NHibernate Reference Documentation

Table of Contents

Quickstart with IIS and Microsoft SQL Server

1.1. Getting started with NHibernate

First persistent class
Mapping the cat
Playing with cats

2. Architecture

2.1. Overview
Instance states
Contextual Sessions

3. ISessionFactory Configuration

3.1. Programmatic Configuration
Obtaining an ISessionFactory
User provided ADO.NET connection
NHibernate provided ADO.NET connection
3.5. Optional
configuration properties

3.5.1. SQL Dialects
Outer Join Fetching
Custom ICacheProvider
Query Language Substitution

3.6. Logging
Implementing an INamingStrategy
XML Configuration File

4. Persistent Classes

4.1. A simple POCO example

4.1.1. Declare properties for persistent fields
Implement a default constructor
Provide an identifier property (optional)
Prefer non-sealed classes and virtual methods (optional)

4.2. Implementing inheritance
Implementing Equals() and GetHashCode()
Dynamic models
Lifecycle Callbacks
IValidatable callback

5. Basic O/R Mapping

5.1. Mapping declaration

5.1.1. XML Namespace
id generator
Hi/Lo Algorithm
UUID Hex Algorithm
UUID String Algorithm
GUID Algorithms
Identity columns and Sequences
Assigned Identifiers

5.1.5. composite-id
version (optional)
timestamp (optional)
component, dynamic-component
map, set, list, bag

5.2. NHibernate Types

5.2.1. Entities and values
Basic value types
Custom value types
Any type mappings

5.3. SQL quoted identifiers
Modular mapping files
Generated Properties
Auxiliary Database Objects

6. Collection Mapping

6.1. Persistent Collections
Mapping a Collection
Collections of Values and Many-To-Many Associations
6.4. One-To-Many
Lazy Initialization
Sorted Collections
Using an <idbag>
Bidirectional Associations
Ternary Associations
Heterogeneous Associations
Collection examples

7. Component Mapping

7.1. Dependent objects
Collections of dependent objects
Components as IDictionary indices
Components as composite identifiers
Dynamic components

8. Inheritance Mapping

8.1. The Three Strategies

8.1.1. Table per class hierarchy
Table per subclass
Table per subclass, using a discriminator
Mixing table per class hierarchy with table per subclass
Table per concrete class
Table per concrete class, using implicit polymorphism
Mixing implicit polymorphism with other inheritance mappings

8.2. Limitations

9. Manipulating Persistent Data

9.1. Creating a persistent object
Loading an object

9.3.1. Scalar queries
The IQuery interface
Filtering collections
Criteria queries
Queries in native SQL

9.4. Updating objects

9.4.1. Updating in the same ISession
Updating detached objects
Reattaching detached objects

9.5. Deleting persistent objects
Ending a Session

9.7.1. Flushing the Session
Committing the database transaction
Closing the ISession

9.8. Exception handling
Lifecyles and object graphs
Metadata API

10. Read-only entities

10.1. Making persistent entities read-only

10.1.1. Entities of immutable classes
Loading persistent entities as read-only
Loading read-only entities from an HQL query/criteria
10.1.4. Making
a persistent entity read-only

10.2. Read-only affect on property type

10.2.1. Simple properties
Unidirectional associations Unidirectional one-to-one and many-to-one
Unidirectional one-to-many and many-to-many

10.2.3. Bidirectional associations Bidirectional one-to-one
Bidirectional one-to-many/many-to-one
Bidirectional many-to-many

11. Transactions And Concurrency

11.1. Configurations, Sessions and Factories
11.2. Threads
and connections
Considering object identity
Optimistic concurrency control

11.4.1. Long session with automatic versioning
Many sessions with automatic versioning
Customizing automatic versioning
Application version checking

11.5. Session disconnection
Pessimistic Locking
Connection Release Modes
Interceptors and events

12.1. Interceptors
Event system
Batch processing

13.1. Batch inserts
The StatelessSession interface
DML-style operations

14. HQL: The Hibernate Query Language

14.1. Case Sensitivity
The from clause
Associations and joins
The select clause
Aggregate functions
Polymorphic queries
The where clause
The order by clause
The group by clause
HQL examples
Tips & Tricks

15. Criteria Queries

15.1. Creating an ICriteria instance
Narrowing the result set
Ordering the results
Dynamic association fetching
Example queries
Projections, aggregation and grouping
Detached queries and subqueries

16. QueryOver Queries

16.1. Structure of a Query
Simple Expressions
Additional Restrictions
Projection Functions

17. Native SQL

17.1. Using an ISQLQuery
Scalar queries
Entity queries
Handling associations and collections
Returning multiple entities Alias and property references
Returning non-managed entities
Handling inheritance

17.2. Named SQL queries

17.2.1. Using return-property to explicitly specify column/alias names
17.2.2. Using
stored procedures for querying Rules/limitations for using stored procedures

17.3. Custom SQL for create, update and delete
17.4. Custom SQL for

18. Filtering data

18.1. NHibernate filters

19. Improving performance

19.1. Fetching strategies

19.1.1. Working with lazy associations
Tuning fetch strategies
Single-ended association proxies
Initializing collections and proxies
Using batch fetching
Using subselect fetching

19.2. The Second Level Cache

19.2.1. Cache mappings
Strategy: read only
Strategy: read/write
Strategy: nonstrict read/write

19.3. Managing the caches
The Query Cache
Understanding Collection performance

19.5.1. Taxonomy
Lists, maps, idbags and sets are the most efficient collections to update
Bags and lists are the most efficient inverse collections
One shot delete

19.6. Batch updates
Multi Query
Multi Criteria

20. Toolset Guide

20.1. Schema Generation

20.1.1. Customizing the schema
Running the tool
Using Ant

20.2. Code Generation

21. Example: Parent/Child

21.1. A note about collections
Bidirectional one-to-many
Cascading lifecycle
Using cascading Update()

22. Example: Weblog Application

22.1. Persistent Classes
NHibernate Mappings
NHibernate Code

23. Example: Various Mappings

23.1. Employer/Employee

24. Best Practices
NHibernateContrib Documentation


25.1. How to use a cache?
Prevalence Cache Configuration
SysCache Configuration
SysCache2 Configuration

25.4.1. Table-based Dependency
Command-Based Dependencies
Aggregate Dependencies
Additional Settings

26. NHibernate.Mapping.Attributes

26.1. What’s new?
How to use it?
Known issues and TODOs
Developer Notes


Working with object-oriented software and a relational database can be cumbersomen and time consuming in today’s enterprise environments. NHibernate is an object/relational mapping tool for .NET environments. The term object/relational mapping (ORM) refers to the technique of mapping a data representation from an object model to a relational data model with a SQL-based schema.

NHibernate not only takes care of the mapping from .NET classes to database tables (and from .NET data types to SQL data types), but also provides data query and retrieval facilities and can significantly reduce development time otherwise spent with manual data handling in SQL and ADO.NET.

NHibernate’s goal is to relieve the developer from 95 percent of common data persistence related programming tasks. NHibernate may not be the best solution for data-centric applications that only use stored-procedures to implement the business logic in the database, it is most useful with object-oriented domain models and business logic in the .NET-based middle-tier. However, NHibernate can certainly help you to remove or encapsulate vendor-specific SQL code and will help with the common task of result set translation from a tabular representation to a graph of objects.

If you are new to NHibernate and Object/Relational Mapping or even .NET Framework, please follow these steps:

  1. Read Chapter 1, Quickstart with IIS and Microsoft SQL Server for a 30 minute tutorial, using Internet Information Services (IIS) web server.

  2. Read Chapter 2, Architecture to understand the environments where NHibernate can be used.

  3. Use this reference documentation as your primary source of information. Consider reading Hibernate in Action (java http://www.manning.com/bauer/) or NHibernate in Action (http://www.manning.com/kuate/) or NHibernate 3.0 Cookbook (https://www.packtpub.com/nhibernate-3-0-cookbook/book) or
    NHibernate 2 Beginner’s Guide (http://www.packtpub.com/nhibernate-2-x-beginners-guide/book)if you need more help with application design or if you prefer a step-by-step tutorial. Also visit http://nhibernate.sourceforge.net/NHibernateEg/ for NHibernate tutorial with examples.

  4. FAQs are answered on the NHibernate users group.

  5. The Community Area on the NHibernate website is a good source for design patterns and various integration solutions (ASP.NET, Windows Forms).

If you have questions, use the NHibernate user forum. We also provide a JIRA issue trackings system for bug reports and feature requests. If you are interested in the development of NHibernate, join the developer mailing list. If you are interested in translating this documentation into your language, contact us on the developer mailing list.

Chapter 1. Quickstart with IIS and Microsoft SQL Server

1.1. Getting started with NHibernate

This tutorial explains a setup of NHibernate 2.1.0 within a Microsoft environment. The tools used in this tutorial are:

  1. Microsoft Internet Information Services (IIS) – web server supporting ASP.NET.
  2. Microsoft SQL Server 2005 – the database server. This tutorial uses the desktop edition (SQL-EXPRESS), a free download from Microsoft. Support for other databases is only a matter of changing the NHibernate SQL dialect and driver configuration.
  3. Microsoft Visual Studio .NET (at leats 2005) – the development environment.

First, we have to create a new Web project. We use the name QuickStart,the project web virtual directory will http://localhost/QuickStart. In the project, add a reference to NHibernate.dll. Visual Studio will automatically copy the library and its dependencies to the project output directory. If you are using a database other than SQL Server, add a reference to the driver assembly to your project.

We now set up the database connection information for NHibernate. To do this, open the file Web.config automatically generated for your project and add configuration elements according to the listing below:

<?xml version="1.0" encoding="utf-8" ?>
  <!-- Add this element -->

    <!-- Add this element -->
        <property name="dialect">
        <property name="connection.provider">
        <property name="connection.connection_string">
initial catalog=quickstart;Integrated Security=True
        <property name="proxyfactory.factory_class">
        <mapping assembly="QuickStart" />

    <!-- Leave the system.web section unchanged -->

The <configSections> element contains definitions
of sections that follow and handlers to use to process their content. We declare
the handler for the configuration section here. The <hibernate-configuration>
section contains the configuration itself, telling NHibernate that we will use a
Microsoft SQL Server 2000 database and connect to it through the specified connection
string. The dialect is a required setting, databases differ in their interpretation
of the SQL “standard”. NHibernate will take care of the differences and comes bundled
with dialects for several major commercial and open source databases.

An ISessionFactory is NHibernate’s concept of a single
datastore, multiple databases can be used by creating multiple XML configuration
files and creating multiple Configuration and
objects in your application.

The last element of the <hibernate-configuration>
section declares QuickStart as the name of an assembly
containing class declarations and mapping files. The mapping files contain the metadata
for the mapping of the POCO class to a database table (or multiple tables). We’ll
come back to mapping files soon. Let’s write the POCO class first and then declare
the mapping metadata for it.

1.2. First persistent class

NHibernate works best with the Plain Old CLR Objects (POCOs, sometimes called Plain
Ordinary CLR Objects) programming model for persistent classes. A POCO has its data
accessible through the standard .NET property mechanisms, shielding the internal
representation from the publicly visible interface:

namespace QuickStart
    public class Cat
        private string id;
        private string name;
        private char   sex;
        private float  weight;

        public Cat()

        public virtual string Id
            get { return id; }
            set { id = value; }

        public virtual string Name
            get { return name; }
            set { name = value; }

        public virtual char Sex
            get { return sex; }
            set { sex = value; }

        public virtual float Weight
            get { return weight; }
            set { weight = value; }

NHibernate is not restricted in its usage of property types, all .NET types and
primitives (like string, char
and DateTime) can be mapped, including classes from the
System.Collections namespace. You can map them as values,
collections of values, or associations to other entities. The Id
is a special property that represents the database identifier (primary key) of that
class, it is highly recommended for entities like a Cat.
NHibernate can use identifiers only internally, without having to declare them on
the class, but we would lose some of the flexibility in our application architecture.

