R – Fluent nHibernate: Need help with ManyToMany Self-referencing mapping

What you're looking for is a of a set of elements, which in standard hbm mapping is:

<set name="Roles" table="UserRoles">
  <key column="UserID" />
  <element column="Role" />
</set>

For Fluent NHibernate you can map this like so:

HasMany<string>(x => x.Roles)
  .AsElement("Role");

You may need to also specify the key name using WithKeyColumn(string).

I believe you need a IList<Person> Contacts {get;set;} and a type (discriminator) so that you know what to cast it to. Maybe it's a hack but that's how I would do it.

EDIT: Let's say your entities would look like below.

public abstract class Person
{
    public virtual int Id { get; private set; }
    public virtual string FirstName { get; set; }
    public virtual string LastName { get; set; }
}

public class Customer : Person
{
    public virtual string Foo { get; set; }
}

public class Contact : Person
{
    public virtual string Bar { get; set; }
}

The easy (hacky) way of solving your problem would be a discriminator and move everything to the same table. The mapping for this could look like:

public class PersonMap : ClassMap<Person>
{
    public PersonMap()
    {
        Id(x => x.Id).GeneratedBy.Native();
        Map(x => x.FirstName).Length(40);
        Map(x => x.LastName).Length(100);
        DiscriminateSubClassesOnColumn("TypeOfPerson");
    }
}

public class CustomerMap : SubclassMap<Customer>
{
    public CustomerMap()
    {
        Map(x => x.Foo);
    }
}

public class ContactMap : SubclassMap<Contact>
{
    public ContactMap()
    {
        Map(x => x.Bar);
    }
}

If you generate your schema based on that model you end up with:

create table [Person] (
  Id                 INT   IDENTITY   NOT NULL,
  BillingDetailsType NVARCHAR(255)   not null,
  FirstName          NVARCHAR(40)   null,
  LastName           NVARCHAR(100)   null,
  Foo                NVARCHAR(255)   null,
  Bar                NVARCHAR(255)   null,
    primary key ( Id ))

I do understand that this is maybe not ideal certainly so if your boss (like mine) is a database guru that refuses to leave the database in such a state. All you need to do then is to remove the DiscriminateSubClassesOnColumn("TypeOfPerson"); and you should instead end up with a table structure like below should you generate the schema from your code:

-- statement #1
create table [Person] (
  Id        INT   IDENTITY   NOT NULL,
  FirstName NVARCHAR(40)   null,
  LastName  NVARCHAR(100)   null,
    primary key ( Id ))

-- statement #2
create table [Customer] (
  Person_id INT   not null,
  Foo       NVARCHAR(255)   null,
    primary key ( Person_id ))

-- statement #3
create table [Contact] (
  Person_id INT   not null,
  Bar       NVARCHAR(255)   null,
    primary key ( Person_id ))

-- statement #4
alter table [Customer]
 add constraint FKFE9A39C0B58BA2A5 foreign key ( Person_id ) references [Person]

With this last schema you need to figure out a way to determine which subclass the selected Person really is if you ever need an IList<Person> but I am sure you can figure that out. :) Hope this helps!