I've been reading about MVC in which the authors suggest that testability is one of the major strengths of MVC. They go to compare it with ASP.NET WebForms and how difficult it is to test the code behind in WebForms.
I do understand it's difficult but can someone explain how unit tests were written to test code behind logic in the old days?
In the old days I tested aspnet webforms using the pattern Model View Presenter. I was able to test code with this pattern because I abstracted the conditional logic / loops / etc into a separate class that didn't live inside the webforms framework.
What was left in the webforms codebehind was nothing more than a few properties and a call in the page load to init the presenter class itself.
Then each event handler would simply pass off the work to the presenter class.
I've spent a great deal of time with this pattern and found that it makes things much more test friendly but it's a great deal of work compared to aspnet mvc
The code behind are plain methods in a class (the only difference with another class is that this class inherits from the Page object)
So it is testable.
most problems arise because the methods were tightly coupled to web.ui controls like the grid; they were not so easy to fake.
If you didn't fake the UI controls, you were also testing the inner workings of the UI controls which is a little overdone.
Related
(Apologies for the vagueness of this question. It is the result of some muddy thinking on my part.)
I am designing a small website with an MSSQL DB. I have just stumbled across the ObjectDataSource class. I think I understand it, but I don't see how it relates to the traditional MVC design pattern, if at all.
The class, according to Microsoft, is designed to sit between the UI (View) and the database, and as such, I would think of it as a Controller. It does indeed map UI actions to behaviours, but it also contains code for writing to the database, which is not something I would expect a Controller to do (but I may be wrong).
It seems to me that it is both the Controller and the Model, but I get the feeling I am not seeing something here.
That's a lot of words to say I am very confused. I would be very grateful to anyone who can point me at anything that might help me to understand what I am missing here.
The purpose of the ObjectDataSource is to intermediate between the view and some code that retrieves your data, not to the database directly. It is not a controller, but something that would call your code to get the data. A better description would be from an MVP scenario, where in MVP the view can manage itself, and the ObjectDataSource is a control that makes that management easier. However, the ObjectDataSource may make it harder to work with a traditional MVP or MVC design implementation out of the box, depending on the framework setup.
I personally don't like this control; I never use it as I'd rather code it so it's easier to see how the binding takes place, in addition to when the binding takes place. My personal opinion.
ObjectDataSource is not used in asp.net MVC if that's what you mean "traditional MVC design pattern". It's used primarily in WebForms development.
It's usually dropped onto a webform component and configured via the UI, although it doesn't have to be. It provides an abstraction that allows other WebForms controls to interact with an object model in a manner that is similar to the way in which these controls work with Databases. This object model may ultimately come from a database, but it allows you to insert your own database methodology in between.
WebForms does not lend itself to the traditional MVC pattern due to some technical issues which complicate using it in that way. A cleaner pattern for use with WebForms is the MVP (Model View Presenter) pattern.
I am quite amused by the MVP Pattern http://webformsmvp.com/
How ever I have certain doubts on the implementation part.
Is it necessary that to use this pattern, I have to implement user controls?
I always thought that user controls are created if we need reuse of controls across pages.
Say if I want to implement and test MVP pattern,I should break my simple page with controls to a lot of user controls so that I can apply the MVP pattern?
What if I have quite a lot of standalone components in my webpage?
Is there any gap in my understanding of MVP?
Help.
It is not necessary to implement user controls to use this pattern, in short. Very briefly, you need a view (could be aspx or wpf or winform or console, etc.), a presenter that'll read from/listen to events from the view, make a call to the model and finally populate view with the right data, that's what MVP pattern is.
edit: this example is simple enough.
You don't have to use user control to use MVP pattern. MVP is GUI pattern that helps you to separate your concern.
For example, You write ASP.NET web page to calculate two numbers, with out MVP or MVC or any other GUI pattern, you would write all this logic in your code behind file which is very hard to test. If you want to test it, then you are bringing lot of extra baggages like ASP.NET framework.
In other hand, you write this app using MVP, you would do this.
View => Dumbest in all three. Doesnt have any or minimun logic. So you dont have to unit test it. It simply "tells" the presenter something happened and does what presenter asks.
Presenter => Controls the flow
Model => Business Logic/persistent logic.
I'm the author of the Web Forms MVP project you mention. This answer is specific to that library (which is just one implementation of the MVP pattern).
No, you do not need to use user controls. If you want, you can make your page inherit from MvpPage and it will then work with a presenter itself.
We recommend that you do use user controls though, even if you aren't using the control multiple times in your site. This lets you keep the view, view model and presenter logic nice and bundled as a logical unit for a particular feature. Pages are then used purely for laying out controls.
I don't have real understanding of the MVC model or Architecture yet but what I read and see is that it separated the 'Concerns' that is the presentation UI and logical code, right? But I know that WebForms Architecture also has a code behind model which then separates the code and UI.
Is there something else also in MVC which further separates the stuff around?
