Creating Entity Model with Cross Referencing Table - EntityFramework 4.1

Question

Here is my scenario (tables):

Departments
====================
Id (int)
Name (varchar)

Employees
====================
Id (int)
Name (varchar)

EmployeesDepartmentXREFs
====================
EmployeeId (int)
DepartmentId (int)

When I import these 3 tables into EDMX model, I get 2 entities: Employee and Department. Entity Employee has navigational property Departments and entity Department has navigational property Employees.

And this is all fine. I actually like it this way.

But, problem starts when I want to add more fields in the XREF table, for example, I would like to have DateCreated, UserId fields to record when change is made and by who. If I add these two fields in XREF table, then my EDMX model will show 3 entities instead of two. It is not big issue, but since I am already done with coding, I am sort of looking for easy solution (to avoid coding).

Question: Is it possible to add new fields into XREF (cross reference) table and to keep only two entities? If yes, how would I update my datetime and userid information?

I assume that XREF entity has to be present in order to be able to update it (with date and user id), but I really like having navigational properties: Departments and Employees, versus having navigational properties EmployeesDepartmentXREFs.

Is there a middle ground for this? What is the best practice?

Thanks.

Answer 1

I am not entirely sure if this will work for model first, but you could use the NotMappedAttribute. I am especially not sure if this will fix the third entity from showing up, but it is definitely a path to look down.

UPDATE

After looking around, I found this MSDN forum question, which sounds like it is before code-first (which I do believe allows more than two columns in the mapping table via the fluent API) and fits your problem.

You could try the fluent API, but again, I am not sure if this will fix anything for model-first. Here is a good stack overflow question that shows how to use the fluent api to map the columns together.

Last, Here is a good article on your problem.

Hopefully, this helps/clarifies things for you.

Answer 2

You could create a readonly helper property:

public class Department
{
    //...
    public ICollection<EmployeesDepartment> EmployeesDepartments { get; set }

    // because here is only a getter the property is ignored for mapping to DB
    public IEnumerable<Employee> Employees
    {
        get { return EmployeesDepartments.Select(ed => ed.Employee); }
    }
}

But this requires that you either have already loaded the EmployeesDepartments including the Employee from the database or that you use lazy loading (mark navigation properties as virtual) and the context the department has been loaded in is not yet disposed. Also, this property is useless for adding or removing relationships between Department and Employee.

Anyway, it requires that you write custom code. There is no way to introduce a property with automatic getter and setter and specify a mapping which would somehow "skip" your intermediate entity. Your model just does not have a many-to-many relationship anymore when you add properties to the join table and best practice is to consider it as two one-to-many relationships.