No special interface has to be implemented for persistent classes nor do we have
to subclass from a special root persistent class. NHibernate also doesn’t use any
build time processing, such as IL manipulation, it relies solely on .NET reflection
and runtime class enhancement (through Castle.DynamicProxy library). So, without
any dependency in the POCO class on NHibernate, we can map it to a database table.

For the above mentioned runtime class enhancement to work, NHibernate requires that
all public properties of an entity class are declared as virtual.

1.3. Mapping the cat

The Cat.hbm.xml mapping file contains the metadata required
for the object/relational mapping. The metadata includes declaration of persistent
classes and the mapping of properties (to columns and foreign key relationships
to other entities) to database tables.

Please note that the Cat.hbm.xml should be set to an embedded resource.

<?xml version="1.0" encoding="utf-8" ?>
<hibernate-mapping xmlns="urn:nhibernate-mapping-2.2"
    namespace="QuickStart" assembly="QuickStart">

    <class name="Cat" table="Cat">

    <!-- A 32 hex character is our surrogate key. It's automatically
            generated by NHibernate with the UUID pattern. -->
       <id name="Id">
         <column name="CatId" sql-type="char(32)" not-null="true"/>
            <generator class="uuid.hex" />

        <!-- A cat has to have a name, but it shouldn' 
               be too long. -->
        <property name="Name">
            <column name="Name" length="16" not-null="true" />
        <property name="Sex" />
        <property name="Weight" />


Every persistent class should have an identifer attribute (actually, only classes
representing entities, not dependent value objects, which are mapped as components
of an entity). This property is used to distinguish persistent objects: Two cats
are equal if catA.Id.Equals(catB.Id) is true, this concept
is called database identity. NHibernate comes
bundled with various identifer generators for different scenarios (including native
generators for database sequences, hi/lo identifier tables, and application assigned
identifiers). We use the UUID generator (only recommended for testing, as integer
surrogate keys generated by the database should be prefered) and also specify the
column CatId of the table Cat
for the NHibernate generated identifier value (as a primary key of the table).

All other properties of Cat are mapped to the same table.
In the case of the Name property, we mapped it with an
explicit database column declaration. This is especially useful when the database
schema is automatically generated (as SQL DDL statements) from the mapping declaration
with NHibernate’s SchemaExport tool. All
other properties are mapped using NHibernate’s default settings, which is what you
need most of the time. The table Cat in the database looks
like this:

 Column |     Type     | Modifiers
 CatId  | char(32)     | not null, primary key
 Name   | nvarchar(16) | not null
 Sex    | nchar(1)     |
 Weight | real         |

You should now create the database and this table manually, and later read Chapter 20, Toolset Guide
if you want to automate this step with the SchemaExport tool. This tool can create
a full SQL DDL, including table definition, custom column type constraints, unique
constraints and indexes. If you are using SQL Server, you should also make sure
the ASPNET user has permissions to use the database.

1.4. Playing with cats

We’re now ready to start NHibernate’s ISession. It is the
persistence manager interface, we use it
to store and retrieve Cats to and from the database. But
first, we’ve to get an ISession (NHibernate’s unit-of-work)
from the ISessionFactory:

ISessionFactory sessionFactory =
            new Configuration().Configure().BuildSessionFactory();

An ISessionFactory is responsible for one database and
may only use one XML configuration file (Web.config or
hibernate.cfg.xml). You can set other properties (and even
change the mapping metadata) by accessing the Configuration
before you build the ISessionFactory
(it is immutable). Where do we create the ISessionFactory
and how can we access it in our application?

An ISessionFactory is usually only built once, e.g. at
startup inside Application_Start event handler. This also
means you should not keep it in an instance variable in your ASP.NET pages, but
in some other location. Furthermore, we need some kind of
, so we can access the ISessionFactory
easily in application code. The approach shown next solves both problems: configuration
and easy access to a ISessionFactory.

We implement a NHibernateHelper helper class:

using System;
using System.Web;
using NHibernate;
using NHibernate.Cfg;

namespace QuickStart
    public sealed class NHibernateHelper
        private const string CurrentSessionKey = 
        private static readonly ISessionFactory sessionFactory;

        static NHibernateHelper()
            sessionFactory = new Configuration()

        public static ISession GetCurrentSession()
            HttpContext context = HttpContext.Current;
            ISession currentSession = context
                                         as ISession;

            if (currentSession == null)
                currentSession = sessionFactory.OpenSession();
                context.Items[CurrentSessionKey] = currentSession;

            return currentSession;

        public static void CloseSession()
            HttpContext context = HttpContext.Current;
            ISession currentSession = context
                                           as ISession;

            if (currentSession == null)
                // No current session


        public static void CloseSessionFactory()
            if (sessionFactory != null)

This class does not only take care of the ISessionFactory
with its static attribute, but also has code to remember the ISession
for the current HTTP request.

An ISessionFactory is threadsafe, many threads can access
it concurrently and request ISessions. An
is a non-threadsafe object that represents a single unit-of-work
with the database. ISessions are opened by an
and are closed when all work is completed:

ISession session = NHibernateHelper.GetCurrentSession();

ITransaction tx = session.BeginTransaction();

Cat princess = new Cat();
princess.Name = "Princess";
princess.Sex = 'F';
princess.Weight = 7.4f;



In an ISession, every database operation occurs inside
a transaction that isolates the database operations (even read-only operations).
We use NHibernate’s ITransaction API to abstract from the
underlying transaction strategy (in our case, ADO.NET transactions). Please note
that the example above does not handle any exceptions.

Also note that you may call NHibernateHelper.GetCurrentSession();
as many times as you like, you will always get the current ISession
of this HTTP request. You have to make sure the ISession
is closed after your unit-of-work completes, either in Application_EndRequest
event handler in your application class or in a HttpModule
before the HTTP response is sent. The nice side effect of the latter is easy lazy
initialization: the ISession is still open when the view
is rendered, so NHibernate can load unitialized objects while you navigate the graph.

NHibernate has various methods that can be used to retrieve objects from the database.
The most flexible way is using the Hibernate Query Language (HQL), which is an easy
to learn and powerful object-oriented extension to SQL:

ITransaction tx = session.BeginTransaction();

IQuery query = session.CreateQuery(
          "select c from Cat as c where c.Sex = :sex");
query.SetCharacter("sex", 'F');
foreach (Cat cat in query.Enumerable())
    Console.Out.WriteLine("Female Cat: " + cat.Name);


NHibernate also offers an object-oriented query by criteria
API that can be used to formulate type-safe queries. NHibernate of course uses IDbCommands and parameter binding for all SQL communication
with the database. You may also use NHibernate’s direct SQL query feature or get
a plain ADO.NET connection from an ISession in rare cases.

1.5. Finally

We only scratched the surface of NHibernate in this small tutorial. Please note
that we don’t include any ASP.NET specific code in our examples. You have to create
an ASP.NET page yourself and insert the NHibernate code as you see fit.

Keep in mind that NHibernate, as a data access layer, is tightly integrated into
your application. Usually, all other layers depend on the persistence mechanism.
Make sure you understand the implications of this design.

Chapter 2. Architecture

2.1. Overview

A (very) high-level view of the NHibernate architecture:


This diagram shows NHibernate using the database and configuration data to provide
persistence services (and persistent objects) to the application.

We would like to show a more detailed view of the runtime architecture. Unfortunately,
NHibernate is flexible and supports several approaches. We will show the two extremes.
The “lite” architecture has the application provide its own ADO.NET connections
and manage its own transactions. This approach uses a minimal subset of NHibernate’s

The “full cream” architecture abstracts the application away from the underlying
ADO.NET APIs and lets NHibernate take care of the details.

Heres some definitions of the objects in the diagrams:

ISessionFactory (NHibernate.ISessionFactory)

A threadsafe (immutable) cache of compiled mappings for a single database. A factory
for ISession and a client of IConnectionProvider.
Might hold an optional (second-level) cache of data that is reusable between transactions,
at a process- or cluster-level.

ISession (NHibernate.ISession)

A single-threaded, short-lived object representing a conversation between the application
and the persistent store. Wraps an ADO.NET connection. Factory for
. Holds a mandatory (first-level) cache of persistent objects,
used when navigating the object graph or looking up objects by identifier.

Persistent Objects and Collections

Short-lived, single threaded objects containing persistent state and business function.
These might be ordinary POCOs, the only special thing about them is that they are
currently associated with (exactly one) ISession. As soon
as the Session is closed, they will be detached and free
to use in any application layer (e.g. directly as data transfer objects to and from

Transient Objects and Collections

Instances of persistent classes that are not currently associated with a
. They may have been instantiated by the application and not (yet)
persisted or they may have been instantiated by a closed ISession.

ITransaction (NHibernate.ITransaction)

(Optional) A single-threaded, short-lived object used by the application to specify
atomic units of work. Abstracts application from underlying ADO.NET transaction.
An ISession might span several ITransactions
in some cases.

IConnectionProvider (NHibernate.Connection.IConnectionProvider)

(Optional) A factory for ADO.NET connections and commands. Abstracts application
from the concrete vendor-specific implementations of IDbConnection
and IDbCommand. Not exposed to application, but can be
extended/implemented by the developer.

IDriver (NHibernate.Driver.IDriver)

(Optional) An interface encapsulating differences between ADO.NET providers, such
as parameter naming conventions and supported ADO.NET features.

ITransactionFactory (NHibernate.Transaction.ITransactionFactory)

(Optional) A factory for ITransaction instances. Not exposed
to the application, but can be extended/implemented by the developer.

Given a “lite” architecture, the application bypasses the ITransaction/ITransactionFactory and/or IConnectionProvider
APIs to talk to ADO.NET directly.

2.2. Instance states

An instance of a persistent classes may be in one of three different states, which
are defined with respect to a persistence context.
The NHibernate ISession object is the persistence context:


The instance is not, and has never been associated with any persistence context.
It has no persistent identity (primary key value).


The instance is currently associated with a persistence context. It has a persistent
identity (primary key value) and, perhaps, a corresponding row in the database.
For a particular persistence context, NHibernate guarantees
that persistent identity is equivalent to CLR identity (in-memory location of the


The instance was once associated with a persistence context, but that context was
closed, or the instance was serialized to another process. It has a persistent identity
and, perhaps, a corrsponding row in the database. For detached instances, NHibernate
makes no guarantees about the relationship between persistent identity and CLR identity.

2.3. Contextual Sessions

Most applications using NHibernate need some form of “contextual” sessions, where
a given session is in effect throughout the scope of a given context. However, across
applications the definition of what constitutes a context is typically different;
and different contexts define different scopes to the notion of current.

Starting with version 1.2, NHibernate added the ISessionFactory.GetCurrentSession()
method. The processing behind ISessionFactory.GetCurrentSession()
is pluggable. An extension interface (NHibernate.Context.ICurrentSessionContext)
and a new configuration parameter (hibernate.current_session_context_class)
have been added to allow pluggability of the scope and context of defining current

See the API documentation for the NHibernate.Context.ICurrentSessionContext
interface for a detailed discussion of its contract. It defines a single method,
CurrentSession(), by which the implementation is responsible
for tracking the current contextual session. Out-of-the-box, NHibernate comes with
several implementations of this interface:

  • NHibernate.Context.ManagedWebSessionContext – current sessions
    are tracked by HttpContext. However, you are responsible
    to bind and unbind an ISession instance with static methods
    on this class, it never opens, flushes, or closes an ISession

  • NHibernate.Context.CallSessionContext – current sessions
    are tracked by CallContext. You are responsible to bind
    and unbind an ISession instance with static methods of
    class CurrentSessionContext .

  • NHibernate.Context.ThreadStaticSessionContext – current
    session is stored in a thread-static variable. This context only supports one session
    factory. You are responsible to bind and unbind an ISession
    instance with static methods of class CurrentSessionContext.

  • NHibernate.Context.WebSessionContext – analogous to ManagedWebSessionContext above, stores the current session
    in HttpContext. You are responsible to bind and unbind
    an ISession instance with static methods of class

  • NHibernate.Context.WcfOperationSessionContext – current
    sessions are tracked by WCF OperationContext. You need
    to register the WcfStateExtension extension in WCF. You
    are responsible to bind and unbind an ISession instance
    with static methods of class CurrentSessionContext.

