Skip to content
This repository

Fluent mapping is the namesake mapping style that Fluent NHibernate uses. It's a fluent interface that allows you to map your entities completely in code, with all the compile-time safety and refactorability that brings.

The getting started guide has a good introduction to mapping with the fluent interface, and the detail provided here builds upon that introduction.

ClassMap

ClassMap is the basis of all your mappings, you derive from this to map anything.

public class PersonMap : ClassMap<Person>
{
  public PersonMap()
  {
  }
}

You map your entities' properties inside the constructor.

'Syntax note:' Every mapping inside a ClassMap is built using lambda expressions, which allow us to reference the properties on your entities without sacrificing compile-time safety. The lambdas typically take the form of x => x.Property. The x on the left is the parameter declaration, which will implicitly be of the same type as the entity being mapped, while the x.Property is accessing a property on your entity (coincidentally called "Property" in this case). You'll quickly get used to these lambdas, as they're used everywhere in Fluent NHibernate.

Once you've declared your ClassMap you're going to need to map the properties on your entity. There are several methods available that map your properties in different ways, and each one of those is a chainable method that you can use to customise the individual mapping.

Id

Every mapping requires an Id of some kind. The Id is mapped using the Id method, which takes a lambda expression that accesses the property on your entity that will be used for the Id. Depending on the return type of the property accessed in the lambda, Fluent NHibernate will make some assumptions about the kind of identifier you're using. For example, if your Id property is an int, an identity column is assumed. Similarly, if you use a Guid then a Guid Comb is assumed.

Id(x => x.Id);

Note that the property you supply to Id can have any name; it does not have to be called "Id," as shown here.

That's the most common scenario for mapping your Id. Customisations can be done by chaining methods off the Id call. For example, if the column to which the Id property was to be mapped were not called Id, we could use the Column method to specify the name. For explicitly specifying the identity generator, you could use the GeneratedBy property.

Id(x => x.Id)
  .Column("PersonId")
  .GeneratedBy.Assigned();

In this example we're specifying that the Id property is mapped to a PersonId column in the database, and it's using an assigned generator.

Fluent NHibernate's interface is designed for discoverability. Everything you need should be easy to find "under" the declaring method using method chains.

Properties

Property mappings make up a large amount of any mapped domain. Fortunately, they're just as easy to create as identities, except we use the Map method.

Map(x => x.FirstName);

That's all you need for most situations. Fluent NHibernate knows what the return type of your property is, and assumes that the column it's mapping against will have the same name as the property itself. There are numerous customisations available through methods chained from the Map call. For example, if you're generating your schema you may want to specify whether the column itself is nullable, you can do that by using the Nullable method and (optionally) the Not operator property.

// nullable
Map(x => x.FirstName)
  .Nullable();

// not nullable
Map(x => x.FirstName)
  .Not.Nullable();

If you need to map private properties, you can explore your options fluent mapping private properties.

Access strategies

Access strategies are NHibernate's way of knowing how to get or set the values in your entity. Fluent NHibernate can detect what access strategy to use, so most of the time you won't need to specify it.

For example, private field via public property:

  private string name;

  public string Name
  {
    get { return name; }
  }

  Map(x => x.Name); // access="nosetter.lowercase"

For example, read-only auto-property:

public string Name { get; private set; }

Map(x => x.Name); // access="backfield"

Relationships

Except in the most basic of scenarios, you'll need to map relationships among entities. These will typically take the form of many-to-one, one-to-many, and many-to-many relationships. For better or worse, we tend not to refer to these by their database design names: we aren't database administrators, after all. Instead, we refer to them as References, HasMany, and HasManyToMany, respectively. We'll go into each in more detail next.

References / many-to-one

References is for creating many-to-one relationships between two entities, and is applied on the "many side." You're referencing a single other entity, so you use the References method. #HasMany / one-to-many is the "other side" of the References relationship, and gets applied on the "one side."

Let's map a relationship between a book and its author.

public class Book
{
  public Author Author { get; set; }
}

public class Author
{
  public IList<Book> Books { get; set; }
}

In domain terms, we have an Author which may be associated with any number of Books, and Books, each of which can be associated with a single Author.

In database terms, we'd have a book table with a foreign key column referencing the primary key of an author table.

To create the references relationship in your Book #ClassMap, add the following call in the BookMap constructor:

References(x => x.Author);

That's it, you've now created a references relationship between Book and Author. The foreign-key is assumed to be named Author_id in your book table, unless you have a different Conventions in place or override the name by using the Column method.

As with all other fluent mappings, you can chain calls to customise the reference relationship. For example if you wanted to specify the cascade strategy you'd use the Cascade property.

