Question about Interfaces and DI? - asp.net-mvc

I am using the Service/Repository/EF/POCO pattern in a MVC app, and had a couple questions about the Interfaces.
1) Should I make an Interface per Service?
2) Should I make an Interface per Repository?
Or, should I have a generic interface per layer (IService(Of T), IRepository(Of T)).
What I dont understand is how in the controller say, it takes a IService(Of Category) interface in it's constructor, how do I implements the methods in the concrete class?
Public Class HomeController
Inherits System.Web.Mvc.Controller
Private _Service As IService(Of Category)
Public Sub New(ByVal Service As IService(Of Category))
_Service = Service
End Sub
Function Index() As ActionResult
Return View()
End Function
End Class
The _Service does not have the methods of my concrete CategoryService class?
Make any sense?

Use concrete interface for service. If your services can be described by generic interface you most probably don't need them at all. Generic interface is often used for repositories because repositories usually offer same core methods.

For myself, I use a generic session object that is strongly type, this class is in my domain project witch contains all my domain classes. You should take a look at this post : http://weblogs.asp.net/scottgu/archive/2010/07/16/code-first-development-with-entity-framework-4.aspx using Code-First approach.
Hope it helps!
Here my code for my Session class :
public class EFSession : ISession
{
DbContext _context;
public EFSession(DbContext context)
{
_context = context;
}
public void CommitChanges()
{
_context.SaveChanges();
}
public void Delete<T>(System.Linq.Expressions.Expression<Func<T, bool>> expression) where T : class, new()
{
var query = All<T>().Where(expression);
foreach (var item in query)
{
Delete(item);
}
}
public void Delete<T>(T item) where T : class, new()
{
_context.Set<T>().Remove(item);
}
public void DeleteAll<T>() where T : class, new()
{
var query = All<T>();
foreach (var item in query)
{
Delete(item);
}
}
public void Dispose()
{
_context.Dispose();
}
public T Single<T>(System.Linq.Expressions.Expression<Func<T, bool>> expression) where T : class, new()
{
return All<T>().FirstOrDefault(expression);
}
public IQueryable<T> All<T>() where T : class, new()
{
return _context.Set<T>().AsQueryable<T>();
}
public void Add<T>(T item) where T : class, new()
{
_context.Set<T>().Add(item);
}
public void Add<T>(IEnumerable<T> items) where T : class, new()
{
foreach (var item in items)
{
Add(item);
}
}
/// <summary>
/// Do not use this since we use EF4, just call CommitChanges() it does not do anything
/// </summary>
/// <typeparam name="T"></typeparam>
/// <param name="item"></param>
public void Update<T>(T item) where T : class, new()
{
//nothing needed here
}
}

Related

Should I use singleton for DAL and Service classes?

My dal and service classes are as follows. I use Ninject to inject dependencies.
public interface IEntityRepository<T> where T : class, IEntity, new()
{
ICollection<T> GetAll(Expression<Func<T, bool>> filter = null);
T Get(Expression<Func<T, bool>> filter);
T Add(T entity);
T Update(T entity);
void Delete(T entity);
}
public class EfEntityRepositoryBase<TEntity, TContext> : IEntityRepository<TEntity>
where TEntity : class, IEntity, new()
where TContext : DbContext, new()
{
public virtual ICollection<TEntity> GetAll(Expression<Func<TEntity, bool>> filter = null)
{
using (var context = new TContext())
{
return (filter == null ? context.Set<TEntity>() : context.Set<TEntity>().Where(filter)).ToList();
}
}
public virtual TEntity Get(Expression<Func<TEntity, bool>> filter)
{
using (var context = new TContext())
{
return context.Set<TEntity>().SingleOrDefault(filter);
}
}
public virtual TEntity Add(TEntity entity)
{
using (var context = new TContext())
{
var addedEntity = context.Entry(entity);
addedEntity.State = EntityState.Added;
context.SaveChanges();
return entity;
}
}
public virtual TEntity Update(TEntity entity)
{
using (var context = new TContext())
{
var updatedEntity = context.Entry(entity);
updatedEntity.State = EntityState.Modified;
context.SaveChanges();
return entity;
}
}
public virtual void Delete(TEntity entity)
{
using (var context = new TContext())
{
var deletedEntity = context.Entry(entity);
deletedEntity.State = EntityState.Deleted;
context.SaveChanges();
}
}
}
public interface ICallService
{
}
public class CallManager : ICallService
{
}
public interface ICallDal : IEntityRepository<Call>
{
}
public class EfCallDal : EfEntityRepositoryBase<Call, DatabaseContext>, ICallDal
{
}
public class BusinessModule : NinjectModule
{
public override void Load()
{
Bind<ICallService>().To<CallManager>().InSingletonScope();
Bind<ICallDal>().To<EfCallDal>();
}
}
What are the advantages or disadvantages of using Singleton Scope in dal and service classes? Would it be right to use it according to your experience?
I'm also curious about the dependency injection of the DbContext class.
Bind<DbContext>().To<MyContext>().InSingletonScope();
I think using singleton for the context class is risky. Is not it?
What are the advantages or disadvantages of using Singleton Scope in
dal and service classes?
Advantages :
you instantiate only one object, gain in CPU and memory.
You can share state (which would be a huge disavantage if not controlled)
Disadvantages :
the object graph must be threadsafe (it is not the case of the DbContext)
the objects in the object graph must be stateless, unless you want the state being shared by all your objects
In practice this is not a good idea and it will a big source of problems.
As you seem to be in a web context (Asp.Net MVC), you should bind most of your objects InRequestScope.
Avoid using new DbContext. Your context should be bound and injected as constructor argument. Otherwise you miss the point of Dependency injection.
Once you have understood the mechanics of scoping, you will be able to play with singletons, and scoped factories and the like.

