Introduction
In this article we will try to see what is Repository and Unit of Work Pattern in an ASP.NET MVC application.
We will also implement a small rudimentary sample application to understand the same.
Background
Possibility of using ORMs in our application saves us from a lot of code that needs to be written in
order to create our entities and data access logic. But using ORMs like entity framework sometimes
lead to scattered data access logic/predicates in various place in code.
Repository and Unit of work pattern provides a clean way to access data and at the same time
maintain the test-ablility of the application. Let us try to understand this by implementing a simple
ASP.NET MVC application.
Using the code
Let us first try to create a simple database on which we will be performing CRUD operations. We will
define a simple tables in the database as:
Now with the database/table in created, we will go ahead and generate the ADO.NET entity data Model for these
tables in our application. The generated entities will look like:
Performing Simple Data Access
Now we have the entity framework ready to be used in our application. We can very well use the
Context class in each controller to perform database operations. Let us try to see this by trying to retrieve the data in our Index action of HomeController.
public ActionResult Index()
{
List<Book> books = null;
using (SampleDatabaseEntities entities = new SampleDatabaseEntities())
{
books = entities.Books.ToList();
}
return View(books);
}
And when we try to run this application, we will see that it is getting the data from the database as:
Note: We will not be doing other CRUD operations here because they can be done on same lines very easily.
To visualize the above implementation:
Now there is nothing wrong from the code and functionality perspective in doing this. But there are two
problems in this approach.
- The Data access code is scattered across the application and this is a maintenance nightmare.
- The
Action in the Controller is creating the Context inside itself. This makes this function non testable using
dummy data and we can never be able to verify the results unless we use test data.
Note: If the second point is not clear then it is recommended to read about Test Driven Development using
MVC. We cannot discuss it in this article otherwise the article will become digressing.
Creating a Repository
Now how can we solve the problem. We can solve the problem by moving all the data access code
of entity framework in one place. So let us define a class that will contain all the data access logic for
Books
table.
But before creating this class, let us also think about the second problem for an instance. If we
create a simple interface defining the contract for accessing the books data and then implement this interface
in our proposed class, we will have one benefit. We can then have another class implementing the same interface
but playing around with the dummy data. Now as long as the controller is using the Interface our test projects
can pass the dummy data class and our controller will not complain.
So let us first define the contract for accessing books data.
interface IBooksRepository
{
List<Book> GetAllBooks();
Book GetBookById(int id);
void AddBook(Book book);
void UpdateBook(int id, Book book);
void DeleteBook(Book book);
void Save();
}
And the implementation of this class will contain the actual logic to perform the CRUD operations on the
Books table.
public class BooksRepository : IBooksRepository, IDisposable
{
SampleDatabaseEntities entities = new SampleDatabaseEntities();
#region IBooksRepository Members
BooksRepository()
{
entities = new SampleDatabaseEntities();
}
public List<Book> GetAllBooks()
{
return entities.Books.ToList();
}
public Book GetBookById(int id)
{
return entities.Books.SingleOrDefault(book => book.ID == id);
}
public void AddBook(Book book)
{
entities.Books.AddObject(book);
}
public void UpdateBook(int id, Book book)
{
Book b = GetBookById(id);
b = book;
}
public void DeleteBook(Book book)
{
entities.Books.DeleteObject(book);
}
public void Save()
{
entities.SaveChanges();
}
#endregion
#region IDisposable Members
public void Dispose()
{
Dispose(true);
GC.SuppressFinalize(this);
}
protected virtual void Dispose(bool disposing)
{
if (disposing == true)
{
entities = null;
}
}
~BooksRepository()
{
Dispose(false);
}
#endregion
}
Now let us create a simple Controller in which we will have the reference to this class being used
perform the CRUD operations on Books table.
public class BooksController : Controller
{
private IBooksRepository booksRepository = null;
public BooksController()
:this(new BooksRepository())
{
}
public BooksController(IBooksRepository bookRepo)
{
this.booksRepository = bookRepo;
}
public ActionResult Index()
{
List<Book> books = booksRepository.GetAllBooks();
return View(books);
}
}
Now here in this above code when the application runs the default parameter-less constructor will run which
will create a BooksRepository object and it will be used in the class. The result of which is that the application
will be able to work with actual data from the database.
