I'm currently writing a data access layer for an application. The access layer makes extensive use of linq classes to return data. Currently in order to reflect data back to the database I've added a private data context member and a public save method. The code looks something like this:
private DataContext myDb;
public static MyClass GetMyClassById(int id)
{
DataContext db = new DataContext();
MyClass result = (from item in db.MyClasss
where item.id == id
select item).Single();
result.myDb = db;
return result;
}
public void Save()
{
db.SubmitChanges();
}
That's a gross over simplification but it gives the general idea. Is there a better way to handle that sort of pattern? Should I be instantiating a new data context every time i want to visit the db?
Treat your datacontext as a resource. And the rule of using resource says
DataContext is pretty lightweight and is intended for unit of work application as you are using it. I don't think that I would keep the DataContext in my object, however. You might want to look at repository patterns if you aren't going to use the designer generated code to manage your business objects. The repository pattern will allow you to work with your objects detached from the data context, then reattach them before doing updates, etc.
Personally, I'm able to live with the DBML designer generated code for the most part, with partial class implementations for my business and validation logic. I also make the designer-generated data context abstract and inherit from it to allow me to intercept things like stored-procedure and table-valued function methods that are added directly to the data context and apply business logic there.
A pattern that I've been using in ASP.NET MVC is to inject a factory class that creates appropriate data contexts as needed for units of work. Using the factory allows me to mock out the data context reasonably easy by (1) using a wrapper around the existing data context class so that it's mockable (mock the wrapper since DataContext is not easily mockable) and (2) creating Fake/Mock contexts and factories to create them. Being able to create them at will from a factory makes it so that I don't have to keep one around for long periods of time.
It actually doesn't matter too much. I asked Matt Warren from the LINQ to SQL team about this a while ago, and here's the reply:
But basically you don't really need to dispose of them in most cases - and that's by design. I personally prefer to do so anyway, as it's easier to follow the rule of "dispose of everything which implements IDisposable" than to remember a load of exceptions to it - but you're unlikely to leak a resource if you do forget to dispose of it.