Repository Pattern in Entity framework 4 when should we dispose? - entity-framework-4

New to EF and I have noticed that using a repository pattern can really simplify things and will allow me to do some mocking too.So far so good.
My Question
A typical usage of the objectContext is to destroy as soon as possible see below
using (var context = new SchoolEntities())
{
context.AddToDepartments(department);
context.SaveChanges();
}
Using the Repository pattern I have noticed that nobody actually uses the "Using Pattern" eg
using (var repository= new Repository<Student>(new MyContext))
{
repository.Add(myStudentEntity)
repository.SaveChanges();
}
Should the idea be that we should dispose of the context as soon as possible otherwise the memory might leak or get very big?
Can anyone clarify? Thanks a lot.

Yes you should dispose context even if you are using repository. It is not clear what advantage does your Repository implementation give you because you are still providing ObjectContext as constructor's parameter, aren't you?
IMO the main reason for using Repository and custom UnitOfWork is persistance ignorance = hidding EF code from upper application layers because ObjectContext + ObjectSet themselves are implementation of repository and unit of work patterns.
If I'm using repository, I'm always wrapping whole EF code, so the public interface of my repository doesn't provide any information about EF related infrastructure. In that case it is up to me how I deal with ObjectContext.
For easy straight forward CRUD scenarios, I can wrap context creation and disposing into each repository method. In more complex scenarios I'm using additional class - UnitOfWork (UoW), which wraps context creation and disposing and it triggers saving changes into database. It also acts as factory for all repositories and passes instance of created context into repositories' constructors.
Most of the time I'm programming services or web applications so I'm dealing with detached objects. I'm always using single UoW instance for request processing. So the UoW is created at the beginning of request processing and released at the end of request processing. In case of WinForms / WPF applications and attached objects I think the good idea is to have UoW / ObjectContext instance "per form" - there is article describing this approach with NHibernate session (same as EF ObjectContext) in MSDN magazine.
Some starting implementation of UnitOfWork and Repository patterns:
Context holder and abstract factory for repositories
public interface IUnitOfWork
{
IRepository<MyEntity> MyEntityRepository { get; }
// Repositories for other entities
SaveChanges();
}
Repository for detached entities
public interface IRepository<T> where T : class
{
IQueryable<T> GetQuery();
void Insert(T entity);
void Delete(T entity);
// In very complex scenarios with big object graphs you will probably give up
// using detached approach and you will always load your entities from DB before
// deleting or updating them. In such case you will not need Update method at all.
void Update(T entity);
}
Disposable implementation of UnitOfWork wrapping Enitity framework
public class UnitOfWork : IUnitOfWork, IDisposable
{
private ObjectContext _context = null;
public UnitOfWork(string connectionString)
{
if (String.IsNullOrEmpty(connectionString)) throw new ArgumentNullException("connectionString");
_context = new ObjectContext(connectionString);
}
private IRepository<MyEntity> _myEntityRepository;
public IRepository<MyEntity> MyEntityRepository
{
get
{
return _myEntityRepository ?? (_myEntityRepository = new GeneralRepository<MyEntity>(_context));
}
}
public void SaveChanges()
{
_context.SaveChanges();
}
public void Dispose()
{
Dispose(true);
GC.SuppressFinalize(this);
}
protected virtual void Dispose(bool disposing)
{
if (disposing)
{
if (_context != null)
{
_context.Dispose();
_context = null;
}
}
}
}
Base repository implementation
public class GeneralRepository<T> : IRepository<T> where T : class
{
private ObjectSet<T> _set;
private ObjectContext _context;
public GeneralRepository(ObjectContext context)
{
if (context == null) throw new ArgumentNullException("context");
_context = context;
_set = context.CreateObjectSet<T>();
}
// Override this method for example if you need Includes
public virtual IQueryable<T> GetQuery()
{
return _set;
}
// Override following methods if you are working with object graphs.
// Methods do not execute operations in database. It is responsibility of
// UnitOfWork to trigger the execution
public virtual void Insert(T entity)
{
if (entity == null) throw new ArgumentNullException("entity");
_set.AddObject(entity);
}
// These impelementations are for detached scenarios like web application
public virtual void Delete(T entity)
{
if (entity == null) throw new ArgumentNullException("entity");
_set.Attach(entity);
_set.DeleteObject(entity);
}
public virtual void Update(T entity)
{
if (entity == null) throw new ArgumentNullException("entity");
_set.Attach(entity);
_context.ObjectStateManager.ChangeObjectState(entity, EntityState.Modified);
}
}
Usage when selecting data
using (var uow = new UnitOfWork(connectionString))
{
var entity = uow.MyEntitiesRepository.GetQuery().Single(e => e.Id == 1);
// Do something with entity
}
Usage when modifing data
using (var uow = new UnitOfWork(connectionString))
{
uow.MyEntitiesRepository.Update(entity);
uow.SaveChanges();
}

