I downloaded the Hanselminutes podcast nr 202, and only by reading in the title the word “WebformsMVC”, my interest was drawn.
I googled for it and found http://webformsmvp.com. A framework for using the MVP pattern in Webforms. I see the  positive side, familiarity if you know webforms, using your control-libraries, but still get the benefits of testability and the benefits you get with a decoupled architecture.

Also want to know more? Take a look at the featureset here.

My thoughts exactly…
Henry Cordes

Be the first to rate this post

  • Currently 0/5 Stars.
  • 1
  • 2
  • 3
  • 4
  • 5

In my previous post Multiple search criteria searching using Linq to SQL I talked about a way to implement multiple search criteria queries using LINQ to SQL.
Because I am doing some work using the ASP.NET MVC Framework, I looked into patterns to make a loosely coupled data layer. Ofcourse I checked out Rob Connery’s blog, he created the MVC Storefront (now Kona) the code can be found on MVC Sample Apps on Codeplex. Rob is leveraging the Repository pattern, this pattern provides dependency-free access to data of any type. I saw the screen cast where Ayende Rahien talks about ‘Filters and Pipes’.
Rob implements the filters in the MVC Storefront, I really like this approach, better than the approach in my previous post Multiple search criteria searching using Linq to SQL because it’s much cleaner, it is possible to ‘chain’ multiple criteria, but every criteria has it’s own extension method, thus following the single responsibility principle.
In my previous post in some cases I did too much in one method, I build up an IQueryable<post> for four search criteria, which breaks the Single Responsibility principle for one.

Using filters, which are extension methods that specify a filter on an IQueryable of something, makes it possible to let the calling party build up whatever they need. I know this can sound confusing, but I feel the following code example explains much better.

In my previous post, I called a few methods, and build up a IQueryable that way to satisfy every search criteria in the query.

   1:  [TestMethod]
   2:  public void Search_For_mvc_in_Title_And_Use_Paging_Test()
   3:  {
   4:   
   5:      MvcBlogDataContext repository = new MvcBlogDataContext("Data Source=.;Initial Catalog=MvcBlog;Persist Security Info=True;");
   6:   
   7:      var postsQuery = from p in repository.Posts
   8:                      select p;
   9:   
  10:      postsQuery = GetPostsQuery(postsQuery, "mvc", "", "", null);
  11:      postsQuery = GetPostsPagingQuery(postsQuery, 0, 5);
  12:   
  13:      List<Post> posts = postsQuery.ToList();
  14:   
  15:      Assert.IsNotNull(posts, "DataContext did not return posts when searching for 'mvc' in title");
  16:      Assert.AreEqual(posts.Count, 2, "DataContext did not return 2 posts when searching for 'mvc' in title");
  17:  }

Listing 1

Consuming filters
Would not it be cool if we could use a fluent interface-like way to query the data, so it will be obvious to what we want to query?
Something like the code (also in a Unit test manner like the previous post) in listing 2:

   1:  [TestMethod]
   2:  public void Search_For_mvc_in_Title_With_Paging_Test()
   3:  {
   4:      List<Post> posts = GetPosts().WithTitleLike("mvc")
   5:                                   .WhereTagsContain("")
   6:                                   .WhereBodyContains("")
   7:                                   .WithPaging(0, 5)
   8:                                   .ToList();
   9:   
  10:      Assert.IsNotNull(posts, "DataContext did not return posts when searching for 'mvc' in title");
  11:      Assert.AreEqual(posts.Count, 2, "DataContext did not return 2 posts when searching for 'mvc' in title");
  12:  }

Listing: 2

 

I think it is clear that the code in listing 2 is far more readable, than the code in listing 1. Another advantage is that it is easy to reuse every part of the query whenever it is needed.

Get all()
First I create a method that returns all Posts present in the database as an IQueryable<Post>. By the way, I am not using Dependency Injection, or Inversion of Control in this example, because it does not help in explaining the filters concept. BUT I think that when using this technique in a real world application, it is  a good thing to use IoC (StructureMap, Windsor, Ninject, Unity, whatever…).

