Onion Architecture in MVC Applications

Introduction
Pre-Requisites
This article is meant for my fellow developers with knowledge of ASP.NET MVC and are wondering what architecture to choose. I am trying to write this article to be as simple and easy to understand and implement.

Let's Start
The Onion Architecture was introduced by Jeffery Palermo in 2008 with the intent to make applications loosely coupled and with proper separation among the folders and the various areas of concern in the application. This makes the development easier, the testing of the application easier and the maintenance becomes easier. During the initial stage of development, the Software Requirement Specifications (SRS) is made and proper planning is done regarding what should be the approach, what technologies are to be used and then is done. The most difficult part is choosing the proper architecture so that the maintenance becomes easier. Here, the points that are kept in mind are:

1. All code depends on layers closer to the center.
   
   
2. Domain Models will be at the Center or the Core.
   
   
3. The inner layer defines the interfaces whereas the outer layer implements these interfaces members.
   
   
4. There is a layered behavior around the domain.
   
   
5. The infrastructure for containing the data and the service implementation should be pushed to the edge. Along with the infrastructure, the UI concerns are also pushed to the edge.

Background
There are many architectures used in web applications, but the question is how to choose the architecture that would help attain loose coupling that is most essential. Loose Coupling depends on a separation of concerns. That means that each layer will be independent of each other. What is tightly coupled and loosely coupled?


Yes, exactly as you are thinking my dear readers. But still let me discuss the difference quickly. A tightly coupling is where the one object/entity needs to have knowledge of other objects or we can say they depend largely on the interfaces where the service methods are declared. This can be avoided in small applications, but in large applications, these terms are required to be kept in mind, else it may lead to chaos.


A loose coupling, yes is the opposite, where there is very less dependency amongst the objects and the interfaces for which there is a chance of a clean separation of concerns and proper flexibility in the applications since it makes the framework more stable and enables proper maintainability. And it makes the developers happy.



The following shows the Onion Architecture at glance:



In the preceding image, as we can see, the Core is the Domain model. This layer contains the POCO entities.

Domain objects are:

 1.Encapsulates application business logic and rules.
 2.Maintains any state that is required.
 3.Does not depend on external infrastructure concerns.
In this article and demo I have added the Repository interfaces in the Core.

The Layers above in Brown has the Service Interfaces. The layer in Green is the implementation layer, or the Infrastructure Layer, where the Repositories and the Services methods are implemented. The Error Logging (especially NLog) is used. Also the Dependency Injection is implemented here to inject the dependencies into the controllers.

The layer in Blue (the outer layer) has the testing and the User Interfaces.

Thus, this was a simple description of the architecture, but the following diagram explains it better:



Here we can see there is no transitive dependency among the Test, UI and the Data Access that seems better for the Unit testing in MVC applications. This is the Layered Onion Architecture that proves to make an application loosely coupled.

The application service implementation as we can see is in a separate layer and the dependency finally is on the Core Domain. The Green arrows in the diagram represents the dependencies. Now let's look at a sample code. The demo is provided in the article for download.

Using the code
In this demo we will see how simple the folder structure of the solution is. Let's see the structure:



This is how simple the solution of the project would look like if we use the Onion Architecture. Now let's look at an expanded project:



Let's discuss each of the folders one by one.

The Core


This is how the core looks like. The core folder contains a class library project, that has interfaces both for Repositories and Services and the Model with the a .tt (T4 template) file that is autogenerated containing the POCO entities since I have used the Database First approach here. An important thing to note here is the .edmx file that is generated by the use of the Database First approach that contains the .tt file. To move that to the Core folder from the .edmx, cut the .tt file and paste that into the folder you want to, here Core -> Models. Only doing this does not end the topic, we need to specify the physical path of that file as shown in the diagram below:



As you can see in the image, there is a folder called Interface. As the name suggests, this contains the Repositories and Service Interfaces.

Infrastructure


This folder contains more than one project and contains the integral part of an application. The first project in this solution is a class library project with the first folder Data containing the .edmx file. An .edmx file, as defined in the MSDN, is a conceptual model and a storage model and the relationships among them. This also contains the Context classes and the .tt file, but regarding the .tt file we have already discussed the core since we have moved this file to the core project. The first project also contains the Repository and Service implementations as specified in the interface.