Related

EntityFramework DbContext in static method per request

How Entity Framework DbContext works in static method per web request? is there any performance issue? I have 2 classes below . Which one is better to use ? or bad? Do we need get instance all pear request ? is there any issue?
public class MyClassA
{
public static Product GetProduct(int Id)
{
using(MyContext myContext = new MyContext())
return myContext.Products.Where(x = > x.Id == Id).SingleOrDefault();
}
}
public class MyClassB
{
MyContext myContext = new MyContext()
public static Product GetProduct(int Id)
{
return myContext.Products.Where(x = > x.Id == Id).SingleOrDefault();
}
}
Call in Controller
public class DefaultController : ControllerBase
{
public Product GetProductStaticMethodinMyClassA(int Id)
{
return MyClassA.GetProduct(Id);
}
public Product GetProductStaticMethodinMyClassB(int Id)
{
return MyClassB.GetProduct(Id);
}
public Product GetProductinRequlurUse(int Id)
{
MyContext myContext = new MyContext();
return myContext.Products.Where(x => x.Id == Id).SingleOrDefault();
}
}
You need to create one DbContext per HTTP request... i.e. do not share the same DbContext between multiple requests.
C# Garbage Collector would automatically dispose of DbContext once it goes out of scope, so you don't have to put it in a using block, but having said that, most people do use using block and MS seems to encourage it:
The lifetime of the context begins when the instance is created and
ends when the instance is either disposed or garbage-collected. Use
using if you want all the resources that the context controls to be
disposed at the end of the block. When you use using, the compiler
automatically creates a try/finally block and calls dispose in the
finally block.
public void UseProducts()
{
using (var context = new ProductContext())
{
// Perform data access using the context
}
}
About usage of DbContext in your controller, your question is about design and opinion based... if you follow DDD principles, DbContext goes into Infrastructure Layer and your Controllers belongs to Presentation Layers... so you would not use DbContext directly in the controller at the first place.
If you want to use DbContext inside your controller, it is better to inject it, than to initialize it in the controller, initializing DbContext inside a controller violates SRP, because it is not controller's concern to initialize DbContext.

Entity Framework DbContext Lifetime in ASP.NET MVC Using Ninject?