   1:  public IQueryable<Post> GetPosts()
   2:  {
   3:      var postsQuery = from p in repository.Posts
   4:      select p;
   5:      return postsQuery;
   6:  }

Listing: 3

Extension methods
Leveraging extension methods, a .NET Framework 3.0 feature in combination with the IQueryable<T> interface, it is possible to create the filters. The class needs to be static and public, the extension methods also need to be static and public.
The first parameter specifies which type the method operates on and needs to be preceded by the ‘this’ modifier.

   1:  public static class PostFilters
   2:  {
   3:      public static IQueryable<Post> WithTitleLike(this IQueryable<Post> postsQuery,
   4:                                             string title)
   5:      {
   6:          if (!string.IsNullOrEmpty(title))
   7:              postsQuery = postsQuery.Where(p => p.Title.Contains(title));
   8:   
   9:          return postsQuery;
  10:      }
  11:   
  12:      public static IQueryable<Post> WhereTagsContain(this IQueryable<Post> postsQuery,
  13:                                                string tags)
  14:      {
  15:          if (string.IsNullOrEmpty(tags))
  16:              return postsQuery;
  17:   
  18:          return postsQuery.Where(p => p.Tags.Contains(tags));
  19:      }
  20:   
  21:      public static IQueryable<Post> WhereBodyContains(this IQueryable<Post> postsQuery,
  22:                                                 string bodyText)
  23:      {
  24:          if (!string.IsNullOrEmpty(bodyText))
  25:              return postsQuery;
  26:   
  27:          return postsQuery.Where(p => p.Body.Contains(bodyText));
  28:      }
  29:   
  30:      public static IQueryable<Post> IsCreatedOn(this IQueryable<Post> postsQuery,
  31:                                           DateTime? createdOn)
  32:      {
  33:          if (!createdOn.HasValue && createdOn.Value == DateTime.MinValue)
  34:              return postsQuery;
  35:   
  36:          return postsQuery.Where(p => p.CreatedOn.Value.Date == createdOn.Value.Date);
  37:      }
  38:   
  39:      public static IQueryable<Post> WithPaging(this IQueryable<Post> postsQuery,
  40:                                          int? startRow,
  41:                                          int? rowCount)
  42:      {
  43:          if ((!startRow.HasValue) && (!rowCount.HasValue || rowCount.Value == 0))
  44:             return  postsQuery;
  45:   
  46:          return postsQuery.Skip((int)startRow).Take((int)rowCount);
  47:      }
  48:  }

Listing: 3

In listing 3 it is obvious that every method has its own responsibility, it is easily maintainable and very readable. I think this is an elegant solution for the problem I tried to solve in my previous post, I found this better technique and want to share.

Henry Cordes
My thoughts exactly…

Currently rated 5.0 by 6 people

  • Currently 5/5 Stars.
  • 1
  • 2
  • 3
  • 4
  • 5

In a recent post, I talked about a project where the requirement is to create modules that can be used in an ASP.NET webapplication and in Sharepoint 2007, where I wanted to have separations of concerns. The way I solved this is: creating ASP.NET UserControl libraries (inside their own ASP.NET Webapplication project) that we package inside Sharepoint webparts. During development we test and debug the modules in a separate ASP.NET webapplication. We use pre-build events to copy the ascx files from the UserControl libraries to this webapplication and because we set a reference to the  webapplications, the dll’s are available already. Scott Guthrie has a tutorial on his website that explains how we do this. 

In the post I said I wanted to leverage Unity as our Inversion of Control Container and I was thinking about using the WCSF, because I wanted to use the MVP design pattern.
After some prototyping, I came to the conclusion using WCSF was not practical for our purpose.  Still I wanted to use the MVP pattern, because of unit-testing and loose coupling needs. I could not use MVC, because of the Sharepoint requirements.

The Model-View-Presenter pattern is different from MVC or Model-View-Controller in the sense that with MVC the Controller is where the request comes in and the Controller is where you control (what’s in a name…) your View and what you get from the Model.
With MVP the request comes in at the View, the View than delegates it to the Presenter that will get data from the Model and returns it back to the View.