The hibernate.current_session_context_class configuration
parameter defines which NHibernate.Context.ICurrentSessionContext
implementation should be used. Typically, the value of this parameter would just
name the implementation class to use (including the assembly name); for the out-of-the-box
implementations, however, there are corresponding short names: “managed_web”, “call”,
“thread_static”, “web” and “wcf_operation”, respectively.

Chapter 3. ISessionFactory Configuration

Because NHibernate is designed to operate in many different environments, there
are a large number of configuration parameters. Fortunately, most have sensible
default values and NHibernate is distributed with an example App.config
file (found in src\NHibernate.Test) that shows the various
options. You usually only have to put that file in your project and customize it.

3.1. Programmatic Configuration

An instance of NHibernate.Cfg.Configuration represents
an entire set of mappings of an application’s .NET types to a SQL database. The
Configuration is used to build an (immutable)
. The mappings are compiled from various XML mapping files.

You may obtain a Configuration instance by instantiating
it directly. Heres an example of setting up a datastore from mappings defined in
two XML configuration files:

Configuration cfg = new Configuration()

An alternative (sometimes better) way is to let NHibernate load a mapping file from
an embedded resource:

Configuration cfg = new Configuration()

Then NHibernate will look for mapping files named NHibernate.Auction.Item.hbm.xml
and NHibernate.Auction.Bid.hbm.xml embedded as resources
in the assembly that the types are contained in. This approach eliminates any hardcoded

Another alternative (probably the best) way is to let NHibernate load all of the
mapping files contained in an Assembly:

Configuration cfg = new Configuration()
    .AddAssembly( "NHibernate.Auction" );

Then NHibernate will look through the assembly for any resources that end with .hbm.xml. This approach eliminates any hardcoded filenames
and ensures the mapping files in the assembly get added.

If a tool like Visual Studio .NET or NAnt is used to build the assembly, then make
sure that the .hbm.xml files are compiled into the assembly
as Embedded Resources.

A Configuration also specifies various optional properties:

IDictionary props = new Hashtable();
Configuration cfg = new Configuration()

A Configuration is intended as a configuration-time object,
to be discarded once an ISessionFactory is built.

3.2. Obtaining an ISessionFactory

When all mappings have been parsed by the Configuration,
the application must obtain a factory for ISession instances.
This factory is intended to be shared by all application threads:

ISessionFactory sessions = cfg.BuildSessionFactory();

However, NHibernate does allow your application to instantiate more than one ISessionFactory. This is useful if you are using more than
one database.

3.3. User provided ADO.NET connection

An ISessionFactory may open an ISession
on a user-provided ADO.NET connection. This design choice frees the application
to obtain ADO.NET connections wherever it pleases:

IDbConnection conn = myApp.GetOpenConnection();
ISession session = sessions.OpenSession(conn);

// do some data access work

The application must be careful not to open two concurrent ISessions
on the same ADO.NET connection!

3.4. NHibernate provided ADO.NET

Alternatively, you can have the ISessionFactory open connections
for you. The ISessionFactory must be provided with ADO.NET
connection properties in one of the following ways:

  1. Pass an instance of IDictionary mapping property names
    to property values to Configuration.SetProperties().

  2. Add the properties to a configuration section in the application configuration file.
    The section should be named nhibernate and its handler
    set to System.Configuration.NameValueSectionHandler.

  3. Include <property> elements in a configuration section
    in the application configuration file. The section should be named
    and its handler set to NHibernate.Cfg.ConfigurationSectionHandler.
    The XML namespace of the section should be set to urn:nhibernate-configuration-2.2.

  4. Include <property> elements in hibernate.cfg.xml
    (discussed later).

If you take this approach, opening an ISession is as simple

ISession session = sessions.OpenSession(); // open a new Session
// do some data access work, an ADO.NET connection will be 
   used on demand

All NHibernate property names and semantics are defined on the class
. We will now describe the most important settings
for ADO.NET connection configuration.

NHibernate will obtain (and pool) connections using an ADO.NET data provider if
you set the following properties:

Table 3.1. NHibernate ADO.NET Properties

Property name Purpose
The type of a custom

if the Provider is built into NHibernate, or full.classname.of.ConnectionProvider,
if using an implementation of IConnectionProvider
not included in NHibernate.

The type of a custom IDriver, if using

full.classname.of.Driver if the Driver is built into NHibernate,
or full.classname.of.Driver, assembly if using an implementation
of IDriver not included in NHibernate.

This is usually not needed, most of the time the dialect
will take care of setting the IDriver using a sensible
default. See the API documentation of the specific dialect for the defaults.

Connection string to use to obtain the connection.
The name of the connection string (defined in <connectionStrings>
configuration file element) to use to obtain the connection.
connection.isolation Set the ADO.NET transaction isolation level. Check System.Data.IsolationLevel
for meaningful values and the database’s documentation to ensure that level is supported.

eg. Chaos, ReadCommitted, ReadUncommitted,
RepeatableRead, Serializable, Unspecified

Specify when NHibernate should release ADO.NET connections. See Section 11.7, “Connection Release Modes”.

eg. auto (default) |
| after_transaction

Note that this setting only affects ISessions returned
from ISessionFactory.OpenSession. For ISessions
obtained through ISessionFactory.GetCurrentSession, the
ICurrentSessionContext implementation configured for use
controls the connection release mode for those ISessions.
See Section 2.3, “Contextual Sessions”.

command_timeout Specify the default timeout of IDbCommands generated by
adonet.batch_size Specify the batch size to use when batching update statements. Setting this to 0
(the default) disables the functionality. See Section 19.6, “Batch updates”.

This is an example of how to specify the database connection properties inside a

<?xml version="1.0" encoding="utf-8" ?>
		<section name="hibernate-configuration"

		    <property name="dialect">
		    <property name="connection.connection_string">
			Server=(local);initial catalog=theDb;
                        Integrated Security=SSPI
		    <property name="connection.isolation">
                    <property name="proxyfactory.factory_class">

    <!-- other app specific config follows -->

NHibernate relies on the ADO.NET data provider implementation of connection pooling.

You may define your own plugin strategy for obtaining ADO.NET connections by implementing
the interface NHibernate.Connection.IConnectionProvider.
You may select a custom implementation by setting connection.provider_class.

3.5. Optional configuration properties

There are a number of other properties that control the behaviour of NHibernate
at runtime. All are optional and have reasonable default values.

System-level properties can only be set manually by setting static properties of
NHibernate.Cfg.Environment class or be defined in the <nhibernate> section of the application configuration
file. These properties cannot be set using Configuration.SetProperties
or be defined in the <hibernate-configuration> section
of the application configuration file.

Table 3.2. NHibernate Configuration Properties

Property name Purpose
dialect The classname of a NHibernate Dialect – enables
certain platform dependent features.

eg. full.classname.of.Dialect, assembly

default_schema Qualify unqualified tablenames with the given schema/tablespace in generated SQL.


max_fetch_depth Set a maximum “depth” for the outer join fetch tree for single-ended associations
(one-to-one, many-to-one). A 0 disables default outer join

eg. recommended values between 0
and 3

use_reflection_optimizer Enables use of a runtime-generated class to set or get properties of an entity or
component instead of using runtime reflection (System-level property). The use of
the reflection optimizer inflicts a certain startup cost on the application but
should lead to better performance in the long run. You can not set this property
in hibernate.cfg.xml or <hibernate-configuration>
section of the application configuration file.

eg. true |

bytecode.provider Specifies the bytecode provider to use to optimize the use of reflection in NHibernate.
Use null to disable the optimization completely,
to use lightweight code generation.

eg. null |

cache.provider_class The classname of a custom ICacheProvider.

eg. classname.of.CacheProvider, assembly

Optimize second-level cache operation to minimize writes,
at the cost of more frequent reads (useful for clustered caches).

eg. true |

cache.use_query_cache Enable the query cache, individual queries still have to be
set cacheable.

eg. true |

The classname of a custom IQueryCacheFactory interface,
defaults to the built-in StandardQueryCacheFactory.

eg. classname.of.QueryCacheFactory,

cache.region_prefix A prefix to use for second-level cache region names.

eg. prefix

query.substitutions Mapping from tokens in NHibernate queries to SQL tokens (tokens might be function
or literal names, for example).

eg. hqlLiteral=SQL_LITERAL, hqlFunction=SQLFUNC

show_sql Write all SQL statements to console.

eg. true |

hbm2ddl.auto Automatically export schema DDL to the database when the ISessionFactory
is created. With create-drop, the database schema will
be dropped when the ISessionFactory is closed explicitly.

eg. create |

hbm2ddl.keywords Automatically import reserved/keywords from the database
when the ISessionFactory is created.

none : disable any operation regarding RDBMS KeyWords

keywords : imports all RDBMS KeyWords where the
can provide the implementation of IDataBaseSchema.

auto-quote : imports all RDBMS KeyWords and auto-quote
all table-names/column-names .

eg. none |
| auto-quote

use_proxy_validator Enables or disables validation of interfaces or classes specified as proxies. Enabled
by default.

eg. true |

The classname of a custom ITransactionFactory implementation,
defaults to the built-in AdoNetWithDistributedTransactionFactory.

eg. classname.of.TransactionFactory,

3.5.1. SQL Dialects

You should always set the dialect property to the correct
NHibernate.Dialect.Dialect subclass for your database.
This is not strictly essential unless you wish to use native
or sequence primary key generation or pessimistic locking
(with, eg. ISession.Lock() or IQuery.SetLockMode()).
However, if you specify a dialect, NHibernate will use sensible defaults for some
of the other properties listed above, saving you the effort of specifying them manually.

Table 3.3. NHibernate SQL Dialects (dialect)

RDBMS Dialect
DB2 DB2Dialect  
DB2 for iSeries (OS/400) DB2400Dialect  
Ingres IngresDialect  
PostgreSQL PostgreSQLDialect  
PostgreSQL 8.1 PostgreSQL81Dialect This dialect supports
available in PostgreSQL
PostgreSQL 8.2 PostgreSQL82Dialect This dialect supports
keyword in
available in PostgreSQL
MySQL 3 or 4 MySQLDialect  
MySQL 5 MySQL5Dialect  
Oracle Oracle8iDialect  
Oracle 9 Oracle9iDialect  
Oracle 10g Oracle10gDialect  
Sybase Adaptive Server Enterprise 15 SybaseASE15Dialect  
Sybase Adaptive Server Anywhere 9 SybaseASA9Dialect  
Sybase Adaptive Server Anywhere 10 SybaseASA10Dialect Deprecated. Use the Sybase SQL
Anywhere 10 Dialect instead.
Sybase SQL Anywhere 10 SybaseSQLAnywhere10Dialect  
Sybase SQL Anywhere 11 SybaseSQLAnywhere11Dialect  
Microsoft SQL Server 7 MsSql7Dialect  
Microsoft SQL Server 2000 MsSql2000Dialect  
Microsoft SQL Server 2005 MsSql2005Dialect  
Microsoft SQL Server 2008 MsSql2008Dialect  
Microsoft SQL Server Compact Edition MsSqlCeDialect  
Firebird FirebirdDialect Set driver_class


for Firebird ADO.NET provider 2.0.
SQLite SQLiteDialect Set driver_class

for System.Data.SQLite provider
for .NET 2.0.

Additional dialects may be available in the NHibernate.Dialect namespace.

3.5.2. Outer Join Fetching

If your database supports ANSI or Oracle style outer joins,
outer join fetching
might increase performance by limiting the number
of round trips to and from the database (at the cost of possibly more work performed
by the database itself). Outer join fetching allows a graph of objects connected
by many-to-one, one-to-many or one-to-one associations to be retrieved in a single

By default, the fetched graph when loading an objects ends at leaf objects, collections,
objects with proxies, or where circularities occur.

For a particular association, fetching may
be configured (and the default behaviour overridden) by setting the
attribute in the XML mapping.

Outer join fetching may be disabled globally
by setting the property max_fetch_depth to
. A setting of 1 or higher enables outer join
fetching for one-to-one and many-to-one associations which have been mapped with

Section 19.1, “Fetching strategies”
for more information.