I have the following unit of work pattern set up for an MVC 5 application using Entity Framework. The unit of work has all the repos defined as follows so that they are all using the same dbcontext and it has one save method to co-ordinate the transaction using the same context:
public class UnitOfWork : IUnitOfWork
{
private readonly ApplicationDbContext _context;
public IProductRepository ProductRepository { get; private set; }
public ICustomerRepository CustomerRepository { get; private set; }
// Other reposistories
public UnitOfWork(ApplicationDbContext context)
{
_context = context;
ProductRepository = new ProductRepository(_context);
CustomerRepository = new CustomerRepository(_context);
// Other reposistories
}
public void Complete()
{
_context.SaveChanges();
}
}
This is an example of my repo. The reason for using repos is for code re-use so that I'm not duplicating queries inside different controllers.
public class ProductRepository : IProductRepository
{
private readonly ApplicationDbContext _context;
public ProductRepository(ApplicationDbContext context)
{
_context = context;
}
public Product GetProduct(int productId)
{
return _context.Ticket.SingleOrDefault(p => p.Id == productId);
}
public void Add(Product product)
{
_context.Product.Add(product);
}
// Other methods
}
I inject the unit of work class in my controller as follows using Ninject:
public class ProductsController : Controller
{
private readonly IUnitOfWork _unitOfWork;
private readonly IFileUploadService _FileUploadService;
public ProductsController(IUnitOfWork unitOfWork,
IFileUploadService fileUploadService)
{
_unitOfWork = unitOfWork;
_FileUploadService = fileUploadService;
}
[HttpPost]
[ValidateAntiForgeryToken]
public ActionResult Create(CreateEditProductViewModel viewModel)
{
var product = new Product
{
// Do stuff
};
_unitOfWork.ProductRepository.Add(product);
// Call file upload service
_fileUploadService.Upload();
_unitOfWork.Complete();
}
}
This unit of work set up works fine if all I'm using are repos that are defined in the unit of work class. But now I want to use a service class to process some additional application logic and then the unit of work is committed in the controller action. If I define the class as follows it will be using a different instance of the context, In which case how would you co-ordinate a transaction where the service layers is ending up with a different context?
public class FileUploadService : IFileUploadService
{
private readonly IUnitOfWork _unitOfWork;
public FileUploadService(IUnitOfWork unitOfWork)
{
_unitOfWork = unitOfWork;
}
public uploadResult Upload()
{
// Some stuff
var uploadedFile = new UploadedFile
{
//some stuff
};
_unitOfWork.UploadedFileRepository.Add(uploadedFile);
}
}
I've done quite a bit of research online and I'm unable to find any resource that provides a practical example to solve this problem. I've read quite a bit of stuff on ditching unit of work and repos and simply using entity frameworks dbset. However as explained above the purpose of using
repos is to consolidate queries. My questions is how do I co-ordinate the unit of work with a service class.
I would like the service to use the same context so that it can access the repositories it needs to work with, and let the controller (client code) commit the operation when it see fits.
* UPDATE *
In my DI Container I resolve all interfaces using the following snippet:
private static IKernel CreateKernel()
{
RegisterServices(kernel);
kernel.Bind<IUnitOfWork>().To<UnitOfWork>().InRequestScope();
// default binding for everything except unit of work
kernel.Bind(x => x.FromAssembliesMatching("*")
.SelectAllClasses()
.Excluding<UnitOfWork>()
.BindDefaultInterface());
return kernel;
}
Would adding the line kernel.Bind<IUnitOfWork>().To<UnitOfWork>().InRequestScope(); ensure that no more than one ApplicationDbContext is created, even if the request ends up hitting multiple controllers or service layers that all require an IUnitOfWork (ApplicationDbContext)?
If you are using MVC, then your unit of work is your web request. If I were you I'd ditch the UOW implementation and just make sure you dbcontext is instantiated in the Application_BeginRequest. Then I'd stuff it into the HttpContext for safe keeping. On Application_EndRequest, I dispose of the DbContext.
I would move the save to your repository.
I'd create a [Transaction] attribute that would maintain a TransactionScope something like this:
[AttributeUsage(AttributeTargets.Method | AttributeTargets.Class)]
public class TransactionAttribute : ActionFilterAttribute
{
private TransactionScope Transaction { get; set; }
public override void OnActionExecuting(ActionExecutingContext filterContext)
{
Transaction = new TransactionScope( TransactionScopeOption.Required);
}
public override void OnActionExecuted(ActionExecutedContext filterContext)
{
if (filterContext.Exception == null)
{
Transaction.Complete();
return;
}
Transaction.Dispose();
}
}
You can then just tag your controller methods with [Transaction].
I'm just spitballing here, but I do something similar with NHibernate instead of EF and it works out nicely for me.
The InRequestScope() will create a new instance of the bound type on every new web request, and at the end of that web request, it will Dispose that instance if it is disposable.
I am not sure how are you passing the ApplicationDbContext into your UnitOfWork. I am assuming that you use Ninject for this injection too. Just make sure that you bind your ApplicationDbContext using the InRequestScope()Bind.To().InRequestScope();.
This way, your ApplicationDbContext instance will be created once per request and disposed at the end.
Also, the use of InRequestScope is for types that are disposable, so you can also release resoruces in the Dispose method of your UnitOfWork method too.