Now from our test project we will call the parameterized constructor with an object of the dummy class containing
dummy data. The benefit of which is that we should be able to test and verify the controller classes using the dummy
data.
Lets run the application to see the output
Note: We will not be doing other CRUD operations here because they can be done on same lines very easily.
Lets try to visualize this version of implementation
Having Multiple Repositories
Now imagine the scenario where we have multiple tables in the database. Then we need to create
multiple repositories in order to map the domain model to the data model. Now having multiple repository
classes poses on problem.
The problem is regarding the ObjectContext object. If we create multiple
repositories, should they contain their ObjectContext separately? We know that using multiple instances of
ObjectContext
object simultaneously can be a problem so should we really allow each repository to contain their own instances?
To solve this problem. Why to let each Repository class instance have its own instance of the ObjectContext.
Why not create the instance of ObjectContext in some central location and then pass this instance to the repository
classes whenever they are being instantiated. Now this new class will be called as UnitOfWork and this class will
be responsible for creating the ObjectContext nstance and handing over all the repository instances to the controllers.
Unit Of Work
So let us create a separate Repository to which will be used via UnitOfWork class and the ObjectContext will
be passed to this class from outside.
public class BooksRepositoryEn
{
SampleDatabaseEntities entities = null;
public BooksRepositoryEn(SampleDatabaseEntities entities)
{
this.entities = entities;
}
public List<Book> GetAllBooks()
{
return entities.Books.ToList();
}
public Book GetBookById(int id)
{
return entities.Books.SingleOrDefault(book => book.ID == id);
}
public void AddBook(Book book)
{
entities.Books.AddObject(book);
}
public void UpdateBook(int id, Book book)
{
Book b = GetBookById(id);
b = book;
}
public void DeleteBook(Book book)
{
entities.Books.DeleteObject(book);
}
public void Save()
{
entities.SaveChanges();
}
}
Now this Repository class is taking the ObjectContext object from outside(whenever it is being created).
Also, we don't need to implement IDisposable here because this class is not creating the instance and so its
not this class's responsibility to dispose it.
Now if we have to create multiple repositories, we can simply have all the repositories take the
ObjectContext
object at the time of construction. Now let us see how the
UnitOfWork class creates the repository and passes it
on to the Controller.
public class UnitOfWork : IDisposable
{
private SampleDatabaseEntities entities = null;
public UnitOfWork()
{
entities = new SampleDatabaseEntities();
BooksRepository = new BooksRepositoryEn(entities);
}
public UnitOfWork(IBooksRepository booksRepo)
{
BooksRepository = booksRepo;
}
public IBooksRepository BooksRepository
{
get;
private set;
}
#region IDisposable Members
public void Dispose()
{
Dispose(true);
GC.SuppressFinalize(this);
}
protected virtual void Dispose(bool disposing)
{
if (disposing == true)
{
entities = null;
}
}
~UnitOfWork()
{
Dispose(false);
}
#endregion
}
Now we have a parameter-less constructor which will be called from controller default constructor i.e.
whenever our page runs. We also have a parameterized constructor which will be created from test
project and passed on to the controllers parameterized constructors.
The dispose pattern is now implemented by the UnitOfWork class because now it is responsible for
creating the ObjectContext so it should be the one disposing it.
Let us look at the implementation of the Controller class now.
public class BookEnController : Controller
{
private UnitOfWork unitOfWork = null;
public BookEnController()
: this(new UnitOfWork())
{
}
public BookEnController(UnitOfWork uow)
{
this.unitOfWork = uow;
}
public ActionResult Index()
{
List<Book> books = unitOfWork.BooksRepository.GetAllBooks();
return View(books);
}
}
Now the test-ablity of this controller is still maintained by having the combination of default
and parameterized constructor. Also, the data access code is now centralized in one place with the possibility
of having multiple repository classes being instantiated at the same time. Let us run the application.
Note: We will not be doing other CRUD operations here because they can be done on same lines very easily.
And finally let us visualize our implementation with Unit of Work in place.
Point of interest
In this article we saw what is Repository and Unit of work pattern. We have also seen a rudimentary implementation
for the same in an ASP.NET MVC application. The next step to the project would be to convert all the
repository classes into one generic repository so that we don't need to create multiple repository classes.
I hope this has been informative.
History
-
12 April 2013: First version.