The main and the best part is the second project in the infrastructure, in other words the Dependency Injection, is an integral part when we want the Separation of concerns/Loose coupling in an application. In the preceding figure, we have two classes, one for each of the Repository Module and the Service Module. In the demo project, I have used Ninject. On adding Ninject for MVC from the Nuget Package, it adds NinjectWebCommon.cs to the App_Start folder that looks as in the following snippet:

1. using System.Collections.Generic;  
2. using DemoStart.Core.Interfaces;  
3. using DemoStart.Infrastructure.DependencyInjection;  
4. using DemoStart.Infrastructure.Logging;  
5. using DemoStart.Infrastructure.Services;  
6. using Ninject.Modules;  
7. [assembly: WebActivatorEx.PreApplicationStartMethod(typeof(DemoStart.App_Start.NinjectWebCommon), "Start")]  
8. [assembly: WebActivatorEx.ApplicationShutdownMethodAttribute(typeof(DemoStart.App_Start.NinjectWebCommon), "Stop")]  
9.   
10. namespace DemoStart.App_Start  
11. {  
12.     using System;  
13.     using System.Web;  
14.   
15.     using Microsoft.Web.Infrastructure.DynamicModuleHelper;  
16.   
17.     using Ninject;  
18.     using Ninject.Web.Common;  
19.   
20.     public static class NinjectWebCommon   
21.     {  
22.         private static readonly Bootstrapper bootstrapper = new Bootstrapper();  
23.   
24.         ///   
25.         <summary> /// Starts the application /// </summary>  
26.           
27.         public static void Start()   
28.         {   
29.             DynamicModuleUtility.RegisterModule(typeof(OnePerRequestHttpModule));   
30.             DynamicModuleUtility.RegisterModule(typeof(NinjectHttpModule));   
31.             bootstrapper.Initialize(CreateKernel);   
32.         }   
33.           
34.         <summary> /// Stops the application. /// </summary>  
35.           
36.         public static void Stop()   
37.         {   
38.             bootstrapper.ShutDown();      
39.         } ///  
40.           
41.         <summary> /// Creates the kernel that will manage your application. /// </summary>  
42.           
43.         /// <returns>The created kernel.</returns>   
44.           
45.         private static IKernel CreateKernel()   
46.         {   
47.             var kernel = new StandardKernel();   
48.             try   
49.             {   
50.                 kernel.Bind<func<ikernel>>().ToMethod(ctx => () => new Bootstrapper().Kernel);   
51.                 kernel.Bind<ihttpmodule>().To<httpapplicationinitializationhttpmodule>();   
52.                 RegisterServices(kernel);   
53.                 return kernel;   
54.             }   
55.             catch   
56.             {   
57.                 kernel.Dispose();   
58.                 throw;   
59.             }   
60.         }   
61.           
62.         <summary> /// Load your modules or register your services here! /// </summary>  
63.           
64.         ///The kernel.   
65.           
66.         private static void RegisterServices(IKernel kernel)   
67.         {   
68.             var modules = new List<ininjectmodule>   
69.             {   
70.                 new RepositoryModule(), new ServiceModule()   
71.             };   
72.               
73.             kernel.Load(modules);   
74.             kernel.Bind<iloggingservice>().To<loggingservice>();   
75.         }   
76.     }   
77. } 

Once we have added the Ninject for the MVC to our project, the next big thing to do is to bind the interfaces of the repositories and services to the Implementations such as in the following snippets.

For RepositoryModule.cs

1. public class RepositoryModule :NinjectModule  
2. {  
3.     public override void Load()  
4.     {  
5.         // BINDINGS..  
6.         Bind<idemointerfacerepository>().To<demointerfaceimplementationrepository>();  
7.     }  
8. } 

 For ServiceModule.cs

1. public class ServiceModule : NinjectModule  
2. {  
3.    public override void Load()  
4.    {  
5.        // BINDINGS..  
6.        Bind<idemointerfaceservice>().To<demointerfaceimplementationservice>();  
7.         
8.    }  
9. } 

Now, you will be wondering why we the Repository and Service. The Service methods are the Bridge/Flyover between the Controller/Business Logic and the Repository implementation, for making it more loosely coupled since there will be another layer between the Business logic and the Data access layer.

The next project is the Logging project that is the error logging that helps in logging the errors/exceptions to the database. Here in this demo project, I have used NLog.

Dependencies
In this architecture, the Web project has its dependencies on the other two projects, in other words the Core and the Infrastructure.

The Infrastructure project depends on the Core and the Core is independent of every other. As in the diagram for the Onion Architecture we have proven that the Core remains in the Core/Center and there is no transitive dependency amongst the UI and the Test with the data Access Layer. Check the images below:



You can check the dependencies among the projects and the Core that is independent.

Points of Interest
Finally, straight from the words of Jeffery Palermo, the father of this concept:

1.The application is built around an independent object model.
2.The layers inside define the interfaces (the core) and the outer layers implement them.
3.The coupling direction is towards the center, as in an Onion, as the name suggests.

Thus, the Onion Architecture helps decouple the Infrastructure and the Business (Controller) and the User Interface (Views) without getting into the OOP concept or has no new concepts with regards to the domain-driven approach.

This is thus a very simple and general, but very effective architecture. I hope readers will love this.

References

Jeffery Palmero.
Peel Onion Archi.