Model View Presenter pattern Pic 1.: Model View Presenter pattern

The solution I finally found to be working really good is using the MVP design pattern and Unity in the following manner.

Model View Presenter UserControl diagramPic 2.:Model View Presenter Class Diagram 

As you can see in picture 2 the Model, View and Presenter all are implementing an interface. Practicing Interface based programming sets the door wide open for leveraging an Inversion of Control container, or IoC container. In my case I am using Unity, because I am doing this project for a client who is heavily into using Microsoft tools. All Presenters are derived from the abstract class Presenter<TView>.


Pic 3.: Interfaces used in MVP diagram

Why the interfaces?
Using Interfaces for all these parts of our component makes it easy to use an IoC container. Our objects are loosely coupled and we can switch out the concrete implementations easily, because we program against the interfaces. Say we want to test the Presenter and to avoid having a dependency on a data store or some service we want to switch the Model out for an implementation that returns the same hard coded objects each time. all we have to do is create this ‘mock’ Model that implements the IZaakModel interface (on Picture 3) and we are good to go.
To make this even more appealing: if we use an IoC container we only have to change the configuration in the container to start to use this other Model…

Here a listing of the Presenter of picture 2.

   1:  using System;
   2:  using Microsoft.Practices.Unity;
   3:  using HC.Interfaces.Zaak;
   4:  using HC.Mvp;
   5:  using HC.Domain;
   6:   
   7:  namespace HC.Presenters.Zaak
   8:  {
   9:      public class ZaakDetailPresenter: Presenter<IZaakDetailsView>, IZaakDetailsPresenter
  10:      {
  11:          private Zaak CurrentZaak{ get; set;}
  12:   
  13:          [Dependency]
  14:          public IZaakService Service { get; set;}
  15:   
  16:          private IZaakModel _model;
  17:          public ZaakDetailPresenter([Dependency] IZaakModel model)
  18:          {
  19:              _model = model;
  20:              _model.CurrentZaakChanged += new EventHandler<DataEventArgs<Zaak>>(_model_CurrentZaakChanged);
  21:          }
  22:   
  23:          void _model_CurrentZaakChanged(object sender, DataEventArgs<Zaak> e)
  24:          {
  25:              CurrentZaak = e.Data;
  26:              View.ShowCurrentZaakOnView(CurrentZaak);
  27:          }
  28:   
  29:          public override void OnViewInitialized()
  30:          {
  31:              base.OnViewInitialized();
  32:          }
  33:   
  34:          public override void OnViewLoaded()
  35:          {
  36:              View.ShowCurrentZaakOnView(CurrentZaak);
  37:          }
  38:   
  39:      }
  40:  }

Listing 1: Presenter

 

Because the Presenter (in listing 1) is derived from the Presenter<TView> abstract class, it is necessary to tell the Presenter of what type its View will be. Because we  use the Interface again, all we have to do is let Unity (our IoC container) do the work to supply the concrete implementation for this Interface. On line 13 of listing 1 we see the Dependency attribute, in this case the attribute will be used by Unity to resolve the concrete implementation of the IZaakService interface for the property (Service) on which the attribute is decorated that is configured.

The Model
The Model from picture 1 implements the Interface IZaakModel, and is responsible for getting the data if the Presenter asks for it and for letting the Presenter know that it has got this data and handing it over to the Presenter.