Generic repository pattern and multiple selects

I am trying to learn the repository pattern and looking at a generic repository I cannot see how to handle customized select statements. For example, using this article the author uses a select by ID and a select all.
public interface IGenericRepository<T> where T:class
{
IEnumerable<T> SelectAll();
T SelectByID(object id);
void Insert(T obj);
void Update(T obj);
void Delete(object id);
void Save();
}
Later the article the IGenericRepository interface is implemented using Northwind. Then that is used to create a Customer controller.
public class CustomerController : Controller
{
private IGenericRepository<Customer> repository = null;
public CustomerController()
{
this.repository = new GenericRepository<Customer>();
}
...
This would handle selecting a list of all Customers or for one Customer by ID but where I get stuck is some more real world examples like "select all Customers for a client" or "select all Customers for a region". Plus, you could have another controller based on a different entity that would filter on different attributes. I assume I'm missing something basic. If the user interface needed to present the Customer entity (or any other entity) by various filters, how would this be done by sticking with one generic repository?
Here you go; to handle any select criteria apart from the Id, you can add Where method
like below
public interface IGenericRepository<T> where T:class
{
IEnumerable<T> SelectAll();
T SelectByID(object id);
IEnumerable<T> Where(Expression<Func<T,bool>> predicate)// this one
void Insert(T obj);
void Update(T obj);
void Delete(object id);
void Save();
}
Now in the Where method implementation do it like this
public IEnumerable<T> Where(Expression<Func<T,bool>> predicate)
{
return _objectSet.Where(predicate).AsEnumerable();
}
Here _objectSet in created in repository constructor like this :
public Repository(ObjectContext context)
{
_context = context;
_objectSet = _context.CreateObjectSet<T>();
}
public CustomerController()
{
_context = new NorthwindEntities();
_reporsitory = new Repository<Customer>(_context);
}
Use of Where method like
reporsitory.Where(c=>c.Country=="Canada").ToList();
For full reference see this project on codeplex (download /browse source code)
https://efgenericrepository.codeplex.com/SourceControl/latest
I think the implementation of the GenericRepository should somehow be able to return the IQueryable of current entity, like adding Get() method.
protected IQueryable<T> Get() // Notice that the access modifier is protected.
{
return table;
}
Then you could just create a derived class from GenericRepository and add a select method that accepts the Filter class.
public class CustomerRepository : GenericRepository<Customer>
{
public IEnumerable<T> SelectAll(CustomerFilter filter){ .. }
}
The filter class contains 2 filters.
public class CustomerFilter
{
public int? ClientId { get; set; }
public int? RegionId { get; set; }
}
Then the SelectAll implementation would be.
public IEnumerable<T> SelectAll(CustomerFilter filter)
{
var query = Get();
if (filter == null)
{
return query;
}
if (filter.ClientId.HasValue)
{
query = query.Where(q => q.ClientId == filter.ClientId.Value);
}
if (filter.RegionId.HasValue)
{
query = query.Where(q => q.RegionId == filter.RegionId.Value);
}
return query;
}
In the controller, calling it like.
public ActionResult Get(int? clientId, int? regionId)
{
var filter = new CustomerFilter { ClientId = clientId, RegionId = regionId };
var customers = _repository.SelectAll(filter);
return View();
}
You might need to see this post as your reference.
An approach I've seen in one asp.net mvc based mission critical app, is to use the generic interface as defined in the question. Then there is an abstract class that implements that interface. And there is one more repository class that inherits the abstract class, which has all methods specific to that class.
public interface IGenericRepository<T> where T:class
{
...
}
public abstract class GenericRepository<T> : IGenericRepository where T:class
{
...
}
And the CustomerRepository class
public class CustomerRepository : GenericRepository<Customer>
{
//add method specific to Customer like select Customers in specific country
}
And in the controller
public class CustomerController : Controller
{
private CustomerRepository repository = null;
public CustomerController()
{
this.repository = new CustomerRepository();
}
...

