I'm trying to write my Domain Model as persistence-ignorant as possible. The only thing I'm doing right now is marking every property and method virtual, as NHibernate requires that for lazy-loading.
In my domain model assembly I define some repository interfaces:
public interface IRepository<TEntity> where TEntity : EntityBase {
TEntity Get(int id);
/* ... */
}
public interface IProductRepository : IRepository<Product> { ... }
Then I have a data assembly. This one will reference NHibernate, it knows about its existence. This is the assembly that implements those repository interfaces:
public abstract class Repository<TEntity> : IRepository<TEntity> {
public TEntity Get(ind id) { ... }
/* ... */
}
public class ProductRepository : Repository<Product>, IProductRepository {
/* ... */
}
and so on.
Now I wanted to implement a transaction functionality to my repositories. To do so, I would add a BeginTransaction method on my IRepository interface. However, I cannot define its return type as NHibernate.ITransaction, since I want to keep the domain model persistence-ignorant, and not be forced to reference NHibernate's assembly from my domain model assembly.
What would you do?
Would you simply implement a void BeginTransaction(), a void Commit(), and a void RollBack() methods on the interface, and let the repository implementation manage the ITransaction object internally?
Or would you find a way to expose the ITransaction object to let the client manage the transaction directly with it, instead of using repository's methods?
Thanks!
You can take a look at the Sharp Architecture which has already implemented everything you talk about, including generic repositories with transactions support. The solution there is that IRepository has DbContext property which encapsulates transactions (it's actually an interface).
This is the first of the options that you described (custom transactions interface which hides NHibernate). And it works well.
I guess you can even re-use S#arp code regardless if you intend to use the full framework.
IMO Transactions should always start and end in business logic, in other words the transaction should start in the service layer not the repository layer and the repository should enlist it's self in the transaction, ideally this would be done implicitly.
Now if you're using NH then if your service and repositories share the same 'session' (which they should) then you can call 'BeginTransaction' in the service layer and commit or roll back as required:
Eg, imagine this a method on a service:
public void RegisterCustomer(Customer customer)
{
try
{
using(var transaction = _session.BeginTransaction())
{
_customerRepository.Save(customer);
_customerSurveyRepository.Save(customerSurvey);
// DO What ever else you want...
transaction.Commit();
}
}
catch (Exception exn)
{
throw new AMException(FAILED_REGISTRATION, exn);
}
}
How the repositories obtain a reference to the same Session can be solved by injecting in the constructors or by using a the SessionFactory to obtain the current session...
Related
In DDD, is the application layer who uses the repository to get the data from database, call the methods of the domain and then call the repository to persists the data. Something like that:
public void MyApplicationService()
{
Order myOrder = _orderRepository.Get(1);
myOrder.Update(data);
_orderRepository.Commit();
}
In this example the repository is a class variable that it is instantiate in the constructor of the service, so its life is the life of the class.
But I am wondering if it wouldn't be better to instantiate a repository for each action that I want to do, to have a shorter life, because if not, if I use the class for many actions, the repository will have many entities that perhaps it will not need more.
So I was thinking in a solution like this:
public void MyApplicationService()
{
OrderRepository myOrderRepository = new OrderRepository(_options);
Order myOrder = myOrderRepository.GetOrder(1);
myOrder.Update(data);
myOrderRepository.Commit();
myOrderRepository.Dispose();
}
So a new instance each time I need to do the action.
So in sumary, I would like to know about the differents solutions and the advantages and disadvanges to decide the lifespan of the repository.
Thanks.
The recommended lifespan of the repository is one business transaction.
Your second patch of code is correct in that aspect, however it has one drawback: you have created a strong dependency between the ApplicationService and OrderRepository classes. With your code, you are not able to isolate both class in order to unit test them separately. Also, you need to update the ApplicationService class whenever you change the constructor of the OrderRepository. If OrderRepository requires parameters to construct, then you have to construct them (which implies to reference their type and base types), despite this being an implementation detail of OrderRepository (needed for data persistence store access) and not needed for your application service layer.