In NHibernate 1.0, outer-join attribute could be used to
achieve a similar effect. This attribute is now deprecated in favor of

3.5.3. Custom

You may integrate a process-level (or clustered) second-level cache system by implementing
the interface NHibernate.Cache.ICacheProvider. You may
select the custom implementation by setting cache.provider_class.
See the
Section 19.2, “The Second Level Cache”
for more details.

3.5.4. Query Language

You may define new NHibernate query tokens using query.substitutions.
For example:

query.substitutions true=1, false=0

would cause the tokens true and false
to be translated to integer literals in the generated SQL.

query.substitutions toLowercase=LOWER

would allow you to rename the SQL LOWER function.

3.6. Logging

NHibernate logs various events using Apache log4net.

You may download log4net from http://logging.apache.org/log4net/.
To use log4net you will need a log4net configuration section
in the application configuration file. An example of the configuration section is
distributed with NHibernate in the src/NHibernate.Test

We strongly recommend that you familiarize yourself with NHibernate’s log messages.
A lot of work has been put into making the NHibernate log as detailed as possible,
without making it unreadable. It is an essential troubleshooting device. Also don’t
forget to enable SQL logging as described above (show_sql),
it is your first step when looking for performance problems.

3.7. Implementing an

The interface NHibernate.Cfg.INamingStrategy allows you
to specify a “naming standard” for database objects and schema elements.

You may provide rules for automatically generating database identifiers from .NET
identifiers or for processing “logical” column and table names given in the mapping
file into “physical” table and column names. This feature helps reduce the verbosity
of the mapping document, eliminating repetitive noise (TBL_
prefixes, for example). The default strategy used by NHibernate is quite minimal.

You may specify a different strategy by calling Configuration.SetNamingStrategy()
before adding mappings:

ISessionFactory sf = new Configuration()

NHibernate.Cfg.ImprovedNamingStrategy is a built-in strategy
that might be a useful starting point for some applications.

3.8. XML Configuration File

An alternative approach is to specify a full configuration in a file named
. This file can be used as a replacement for the
or <hibernate-configuration>
sections of the application configuration file.

The XML configuration file is by default expected to be in your application directory.
Here is an example:

<?xml version='1.0' encoding='utf-8'?>
<hibernate-configuration xmlns="urn:nhibernate-configuration-2.2">

    <!-- an ISessionFactory instance -->

        <!-- properties -->
        <property name="connection.provider">
        <property name="connection.driver_class">
         Server=localhost;initial catalog=nhibernate; 
         User Id=;Password=
        <property name="show_sql">
        <property name="dialect">

        <!-- mapping files -->
            assembly="NHibernate.Auction" />
            assembly="NHibernate.Auction" />



Configuring NHibernate is then as simple as

              ISessionFactory sf = new Configuration().Configure()

You can pick a different XML configuration file using

ISessionFactory sf = new Configuration()

Chapter 4. Persistent Classes

Persistent classes are classes in an application that implement the entities of
the business problem (e.g. Customer and Order in an E-commerce application). Persistent
classes have, as the name implies, transient and also persistent instance stored
in the database.

NHibernate works best if these classes follow some simple rules, also known as the
Plain Old CLR Object (POCO) programming model.

4.1. A simple POCO example

Most .NET applications require a persistent class representing felines.

using System;
using Iesi.Collections;

namespace Eg
    public class Cat
        long id; // identifier
        public virtual long Id
            get { return id; }
            protected set { id = value; }

        public virtual string Name { get; set; } 
        public virtual Cat Mate { get; set; } 
        public virtual DateTime Birthdate { get; set; } 
        public virtual float Weight { get; set; } 
        public virtual Color Color { get; set; } 
        public virtual ISet Kittens { get; set; } 
        public virtual char Sex { get; set; } 

        // AddKitten not needed by NHibernate
        public virtual void AddKitten(Cat kitten)

There are four main rules to follow here:

4.1.1. Declare properties for
persistent fields

Cat declares properties for all the persistent fields.
Many other ORM tools directly persist instance variables. We believe it is far better
to decouple this implementation detail from the persistence mechanism. NHibernate
persists properties, using their getter and setter methods.

Properties need not be declared public –
NHibernate can persist a property with an internal, protected, protected internal
or private visibility.

As shown in the example, both automatic properties and properties with a backing
field are supported.

4.1.2. Implement a default

Cat has an implicit default (no-argument) constructor.
All persistent classes must have a default constructor (which may be non-public)
so NHibernate can instantiate them using Activator.CreateInstance().

4.1.3. Provide an identifier
property (optional)

Cat has a property called Id.
This property holds the primary key column of a database table. The property might
have been called anything, and its type might have been any primitive type,
or System.DateTime. (If your legacy database
table has composite keys, you can even use a user-defined class with properties
of these types – see the section on composite identifiers below.)

The identifier property is optional. You can leave it off and let NHibernate keep
track of object identifiers internally. However, for many applications it is still
a good (and very popular) design decision.

What’s more, some functionality is available only to classes which declare an identifier

  • Cascaded updates (see “Lifecycle Objects”)

  • ISession.SaveOrUpdate()

We recommend you declare consistently-named identifier properties on persistent

4.1.4. Prefer non-sealed classes
and virtual methods (optional)

A central feature of NHibernate, proxies,
depends upon the persistent class being non-sealed and all its public methods, properties
and events declared as virtual. Another possibility is for the class to implement
an interface that declares all public members.

You can persist sealed classes that do not implement an
interface and don’t have virtual members with NHibernate, but you won’t be able
to use proxies – which will limit your options for performance tuning.

4.2. Implementing inheritance

A subclass must also observe the first and second rules. It inherits its identifier
property from Cat.

using System;
namespace Eg
    public class DomesticCat : Cat
        public virtual string Name { get; set; }

4.3. Implementing
and GetHashCode()

You have to override the Equals() and GetHashCode()
methods if you intend to mix objects of persistent classes (e.g. in an

This only applies if these objects are loaded in two different
ISessions, as NHibernate only guarantees identity (
a == b
, the default implementation of Equals())
inside a single ISession!

Even if both objects a and b are
the same database row (they have the same primary key value as their identifier),
we can’t guarantee that they are the same object instance outside of a particular
ISession context.

The most obvious way is to implement Equals()/GetHashCode()
by comparing the identifier value of both objects. If the value is the same, both
must be the same database row, they are therefore equal (if both are added to an
ISet, we will only have one element in the
). Unfortunately, we can’t use that approach. NHibernate will only
assign identifier values to objects that are persistent, a newly created instance
will not have any identifier value! We recommend implementing Equals()
and GetHashCode() using Business
key equality

Business key equality means that the Equals() method compares
only the properties that form the business key, a key that would identify our instance
in the real world (a natural candidate key):

public class Cat

    public override bool Equals(object other)
        if (this == other) return true;
        Cat cat = other as Cat;
        if (cat == null) return false; // null or not a cat

        if (Name != cat.Name) return false;
        if (!Birthday.Equals(cat.Birthday)) return false;

        return true;

    public override int GetHashCode()
            int result;
            result = Name.GetHashCode();
            result = 29 * result + Birthday.GetHashCode();
            return result;


Keep in mind that our candidate key (in this case a composite of name and birthday)
has to be only valid for a particular comparison operation (maybe even only in a
single use case). We don’t need the stability criteria we usually apply to a real
primary key!

4.4. Dynamic models

Note that the following features are currently considered
experimental and may change in the near future.

Persistent entities don’t necessarily have to be represented as POCO classes at
runtime. NHibernate also supports dynamic models (using Dictionaries
of Dictionarys at runtime) . With this approach, you don’t
write persistent classes, only mapping files.

By default, NHibernate works in normal POCO mode. You may set a default entity representation
mode for a particular ISessionFactory using the
configuration option (see Table 3.2, “NHibernate
Configuration Properties”

The following examples demonstrates the representation using Maps
(Dictionary). First, in the mapping file, an entity-name
has to be declared instead of (or in addition to) a class name:


    <class entity-name="Customer">

        <id name="id"
            <generator class="sequence"/>

        <property name="name"

        <property name="address"

        <many-to-one name="organization"

        <bag name="orders"
            <key column="CUSTOMER_ID"/>
            <one-to-many class="Order"/>


Note that even though associations are declared using target class names, the target
type of an associations may also be a dynamic entity instead of a POCO.

After setting the default entity mode to dynamic-map for
the ISessionFactory, we can at runtime work with
of Dictionaries:

using(ISession s = OpenSession())
using(ITransaction tx = s.BeginTransaction())
    // Create a customer
    var frank = new Dictionary<string, object>();
    frank["name"] = "Frank";

    // Create an organization
    var foobar = new Dictionary<string, object>();
    foobar["name"] = "Foobar Inc.";

    // Link both
    frank["organization"] =  foobar;

    // Save both
    s.Save("Customer", frank);
    s.Save("Organization", foobar);


The advantages of a dynamic mapping are quick turnaround time for prototyping without
the need for entity class implementation. However, you lose compile-time type checking
and will very likely deal with many exceptions at runtime. Thanks to the NHibernate
mapping, the database schema can easily be normalized and sound, allowing to add
a proper domain model implementation on top later on.

Entity representation modes can also be set on a per ISession

using (ISession dynamicSession = pocoSession
    // Create a customer
    var frank = new Dictionary<string, object>();
    frank["name"] = "Frank";
    dynamicSession.Save("Customer", frank);
// Continue on pocoSession

Please note that the call to GetSession() using an
is on the ISession API, not the
. That way, the new ISession
shares the underlying ADO connection, transaction, and other context information.
This means you don’t have tocall Flush() and
on the secondary ISession, and also leave
the transaction and connection handling to the primary unit of work.

4.5. Tuplizers

NHibernate.Tuple.Tuplizer, and its sub-interfaces, are
responsible for managing a particular representation of a piece of data, given that
representation’s NHibernate.EntityMode. If a given piece
of data is thought of as a data structure, then a tuplizer is the thing which knows
how to create such a data structure and how to extract values from and inject values
into such a data structure. For example, for the POCO entity mode, the correpsonding
tuplizer knows how create the POCO through its constructor and how to access the
POCO properties using the defined property accessors. There are two high-level types
of Tuplizers, represented by the NHibernate.Tuple.Entity.IEntityTuplizer
and NHibernate.Tuple.Component.IComponentTuplizer interfaces.
IEntityTuplizers are responsible for managing the above
mentioned contracts in regards to entities, while IComponentTuplizers
do the same for components.

Users may also plug in their own tuplizers. Perhaps you require that a
implementation other than System.Collections.Hashtable
be used while in the dynamic-map entity-mode; or perhaps you need to define a different
proxy generation strategy than the one used by default. Both would be achieved by
defining a custom tuplizer implementation. Tuplizers definitions are attached to
the entity or component mapping they are meant to manage. Going back to the example
of our customer entity:

    <class entity-name="Customer">
            Override the dynamic-map entity-mode
            tuplizer for the customer entity
        <tuplizer entity-mode="dynamic-map"

        <id name="id" type="long" column="ID">
            <generator class="sequence"/>

        <!-- other properties -->

public class CustomMapTuplizerImpl 
     : NHibernate.Tuple.Entity.DynamicMapEntityTuplizer
    // override the BuildInstantiator() 
    //method to plug in our custom map...
    protected override IInstantiator 
             NHibernate.Mapping.PersistentClass mappingInfo)
        return new CustomMapInstantiator(mappingInfo);

    private sealed class CustomMapInstantiator 
      : NHibernate.Tuple.DynamicMapInstantiator
        // override the generateMap() 
        //method to return our custom map...
        protected override IDictionary GenerateMap()
            return new CustomMap();

4.6. Lifecycle Callbacks

Optionally, a persistent class might implement the interface ILifecycle
which provides some callbacks that allow the persistent object to perform necessary
initialization/cleanup after save or load and before deletion or update.

