Take the 2-minute tour ×
Programmers Stack Exchange is a question and answer site for professional programmers interested in conceptual questions about software development. It's 100% free, no registration required.

I'm developing a series of repository classes and a UnitOfWork class (plus its IUnitOfWork interface of course). I'm using Castle Windsor, which injects dependencies via constructors.

My business tier has classes such as "CustomerBusinessLogic", with methods like RetrieveCustomer, SaveCustomer, etc. These in turn use the UOW to perform the database operations.

I'm struggling to find how/where the BLL class should get an instance of the UOW. It doesn't feel right injecting it into the ctr, as a unit of work is exactly that - a unit of work! I think an individual method (LoadCustomer, SaveCustomer) should be responsible for instantiating the UOW for the lifetime of its operation. However doing this means resorting to a service locator, which is an anti-pattern.

Suggestions welcome!

share|improve this question

2 Answers 2

I just finished an article today about the repository pattern (with sample implementations).

The thing with most .NET ioc containers is that they support scoping. That is, they can create objects with a limited lifetime. That works very well with HTTP applications since the scope is the same as the lifetime of a HTTP request.

If you use ASP.NET MVC you can combine that with built in features of MVC to trigger the UoW if no errors where detected: http://blog.gauffin.org/2012/06/how-to-handle-transactions-in-asp-net-mvc3/

If you use any other kind of application I usually create a scope by myself (for instance to wrap a command):

using (var scope = MyContainer.CreateChildCointainer())
{
    using (var uow = scope.Resolve<IUnitOfWork>())
    {
        // do something here


        uow.SaveChanges();
    }
}

The thing is that the repositories etc do not have to be aware of the unit of work (if you use databases in .NET). The UoW implementation can make sure that all OR/M or UoW implementation enlists all db commands in the transaction.

share|improve this answer
    
This is the problem I've been finding with examples on the internet - many relate to MVC apps where there is a naturally limited lifetime (i.e. a request). Having read some more articles I'm starting to think that the repository and UoW pattern are overkill, as EF provides much of this already - it's starting to feel like I'm putting one abstraction over another. –  Andrew Stephens Jan 12 '13 at 10:01
    
My primary reason for creating the repository and UoW was to allow unit testing of the BLL classes using mock UoWs. I guess it also isolates the BLL from the data layer to some extent, but as is usually the case, it's very unlikely that you are ever going to switch from EF to some other framework! I'm now thinking of keeping things simple and just using the entity context directly within my BLL classes. I've found an article on how to extract an interface to enable mocking of the context for testing purposes. Thanks for your reply though, it was an interesting read. –  Andrew Stephens Jan 12 '13 at 10:04
    
You don't create abstractions to be able to switch framework (like from EF to something other). You create abstractions to make the code less complex (and therefore less error prone). As for unit of work. Most applications do have clear scopes. In WPF/Winforms it's when the users execute a command by pressing a button. –  jgauffin Jan 13 '13 at 16:38

What you're describing is a really common pattern where instead of injecting the object of interest you want to inject an object that can create instances of the object of interest. You might call it a factory. If you are using Autofac it is baked in by simply injecting Func<T> (to create instances of T). I'm pretty sure Castle Windsor has this same feature.

You may need to consider the boundaries of your unit of work further. Your question implies that a unit of work falls within the boundaries of, say, SaveCustomer but ask yourself what if I reuse that method within a larger unit of work. What are the boundaries of your unit of work then, and does your injection strategy and unit of work implementation handle that correctly.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.