Entity Framework mapping two table columns to same related table Key

Question

I'm in a situation where I have a table called Elements. Now I'm creating a table called Divergences that will store basically pairs of Elements. The purpose of Divergence is to check if two Elements have diverging answers.

Element               Divergence
---------             ----------
ElementId             ElementId1
                      ElementId2

In the above table schema, ElementId1 and ElementId2 are Foreign Keys mapping to ElementId in Elements table and form the composite primary key for the Divergence table.

I use Database First approach where I create the tables in SQL Server Management Studio and later I do a Update Model from Database... in Entity Framework Designer.

The problem I'm facing is that when EF Designer generates the model it creates 2 sets ICollection<Element> inside the Element class namely Elements and Elements1.

This doesn't give me a Divergences DbSet.

What I'd want to have is a Divergence class where I'd do something like this in code:

Divergence d = new Divergence();
d.Element1 = element1;
d.Element2 = element2;

Database.Divergences.Add(d);
Database.SaveChanges();

and later on:

Element e = Database.Elements.Single(e => e.ElementId == 7);

var divergences = e.Divergences;

I tried adding a new column to the Divergence table like this:

Element               Divergence
---------             ------------
ElementId             DivergenceId
                      ElementId1
                      ElementId2

This correctly leads to a 1 <-> * relationship in Entity Framework Designer. I finally get a Divergences DbSet but DivergenceId property is useless in the code and I still get the 2 sets in the Element class. It's important to note that ElementId1 and ElementId2 still form the composite primary key.

What do you think is the correct way of mapping this specific situation? Thanks for the input.

Answer 1

Instead of...

Divergence d = new Divergence();
d.Element1 = element1;
d.Element2 = element2;

Database.Divergences.Add(d);
Database.SaveChanges();

... you could actually use:

element1.Elements = new List<Element>();
// if that collection isn't already instantiated, for example in the constructor
element1.Elements.Add(element2);

Database.SaveChanges();

This will create exactly the same SQL INSERT statements to the link table without the need of having a Divergence entity. (Change tracking will recognize that you changed the relationship by adding an item to the collection and infer the necessary SQL commands from this change. element1 and element2 must be attached to the context in state Unchanged, but that is also required for your original code in order to work correctly.)

Also, instead of ...

Element e = Database.Elements.Single(e => e.ElementId == 7);
var divergences = e.Divergences;

... you can fetch the columns from the Divergences table like so:

var divergences = Database.Elements.Where(e => e.ElementId == 7)
    .SelectMany(e1 => e1.Elements.Select(e2 => new
    {
        ElementId1 = e1.ElementId,
        ElementId2 = e2.ElementId,
    }))
    .ToList();

So, you will get your desired results without the Divergence entity and honestly I would use it this way. I'm not aware of a way to force EF to create that entity with database first approach, unless you introduce some artificial additional column like you did. If there is a way it is probably a hack or more difficult to achieve and maintain than just working with EF's default, that is without a Divergence entity.

However, you could consider to remove one of the collections (just delete it in the model surface). Having both is a bit confusing in this model in my opinion. Or at least rename them to SourceElements and TargetElements for example :)