   1:  using System;
   2:  using Microsoft.Practices.Unity;
   3:  using HC.Domain;
   4:  using HC.Mvp;
   5:  using HC.Interfaces.Zaak;
   6:   
   7:  namespace HC.Models
   8:  {
   9:      public class ZaakModel : IZaakModel
  10:      {
  11:          public event EventHandler<DataEventArgs<Zaak>> CurrentZaakChanged;
  12:   
  13:          private IZaakService _datasource;
  14:          private Zaak _CurrentZaak;
  15:   
  16:          public ZaakModel([Dependency] IZaakService datasource)
  17:          {
  18:              _datasource = datasource;
  19:          }
  20:   
  21:          #region IZaakModel Members
  22:          public void LoadCurrentZaakById(int zaakId)
  23:          {
  24:              CurrentZaak = _datasource.GetZaakById(zaakId);
  25:          }
  26:   
  27:          public Zaak CurrentZaak
  28:          {
  29:              get { return _CurrentZaak; }
  30:              set
  31:              {
  32:                  _CurrentZaak = value;
  33:                  OnCurrentZaakChanged(new DataEventArgs<Zaak>(_CurrentZaak));
  34:              }
  35:          }
  36:   
  37:          public virtual void OnCurrentZaakChanged(DataEventArgs<ZaakExtended> e)
  38:          {
  39:              if (CurrentZaakChanged != null)
  40:                  CurrentZaakChanged(this, e);
  41:          }
  42:          #endregion
  43:      }   
  44:  }

Listing 2.: The Model

In listing 2 the Model is listed, is also depending on Unity for the resolving of the concrete implementations for interfaces. It implements IZaakModel  and thus implements the LoadCurrentZaakById method and the CurrentZaak property. When the  LoadCurrentZaakById (line 22, listing 2) is called it fills the CurrentZaak property calling a method on an object that implements the IZaakService interface (the private _datasource variable that is set by the ZaakModel constructor using Unity).

When the CurrentZaak property is filled the setter (line 30, listing 2) is used and in the setter the CurrentZaakChanged event is raised. Every object that subscribes to this event will recieve the change.

In listing 1 we can see on line 20 the Presenter has attached the event and will recieve this change. On line 26 of listing 1 the ShowCurrentZaakOnView method of the View is called.

The View
In ASP.NET a request will load an aspx page, the view is where the request starts, the view than delegates responsibility to the Presenter.

   1:  using System;
   2:  using Microsoft.Practices.Unity;
   3:  using HC.Interfaces.Zaak;
   4:  using HC.Mvp;
   5:  using HC.Domain;
   6:   
   7:  namespace HC.ZaakModule
   8:  {
   9:      public partial class ZaakDetails : System.Web.UI.UserControl,IZaakDetailsView
  10:      {
  11:          IZaakDetailsPresenter _presenter;
  12:          public event EventHandler UserControlLoaded;
  13:          [Dependency]
  14:          public IZaakDetailsPresenter Presenter
  15:          {
  16:              get
  17:              {
  18:                  return _presenter;
  19:              }
  20:              set
  21:              {
  22:                  if (value == null)
  23:                      throw new ArgumentNullException("value is null");
  24:              
  25:                  _presenter = value;
  26:                  _presenter.View = this;
  27:              }
  28:          }
  29:      
  30:          protected void Page_Load(object sender, EventArgs e)
  31:          {
  32:              if (!this.IsPostBack)
  33:              {
  34:                  this._presenter.OnViewInitialized();
  35:              }
  36:              this._presenter.OnViewLoaded();
  37:          }
  38:          
  39:          #region Control Init (Unity)
  40:          protected override void OnInit(EventArgs e)
  41:          {
  42:              IUnityContainer container = Container.Instance;
  43:              _presenter = container.Resolve<IZaakDetailsPresenter>();
  44:              _presenter.View = this;
  45:              base.OnInit(e);
  46:          }
  47:          #endregion
  48:      
  49:          #region IZaakDetailsView Members
  50:          public void ShowCurrentZaakOnView(Zaak zaak)
  51:          {
  52:              DetailsAlgemeenControl.CurrentZaak = zaak;
  53:              DetailsVoortgangControl.CurrentZaak = zaak;
  54:          }
  55:          #endregion
  56:      }
  57:  }

Listing 3.: The View

On line 50 of listing 3 the OnInit method of the UserControl (our view) is listed. Inside the body of this method we can see that we instantiate a variable named container (type IUnityContainer) and put a Container.Instance in it. On the next line (line 53) we fill the private member _presenter (the private member that is used in our property Presenter) by letting the container resolve the interface  IZaakDetailsPresenter. On the next line (line 54) we tell the Presenter that it’s View is this control. We have to do this here, because the View is where the request comes in. So this is the first component of our App that is loaded. Because my requirement is to create modules that can be used in Sharepoint or in ASP.NET, I cannot use the Global.asax to resolve the concrete implementations.