Ninject : Constructor parameter

I am using Ninject together with ASP.NET MVC 4. I am using repositories and want to do constructor injection to pass in the repository to one of the controllers.
This is my Repository interface:
public interface IRepository<T> where T : TableServiceEntity
{
void Add(T item);
void Delete(T item);
void Update(T item);
IEnumerable<T> Find(params Specification<T>[] specifications);
IEnumerable<T> RetrieveAll();
void SaveChanges();
}
The AzureTableStorageRepository below is an implementation of IRepository<T>:
public class AzureTableRepository<T> : IRepository<T> where T : TableServiceEntity
{
private readonly string _tableName;
private readonly TableServiceContext _dataContext;
private CloudStorageAccount _storageAccount;
private CloudTableClient _tableClient;
public AzureTableRepository(string tableName)
{
// Create an instance of a Windows Azure Storage account
_storageAccount = CloudStorageAccount.Parse(ConfigurationManager.ConnectionStrings["StorageConnectionString"].ConnectionString);
_tableClient = _storageAccount.CreateCloudTableClient();
_tableClient.CreateTableIfNotExist(tableName);
_dataContext = _tableClient.GetDataServiceContext();
_tableName = tableName;
}
Note the tableName parameter needed because I was using a generic table repository to persist data to Azure.
And finally I have the following controller.
public class CategoriesController : ApiController
{
static IRepository<Category> _repository;
public CategoriesController(IRepository<Category> repository)
{
if (repository == null)
{
throw new ArgumentNullException("repository");
}
_repository = repository;
}
Now I want to inject a repository into the controller. So I have created a module that contains the bindings:
/// <summary>
/// Ninject module to handle dependency injection of repositories
/// </summary>
public class RepositoryNinjectModule : NinjectModule
{
public override void Load()
{
Bind<IRepository<Category>>().To<AzureTableRepository<Category>>();
}
}
The loading of the module is done in the NinjectWebCommon.cs
/// <summary>
/// Creates the kernel that will manage your application.
/// </summary>
/// <returns>The created kernel.</returns>
private static IKernel CreateKernel()
{
var kernel = new StandardKernel();
kernel.Bind<Func<IKernel>>().ToMethod(ctx => () => new Bootstrapper().Kernel);
kernel.Bind<IHttpModule>().To<HttpApplicationInitializationHttpModule>();
RegisterServices(kernel);
return kernel;
}
/// <summary>
/// Load your modules or register your services here!
/// </summary>
/// <param name="kernel">The kernel.</param>
private static void RegisterServices(IKernel kernel)
{
// Load the module that contains the binding
kernel.Load(new RepositoryNinjectModule());
// Set resolver needed to use Ninject with MVC4 Web API
GlobalConfiguration.Configuration.DependencyResolver = new NinjectResolver(kernel);
}
The DependencyResolver was created because Ninject's DependencyResolver implements System.Web.Mvc.IDependencyResolver and this cannot be assigned to GlobalConfiguration.Configuration of the WebApi Application.
So with all this in place, the Ninject part is actually injecting the right type in the Controller but Ninject cannot inject the tableName parameter in the constructor of AzureTableRepository.
How would I be able to do this in this case? I have consulted a lot of articles and the ninject documentation to see how I could use parameters, but I cannot seem to get it working.
Any help would be appreciated.
I'd use the WithConstructorArgument() method like...
Bind<IRepository<Category>>().To<AzureTableRepository<Category>>()
.WithConstructorArgument("tableName", "categories");
The rest of the repository design is probably another question. IMHO It seems like a big no no to create a table or do any heavy lifting in a ctor.
Meanwhile I have been playing around with Providers to try and do the trick and it seems to work.
I don't know if this is good idea or if it is overkill but here is what I have done:
I have created a generic provider class:
public abstract class NinjectProvider<T> : IProvider
{
public virtual Type Type { get; set; }
protected abstract T CreateInstance(IContext context);
public object Create(IContext context)
{
throw new NotImplementedException();
}
object IProvider.Create(IContext context)
{
throw new NotImplementedException();
}
Type IProvider.Type
{
get { throw new NotImplementedException(); }
}
}
And then I implemented that one in the AzureTableRepositoryProvider. (T to support having the same repository for multiple entity types.)
public class AzureTableRepositoryProvider<T> : Provider<AzureTableRepository<T>> where T : TableServiceEntity
{
protected override AzureTableRepository<T> CreateInstance(IContext context)
{
string tableName = "";
if (typeof(T).Name == typeof(Category).Name)
{
// TODO Get the table names from a resource
tableName = "categories";
}
// Here other types will be addedd as needed
AzureTableRepository<T> azureTableRepository = new AzureTableRepository<T>(tableName);
return azureTableRepository;
}
}
By using this provider I can pass in the right table name for the repository to work with. But for me, two questions remain:
Is this good practice or could we do things much simpler?
In the NinjectProvider class I have two notImplementedException cases. How could I resolve these? I used sample code from the following link but that does not work as the Provider is abstract and the code does not have a body for the create method... enter link description here