The NHibernate IInterceptor offers a less intrusive alternative,

public interface ILifecycle
        LifecycleVeto OnSave(ISession s);                            (2)
        LifecycleVeto OnUpdate(ISession s);                          (3)
        LifecycleVeto OnDelete(ISession s);                          (4)
        void OnLoad(ISession s, object id);

OnSave – called just before the object is saved or inserted


OnUpdate – called just before an object is updated (when
the object is passed to ISession.Update())


OnDelete – called just before an object is deleted


OnLoad – called just after an object is loaded

OnSave(), OnDelete() and
may be used to cascade saves and deletions of dependent objects.
This is an alternative to declaring cascaded operations in the mapping file. OnLoad() may be used to initialize transient properties
of the object from its persistent state. It may not be used to load dependent objects
since the ISession interface may not be invoked from inside
this method. A further intended usage of OnLoad(),
and OnUpdate() is to store a reference
to the current ISession for later use.

Note that OnUpdate() is not called every time the object’s
persistent state is updated. It is called only when a transient object is passed
to ISession.Update().

If OnSave(), OnUpdate() or OnDelete() return LifecycleVeto.Veto,
the operation is silently vetoed. If a CallbackException
is thrown, the operation is vetoed and the exception is passed back to the application.

Note that OnSave() is called after an identifier is assigned
to the object, except when native key generation is used.

4.7. IValidatable callback

If the persistent class needs to check invariants before its state is persisted,
it may implement the following interface:

public interface IValidatable
        void Validate();

The object should throw a ValidationFailure if an invariant
was violated. An instance of Validatable should not change
its state from inside Validate().

Unlike the callback methods of the ILifecycle interface,
Validate() might be called at unpredictable times. The
application should not rely upon calls to Validate() for
business functionality.

Chapter 5. Basic O/R Mapping

5.1. Mapping declaration

Object/relational mappings are defined in an XML document. The mapping document
is designed to be readable and hand-editable. The mapping language is object-centric,
meaning that mappings are constructed around persistent class declarations, not
table declarations.

Note that, even though many NHibernate users choose to define XML mappings by hand,
a number of tools exist to generate the mapping document, including NHibernate.Mapping.Attributes
library and various template-based code generators (CodeSmith, MyGeneration).

Let’s kick off with an example mapping:

<?xml version="1.0"?>
<hibernate-mapping xmlns="urn:nhibernate-mapping-2.2" assembly="Eg"

        <class name="Cat" table="CATS" discriminator-value="C">
                <id name="Id" column="uid" type="Int64">
                        <generator class="hilo"/>
                <discriminator column="subclass" type="Char"/>
                <property name="BirthDate" type="Date"/>
                <property name="Color" not-null="true"/>
                <property name="Sex" not-null="true" update="false"/>
                <property name="Weight"/>
                <many-to-one name="Mate" column="mate_id"/>
                <set name="Kittens">
                        <key column="mother_id"/>
                        <one-to-many class="Cat"/>
                <subclass name="DomesticCat" discriminator-value="D">
                        <property name="Name" type="String"/>

        <class name="Dog">
                <!-- mapping for Dog could go here -->


We will now discuss the content of the mapping document. We will only describe the
document elements and attributes that are used by NHibernate at runtime. The mapping
document also contains some extra optional attributes and elements that affect the
database schemas exported by the schema export tool. (For example the

5.1.1. XML Namespace

All XML mappings should declare the XML namespace shown. The actual schema definition
may be found in the src\nhibernate-mapping.xsd file in
the NHibernate distribution.

Tip: to enable IntelliSense for mapping and configuration
files, copy the appropriate .xsd files as part of any project
in your solution, (Build Action can be “None”) or as “Solution
Files” or in your "Lib" folder and then add it to the Schemas property of the xml file. You can copy it in
<VS installation directory>\Xml\Schemas
, take care because you
will have to deal with different version of the xsd for different versions of NHibernate.

5.1.2. hibernate-mapping

This element has several optional attributes. The schema
attribute specifies that tables referred to by this mapping belong to the named
schema. If specified, tablenames will be qualified by the given schema name. If
missing, tablenames will be unqualified. The default-cascade
attribute specifies what cascade style should be assumed for properties and collections
which do not specify a cascade attribute. The
attribute lets us use unqualified class names in the query
language, by default. The assembly and
attributes specify the assembly where persistent classes are
located and the namespace they are declared in.

         schema="schemaName"                          (2)
         default-cascade="none|save-update"           (3)
         auto-import="true|false"                     (4)
         assembly="Eg"                                (5)
         namespace="Eg"                               (6)

schema (optional): The name of a database schema.


default-cascade (optional – defaults to
): A default cascade style.


auto-import (optional – defaults to true):
Specifies whether we can use unqualified class names (of classes in this mapping)
in the query language.


assembly and namespace(optional):
Specify assembly and namespace to assume for unqualified class names in the mapping


default-access (optional – defaults to property): The strategy
NHibernate should use for accessing a property value


default-lazy (optional – defaults to true):
Lazy fetching may be completely disabled by setting default-lazy=”false”.

If you are not using assembly and namespace
attributes, you have to specify fully-qualified class names, including the name
of the assembly that classes are declared in.

If you have two persistent classes with the same (unqualified) name, you should
set auto-import="false". NHibernate will throw an exception
if you attempt to assign two classes to the same “imported” name.

5.1.3. class

You may declare a persistent class using the class element:

        table="tableName"                             (2)
        discriminator-value="discriminator_value"     (3)
        mutable="true|false"                          (4)
        schema="owner"                                (5)
        proxy="ProxyInterface"                        (6)
        dynamic-update="true|false"                   (7)
        dynamic-insert="true|false"                   (8)
        select-before-update="true|false"             (9)
        polymorphism="implicit|explicit"              (10)
        where="arbitrary sql where condition"         (11)
        persister="PersisterClass"                    (12)
        batch-size="N"                                (13)
        optimistic-lock="none|version|dirty|all"      (14)
        lazy="true|false"                             (15)
        abstract="true|false"                         (16)

name: The fully qualified .NET class name of the persistent
class (or interface), including its assembly name.


table(optional – defaults to the unqualified class name):
The name of its database table.


discriminator-value (optional – defaults to the class name):
A value that distiguishes individual subclasses, used for polymorphic behaviour.
Acceptable values include null and not


mutable (optional, defaults to true):
Specifies that instances of the class are (not) mutable.


schema (optional): Override the schema name specified by
the root <hibernate-mapping> element.


proxy (optional): Specifies an interface to use for lazy
initializing proxies. You may specify the name of the class itself.


dynamic-update (optional, defaults to false):
Specifies that UPDATE SQL should be generated at runtime
and contain only those columns whose values have changed.


dynamic-insert (optional, defaults to false):
Specifies that INSERT SQL should be generated at runtime
and contain only the columns whose values are not null.


select-before-update (optional, defaults to
): Specifies that NHibernate should never
perform an SQL UPDATE unless it is certain that an object
is actually modified. In certain cases (actually, only when a transient object has
been associated with a new session using update()), this
means that NHibernate will perform an extra SQL SELECT
to determine if an UPDATE is actually required.


polymorphism (optional, defaults to implicit):
Determines whether implicit or explicit query polymorphism is used.


where (optional) specify an arbitrary SQL
condition to be used when retrieving objects of this class


persister (optional): Specifies a custom


batch-size (optional, defaults to 1)
specify a “batch size” for fetching instances of this class by identifier.


optimistic-lock (optional, defaults to
): Determines the optimistic locking strategy.


lazy (optional): Lazy fetching may be completely disabled
by setting lazy="false".


abstract (optional): Used to mark abstract superclasses
in <union-subclass> hierarchies.

It is perfectly acceptable for the named persistent class to be an interface. You
would then declare implementing classes of that interface using the
element. You may persist any inner class. You should specify
the class name using the standard form ie. Eg.Foo+Bar, Eg.
Due to an HQL parser limitation inner classes can not be used in queries in NHibernate

Changes to immutable classes, mutable="false", will not
be persisted. This allows NHibernate to make some minor performance optimizations.

The optional proxy attribute enables lazy initialization
of persistent instances of the class. NHibernate will initially return proxies which
implement the named interface. The actual persistent object will be loaded when
a method of the proxy is invoked. See “Proxies for Lazy Initialization” below.

Implicit polymorphism means that instances
of the class will be returned by a query that names any superclass or implemented
interface or the class and that instances of any subclass of the class will be returned
by a query that names the class itself. Explicit
polymorphism means that class instances will be returned only be queries that explicitly
name that class and that queries that name the class will return only instances
of subclasses mapped inside this <class> declaration
as a <subclass> or <joined-subclass>.
For most purposes the default, polymorphism="implicit",
is appropriate. Explicit polymorphism is useful when two different classes are mapped
to the same table (this allows a “lightweight” class that contains a subset of the
table columns).

The persister attribute lets you customize the persistence
strategy used for the class. You may, for example, specify your own subclass of
NHibernate.Persister.EntityPersister or you might even
provide a completely new implementation of the interface NHibernate.Persister.IClassPersister
that implements persistence via, for example, stored procedure calls, serialization
to flat files or LDAP. See NHibernate.DomainModel.CustomPersister
for a simple example (of “persistence” to a Hashtable).

Note that the dynamic-update and dynamic-insert
settings are not inherited by subclasses and so may also be specified on the <subclass> or <joined-subclass>
elements. These settings may increase performance in some cases, but might actually
decrease performance in others. Use judiciously.

Use of select-before-update will usually decrease performance.
It is very useful to prevent a database update trigger being called unnecessarily.

If you enable dynamic-update, you will have a choice of
optimistic locking strategies:

  • version check the version/timestamp columns

  • all check all columns

  • dirty check the changed columns

  • none do not use optimistic locking

We very strongly recommend that you use version/timestamp
columns for optimistic locking with NHibernate. This is the optimal strategy with
respect to performance and is the only strategy that correctly handles modifications
made outside of the session (ie. when ISession.Update()
is used). Keep in mind that a version or timestamp property should never be null,
no matter what unsaved-value strategy, or an instance will
be detected as transient.

Beginning with NHibernate 1.2.0, version numbers start with 1, not 0 as in previous
versions. This was done to allow using 0 as unsaved-value
for the version property.

5.1.4. id

Mapped classes must declare the primary key
column of the database table. Most classes will also have a property holding the
unique identifier of an instance. The <id> element
defines the mapping from that property to the primary key column.

        type="typename"                          (2)
        column="column_name"                     (3)
        unsaved-value="any|none|null|id_value"   (4)

        <generator class="generatorClass"/>

name (optional): The name of the identifier property.


type (optional): A name that indicates the NHibernate type.


column (optional – defaults to the property name): The
name of the primary key column.


unsaved-value (optional – defaults to a “sensible” value):
An identifier property value that indicates that an instance is newly instantiated
(unsaved), distinguishing it from transient instances that were saved or loaded
in a previous session.


access (optional – defaults to property):
The strategy NHibernate should use for accessing the property value.

If the name attribute is missing, it is assumed that the
class has no identifier property.

The unsaved-value attribute is almost never needed in NHibernate

There is an alternative <composite-id> declaration
to allow access to legacy data with composite keys. We strongly discourage its use
for anything else. generator

The required generator names a .NET class used to generate
unique identifiers for instances of the persistent class.

The generator can be declared using the <generator>
child element. If any parameters are required to configure or initialize the generator
instance, they are passed using <param> elements.

<id name="Id" type="Int64" column="uid" unsaved-value="0">
        <generator class="NHibernate.Id.TableHiLoGenerator">
                <param name="table">uid_table</param>
                <param name="column">next_hi_value_column</param>

If no parameters are required, the generator can be declared using a
attribute directly on the <id>
element, as follows:

   <id name="Id" type="Int64" column="uid" 
     unsaved-value="0" generator="native" />

All generators implement the interface NHibernate.Id.IIdentifierGenerator.
This is a very simple interface; some applications may choose to provide their own
specialized implementations. However, NHibernate provides a range of built-in implementations.
There are shortcut names for the built-in generators:


generates identifiers of any integral type that are unique only when no other process
is inserting data into the same table. Do not use in a cluster.


supports identity columns in DB2, MySQL, MS SQL Server and Sybase. The identifier
returned by the database is converted to the property type using
. Any integral property type is thus supported.


uses a sequence in DB2, PostgreSQL, Oracle or a generator in Firebird. The identifier
returned by the database is converted to the property type using
. Any integral property type is thus supported.


uses a hi/lo algorithm to efficiently
generate identifiers of any integral type, given a table and column (by default
hibernate_unique_key and next_hi
respectively) as a source of hi values. The hi/lo algorithm generates identifiers
that are unique only for a particular database. Do not use
this generator with a user-supplied connection.