For these reasons, most of modern program development rely on a pattern called Dependency Injection (DI). With DI, you specify that your ApplicationService class depends on an instance of the OrderRepository class, or better, an interface IOrderRepository whom the OrderRepository class implements. The dependency is declared by adding a parameter in the ApplicationService constructor:
public interface IOrderRepository : IDisposable
{
Order GetOrder(int id);
void Commit();
}
public class ApplicationService
{
private readonly OrderRepository orderRepository;
public ApplicationService(IOrderRepository orderRepository)
{
this.orderRepository = orderRepository ?? throw new ArgumentNullException(nameof(orderRepository));
}
public void Update(int id, string data)
{
Order myOrder = orderRepository.Get(id);
myOrder.Update(data);
orderRepository.Commit();
}
}
Now the DI library is responsible to construct OrderRepository and inject the instance in the ApplicationService class. If OrderRepository has its own dependencies, the library will resolve them first and construct the whole object graph so you don't have to do that yourself. You simply need to tell your DI library what specific implementation you want for each referenced interface. For example in C#:
public IServiceCollection AddServices(IServiceCollection services)
{
return services.AddScoped<IOrderRepository,OrderRepository>();
}
When unit testing your code, you can replace the actual implementation of OrderRepository with a mock object, such as Mock<IOrderRepository> or your own MockOrderRepository implementation. The code under test is then exactly the code in production, all wiring being done by the DI framework.
Most modern DI libraries have support for object lifetime management, including transient (always resolve a new object), singleton (always reuse the same object), or scoped (each scope has a single instance). The latter is what is used to isolate objects instance per business transaction, using a singleton ScopeFactory to create scopes whenever you start a business transaction:
public class UpdateOrderUseCase : UseCase
{
private readonly IScopeFactory scopeFactory;
public UpdateOrderUseCase(IScopeFactory scopeFactory) // redacted
public void UpdateOrder(int id, string data)
{
using var scope = scopeFactory.CreateScope();
var orderRepository = scope.GetService<IOrderRepository>();
var order = orderRepository.Get(id);
order.Update(data);
orderRepository.Commit();
// disposing the scope will also dispose the object graph
}
}
When you implement a REST service, that transaction usually corresponds to one HTTP request. Modern frameworks, such as asp.net core, will automatically create scopes per HTTP request and use that to resolve your dependency graph later in the framework internals. This means you don't even have to handle the ScopeFactory yourself.
According COMPOSITION ROOT pattern, I must to construct all dependencies graph as close as possible to the application's entry point.
My architecture is plugin oriented. So, if someone wants to extend my base system he can.
For example, in my base system I have this structure:
View Layer
Services Layer
Data Access Layer
Model Layer
In DAL, I expose some classes like:
IRepository
NHibernateRepository
ProductRepository
So, I'd like if a plugin wants to extend my base Product class to ExtendedProduct, and then create ExtendedProductRepository that inherits from NHibernateRepository.
The question is:
How can instantiate from my base system an instance of NHibernateRepository using NInject?
So, I know the first thing to do is to construct the graph dependencies:
using (var kernel = new StandardKernel())
{
kernel.Bind(b => b.FromAssembliesMatching("*")
.SelectAllClasses()
.InheritedFrom<IRepository>()
.BindAllInterfaces());
}
However, I'm figuring out that when I execute something like:
kernel.GetAll<IRepository>()
It's going to return me a ProductRepository instance, and another ProductExtendedRepository under two IRepository objects.
So, how I can save a ProductExtended object from my base system...?
Another question, would be, how could I inject a object instance in my plugins, or, how can plugins autoinject some instance of base system assembly?
Thanks for all.
I'll appreciate a lot some help.
I use this pattern for my NHibernate based projects:
public interface IRepository<T> : IQueryable<T>
{
T Get(int id);
void Save(T item);
void Delete(T item);
}
public class NHibernateRepository<ModelType> : IRepository<ModelType>
where ModelType : class
{
// implementation
}
then...
public interface IProductRepository : IRepository<Product>
{
// product specific data access methods
}
public class ProductRepository : NHibernateRepository<Product>, IProductRepository
{
// implementation
}
... and in Ninject Module:
Bind(typeof(IRepository<>)).To(typeof(NHibernateRepository<>));
Bind<IProductRepository>().To<ProductRepository>();
then you can either request the base functionality like:
public Constructor(IRepository<Product> repo) { ... }
or specific product repository functionality:
public Constructor(IProductRepository repo) { ... }
your plugins can either get the base functionality and won't have to register anything:
public PluginConstructor(IRepository<ProductExtended> repo { ... }
or create their own repositories and register them in a Ninject module.
Thanks dave.
It's perfect. I'll try it.