Ninject dependency injection into ASP.NET MVC controller where repository type is known only at runtime

I've set up DI with Ninject in my ASP.NET MVC application like this
Bind<IRepository>().To<XmlDefaultRepository>().WhenInjectedInto(typeof(PageController)).WithConstructorArgument("contentType", ContentType.Page);
Bind<IRepository>().To<XmlDefaultRepository>().WhenInjectedInto(typeof(WidgetController)).WithConstructorArgument("contentType", ContentType.Page);
Bind<IRepository>().To<XmlDefaultRepository>().WhenInjectedInto(typeof(SectionController)).WithConstructorArgument("contentType", ContentType.Section);
Bind<IRepository>().To<XmlDefaultRepository>().WhenInjectedInto(typeof(WidgetZoneController)).WithConstructorArgument("contentType", ContentType.WidgetZone);
XmlDefaultRepository implements IRepository and it contains a constructor that takes contentType parameter which I use for persistance path generation.
Now I have a ServicesController which does not have a default type by itself - it is just a controller that provides JSON data (JSON actions) to consumers from jQuery.
This is how it looks now:
public class ServicesController : ContentController
{
public ActionResult ContentSlugs(string contentType, string q, int limit)
{
IRepository _repository = new XmlDefaultRepository(contentType); // this instantiation depends on contentType provided by JSON GET request at runtime and I want to somehow replace it with DI
return Json(_repository.GetSlugsForContentType(limit, q), JsonRequestBehavior.AllowGet);
}
}
And this is how other controllers look like (DI is done thru constructor injection here):
public class SectionController : ContentController
{
private IRepository ContentRepository;
public SectionController(IRepository repository)
{
ContentRepository = repository;
}
}
How can I get rid of "new XmlDefaultRepository(contentType)" dependency in ServicesController?
I've solved this problem by implementing method
_repository.GetSlugsForContentType(contentType, limit, q)
and creating a parameterless default constructor on XmlDefaultRepository implementation of IRepository.
This is the DI way of solving this issue with a Repository factory.
// ServicesController.cs
// ninject factory example, see IRepositoryFactory interface and its XmlRepositoryFactory implementation
[Inject]
public IRepositoryFactory RepositoryFactory
{
set
{
factory = value;
}
}
private IRepositoryFactory factory;
public ActionResult ContentSlugsThruFactory(string contentType, string q, int limit)
{
IRepository _repository = factory.Create(contentType);
return Json(_repository.GetSlugsForContentType(limit, q), JsonRequestBehavior.AllowGet);
}
// IRepositoryFactory.cs
public interface IRepositoryFactory
{
IRepository Create(string contentType);
}
// XmlRepositoryFactory.cs
public class XmlRepositoryFactory : IRepositoryFactory
{
public IRepository Create(string contentType)
{
return XmlDefaultRepository.Create(contentType);
}
}
// XmlDefaultRepository.cs
public static XmlDefaultRepository Create(ContentType contentType)
{
return new XmlDefaultRepository(contentType);
}
public static XmlDefaultRepository Create(string contentType)
{
return new XmlDefaultRepository(contentType);
}
// global.asax.cs
Bind<IRepositoryFactory>().To<XmlRepositoryFactory>().InSingletonScope();