References(x => x.Author)
  .Column("AuthorId")
  .Cascade.All();

HasMany / one-to-many

HasMany is probably the most common collection-based relationship you're going to use. HasMany is the "other side" of a #References / many-to-one relationship, and gets applied on the "one side" (one author has many books). Now let's map the author side of the relationship we started above. We need to join into the books table returning a collection of any books associated with that author.

public class Author
{
  public IList<Book> Books { get; set; }
}

We use the HasMany method in the AuthorMap constructor to map this side of the relationship:

HasMany(x => x.Books);

As with References, the foreign-key defaults to Author_id, and you can override it with the KeyColumn method or change the default behaviour with a Conventions.

There are a few different types of collections you can use, and they're all available under the HasMany call.

Collection types

Fluent NHibernate does its best to detect the collection type for your collections. For example, an ISet<T> will be mapped as a Set for NHibernate.

As of release 1.2, Fluent NHibernate can also detect collection types for collections that are exposed by an IEnumerable<T>.

For example:

public IList<Child> Children { get; set; }

HasMany(x => x.Children); // <bag />


private ISet<Child> _children;

public IEnumerable<Child> Children
{
  get { return _children; }
}

HasMany(x => x.Children); // <set access="nosetter.camelcase-underscore" />


private ISet<Child> _children;

public IEnumerable<Child> GetChildren()
{
  return _children;
}

HasMany(x => x.GetChildren()); // <set name="_children" access="field" />

HasManyToMany / many-to-many

HasManyToMany works exactly the same as #HasMany / one-to-many, except the underlying database structure to which it maps is different.

HasManyToMany(x => x.Books);

There are a few different types of collections you can use, and they're all available under the HasManyToMany call.

HasOne / one-to-one

HasOne is usually reserved for a special case. Generally, you'd use a #References / many-to-one relationship in most situations (see: I think you mean a many-to-one). If you really do want a one-to-one, then you can use the HasOne method.

HasOne(x => x.Cover);

To specify the foreign key, you can use the PropertyRef() method. Additionally, if you are using References() to specify the bi-directional relationship on the other end of the relationship, be sure to use the .Unique() mapping modifier to effectively specify that this is a one-to-one relationship.

In the example below, we are mapping the relationship between a Car entity and a SteeringWheel entity. For our business rules, a Car will only ever have one SteeringWheel so we would map it like so:

public class CarMap : ClassMap<Car>
{
    public CarMap()
    {
        Table( "Vehicles.dbo.Car" );

        Id( x => x.CarId );
        Map( x => x.Name );
        Map( x => x.Year );

        HasOne( x => x.SteeringWheel ).PropertyRef( x => x.Car);
    }
}

public class SteeringWheelMap : ClassMap<SteeringWheel>
{
    public SteeringWheelMap()
    {
        Table( "Vehicles.dbo.SteeringWheel" );

        Id( x => x.SteeringWheelId );
        Map( x => x.Diameter );
        Map( x => x.Color );

        References( x => x.Car, "CarId" ).Unique();
    }
}

Any

Any mappings are another special case, and you really only should use them if you know what you're doing. To quote the NHibernate documentation:

The <any> mapping element defines a polymorphic association to classes from multiple tables. This type of mapping always requires more than one column. The first column holds the type of the associated entity. The remaining columns hold the identifier. It is impossible to specify a foreign key constraint for this kind of association, so this is most certainly not meant as the usual way of mapping (polymorphic) associations. You should use this only in very special cases (eg. audit logs, user session data, etc).

There are three things you need to provide to be able to map using an Any:

  • A column that holds the type of the entity,
  • At least one column holding the identifier value, and
  • A type for the identifier itself.

You can specify these using the EntityTypeColumn, EntityIdentifierColumn, and IdentityType methods respectively.

ReferencesAny(x => x.Author)
  .EntityTypeColumn("Type")
  .EntityIdentifierColumn("Id")
  .IdentityType<int>();

Components

Components are a clever way of mapping a normalised data model into a more reasonable object model. You may have a customer table that has a series of address columns, ideally you'd want the address columns to be mapped into an Address object, rather than just being properties on a Customer; you can do that with components.

The Component method takes two parameters, rather than just one like the rest of the methods you've seen so far. The first parameter is a property accessor lambda, like all the other methods, and the second one is another lambda (quite often referred to as a nested-closure in these situations) that defines a new scope for defining the mappings of that particular sub-part (in this case the component).

Component(x => x.Address, m =>
{
  m.Map(x => x.Number);
  m.Map(x => x.Street);
  m.Map(x => x.PostCode);
});

As you can see, the first parameter references the Address property on our entity, which is the property that holds our component. The second property is where we define what makes up the component. The main difference in this lower scope is that you have to use the m instance to access the available methods; you don't have to call it m, but we are for brevity.

In this case we've mapped a component (stored in the Address property), that's made up of three properties (Number, Street, and PostCode); these properties are stored in the same table as the parent entity in a normalised fashion.