The ascx, or html is very straightforward, so i am not showing it, because I feel it is in the way of what i want to tell. I hope it is clear enough without it.

The Container
So how can
the container variable that is of Type IUnityContainer, an interface, on line 52 know what concrete class to Resolve? and where does Container come from. What does Container.Instance do?
In a class called container I create a UnityContainer using the Singleton design pattern:

   1:  using System;
   2:  using Microsoft.Practices.Unity;
   3:  using HC.Models;
   4:  using HC.Interfaces.Zaak;
   5:  using HC.Services;
   6:  using HC.Presenters.Zaak;
   7:   
   8:  namespace HC.ZaakModule
   9:  {
  10:      public class Container
  11:      {
  12:   
  13:          private static IUnityContainer instance;
  14:   
  15:          private Container() { }
  16:   
  17:          public static IUnityContainer Instance
  18:          {
  19:              get
  20:              {
  21:                  if (instance == null)
  22:                  {
  23:                      instance = new UnityContainer();
  24:   
  25:                      instance.RegisterType<IZaakModel, ZaakModel>(new ContainerControlledLifetimeManager())
  26:                          .RegisterType<IZaakService, ZaakService>()
  27:                          .RegisterType<IZaakView, Zaken>()
  28:                          .RegisterType<IZaakPresenter, ZaakPresenter>()
  29:                          .RegisterType<IZaakListView, ZaakList>()
  30:                          .RegisterType<IZaakSearchPresenter, ZaakSearchPresenter>()
  31:                          .RegisterType<IZaakSearchView, ZaakSearch>()
  32:                          .RegisterType<IZaakListPresenter, ZaakListPresenter>()
  33:                          .RegisterType<IZaakDetailsView, ZaakDetails>()
  34:                          .RegisterType<IZaakDetailsPresenter, ZaakDetailPresenter>();
  35:                  }
  36:                  return instance;
  37:              }
  38:          }
  39:      }
  40:  }

Listing 4.: The Container

 

Singleton
Using the singleton design pattern
the registering of the concrete implementation to the interfaces will be done only once, as you can see in the module I am building there are more components that follow the same pattern, so using a singleton was in my opinion the way to go in this case.
I choose to use a simple version of the singleton, I make a class called container, this class has a private static variable called instance (of type IUnityContainer).
The class has a private constructor, so the class cannot be directly created. Furthermore the class has a public static property (type IUnityContainer) that can only get a single instance of a IUnityContainer, because in the getter we check if the private member called instance (type IUnityContainer) is null or has been filled already, if it is filled the instance it contains is returned, but if it is null a new UnityContainer is instantiated and using a Fluent Interface all Interfaces and their concrete implementations are registered inside it before we pass it to the instance variable, after that the instance it contains is returned.

On line 25 of listing 4 the RegisterType has a parameter and the following RegisterType methods of the fluent interface have not. With new ContainerControlledLifeManager() as a parameter into the RegisterType method we tell Unity to resolve the concrete implementation as a singleton. In the case of the Model I need it to be a singleton, because the model is the only place some state need to be held in my architecture.

Configuration file
Configuring the registration in a  configuration file is also possible, in my case I did not wanted to do it, but I can understand that in some cases that is exactly what you want.

Conclusion
Ofcourse a lot more has to be considered when using this architecture, but I hope this article helps in understanding how you can use Dependency injection or Inversion of control and the Model view Presenter pattern to create maintainable and robust applications.

Henry Cordes
My thoughts exactly…

Currently rated 4.3 by 6 people

  • Currently 4.333333/5 Stars.
  • 1
  • 2
  • 3
  • 4
  • 5

Factory Pattern

Published 11/10/2005 by Henry in C# | Patterns

The challenge:
We got a lot of classes derived from one baseclass. Like a TableRow with controls in it.
Every control got another control in the third row, like:

  • A Textbox
  • A DropdownListBox 
  • A RadioButtonList
  • Multiple Textboxes and a Dropdownlist (Date)
  • Etc.