However, how could I save or get or update (whichever IRepository methods)... an ExtendedProduct instance from my base system?
Think the follow out:
public interface BasePlugin<T> {...}
In another assembly:
public class PluginExtendedProduct : BasePlugin<ExtendedProduct>
{
public PluginExtendedProduct (IRepository<ExtendedProduct> repo { ... }
}
My headache is how to create an instance of (so, ExtendedProduct) in my base system in order to call methods PluginExtendedProduct that uses an IRepository.
I don't know if I'm explaining myself well...
Thanks for all.
I have a pretty standard repository interface:
public interface IRepository<TDomainEntity>
where TDomainEntity : DomainEntity, IAggregateRoot
{
TDomainEntity Find(Guid id);
void Add(TDomainEntity entity);
void Update(TDomainEntity entity);
}
We can use various infrastructure implementations in order to provide default functionality (e.g. Entity Framework, DocumentDb, Table Storage, etc). This is what the Entity Framework implementation looks like (without any actual EF code, for simplicity sake):
public abstract class EntityFrameworkRepository<TDomainEntity, TDataEntity> : IRepository<TDomainEntity>
where TDomainEntity : DomainEntity, IAggregateRoot
where TDataEntity : class, IDataEntity
{
protected IEntityMapper<TDomainEntity, TDataEntity> EntityMapper { get; private set; }
public TDomainEntity Find(Guid id)
{
// Find, map and return entity using Entity Framework
}
public void Add(TDomainEntity item)
{
var entity = EntityMapper.CreateFrom(item);
// Insert entity using Entity Framework
}
public void Update(TDomainEntity item)
{
var entity = EntityMapper.CreateFrom(item);
// Update entity using Entity Framework
}
}
There is a mapping between the TDomainEntity domain entity (aggregate) and the TDataEntity Entity Framework data entity (database table). I will not go into detail as to why there are separate domain and data entities. This is a philosophy of Domain Driven Design (read about aggregates). What's important to understand here is that the repository will only ever expose the domain entity.
To make a new repository for, let's say, "users", I could define the interface like this:
public interface IUserRepository : IRepository<User>
{
// I can add more methods over and above those in IRepository
}
And then use the Entity Framework implementation to provide the basic Find, Add and Update functionality for the aggregate:
public class UserRepository : EntityFrameworkRepository<Stop, StopEntity>, IUserRepository
{
// I can implement more methods over and above those in IUserRepository
}
The above solution has worked great. But now we want to implement deletion functionality. I have proposed the following interface (which is an IRepository):
public interface IDeleteableRepository<TDomainEntity>
: IRepository<TDomainEntity>
{
void Delete(TDomainEntity item);
}
The Entity Framework implementation class would now look something like this:
public abstract class EntityFrameworkRepository<TDomainEntity, TDataEntity> : IDeleteableRepository<TDomainEntity>
where TDomainEntity : DomainEntity, IAggregateRoot
where TDataEntity : class, IDataEntity, IDeleteableDataEntity
{
protected IEntityMapper<TDomainEntity, TDataEntity> EntityMapper { get; private set; }
// Find(), Add() and Update() ...
public void Delete(TDomainEntity item)
{
var entity = EntityMapper.CreateFrom(item);
entity.IsDeleted = true;
entity.DeletedDate = DateTime.UtcNow;
// Update entity using Entity Framework
// ...
}
}
As defined in the class above, the TDataEntity generic now also needs to be of type IDeleteableDataEntity, which requires the following properties:
public interface IDeleteableDataEntity
{
bool IsDeleted { get; set; }
DateTime DeletedDate { get; set; }
}
These properties are set accordingly in the Delete() implementation.
This means that, IF required, I can define IUserRepository with "deletion" capabilities which would inherently be taken care of by the relevant implementation:
public interface IUserRepository : IDeleteableRepository<User>
{
}
Provided that the relevant Entity Framework data entity is an IDeleteableDataEntity, this would not be an issue.
The great thing about this design is that I can start granualising the repository model even further (IUpdateableRepository, IFindableRepository, IDeleteableRepository, IInsertableRepository) and aggregate repositories can now expose only the relevant functionality as per our specification (perhaps you should be allowed to insert into a UserRepository but NOT into a ClientRepository). Further to this, it specifies a standarised way in which certain repository actions are done (i.e. the updating of IsDeleted and DeletedDate columns will be universal and are not at the hand of the developer).