MVC Repository with Unit Of Work, Automapper and Generic Repository

I've been looking at a few blog posts to try and create an appropriate solution for the following requirements but I can't seem to piece them together. Hope fully someone can help.
I've been using Repository pattern with interfaces using Automapper...here's a trimmed down example:
public class BookingRepository : IBookingRepository
{
Entities context = new Entities();
public IEnumerable<BookingDto> GetBookings
{
get { return Mapper.Map<IQueryable<Booking>, IEnumerable<BookingDto>>(context.Bookings); }
}
public BookingDto GetBookingWithProduct(Guid bookingId)
{
return Mapper.Map<BookingDto>(context.Bookings.Include(c => c.Products).SingleOrDefault(c => c.BookingId == bookingId));
}
public void Update(BookingDto bookingDto)
{
var booking = Mapper.Map<Booking>(bookingDto);
context.Entry(booking).State = EntityState.Modified;
}
public void Save()
{
context.SaveChanges();
}
public void Dispose()
{
context.Dispose();
}
}
public interface IBookingRepository : IDisposable
{
IEnumerable<BookingDto> GetBookings { get; }
BookingDto GetBooking(Guid bookingId);
void Update(BookingDto bookingDto);
void Save();
}
With a seperate Repository for a different Entity, for example
public class ProductRepository : IProductRepository
{
Entities context = new Entities();
public IEnumerable<ProductDto> GetProducts
{
get { return Mapper.Map<IQueryable<Product>, IEnumerable<ProductDto>>(context.Products); }
}
public ProductDto GetProductWithDesign(int productId)
{
return Mapper.Map<ProductDto>(context.Products.Include(c => c.Designs).SingleOrDefault(c => c.ProductId == productId));
}
public void Update(ProductDto productDto)
{
var product = Mapper.Map<Product>(productDto);
context.Entry(product).State = EntityState.Modified;
}
public void Save()
{
context.SaveChanges();
}
public void Dispose()
{
context.Dispose();
}
}
public interface IProductRepository : IDisposable
{
IEnumerable<ProductDto> GetProducts { get; }
ProductDto GetProduct(int productId);
void Update(ProductDto productDto);
void Save();
}
Then in my Controller I'm using the repositories as so:
public class HomeController : Controller
{
private readonly IBookingRepository bookingRepository;
private readonly IProductRepository productRepository;
public HomeController() : this(new BookingRepository(), new ProductRepository()) { }
public HomeController(IBookingRepository bookingRepository, IProductRepository productRepository)
{
this.bookingRepository = bookingRepository;
this.productRepository = productRepository;
}
protected override void Dispose(bool disposing)
{
base.Dispose(disposing);
if (disposing && this.bookingRepository != null)
this.bookingRepository.Dispose();
if (disposing && this.productRepository != null)
this.productRepository.Dispose();
}
}
So now I'm hoping to create a Unit Of Work to abstract these repositories and share the context and also create a generic repository for the duplicated actions (Save and Update) bearing in mind I'm passing in Dtos and Mapping to Entity objects. I'm having difficulty understanding how to knit it all together.
Additionally, I've seen this post
Repository pattern with generics and DI
which states "You should not have other repository interfaces besides your generic repository" and that custom queries "deserve their own (generic) abstraction:" which is adding another complication to my overworked brain as my repositories will have custom queries that return complex linked objects using Include Statements as Lazy Loading is disabled.
So I'm prepared to be shot down and told that I'm going about this the wrong way but would be grateful for any direction given.
Thanks in advance.
Don't use generic repositories. They are all leaky abstractions. Ask yourself, what benefit to you get by using an abstraction that doesn't really abstract away something? You could use your OR/M directly in those cases.
What I means is that anything that exposes IQueryable<T> forces the user to learn about the weaknesses that the underlying OR/M has. Examples: How do the orm handle lazy loading? How do I eagerly load related entities? How do I create a IN clause?
If you truly want to use the repository pattern either use it together with the specification pattern (you can keep on using a generic repository then) or create repositories that are specific for each root aggregate.
I've blogged about it: http://blog.gauffin.org/2013/01/repository-pattern-done-right/
What I usually do in this case is to create a Base abstract Repository class like this:
public abstract class BaseRepository<T> : IRepository<T>
{
Entities context = new Entities();
public virtual T GetAll()
{
return context.Set<T>();
}
// Add base implementation for normal CRUD here
}
If you don't need special queries then you don't need to create special interface and classes (but you can of course, to improve readability). So you will use, for example:
var bookingsRepo = new BaseRepository<BookingsDto>();
var allBookings = bookingsRepo.GetAll();
If you need some special queries, you create an interface that extends the base interface:
public interface IProductRepository : IRepository<Product>
{
Product GetSpecialOffer();
}
Then create your class:
public class ProductRepository : BaseRepository<Product>, IProductRepository
{
public Product GetSpecialOffer()
{
// your logic here
}
}
That way you only specify a minimal number of special cases while relying on the Base abstract implementation for all things normal.
I added virtual to the base methods because I always like to give derived class the ability to override stuff...

