controller documentation.
The name Controller is bound to cause a bit of confusion, which is rather unfortunate. There was some discussion and debate about what to call this object, the idea that people would confuse this with an MVC style controller came up a number of times. In the end, we decided to call this a controller anyways, as the typical use case is to control the workflow and process of an application and / or module.
But the truth is, this is a very generic, multi-purpose object that can serve many different roles in many different scenarios. We are always open to suggestions, with good reason and discussion, on renaming objects to be more descriptive, less confusing, etc. If you would like to suggest a different name, please do so in either the mailing list or the github issues list.
I'm part way through a pretty complex Knockout.js app using the mapping plugin to work with a deep object graph which closely mirrors server side domain objects. I've been refining my patterns as I go to something which works pretty well for my own slightly awkward context but I'd like to know if it's a good / bad overall way to approach MVVM Javascript.
Essentially my page pattern is to have a revealing module function which acts a bit like a controller - it owns the hierarchy of view models and is responsible for detecting changes, Ajaxing changes to the server and using the mapping plugin to update its view model graph with any flow-on changes which may come back in the response as JSON. I've done it this way because my domain is such that a small change in one part of the graph, when validated on the server, may result in changes / removals in distant parts of the graph. When this happens I need a common point at which to re-map the changes, present message dialogs to the user, etc.
All of my view models are instantiable functions and I've designed it so that they know nothing about the page they're used in, or the server (i.e. they don't do their own Ajaxing). The constructor of each view model creates its children via mapping options and each level is passed a reference to its parent. I've implemented a generic dirty flag which most of the view models use, and when a change occurs they pass a reference to themselves up the chain to a "dirty view model" observable at the top, which the module is subscribed to. I know this sounds a bit odd but it seemed the best way to approach it because items at each level are constantly being added and removed so I can't statically subscribe to properties at initialization time. And I don't want to keep removing and re-adding subscriptions each time I re-map my graph (which can get quite big).
From a pure efficiency point of view this isn't the best. The simplest way would be that each view model directly calls a function in the module when it needs to, but that type of coupling has to be wrong. Or I could pass in a reference to the module (or its relevant function) to each view model constructor, and the view model calls that, a bit like Javascript dependency injection. But that just seems too abstract. This is complex enough as it is.
Should I be looking at a higher level framework such as Backbone to sit on top of all this? Is injecting the module reference really too abstract? Or does this way of structuring things basically make sense as it is? I'm keen to hear from anyone who has worked on similarly challenging scenarios as to how you progressed and refined your patterns.
EDIT: I should have clarified that for various reasons, this app works in "save as you go" mode, whereby a change at a given level causes an immediate discrete Ajax post of just that one view model (not including its children) to be sent to the server (which may return a result which represents a change to just about anything else). Despite this annoying need for constant Ajaxing as opposed to pure client side action, Knockout.js has still made my app WAY more elegant, maintainable and scalable than my MVC apps of Olde.
Decoupling your viewmodels and reducing references can be achieved with a pub/sub model, like the one Ryan Niemeyer discusses here.
Ryan also made a Dirty flag for viewmodels, which can be found here.