The main difference between MVC and WebForms is that in WebForms it is the view that receives the request (/foo.aspx), while in MVC it is the controller (/controller) which will manipulate the model and choose the appropriate view to render. Another important difference is that all the HTTP Context stack (Request, Response, Session, ...) has been abstracted behind abstract classes and interfaces in ASP.NET MVC which allows for better separation of concerns and unit testability in isolation. You also have far more control over the generated markup in ASP.NET MVC in contrast to WebForms where the markup is essentially generated by server side controls.
Code-behind is specifically what makes UI tightly coupled to business logic in WebForms -- the code-behind is part of the UI.
Using controllers instead of code-behind is one of the primary ways in which MVC decouples these concerns.
The easiest way to think about this is that the .aspx and code behind are essentially two different views of the same component, the UI. It's completely possible to use code behinds with MVC (if you're using the webforms view engine), but both the view and the code behind are considered part of the UI. The controller is a seperate entity, as is the model.
Code-behind is specifically what makes
UI tightly coupled to business logic
in WebForms -- the code-behind is part
of the UI.
Only if you choose to let it.
Business logic will not be coupled to UI if you choose to have seperate business layer/tier and implment an appropriote UI layer pattern such as MVP/MVC/MVVM.
Id argue a n-layer design with an MVP pattern in webforms can offer even greater seperation of concerns than asp.net MVC but requires alot more upfront design.
Asp.Net MVC forces better seperation out of the box. Its baked into it.
This is good for developers who might not know any better. With webforms it is completly up to the architect/developer to select the level of sepearion required, a double edged sword because if you're experienced with the platform can do some great stuff, but if your new to it or coming from a classic asp style of development you may make a mess of it without some guidance.
Biggest plus I see for asp.net MVC over webforms isnt soc, or testability (as they can be acheived in webforms) it's the ability have better control over markup (if you need it) and the more web centric focus (again, if thats what you need).
I have just started to learn ASP.NET MVC. When comparing ASP.NET MVC with Web Forms one of the main advantage of MVC is always told to be better support of Unit Testing. Can I got a good explanation of how it has better support?
Edit :
If possible please provide example in both.
Asp.Net MVC has better unit testing support for one main reason - the whole architecture is built to use HttpContextBase, HttpRequestBase and HttpResponseBase.
Asp.Net webforms is dependent on HttpContext.Current, which is a singleton that you have no control over - it is set up and passed to your pages as part of the HttpApplication executing the request. In most cases, in order to get the page to execute correctly, you need to execute it in a real HttpContext. Since many of the properties of HttpContext are not settable (like Request and Response), it is extremely difficult to construct fake requests to send to your page objects.
This makes unit testing webforms pages a nightmare, since it couples all your tests to needing all kinds of context setup.
Contrast that to ASP.Net MVC, where you can mock an HttpContext! Now your code doesn't even need a web server to give it context, you can just set up the bits you need, and hand the mocked context to your method.
The ASP.NET page lifecycle makes it incredibly difficult to unit test classes that derive from Page, which starts out with too many responsibilities and becomes a god object when you add application logic to it. Even worse, it has hidden dependencies on static classes and requires a default parameterless constructor, which limits your ability to inject dependencies.
So, to make an ASP.NET WebForms page testable, you need to take all the logic out of your code-behinds and put it in another class - typically a Presenter as in the Model-View-Presenter pattern.
ASP.NET MVC controllers are already separated from their templates and aren't encumbered by the ASP.NET page lifecycle.
Because you can create a controller object in your unit test, call some actions on it, and see the result right away, then you can Assert.IsBlahBlahBlah(); on it.
For example,
[TestMethod]
public void Index()
{
// Arrange
HomeController controller = new HomeController();
// Act
ViewResult result = controller.Index() as ViewResult;
Assert.IsNotNull(result);
}
With that method, you now know that your Index view is returned from the Home controller.
If you want to use ASP.Net WebForms (like i do) and unit tests together, take a look at this:
WebForms MVP on codeplex
Works for me.
It's far more cumbersome to test a code behind page than to test a controller. With the MVC pattern, there's a more logical separation of presentation logic thus making it easier to write tests.
You can do exactly the same thing with Web Forms, it's just that people who don't know better write code that can't be tested this way.
There's no reason to have business logic in a codebehind class. Same with data access logic. Even right there, that lets you test the parts of the application most subject to both bugs and testing.
Some might say that doesn't allow you to test for button clicks and other user-interface events. If you want to do that, then you can go ahead and create your own MVC or MVP or other such pattern, that does use a separate interface for the user interface actions. Then do exactly the same kind of test you'd do if using ASP.NET MVC.
And you still have the problem of not being able to test client-side code.
Slightly OT but you may want to look at Selenium for Unit Testing webpages....
In reference to Lawrences suggestion about Selenium, there is also WatiN.
This is not specific to MVC, but I think MVC definitely helps in that it keeps element ids and classes clean and easier to target from the test.