PROBLEM
A problem with the above design arises when I want to create a repository for some aggregate WITHOUT deletion capabilities, e.g:
public interface IClientRepository : IRepository<Client>
{
}
The EntityFrameworkRepository implementation still requires TDataEntity to be of type IDeleteableDataEntity.
I can ensure that the client data entity model does implement IDeleteableDataEntity, but this is misleading and incorrect. There will be additional fields that are never updated.
The only solution I can think of is to remove the IDeleteableDataEntity generic condition from TDataEntity and then cast to the relevant type in the Delete() method:
public abstract class EntityFrameworkRepository<TDomainEntity, TDataEntity> : IDeleteableRepository<TDomainEntity>
where TDomainEntity : DomainEntity, IAggregateRoot
where TDataEntity : class, IDataEntity
{
protected IEntityMapper<TDomainEntity, TDataEntity> EntityMapper { get; private set; }
// Find() and Update() ...
public void Delete(TDomainEntity item)
{
var entity = EntityMapper.CreateFrom(item);
var deleteableEntity = entity as IDeleteableEntity;
if(deleteableEntity != null)
{
deleteableEntity.IsDeleted = true;
deleteableEntity.DeletedDate = DateTime.UtcNow;
entity = deleteableEntity;
}
// Update entity using Entity Framework
// ...
}
}
Because ClientRepository does not implement IDeleteableRepository, there will be no Delete() method exposed, which is good.
QUESTION
Can anyone advise of a better architecture which leverages the C# typing system and does not involve the hacky cast?
Interestly enough, I could do this if C# supported multiple inheritance (with separate concrete implementation for finding, adding, deleting, updating).
I do think that you're complicating things a bit too much trying to get the most generic solution of them all, however I think there's a pretty easy solution to your current problem.
TDataEntity is a persistence data structure, it has no Domain value and it's not known outside the persistence layer. So it can have fields it won't ever use, the repository is the only one knowing that, it'a persistence detail . You can afford to be 'sloppy' here, things aren't that important at this level.
Even the 'hacky' cast is a good solution because it's in one place and a private detail.
It's good to have clean and maintainable code everywhere, however we can't afford to waste time coming up with 'perfect' solutions at every layer. Personally, for view and persistence models I prefer the quickest and simplest solutions even if they're a bit smelly.
P.S: As a thumb rule, generic repository interfaces are good, generic abstract repositories not so much (you need to be careful) unless you're serializing things or using a doc db.
I have the following implementation and would like some feedback as to whether it makes correct use of NHibernate for sessions and transactions.
public interface IUnitOfWork : IDisposable
{
ISession CurrentSession { get; }
void Commit();
void Rollback();
}
public class UnitOfWork : IUnitOfWork
{
private readonly ISessionFactory _sessionFactory;
private readonly ITransaction _transaction;
public UnitOfWork(ISessionFactory sessionFactory)
{
_sessionFactory = sessionFactory;
CurrentSession = _sessionFactory.OpenSession();
_transaction = CurrentSession.BeginTransaction();
}
public ISession CurrentSession { get; private set; }
public void Dispose()
{
CurrentSession.Close();
CurrentSession = null;
}
public void Commit()
{
_transaction.Commit();
}
public void Rollback()
{
if (_transaction.IsActive) _transaction.Rollback();
}
}
Ninject binding
Bind<IUnitOfWork>().To<UnitOfWork>().InTransientScope();
Bind<ISessionFactory>().ToProvider<NHibernateSessionFactoryProvider>().InSingletonScope();
Bind<IRepository>().To<Repository>().InTransientScope();
Here is an example of the usage:
public class Repository : IRepository
{
private readonly ISessionFactory _sessionFactory;
public Repository(ISessionFactory sessionFactory)
{
_sessionFactory = sessionFactory;
}
public void Add(IObj obj)
{
using (var unitOfWork = new UnitOfWork(_sessionFactory))
{
unitOfWork.CurrentSession.Save(obj);
unitOfWork.Commit();
}
}
}
In my previous implementation I would inject IUnitOfWork into my repository constructor like so
public Repository(IUnitOfWork unitOfWork)
{...
But the Dispose() method would not execute causing a subsequent call to throw this exception: "Cannot access a disposed object. Object name: 'AdoTransaction'."