Inject a dependency into a custom model binder and using InRequestScope using Ninject

I'm using NInject with NInject.Web.Mvc.
To start with, I've created a simple test project in which I want an instance of IPostRepository to be shared between a controller and a custom model binder during the same web request. In my real project, I need this because I'm getting IEntityChangeTracker problems where I effectively have two repositories accessing the same object graph. So to keep my test project simple, I'm just trying to share a dummy repository.
The problem I'm having is that it works on the first request and that's it. The relevant code is below.
NInjectModule:
public class PostRepositoryModule : NinjectModule
{
public override void Load()
{
this.Bind<IPostRepository>().To<PostRepository>().InRequestScope();
}
}
CustomModelBinder:
public class CustomModelBinder : DefaultModelBinder
{
[Inject]
public IPostRepository repository { get; set; }
public override object BindModel(ControllerContext controllerContext, ModelBindingContext bindingContext)
{
repository.Add("Model binder...");
return base.BindModel(controllerContext, bindingContext);
}
}
public class HomeController : Controller
{
private IPostRepository repository;
public HomeController(IPostRepository repository)
{
this.repository = repository;
}
public ActionResult Index(string whatever)
{
repository.Add("Action...");
return View(repository.GetList());
}
}
Global.asax:
protected override void OnApplicationStarted()
{
AreaRegistration.RegisterAllAreas();
RegisterGlobalFilters(GlobalFilters.Filters);
RegisterRoutes(RouteTable.Routes);
ModelBinders.Binders.Add(typeof(string), kernel.Get<CustomModelBinder>());
}
Doing it this way is actually creating 2 separate instances of IPostRepository rather than the shared instance. There's something here that I'm missing with regards to injecting a dependency into my model binder. My code above is based on the first setup method described in the NInject.Web.Mvc wiki but I have tried both.
When I did use the second method, IPostRepository would be shared only for the very first web request, after which it would default to not sharing the instance. However, when I did get that working, I was using the default DependencyResolver as I couldn't for the life of me figure out how to do the same with NInject (being as the kernel is tucked away in the NInjectMVC3 class). I did that like so:
ModelBinders.Binders.Add(typeof(string),
DependencyResolver.Current.GetService<CustomModelBinder>());
I suspect the reason this worked the first time only is because this isn't resolving it via NInject, so the lifecycle is really being handled by MVC directly (although that means I have no idea how it's resolving the dependency).
So how do I go about properly registering my model binder and getting NInject to inject the dependency?
The ModelBinders are reused by MVC for multiple requests. This means they have a longer lifecycle than request scope and therefore aren't allowed to depend on objects with the shorter request scope life cycle.
Use a Factory instead to create the IPostRepository for every execution of BindModel
It's actually really simple to get the Ninject factory extension up and running, but that wasn't clear to me from the existing answers.
The factory extensions plugin is a prerequisite, which can be installed via NUGet:
Install-Package Ninject.Extensions.Factory
You just need the factory injected into your model binder somewhere, eg:
private IPostRepositoryFactory _factory;
public CustomModelBinder(IPostRepositoryFactory factory) {
_factory = factory;
}
Then create an interface for the factory. The name of the factory and the name of the method doesn't actually matter at all, just the return type. (Good to know if you want to inject an NHibernate session but don't want to have to worry about referencing the correct namespace for ISessionFactory, also useful to know if GetCurrentRepository makes what it actually does more clear in context):
public interface IPostRepositoryFactory {
IPostRepository CreatePostRepository();
}
Then, assuming your IPostRepository is already being managed by Ninject correctly, the extension will do everything else for you just by calling the .ToFactory() method.
kernel.Bind<IPostRepository().To<PostRepository>();
kernel.Bind<IPostRepositoryFactory>().ToFactory();
Then you just call your factory method in the code where you need it:
var repo = _factory.CreatePostRepository();
repo.DoStuff();
(Update: Apparently naming your factory function GetXXX will actually fail if the service doesn't already exist in the session. So you do actually have to be somewhat careful with what you name the method.)
I eventually managed to solve it with a factory as suggested. However, I just could not figure out how to accomplish this with Ninject.Extensions.Factory which is what I would've preferred. Here is what I ended up with:
The factory interface:
public interface IPostRepositoryFactory
{
IPostRepository CreatePostRepository();
}
The factory implementation:
public class PostRepositoryFactory : IPostRepositoryFactory
{
private readonly string key = "PostRepository";
public IPostRepository CreatePostRepository()
{
IPostRepository repository;
if (HttpContext.Current.Items[key] == null)
{
repository = new PostRepository();
HttpContext.Current.Items.Add(key, repository);
}
else
{
repository = HttpContext.Current.Items[key] as PostRepository;
}
return repository;
}
}
The Ninject module for the factory:
public class PostRepositoryFactoryModule : NinjectModule
{
public override void Load()
{
this.Bind<IPostRepositoryFactory>().To<PostRepositoryFactory>();
}
}
The custom model binder:
public class CustomModelBinder : DefaultModelBinder
{
private IPostRepositoryFactory factory;
public CustomModelBinder(IPostRepositoryFactory factory)
{
this.factory = factory;
}
public override object BindModel(ControllerContext controllerContext, ModelBindingContext bindingContext)
{
IPostRepository repository = factory.CreatePostRepository();
repository.Add("Model binder");
return base.BindModel(controllerContext, bindingContext);
}
}
The controller:
public class HomeController : Controller
{
private IPostRepository repository;
public HomeController(IPostRepositoryFactory factory)
{
this.repository = factory.CreatePostRepository();
}
public ActionResult Index(string whatever)
{
repository.Add("Action method");
return View(repository.GetList());
}
}
Global.asax to wire up the custom model binder:
protected override void OnApplicationStarted()
{
AreaRegistration.RegisterAllAreas();
RegisterGlobalFilters(GlobalFilters.Filters);
RegisterRoutes(RouteTable.Routes);
ModelBinders.Binders.Add(typeof(string), kernel.Get<CustomModelBinder>());
}
Which in my view, gave me the desired output of:
Model binder
Action method