You can use the “where” parameter to specify the row to use in a table. This is
useful if you want to use a single tabel for your identifiers, with different rows
for each table.


uses a hi/lo algorithm to efficiently generate identifiers of any integral type,
given a named database sequence.


uses System.Guid and its ToString(string
method to generate identifiers of type string. The length of the
string returned depends on the configured format.


uses a new System.Guid to create a byte[]
that is converted to a string.


uses a new System.Guid as the identifier.


uses the algorithm to generate a new System.Guid described
by Jimmy Nilsson in the article http://www.informit.com/articles/article.asp?p=25862.


picks identity, sequence or hilo depending upon the capabilities of the underlying


lets the application to assign an identifier to the object before
is called.


uses the identifier of another associated object. Usually used in conjunction with
a <one-to-one> primary key association. Hi/Lo Algorithm

The hilo and seqhilo generators
provide two alternate implementations of the hi/lo algorithm, a favorite approach
to identifier generation. The first implementation requires a “special” database
table to hold the next available “hi” value. The second uses an Oracle-style sequence
(where supported).

<id name="Id" type="Int64" column="cat_id">
        <generator class="hilo">
                <param name="table">hi_value</param>
                <param name="column">next_value</param>
                <param name="max_lo">100</param>
<id name="Id" type="Int64" column="cat_id">
        <generator class="seqhilo">
                <param name="sequence">hi_value</param>
                <param name="max_lo">100</param>

Unfortunately, you can’t use hilo when supplying your own
IDbConnection to NHibernate. NHibernate must be able to
fetch the “hi” value in a new transaction. UUID Hex Algorithm

<id name="Id" type="String" column="cat_id">
        <generator class="uuid.hex">
            <param name="format">format_value</param>
            <param name="seperator">seperator_value</param>

The UUID is generated by calling Guid.NewGuid().ToString(format).
The valid values for format are described in the MSDN documentation.
The default seperator is - and
should rarely be modified. The format determines if the
configured seperator can replace the default seperator
used by the format. UUID String Algorithm

The UUID is generated by calling Guid.NewGuid().ToByteArray()
and then converting the byte[] into a char[].
The char[] is returned as a String
consisting of 16 characters. GUID Algorithms

The guid identifier is generated by calling
. To address some of the performance concerns with using
Guids as primary keys, foreign keys, and as part of indexes with MS SQL the
can be used. The benefit of using the guid.comb
with other databases that support GUIDs has not been measured. Identity columns and

For databases which support identity columns (DB2, MySQL, Sybase, MS SQL), you may
use identity key generation. For databases that support
sequences (DB2, Oracle, PostgreSQL, Interbase, McKoi, SAP DB) you may use
style key generation. Both these strategies require two SQL queries
to insert a new object.

<id name="Id" type="Int64" column="uid">
        <generator class="sequence">
                <param name="sequence">uid_sequence</param>
<id name="Id" type="Int64" column="uid" unsaved-value="0">
        <generator class="identity"/>

For cross-platform development, the native strategy will
choose from the identity, sequence
and hilo strategies, dependent upon the capabilities of
the underlying database. Assigned Identifiers

If you want the application to assign identifiers (as opposed to having NHibernate
generate them), you may use the assigned generator. This
special generator will use the identifier value already assigned to the object’s
identifier property. Be very careful when using this feature to assign keys with
business meaning (almost always a terrible design decision).

Due to its inherent nature, entities that use this generator cannot be saved via
the ISession’s SaveOrUpdate() method. Instead you have to explicitly specify to
NHibernate if the object should be saved or updated by calling either the
or Update() method of the ISession.

5.1.5. composite-id


        <key-property name="PropertyName" 
            type="typename" column="column_name"/>
        <key-many-to-one name="PropertyName 
             class="ClassName" column="column_name"/>

For a table with a composite key, you may map multiple properties of the class as
identifier properties. The <composite-id> element
accepts <key-property> property mappings and
mappings as child elements.

        <key-property name="MedicareNumber"/>
        <key-property name="Dependent"/>

Your persistent class must override
and GetHashCode() to implement composite
identifier equality. It must also be Serializable.

Unfortunately, this approach to composite identifiers means that a persistent object
is its own identifier. There is no convenient “handle” other than the object itself.
You must instantiate an instance of the persistent class itself and populate its
identifier properties before you can load() the persistent
state associated with a composite key. We will describe a much more convenient approach
where the composite identifier is implemented as a seperate class in Section 7.4, “Components as
composite identifiers”
. The attributes described below apply only to this
alternative approach:

  • name (optional, required for this approach): A property
    of component type that holds the composite identifier (see next section).

  • access (optional – defaults to property):
    The strategy NHibernate should use for accessing the property value.

  • class (optional – defaults to the property type determined
    by reflection): The component class used as a composite identifier (see next section).

5.1.6. discriminator

The <discriminator> element is required for polymorphic
persistence using the table-per-class-hierarchy mapping strategy and declares a
discriminator column of the table. The discriminator column contains marker values
that tell the persistence layer what subclass to instantiate for a particular row.
A restricted set of types may be used: String,
, Int32, Byte,
, Boolean, YesNo,

        type="discriminator_type"      (2)
        force="true|false"             (3)
        insert="true|false"            (4)
        formula="arbitrary SQL expressi(5)on"

column (optional – defaults to class)
the name of the discriminator column.


type (optional – defaults to String)
a name that indicates the NHibernate type


force (optional – defaults to false)
“force” NHibernate to specify allowed discriminator values even when retrieving
all instances of the root class.


insert (optional – defaults to true)
set this to false if your discriminator column is also
part of a mapped composite identifier.


formula (optional) an arbitrary SQL expression that is
executed when a type has to be evaluated. Allows content-based discrimination.

Actual values of the discriminator column are specified by the discriminator-value
attribute of the <class> and <subclass>

The force attribute is (only) useful if the table contains
rows with “extra” discriminator values that are not mapped to a persistent class.
This will not usually be the case.

Using the formula attribute you can declare an arbitrary
SQL expression that will be used to evaluate the type of a row:

    formula="case when CLASS_TYPE in ('a', 'b', 'c') 
              then 0 else 1 end"

5.1.7. version (optional)

The <version> element is optional and indicates that
the table contains versioned data. This is particularly useful if you plan to use
long transactions (see below).

        name="PropertyName"                                (2)
        type="typename"                                    (3)
        access="field|property|nosetter|ClassName"         (4)
        unsaved-value="null|negative|undefined|value"      (5)
        generated="never|always"                           (6)

column (optional – defaults to the property name): The
name of the column holding the version number.


name: The name of a property of the persistent class.


type (optional – defaults to Int32):
The type of the version number.


access (optional – defaults to property):
The strategy NHibernate should use for accessing the property value.


unsaved-value (optional – defaults to a “sensible” value):
A version property value that indicates that an instance is newly instantiated (unsaved),
distinguishing it from transient instances that were saved or loaded in a previous
session. (undefined specifies that the identifier property
value should be used.)


generated (optional – defaults to never):
Specifies that this version property value is actually generated by the database.
See the discussion of Section 5.5, “Generated Properties”.

Version numbers may be of type Int64, Int32,
Int16, Ticks,
, or TimeSpan (or their nullable counterparts
in .NET 2.0).

5.1.8. timestamp (optional)

The optional <timestamp> element indicates that the
table contains timestamped data. This is intended as an alternative to versioning.
Timestamps are by nature a less safe implementation of optimistic locking. However,
sometimes the application might use the timestamps in other ways.

        name="PropertyName"                 (2)
        generated="never|always"            (5)

column (optional – defaults to the property name): The
name of a column holding the timestamp.


name: The name of a property of .NET type
of the persistent class.


access (optional – defaults to property):
The strategy NHibernate should use for accessing the property value.


unsaved-value (optional – defaults to null):
A timestamp property value that indicates that an instance is newly instantiated
(unsaved), distinguishing it from transient instances that were saved or loaded
in a previous session. (undefined specifies that the identifier
property value should be used.)


generated (optional – defaults to never):
Specifies that this timestamp property value is actually generated by the database.
See the discussion of Section 5.5, “Generated Properties”.

Note that <timestamp> is equivalent to
<version type="timestamp">

5.1.9. property

The <property> element declares a persistent property
of the class.

        column="column_name"                (2)
        type="typename"                     (3)
        update="true|false"                 (4)
        insert="true|false"                 (4)
        formula="arbitrary SQL expression"  (5)
        access="field|property|ClassName"   (6)
        optimistic-lock="true|false"        (7)
        generated="never|insert|always"     (8)
        lazy="true|false"                   (9)

name: the name of the property of your class.


column (optional – defaults to the property name): the
name of the mapped database table column.


type (optional): a name that indicates the NHibernate type.


update, insert (optional – defaults to
) : specifies that the mapped columns should be included in SQL
and/or INSERT statements. Setting both
to false allows a pure “derived” property whose value is
initialized from some other property that maps to the same column(s) or by a trigger
or other application.


formula (optional): an SQL expression that defines the
value for a computed property. Computed properties
do not have a column mapping of their own.


access (optional – defaults to property):
The strategy NHibernate should use for accessing the property value.


optimistic-lock (optional – defaults to
): Specifies that updates to this property do or do not require acquisition
of the optimistic lock. In other words, determines if a version increment should
occur when this property is dirty.


generated (optional – defaults to never):
Specifies that this property value is actually generated by the database. See the
discussion of Section 5.5, “Generated Properties”.


lazy (optional – defaults to false):
Specifies that this property is lazy. A lazy property is not loaded when the object
is initially loaded, unless the fetch mode has been overriden in a specific query.
Values for lazy properties are loaded when any lazy property of the object is accessed.

typename could be:

  1. The name of a NHibernate basic type (eg. Int32, String, Char, DateTime,
    Timestamp, Single, Byte[], Object, ...

  2. The name of a .NET type with a default basic type (eg. System.Int16,
    System.Single, System.Char, System.String, System.DateTime, System.Byte[], ...

  3. The name of an enumeration type (eg. Eg.Color, Eg).

  4. The name of a serializable .NET type.

  5. The class name of a custom type (eg. Illflow.Type.MyCustomType).

Note that you have to specify full assembly-qualified
names for all except basic NHibernate types (unless you set assembly
and/or namespace attributes of the <hibernate-mapping>

NHibernate supports .NET 2.0 Nullable types. These types
are mostly treated the same as plain non-Nullable types
internally. For example, a property of type Nullable<Int32>
can be mapped using type="Int32" or type="System.Int32".

If you do not specify a type, NHibernate will use reflection upon the named property
to take a guess at the correct NHibernate type. NHibernate will try to interpret
the name of the return class of the property getter using rules 2, 3, 4 in that
order. However, this is not always enough. In certain cases you will still need
the type attribute. (For example, to distinguish between
NHibernateUtil.DateTime and NHibernateUtil.Timestamp,
or to specify a custom type.)

The access attribute lets you control how NHibernate will
access the value of the property at runtime. The value of the access
attribute should be text formatted as access-strategy.naming-strategy.
The .naming-strategy is not always required.

Table 5.1. Access Strategies

Access Strategy Name Description

The default implementation. NHibernate uses the get/set accessors of the property.
No naming strategy should be used with this access strategy because the value of
the name attribute is the name of the property.


NHibernate will access the field directly. NHibernate uses the value of the
attribute as the name of the field. This can be used when a property’s
getter and setter contain extra actions that you don’t want to occur when NHibernate
is populating or reading the object. If you want the name of the property and not
the field to be what the consumers of your API use with HQL, then a naming strategy
is needed.