ComponentMap

Note: ComponentMap's are incompatible with the automapper. If you're using the automapper then your ComponentMap's will be ignored.

If you have a particular component that occurs regularly in your entities, you can abstract the mappings into a single ComponentMap definition.

Using the example of an Address again, it wouldn't be surprising to find that multiple entities had addresses; perhaps Person, and Company.

public class Address
{
  public int Number { get; set; }
  public string Street { get; set; }
  public string City { get; set; }
  public string PostCode { get; set; }
}

public class Person
{
  public Address Address { get; set; }
}

public class Company
{
  public Address Address { get; set; }
}

Normally you would need to define an inline component mapping for each occurrence of Address, using the Component(property, mapping) method, even though they'd be identical.

public PersonMap()
{
  Component(x => x.Address, m =>
  {
    m.Map(x => x.Number);
    m.Map(x => x.Street);
    m.Map(x => x.City);
    m.Map(x => x.PostCode);
  });
}

public CompanyMap()
{
  Component(x => x.Address, m =>
  {
    m.Map(x => x.Number);
    m.Map(x => x.Street);
    m.Map(x => x.City);
    m.Map(x => x.PostCode);
  });
}

However, using the ComponentMap, you can define your component mapping in a single location.

public class AddressMap : ComponentMap<Address>
{
  public AddressMap()
  {
    Map(x => x.Number);
    Map(x => x.Street);
    Map(x => x.City);
    Map(x => x.PostCode);
  }
}

With that created, you only need to signify that your properties are a component from your entity mappings by using the new Component(property) overload that takes a single parameter (it doesn't take a closure for defining the body like the other Component methods).

public PersonMap()
{
  Component(x => x.Address);
}

public CompanyMap()
{
  Component(x => x.Address);
}

Column names

If you have multiple instances of the same component in an entity, say WorkAddress and HomeAddress, you'll need to supply a prefix for the columns. This is because the column names are, by default, based on the properties within the component, so you'd end up with two Streets, etc...

To supply a prefix, chain a call to ColumnPrefix from your Component(property) call.

'Note:' this only works for ComponentMap-based components.

Component(x => x.WorkAddress)
  .ColumnPrefix("Work");

This would create WorkStreet, WorkCity, WorkPostCode columns.

Subclasses

Note: SubclassMap's are incompatible with the automapper. If you're using the automapper, your SubclassMap's will be ignored. See Auto mapping#Inheritance for guidance on dealing with subclasses with the automapper.

Subclasses work in a very similar way to #ClassMap, in that you create a derived class into which you put your mappings, but here you use SubclassMap instead of ClassMap.

There are two strategies for mapping inheritance hierarchies in Fluent NHibernate: table-per-class-hierarchy and table-per-subclass, the former being a subclass and the latter a joined-subclass. This means that you can define one table to hold all subclasses (with 1 or more columns used to identify the specific type for each row), or you can define separate tables for each subclass.

'Note:' For multiple discriminator columns use a DiscriminateSubClassesOnColumn("").Formula([insert custom sql here])

DiscriminateSubClassesOnColumn("")
  .Formula("case when discriminatorColumnName = ** then 1 else 2 end")

Table-per-subclass is the default mapping for subclasses, so unless you say otherwise you'll have a separate table for each subclass. The parent mapping dictates what the subclass mapping strategy will be, by either specifying or not specifying a discriminator (discriminators are required for table-per-class-hierarchy).

We'll use the following two classes for examples:

public class Parent
{
  public int Id { get; set; }
  public string Name { get; set; }
}

public class Child : Parent
{
  public string AnotherProperty { get; set; }
}

If you wanted to map this as a table-per-subclass, you'd do it like this:

public class ParentMap : ClassMap<Parent>
{
  public ParentMap()
  {
    Id(x => x.Id);
    Map(x => x.Name);
  }
}

public class ChildMap : SubclassMap<Child>
{
  public ChildMap()
  {
    Map(x => x.AnotherProperty);
  }
}

That's all there is to it, Parent and Child are now mapped as subclasses. When Fluent NHibernate finds ChildMap, it knows that it's a subclass of Parent.

If you wanted to do a table-per-class-hierarchy strategy, then you just need to specify the discriminator column in your ClassMap.

public class ParentMap : ClassMap<Parent>
{
  public ParentMap()
  {
    Id(x => x.Id);
    Map(x => x.Name);

    DiscriminateSubClassesOnColumn("type");
  }
}

public class ChildMap : SubclassMap<Child>
{
  public ChildMap()
  {
    DiscriminatorValue(valueRepresentingThisSubclass);
    Map(x => x.AnotherProperty);
  }
}

The only difference is that we're now calling DiscriminateSubclassesOnColumn in ParentMap with a "type" parameter. This parameter specifies the name of the column in the table that identifies which subclass the data in a given row represents.

Something went wrong with that request. Please try again.