Ninject And Connection Strings

I am very new to Ninject and am trying Ninject 2 with MVC and Linq. I have a SqlProductRepository class and all I want to know is what's the best way of passing the connectionstring in the constructor if I am injecting the Repository object in the controller.
public class SqlProductRepository:IProductRepository
{
private Table<Product> productsTable;
public SqlProductRepository(string connectionString)
{
productsTable = (new DataContext(connectionString)).GetTable<Product>();
}
public IQueryable<Product> Products
{
get { return productsTable; }
}
}
This is my ProductController class where I am injecting the Repository:
public class ProductsController : Controller
{
private int pageSize = 4;
public int PageSize { get { return pageSize; } set { pageSize = value; } }
IProductRepository _productsRepository;
[Inject]
public ProductsController(IProductRepository productRepository)
{
_productsRepository = productRepository;
}
public ViewResult List(int page)
{
return View(_productsRepository.Products
.Skip((page - 1) * pageSize)
.Take(pageSize)
.ToList()
);
}
}
Can somebody please guide me regarding this?
You can set it up in your binding
_kernel.Bind<IProductRepository>()
.To<SqlProductRepository>()
.WithConstructorArgument("connectionString",yourConnectionString );
You're doing:
new DataContext(connectionString)
in your code - this is the very newing and binding to classes you're trying to push out of your code by using a DI container. At the very least, consider adding an IConnectionStringSelector interface or something like that. You dont want to have 20 Bind calls for 20 repositories - you want a higher level abstraction than that.
I'd suggest the best solution is that you should be demanding either an IDataContext or an IDataContextFactory in the constructor instead and letting that worry about it.
You could supply the connection string as a constructor argument when binding the SqlProductRepository to the IProductRepository interface.
public class LinqToSqlModule : NinjectModule
{
public override void Load()
{
Bind<IProductRepository>().To<SqlProductRepository>()
.WithConstructorArgument(connectionString, "connectionstring");
}
}
I would suggest a slightly different approach. First of all, you might want to create a binding for the DataContext class in the kernel. You could do so by using a provider class to create your DataContext passing the connection string as an argument to its constructor. Then you bind the DataContext to the DataContextProvider.
public class DataContextProvider : Provider<DataContext>
{
protected override DataContext CreateInstance(IContext context)
{
string connectionString = "connectionstring";
return new DataContext(connectionString);
}
}
public class LinqToSqlModule : NinjectModule
{
public override void Load()
{
Bind<DataContext>().ToProvider<DataContextProvider>();
Bind<IProductRepository>().To<SqlProductRepository>();
}
}
Next modify the constructor of SqlProductRepository class to accept a DataContext object instead.
public class SqlProductRepository : IProductRepository
{
private readonly DataContext context;
public ProductRepository(DataContext context)
{
this.context = context;
}
public IQueryable<Product> Products
{
get { return context.GetTable<Product>(); }
}
}
By the way you don't have to decorate your constructor with the Inject attribute. Ninject will select the constructor with the most parameters by default.
Please refer below code snap:
//Bind the default connection string
public void BindDataContext()
{
ConstructorArgument parameter = new ConstructorArgument("connectionString", "[Config Value]");
Bind<DataContext>().ToSelf().InRequestScope().WithParameter(parameter);
}
//Re-Bind the connection string (in case of multi-tenant architecture)
public void ReBindDataContext(string cn)
{
ConstructorArgument parameter = new ConstructorArgument("connectionString", cn);
Rebind<DataContext>().ToSelf().InRequestScope().WithParameter(parameter);
}
For more information, please visit below link
MVC3, Ninject and Ninject.MVC3 problem

Resources