In our client (A Webb App) we do not want to be bothered with which control we have to instantiate. We want to instantiate and get the right kind of object returned.
For this kind of scenario we can use the Factory Pattern.

A Factory Pattern returns an instance of one of the several classes depending upon the parameters passed to the shared/non-shared factory method. Typically, all these classes are inherited from a common parent class.
Each of them has common methods. But each of them has their own optimization task operation on data. For simplicity I decided to give an example that is more simple than the real life implementation I described above.
I will use a very simple example.

We got an abstract Employee class (must be inherited), which will act as our base class.
The Employee class has two properties:

  • Name (string)
  • Salary (double)

We derive from the Employee class two classes:

  • Developer
  • Manager

Because we do not want the hassle to instantiate another class everytime we want to create a developer instance, or a Manager instance and because there will be more Employee derived objects in the future (Secretary, Architect, etc.), we create a Factory class that does the job for us.
We call it the EmployeeFactory. 

UMLFactory.jpg

 Here is the code for the Employee class, with the derived classes:

   1:  using System;
   2:  namespace FactoryExample
   3:  {
   4:      public abstract class Employee
   5:      {
   6:          #region Private Members
   7:          private string m_Name = string.Empty;
   8:          private double m_Salary = 0;
   9:          #endregion
  10:          #region Public Properties
  11:          public string Name
  12:          {
  13:              get { return m_Name; }
  14:              set { m_Name = value; }
  15:          }
  16:          public double Salary
  17:          {
  18:              get { return m_Salary; }
  19:              set { m_Salary = value; }
  20:          }
  21:          #endregion
  22:      }
  23:   
  24:      public class Developer:  Employee
  25:      {
  26:          public Developer()
  27:          {
  28:              Name = "John Refactorer";
  29:              Salary = 2000;
  30:          }
  31:      }
  32:   
  33:      public class Manager: Employee
  34:      {
  35:          public Manager()
  36:          {
  37:              Name = "Peter PeopleManager";
  38:              Salary = 5000;
  39:          }
  40:      }
  41:  }

Next we have the Factory:

   1:  using System;
   2:  namespace FactoryExample
   3:  {
   4:      public class EmployeeFactory
   5:      {
   6:          public Employee CreateEmployee(int employeeKind)
   7:          {
   8:              Employee EmployeeObject;
   9:              switch (employeeKind)
  10:              {
  11:                  case 0:
  12:                      EmployeeObject = new Developer();
  13:                      break;
  14:                  case 1:
  15:                      EmployeeObject = new Manager();
  16:                      break;
  17:                  default:
  18:                      EmployeeObject = new Developer();
  19:                      break;
  20:              }
  21:              return EmployeeObject;
  22:          }
  23:      }
  24:  }

And last but not least the console app that calls the factory to instantiate the Developer and manager objects.

   1:  using System.Collections;
   2:  using System.Text;
   3:  namespace FactoryExample
   4:  {
   5:      class Program
   6:      {
   7:          static void Main(string[] args)
   8:          {
   9:              EmployeeFactory EmpFactory = new EmployeeFactory();
  10:              Employee emp;
  11:              for (int i = 0; i &amp;lt; 2; i++)
  12:              {
  13:                  emp = EmpFactory.CreateEmployee(i);
  14:                  Console.WriteLine("{0} - ({1}) Name: {2} Salary: {3}",
  15:                                      i,
  16:                                      emp.GetType().Name,
  17:                                      emp.Name,
  18:                                      emp.Salary);
  19:                  emp = null;
  20:              }
  21:              Console.Read();
  22:          }
  23:      }
  24:  }

When we run this code the output in the console looks like this:

 Console_fact.gif

I hope you get the idea, I have some real use for this pattern in the situaton described at the beginning of this post.

Henry Cordes
My thoughts exactly...

Be the first to rate this post

  • Currently 0/5 Stars.
  • 1
  • 2
  • 3
  • 4
  • 5