NHibernate will access the field directly when setting the value and will use the
Property when getting the value. This can be used when a property only exposes a
get accessor because the consumers of your API can’t change the value directly.
A naming strategy is required because NHibernate uses the value of the
attribute as the property name and needs to be told what the name
of the field is.


If NHibernate’s built in access strategies are not what is needed for your situation
then you can build your own by implementing the interface NHibernate.Property.IPropertyAccessor.
The value of the access attribute should be an assembly-qualified
name that can be loaded with Activator.CreateInstance(string assemblyQualifiedName).

Table 5.2. Naming Strategies

Naming Strategy Name Description

The name attribute is converted to camel case to find the
field. <property name="FooBar" ... > uses the field


The name attribute is converted to camel case and prefixed
with an underscore to find the field. <property name="FooBar"
... >
uses the field _fooBar.


The name attribute is converted to camel case and prefixed
with the character m and an underscore to find the field.
<property name="FooBar" ... > uses the field


The name attribute is converted to lower case to find the
Field. <property name="FooBar" ... > uses the field


The name attribute is converted to lower case and prefixed
with an underscore to find the Field. <property name="FooBar"
... >
uses the field _foobar.


The name attribute is prefixed with an underscore to find
the field. <property name="FooBar" ... > uses the
field _FooBar.


The name attribute is prefixed with the character
to find the field. <property name="FooBar" ... >
uses the field mFooBar.


The name attribute is prefixed with the character
and an underscore to find the field. <property name="FooBar"
... >
uses the field m_FooBar.

5.1.10. many-to-one

An ordinary association to another persistent class is declared using a
element. The relational model is a many-to-one association.
(It’s really just an object reference.)

        column="column_name"                               (2)
        class="ClassName"                                  (3)
        cascade="all|none|save-update|delete"              (4)
        fetch="join|select"                                (5)
        update="true|false"                                (6)
        insert="true|false"                                (6)
        property-ref="PropertyNameFromAssociatedClass"     (7)
        access="field|property|nosetter|ClassName"         (8)
        unique="true|false"                                (9)
        optimistic-lock="true|false"                       (10)
        not-found="ignore|exception"                       (11)

name: The name of the property.


column (optional): The name of the column.


class (optional – defaults to the property type determined
by reflection): The name of the associated class.


cascade (optional): Specifies which operations should be
cascaded from the parent object to the associated object.


fetch (optional – defaults to select):
Chooses between outer-join fetching or sequential select fetching.


update, insert (optional – defaults to
) specifies that the mapped columns should be included in SQL
and/or INSERT statements. Setting both
to false allows a pure “derived” association whose value
is initialized from some other property that maps to the same colum(s) or by a trigger
or other application.


property-ref: (optional) The name of a property of the
associated class that is joined to this foreign key. If not specified, the primary
key of the associated class is used.


access (optional – defaults to property):
The strategy NHibernate should use for accessing the property value.


unique (optional): Enable the DDL generation of a unique
constraint for the foreign-key column.


optimistic-lock (optional – defaults to
): Specifies that updates to this property do or do not require acquisition
of the optimistic lock. In other words, dertermines if a version increment should
occur when this property is dirty.


not-found (optional – defaults to exception):
Specifies how foreign keys that reference missing rows will be handled:
will treat a missing row as a null association.

The cascade attribute permits the following values: all, save-update,
, none. Setting a value other than
will propagate certain operations to the associated (child)
object. See “Lifecycle Objects” below.

The fetch attribute accepts two different values:

  • join Fetch the association using an outer join

  • select Fetch the association using a separate query

A typical many-to-one declaration looks as simple as

<many-to-one name="product" class="Product" column="PRODUCT_ID"/>

The property-ref attribute should only be used for mapping
legacy data where a foreign key refers to a unique key of the associated table other
than the primary key. This is an ugly relational model. For example, suppose the
Product class had a unique serial number, that is not the
primary key. (The unique attribute controls NHibernate’s
DDL generation with the SchemaExport tool.)

<property name="serialNumber" unique="true" 
type="string" column="SERIAL_NUMBER"/>

Then the mapping for OrderItem might use:

<many-to-one name="product" property-ref="serialNumber" 

This is certainly not encouraged, however.

5.1.11. one-to-one

A one-to-one association to another persistent class is declared using a

        class="ClassName"                                  (2)
        cascade="all|none|save-update|delete"              (3)
        constrained="true|false"                           (4)
        fetch="join|select"                                (5)
        property-ref="PropertyNameFromAssociatedClass"     (6)
        access="field|property|nosetter|ClassName"         (7)

name: The name of the property.


class (optional – defaults to the property type determined
by reflection): The name of the associated class.


cascade (optional) specifies which operations should be
cascaded from the parent object to the associated object.


constrained (optional) specifies that a foreign key constraint
on the primary key of the mapped table references the table of the associated class.
This option affects the order in which Save() and
are cascaded (and is also used by the schema export tool).


fetch (optional – defaults to select):
Chooses between outer-join fetching or sequential select fetching.


property-ref: (optional) The name of a property of the
associated class that is joined to the primary key of this class. If not specified,
the primary key of the associated class is used.


access (optional – defaults to property):
The strategy NHibernate should use for accessing the property value.

There are two varieties of one-to-one association:

  • primary key associations

  • unique foreign key associations

Primary key associations don’t need an extra table column; if two rows are related
by the association then the two table rows share the same primary key value. So
if you want two objects to be related by a primary key association, you must make
sure that they are assigned the same identifier value!

For a primary key association, add the following mappings to Employee
and Person, respectively.

<one-to-one name="Person" class="Person"/>
<one-to-one name="Employee" class="Employee" constrained="true"/>

Now we must ensure that the primary keys of related rows in the PERSON and EMPLOYEE
tables are equal. We use a special NHibernate identifier generation strategy called

<class name="Person" table="PERSON">
    <id name="Id" column="PERSON_ID">
        <generator class="foreign">
            <param name="property">Employee</param>
    <one-to-one name="Employee"

A newly saved instance of Person is then assigned the same
primar key value as the Employee instance refered with
the Employee property of that Person.

Alternatively, a foreign key with a unique constraint, from Employee
to Person, may be expressed as:

<many-to-one name="Person" class="Person" column="PERSON_ID" 

And this association may be made bidirectional by adding the following to the Person mapping:

<one-to-one name="Employee" class="Employee" property-ref="Person"/>

5.1.12. natural-id

<natural-id mutable="true|false"/>
        <property ... />
        <many-to-one ... />

Even though we recommend the use of surrogate keys as primary keys, you should still
try to identify natural keys for all entities. A natural key is a property or combination
of properties that is unique and non-null. If it is also immutable, even better.
Map the properties of the natural key inside the <natural-id>
element. NHibernate will generate the necessary unique key and nullability constraints,
and your mapping will be more self-documenting.

We strongly recommend that you implement Equals() and GetHashCode() to compare the natural key properties of
the entity.

This mapping is not intended for use with entities with natural primary keys.

  • mutable (optional, defaults to false):
    By default, natural identifier properties as assumed to be immutable (constant).

5.1.13. component, dynamic-component

The <component> element maps properties of a child
object to columns of the table of a parent class. Components may, in turn, declare
their own properties, components or collections. See “Components” below.

        class="ClassName"                                  (2)
        insert="true|false"                                (3)
        upate="true|false"                                 (4)
        access="field|property|nosetter|ClassName"         (5)
        optimistic-lock="true|false">                      (6)
        <property ...../>
        <many-to-one .... />

name: The name of the property.


class (optional – defaults to the property type determined
by reflection): The name of the component (child) class.


insert: Do the mapped columns appear in SQL


update: Do the mapped columns appear in SQL


access (optional – defaults to property):
The strategy NHibernate should use for accessing the property value.


optimistic-lock (optional – defaults to
): Specifies that updates to this component do or do not require acquisition
of the optimistic lock. In other words, determines if a version increment should
occur when this property is dirty.

The child <property> tags map properties of the child
class to table columns.

The <component> element allows a
subelement that maps a property of the component class as
a reference back to the containing entity.

The <dynamic-component> element allows an
to be mapped as a component, where the property names refer
to keys of the dictionary.

5.1.14. properties

The <properties> element allows the definition of
a named, logical grouping of the properties of a class. The most important use of
the construct is that it allows a combination of properties to be the target of
a property-ref. It is also a convenient way to define a
multi-column unique constraint. For example:

      insert="true|false"                                  (2)
      update="true|false"                                  (3)
      optimistic-lock="true|false"                         (4)
      unique="true|false">                                 (5)

      <property .../>
      <many-to-one .../>

name: the logical name of the grouping. It is
an actual property name.


insert: do the mapped columns appear in SQL


update: do the mapped columns appear in SQL


optimistic-lock (optional – defaults to
): specifies that updates to these properties either do or do not require
acquisition of the optimistic lock. It determines if a version increment should
occur when these properties are dirty.


unique (optional – defaults to false):
specifies that a unique constraint exists upon all mapped columns of the component.

For example, if we have the following <properties>

<class name="Person">
      <id name="personNumber" />
      <properties name="name" unique="true" update="false">
          <property name="firstName" />
          <property name="lastName" />
          <property name="initial" />

You might have some legacy data association that refers to this unique key of the
Person table, instead of to the primary key:

<many-to-one name="owner" class="Person" property-ref="name">
        <column name="firstName" />
        <column name="lastName" />
        <column name="initial" />

The use of this outside the context of mapping legacy data is not recommended.

5.1.15. subclass

Finally, polymorphic persistence requires the declaration of each subclass of the
root persistent class. For the (recommended) table-per-class-hierarchy mapping strategy,
the <subclass> declaration is used.

        discriminator-value="discriminator_value"     (2)
        proxy="ProxyInterface"                        (3)
        lazy="true|false"                             (4)

        <property .... />
        <properties .... />

name: The fully qualified .NET class name of the subclass,
including its assembly name.


discriminator-value (optional – defaults to the class name):
A value that distiguishes individual subclasses.


proxy (optional): Specifies a class or interface to use
for lazy initializing proxies.


lazy (optional, defaults to true):
Setting lazy="false" disables the use of lazy fetching.

Each subclass should declare its own persistent properties and subclasses.
and <id> properties are
assumed to be inherited from the root class. Each subclass in a hierarchy must define
a unique discriminator-value. If none is specified, the
fully qualified .NET class name is used.

For information about inheritance mappings, see Chapter 8, Inheritance Mapping.

5.1.16. joined-subclass

Alternatively, a subclass that is persisted to its own table (table-per-subclass
mapping strategy) is declared using a <joined-subclass>

        proxy="ProxyInterface"              (2)
        lazy="true|false"                   (3)

        <key .... >

        <property .... />
        <properties .... />

name: The fully qualified class name of the subclass.


proxy (optional): Specifies a class or interface to use
for lazy initializing proxies.


lazy (optional): Setting lazy="true"
is a shortcut equalivalent to specifying the name of the class itself as the proxy interface.

No discriminator column is required for this mapping strategy. Each subclass must,
however, declare a table column holding the object identifier using the
element. The mapping at the start of the chapter would be re-written

<?xml version="1.0"?>
<hibernate-mapping xmlns="urn:nhibernate-mapping-2.2" assembly="Eg"

        <class name="Cat" table="CATS">
                <id name="Id" column="uid" type="Int64">
                        <generator class="hilo"/>
                <property name="BirthDate" type="Date"/>
                <property name="Color" not-null="true"/>
                <property name="Sex" not-null="true"/>
                <property name="Weight"/>
                <many-to-one name="Mate"/>
                <set name="Kittens">
                        <key column="MOTHER"/>
                        <one-to-many class="Cat"/>
                <joined-subclass name="DomesticCat" 
                    <key column="CAT"/>
                        <property name="Name" type="String"/>

        <class name="Dog">
                <!-- mapping for Dog could go here -->


For information about inheritance mappings, see Chapter 8, Inheritance Mapping.