NHibernateUnitOfWork + ASP.Net MVC

I'm in my first time with DDD, so I'm begginer! So, let's take it's very simple :D
I developed an application using asp.net mvc 2 , ddd and nhibernate. I have a domain model in a class library, my repositories in another class library, and an asp.net mvc 2 application. My Repository base class, I have a construct that I inject and dependency (my unique ISessionFactory object started in global.asax), the code is:
public class Repository<T> : IRepository<T>
where T : Entidade
{
protected ISessionFactory SessionFactory { get; private set; }
protected ISession Session
{
get { return SessionFactory.GetCurrentSession(); }
}
protected Repository(ISessionFactory sessionFactory)
{
SessionFactory = sessionFactory;
}
public void Save(T entity)
{
Session.SaveOrUpdate(entity);
}
public void Delete(T entity)
{
Session.Delete(entity);
}
public T Get(long key)
{
return Session.Get<T>(key);
}
public IList<T> FindAll()
{
return Session.CreateCriteria(typeof(T)).SetCacheable(true).List<T>();
}
}
And After I have the spefic repositories, like this:
public class DocumentRepository : Repository<Domain.Document>, IDocumentRepository
{
// constructor
public DocumentRepository (ISessionFactory sessionFactory) : base(sessionFactory)
{ }
public IList<Domain.Document> GetByType(int idType)
{
var result = Session.CreateQuery("from Document d where d.Type.Id = :IdType")
.SetParameter("IdType", idType)
.List<Domain.Document>();
return result;
}
}
there is not control of transaction in this code, and it's working fine, but, I would like to make something to control this repositories in my controller of asp.net mvc, something simple, like this:
using (var tx = /* what can I put here ? */) {
try
{
_repositoryA.Save(objA);
_repositoryB.Save(objB);
_repositotyC.Delete(objC);
/* ... others tasks ... */
tx.Commit();
}
catch
{
tx.RollBack();
}
}
I've heared about NHibernateUnitOfWork, but i don't know :(, How Can I configure NHibernateUnitOfWork to work with my repositories ? Should I change the my simple repository ? Sugestions are welcome!
So, thanks if somebody read to here! If can help me, I appretiate!
PS: Sorry for my english!
bye =D
Session is NHibernate's unit of work. But you can always create your own abstraction of it.
using (var tx = Session.BeginTransaction) { ...
There is an excellent library called NCommon (source) that provides a great UnitOfWork implementation built right in. Version 1.1 allows you to do something like:
public class Foo
{
private readonly IRepository<Stuff> _repository;
public Foo(IRepository<Stuff> repository)
{
_repository = repository;
}
public void DoSomething()
{
using (var scope = new UnitOfWorkScope())
{
_repository.Save(a);
scope.Commit();
}
}
}
It integrates with the latest NHibernate and even uses NHibernate.Linq to provide some powerful querying features. You have little or nothing to build yourself and it works great out of the box.
Edit:
I elaborated on my example to show the full recommended way to use NCommon in a project with dependency-injection.
You can make the Session on your Repository a public property. Then you can do the following:
using(var tx = _repository.Session.BeginTransaction())
On a somewhat related note, this should all be inside of a service layer, not in your controller. Then the controller should have references to your services.

Resources