First observation: your repository should not commit the unit of work. This defeats the whole point of the unit of work pattern. By immediately saving your changes inside the repository, you're "micro managing" the NHibernate Session.
The unit of work should be referenced higher up the stack, in your application/service layer. This allows you to have application code that performs several actions, potentially on different repositories, and still at the end commit everything at once.
The UnitOfWork class itself looks Ok, though you should ask yourself if you really need it. In NHibernate, the ISession IS your unit of work. Your UnitOfWork class does not seem to add a lot of value (especially since you're exposing the CurrentSession property anyway)
But you do need to think about it's lifetime. I think you have it wrong on this point. Session lifetime management depends on the type of application you're developing: in a web app, you typically want to have a unit of work per request (you might want to google on 'nhibernate session per request'). In a desktop app it's slightly more complicated, you will most of the time want a 'session per screen' or 'conversation per business transaction'.
I have a mostly CRUD type of application, and I implemented the Unit Of Work with Repository pattern, but couldn't really get away from the Session/Transaction split. Sessions and Transactions need different lifetimes. In the desktop world, a Session is usually "per-screen" and a Transaction is "per-user-action".
More information in this excellent article.
So what I ended up with was:
IUnitOfWork -> Wraps session, implements IDisposable
IAtomicUnitOfWork -> Wraps transaction, implements IDisposable
IRepository -> Provides Get, Save, Delete and query access
I made it so that you need an IUnitOfWork to build an IAtomicUnitOfWork and you need an IAtomicUnitOfWork to build an IRepository, so that enforces proper transaction management. That's really all I gained by implementing my own interfaces.
As jeroenh said, you are almost just as well to use ISession and ITransaction but in the end I felt a little better writing all my code against an interface that I defined.
An important part of the answer lies in what you want your transaction sizes to be. Right now (as jeroenh has indicated) the transaction is per method invocation on your repository. This is very small and probably not needed. I created an ASP.MVC application and it uses a transaction size that included everything from a single http request. This could be multiple database reads/updates. I am using the same unit of work and Ninject for IOC. Take a look, maybe something will help with your issues:
http://bobcravens.com/2010/06/the-repository-pattern-with-linq-to-fluent-nhibernate-and-mysql/
http://bobcravens.com/2010/07/using-nhibernate-in-asp-net-mvc/
http://bobcravens.com/2010/09/the-repository-pattern-part-2/
http://bobcravens.com/2010/11/using-ninject-to-manage-critical-resources/
Hope this helps.
Bob
Ok so I 'm just getting into nhibernate (using fluent).
One thing that I love about it is that I can use the Repository pattern (read about it from the nhibernate rhino blog).
Basically using generics, I can create methods that will work accross ALL my database tables.
public interface IRepository<T>
{
T GetById(int id);
ICollection<T> FindAll();
void Add(T entity);
void Remove(T entity);
}
public class Repository<T> : IRepository<T>
{
public ISession Session
{
get
{
return SessionProvider.GetSession();
}
}
public T GetById(int id)
{
return Session.Get<T>(id);
}
public ICollection<T> FindAll()
{
return Session.CreateCriteria(typeof(T)).List<T>();
}
public void Add(T t)
{
Session.Save(t);
}
public void Remove(T t)
{
Session.Delete(t);
}
}
I then inherit the Repository class and I can then add methods that are specific to that entity.
When trying to add an Update method, someone mentioned that the Repository pattern is suppose to act on collections? Am I looking at things incorrectly here? Why can't I create an update method?
I tried adding a update method, but I'm confused as to how I will handle the session and update the database?
I want a single place for all my database access for each entity, so UserRepository will have all basic CRUD and then maybe some other methods like GetUserByEmail() etc.
Don't use the repository pattern - use the UnitOfWork pattern instead, and pass defined query ICriteria to the ISession. Essentially the Repo pattern is wrapping something that doesn't need to be wrapped with NH.
see http://ayende.com/Blog/archive/2009/04/17/repository-is-the-new-singleton.aspx for more info
Perhaps you misheard or someone mispoke - the Repository pattern is supposed to expose collection like behavior, not operate on collections. Just like you can add, remove and search for items in a collection, your repository offers save, delete and search operations that work against your database.
I suggest you download the code for S#arp Architecture. It includes a repository implementation that you can reuse quite easily. If you don't want to take the dependency, at the very least you can spend some time studying their implementation to give you a better idea of how to approach it yourself.