With WatiN you can do the following (example from their site).
[Test]
public void SearchForWatiNOnGoogle()
{
using (var browser = new IE("http://www.google.com"))
{
browser.TextField(Find.ByName("q")).TypeText("WatiN");
browser.Button(Find.ByName("btnG")).Click();
Assert.IsTrue(browser.ContainsText("WatiN"));
}
}
In traditional ASP.NET Web Form applications, UserControls are a great way to encapsulate functionality so that it can be reused. However, UserControls don't fit well into the MVC model. They often make heavy use of ViewState and they blur the seperation of concerns that MVC promotes.
My question is, how do you best bundle a piece of functionality so it can be shared across MVC applications?
As an example, consider a from/to date-selector UserControl that:
allows a user to select two dates, either using a javascript overlay or by typing in day, month and year into seperate fields
can be configured to default to either today and tomorrow's dates or to dates of the developer's choosing
validates the dates that comes back from the user to ensure the from date is before the to date
exposes From and To properties that can be accessed by code-behind
How would I best build something like this in .NET MVC so that I can easily reuse it?
Note that to fully emulate User Control's functionality the MVC component would have to manage the submitted form data and validation - not just the presentation.
In general I would agree that user controls are nice in terms of encapsulating UI stuff, but I don't think too much has really changed in MVC. If I remember right re-using user controls across classic Asp.net projects was a pain and was never really the best way to truly create reusable components. Most UI toolkits that you bought for classic ASP.net didn't give you user controls, they gave you essentially server controls and javascript controls.
In your example, I would probably create or find a jquery (or ur framework of choice) plugin that did what you wanted on the client side. You could also build a C# wrapper around it similar to what Telerik did with some of the jquery UI controls. I do think that the word code-behind and even viewstate will disappear from your vocabulary the more you get into MVC.
If you look at what open source projects are out there for MVC you will get your answer in terms of what you should be doing.
The MVC Contrib app adds a lot of features by creating extension methods and helpers. Their grid control is a typical way to create a reusable component that you could use across projects
Telerik, created some extensions that wrap jquery controls and do asset management.
Finally I think if you look to the future, MVC has areas, which if I interpret it right will give you the ability to break your project apart into multiple smaller projects.
Besides what is already suggested, ASP.NET MVC v2 will have generic templated input controls, see here. You can read how other people do similar techniques, for example, here:
We have
exactly 1 method call for generating a
form element, “Html.InputFor”. As
part of that “InputFor”, it examines
an input specification, that collects
the PropertyInfo, any attributes, the
type, any modifiers called, and
selects an appropriate InputBuilder.
Call InputFor(p => p.Id) and Id is a
GUID? That creates a hidden input
element. Call InputFor(p =>
p.Customer.Address) and Address is a
complex type? That looks for a
partial with the same name of the type
Having considered the helpful answers from others, I will have a go at answering my own question.
It seems to me that the key difficulty with emulating UserControls in MVC is that they crosscut the concerns that MVC aims to seperate. The from/to date selector UserControl in my example incorporates elements of Model, View, Control and interation. UserControls' ability to bundle all this together is exactly the reason that they don't fit well into MVC.
That means that to create a psuedo-UserControl in MVC requires four seperate pieces:
A Model class - in this case an Interval class or similar
A PartialView that knows how to render the Model to HTML
A jQuery script to layer interactivity on top of the PartialView's HTML
A ModelBinder that can deserialise postdata into an instance of the Model class.
The ModelBinder is important because it deals with data coming back from the user. Without it, every Controller that wanted to display a to/from date selector in any of its Views would have to know how to assemble the six postdata fields - and how to cope if they were invalid or some were missing.
Two ways that I can think of. A partial view though this doesn't really transfer well from app to app because you are moving around ascx files. Not a big pain but not my flavour.
I prefer to use WebControls. They are super easy in mvc and all you need to do is reference the library in the project and possibly in your config file and there you go.
I think some of the answers have missed out on the postback functionality of controls. One way you could handle that is to pass any generic information via ViewData when rendering your partial view. That could then post back to its own control, which in turn could redirect to the UrlReferrer.
Its a little messy and use of UrlReferrer poses a security risk. But it is one way around the problem
You can create a jQuery plugin.
As user-controls provided in ASP.NET Webforms, MVC provide a lot of ways to make the controls and code that can be reused in other app.
Using Partials If your partial code have some C# logic and render the html using Razor/aspx code then it's bst to maintain them in razor file.
Write JavaScript Functionality as plugin If you maintain your code and write it as better as it can be used in other app then it would be a huge advantage for you. Next time when you work on other app just open this solution copy it and modify it. Write JavaScript code that can be used as plugin maybe take some more brainstorming.
Write Code As a Separate C# library If some code is too common for every app you make.for example you write a member authentication system or some global function (C#) that are used in every app you made then maintain them in a separate solution so it can be used in other app you made whenever you trying to make a new app in future.