5.1.17. union-subclass

A third option is to map only the concrete classes of an inheritance hierarchy to
tables, (the table-per-concrete-class strategy) where each table defines all persistent
state of the class, including inherited state. In NHibernate, it is not absolutely
necessary to explicitly map such inheritance hierarchies. You can simply map each
class with a separate <class> declaration. However,
if you wish use polymorphic associations (e.g. an association to the superclass
of your hierarchy), you need to use the <union-subclass>

        table="tablename"                   (2)
        proxy="ProxyInterface"              (3)
        lazy="true|false"                   (4)
        subselect="SQL expression"

        <property .... />
        <properties .... />

name: The fully qualified class name of the subclass.


table: The name of the subclass table.


proxy (optional): Specifies a class or interface to use
for lazy initializing proxies.


lazy (optional, defaults to true):
Setting lazy="false" disables the use of lazy fetching.

No discriminator column or key column is required for this mapping strategy.

For information about inheritance mappings, see Chapter 8, Inheritance Mapping.

5.1.18. join

Using the <join> element, it is possible to map properties
of one class to several tables, when there’s a 1-to-1 relationship between the tables.

        schema="owner"                           (2)
        fetch="join|select"                      (3)
        inverse="true|false"                     (4)
        optional="true|false">                   (5)

        <key ... />

        <property ... />

table: The name of the joined table.


schema (optional): Override the schema name specified by
the root <hibernate-mapping> element.


fetch (optional – defaults to join):
If set to join, the default, NHibernate will use an inner
join to retrieve a <join> defined by a class or its
superclasses and an outer join for a <join> defined
by a subclass. If set to select then NHibernate will use
a sequential select for a <join> defined on a subclass,
which will be issued only if a row turns out to represent an instance of the subclass.
Inner joins will still be used to retrieve a <join>
defined by the class and its superclasses.


inverse (optional – defaults to false):
If enabled, NHibernate will not try to insert or update the properties defined by
this join.


optional (optional – defaults to false):
If enabled, NHibernate will insert a row only if the properties defined by this
join are non-null and will always use an outer join to retrieve the properties.

For example, the address information for a person can be mapped to a separate table
(while preserving value type semantics for all properties):

<class name="Person"

    <id name="id" column="PERSON_ID">...</id>

    <join table="ADDRESS">
        <key column="ADDRESS_ID"/>
        <property name="address"/>
        <property name="zip"/>
        <property name="country"/>

This feature is often only useful for legacy data models, we recommend fewer tables
than classes and a fine-grained domain model. However, it is useful for switching
between inheritance mapping strategies in a single hierarchy, as explained later.

5.1.19. map, set, list, bag

Collections are discussed later.

5.1.20. import

Suppose your application has two persistent classes with the same name, and you
don’t want to specify the fully qualified name in NHibernate queries. Classes may
be “imported” explicitly, rather than relying upon auto-import="true".
You may even import classes and interfaces that are not explicitly mapped.

<import class="System.Object" rename="Universe"/>
        class="ClassName"              (1)
        rename="ShortName"             (2)

class: The fully qualified class name of any .NET class,
including its assembly name.


rename (optional – defaults to the unqualified class name):
A name that may be used in the query language.

5.2. NHibernate Types

5.2.1. Entities and values

To understand the behaviour of various .NET language-level objects with respect
to the persistence service, we need to classify them into two groups:

An entity exists independently of any other
objects holding references to the entity. Contrast this with the usual .NET model
where an unreferenced object is garbage collected. Entities must be explicitly saved
and deleted (except that saves and deletions may be cascaded
from a parent entity to its children). This is different from the ODMG model of
object persistence by reachability – and corresponds more closely to how application
objects are usually used in large systems. Entities support circular and shared
references. They may also be versioned.

An entity’s persistent state consists of references to other entities and instances
of value types. Values are primitives, collections,
components and certain immutable objects. Unlike entities, values (in particular
collections and components) are persisted
and deleted by reachability. Since value objects (and primitives) are persisted
and deleted along with their containing entity they may not be independently versioned.
Values have no independent identity, so they cannot be shared by two entities or

All NHibernate types except collections support null semantics if the .NET type
is nullable (i.e. not derived from System.ValueType).

Up until now, we’ve been using the term “persistent class” to refer to entities.
We will continue to do that. Strictly speaking, however, not all user-defined classes
with persistent state are entities. A component
is a user defined class with value semantics.

5.2.2. Basic value types

The basic types may be roughly categorized
into three groups – System.ValueType types,
types, and System.Object types for
large objects. Just like the .NET Types, columns for System.ValueType types can not store null
values and System.Object types can store
null values.

Table 5.3. System.ValueType Mapping Types

NHibernate Type .NET Type Database Type Remarks
AnsiChar Char DbType
- 1 char
Boolean Boolean DbType
Default when
Byte Byte DbType
Default when
Char Char DbType
- 1 char
Default when
DateTime DateTime DbType
– ignores the milliseconds
Default when
LocalDateTime DateTime DbType

ignores the milliseconds
Ensures the
is set to
UtcDateTime DateTime DbType

ignores the milliseconds
Ensures the
is set to
Decimal Decimal DbType
Default when
Double Double DbType
Default when
Guid Guid DbType
Default when
Int16 Int16 DbType
Default when
Int32 Int32 DbType
Default when
Int64 Int64 DbType
Default when
PersistentEnum A Enum The DbType
for the underlying
Do not specify

in the mapping. Instead
specify the Assembly
Qualified Name of
the Enum or let
NHibernate use
to “guess” the Type.
The UnderlyingType
of the Enum is used
to determine the correct
Single Single DbType
Default when no
Ticks DateTime DbType
must be
Default when no
Timestamp DateTime DbType

as specific as
database supports.

must be
TrueFalse Boolean DbType

– 1 char either
‘T’ or ‘F’

must be specified.
YesNo Boolean DbType

– 1 char either
‘Y’ or ‘N’
must be specified.

Table 5.4. System.Object Mapping Types

NHibernate Type .NET Type Database Type Remarks
AnsiString String DbType
must be specified.
CultureInfo Globalization
– 5 chars for culture
Default when no
attribute specified.
Binary Byte[] DbType
Default when no
attribute specified.
Type Type DbType
holding Assembly Qualified Name.
Default when no
attribute specified.
String String DbType
Default when no
attribute specified.

Table 5.5. Large Object Mapping Types

NHibernate Type .NET Type Database Type Remarks
StringClob String DbType
must be specified.
Entire field is read
into memory.
BinaryBlob Byte[] DbType
must be specified.
Entire field is read
into memory.
Serializable Any Object
that is marked with
should be specified.
This is the fallback
type if no NHibernate
Type can be found for the Property.

NHibernate supports some additional type names for compatibility with NHibernate
(useful for those coming over from NHibernate or using some of the tools to generate
hbm.xml files). A type="integer"
or type="int" will map to an Int32
NHibernate type, type="short" to an Int16
NHibernateType. To see all of the conversions you can view the source of static
constructor of the class NHibernate.Type.TypeFactory.

5.2.3. Custom value types

It is relatively easy for developers to create their own value types. For example,
you might want to persist properties of type Int64 to VARCHAR columns. NHibernate does not provide a built-in
type for this. But custom types are not limited to mapping a property (or collection
element) to a single table column. So, for example, you might have a property Name { get; set; } of type String
that is persisted to the columns FIRST_NAME,

To implement a custom type, implement either NHibernate.UserTypes.IUserType
or NHibernate.UserTypes.ICompositeUserType and declare
properties using the fully qualified name of the type. Check out
to see the kind of things that
are possible.

<property name="TwoStrings" type="NHibernate.DomainModel.DoubleStringType, 
    <column name="first_string"/>
    <column name="second_string"/>

Notice the use of <column> tags to map a property
to multiple columns.

The ICompositeUserType, IEnhancedUserType,
INullableUserType, IUserCollectionType,
and IUserVersionType interfaces provide support for more
specialized uses.

You may even supply parameters to an IUserType in the mapping
file. To do this, your IUserType must implement the NHibernate.UserTypes.IParameterizedType interface. To supply
parameters to your custom type, you can use the <type>
element in your mapping files.

<property name="priority">
    <type name="MyCompany.UserTypes.DefaultValueIntegerType">
        <param name="default">0</param>

The IUserType can now retrieve the value for the parameter
named default from the IDictionary
object passed to it.

If you use a certain UserType very often, it may be useful
to define a shorter name for it. You can do this using the <typedef>
element. Typedefs assign a name to a custom type, and may also contain a list of
default parameter values if the type is parameterized.

<typedef class="MyCompany.UserTypes.DefaultValueIntegerType" 
    <param name="default">0</param>
<property name="priority" type="default_zero"/>

It is also possible to override the parameters supplied in a typedef on a case-by-case
basis by using type parameters on the property mapping.

Even though NHibernate’s rich range of built-in types and support for components
means you will very rarely need to use a
custom type, it is nevertheless considered good form to use custom types for (non-entity)
classes that occur frequently in your application. For example, a
class is a good candidate for an ICompositeUserType,
even though it could easily be mapped as a component. One motivation for this is
abstraction. With a custom type, your mapping documents would be future-proofed
against possible changes in your way of representing monetary values.

5.2.4. Any type mappings

There is one further type of property mapping. 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,

<any name="AnyEntity" id-type="Int64" 
    <column name="table_name"/>
    <column name="id"/>

The meta-type attribute lets the application specify a
custom type that maps database column values to persistent classes which have identifier
properties of the type specified by id-type. If the meta-type
returns instances of System.Type, nothing else is required.
On the other hand, if it is a basic type like String or
Char, you must specify the mapping from values to classes.

<any name="AnyEntity" id-type="Int64" meta-type="String">
    <meta-value value="TBL_ANIMAL" class="Animal"/>
    <meta-value value="TBL_HUMAN" class="Human"/>
    <meta-value value="TBL_ALIEN" class="Alien"/>
    <column name="table_name"/>
    <column name="id"/>
        id-type="idtypename"                               (2)
        meta-type="metatypename"                           (3)
        cascade="none|all|save-update"                     (4)
        access="field|property|nosetter|ClassName"         (5)
        optimistic-lock="true|false"                       (6)
        <meta-value ... />
        <meta-value ... />
        <column .... />
        <column .... />

name: the property name.


id-type: the identifier type.


meta-type (optional – defaults to Type):
a type that maps System.Type to a single database column
or, alternatively, a type that is allowed for a discriminator mapping.


cascade (optional – defaults to none):
the cascade style.


access (optional – defaults to property):
The strategy NHibernate should use for accessing the property value.


optimistic-lock (optional – defaults to
): Specifies that updates to this property do or do not require acquisition
of the optimistic lock. In other words, define if a version increment should occur
if this property is dirty.

5.3. SQL quoted identifiers

You may force NHibernate to quote an identifier in the generated SQL by enclosing
the table or column name in backticks in the mapping document. NHibernate will use
the correct quotation style for the SQL Dialect (usually
double quotes, but brackets for SQL Server and backticks for MySQL).

<class name="LineItem" table="`Line Item`">
    <id name="Id" column="`Item Id`"/&gt
       ;<generator class="assigned"/>
    <property name="ItemNumber" column="`Item #`"/>

5.4. Modular mapping files

It is possible to define subclass and joined-subclass
mappings in seperate mapping documents, directly beneath hibernate-mapping.
This allows you to extend a class hierachy just by adding a new mapping file. You
must specify an extends attribute in the subclass mapping,
naming a previously mapped superclass. Use of this feature makes the ordering of
the mapping documents important!

        <subclass name="Eg.Subclass.DomesticCat, Eg"
            extends="Eg.Cat, Eg" discriminator-value="D">
             <property name="name" type="string"/>

5.5. Generated Properties

Generated properties are properties which have their values generated by the database.
Typically, NHibernate applications needed to Refresh objects
which contain any properties for which the database was generating values. Marking
properties as generated, however, lets the application delegate this responsibility
to NHibernate. Essentially, whenever NHibernate issues an SQL INSERT or UPDATE for
an entity which has defined generated properties, it immediately issues a select
afterwards to retrieve the generated values.

Properties marked as generated must additionally be non-insertable and non-updateable.
Only Section 5.1.7, “version (optional)”,

Section 5.1.8, “timestamp (optional)”
, and Section 5.1.9, “property” can be marked
as generated.

never (the default) – means that the given property value
is not generated within the database.

insert – states that the given property value is generated
on insert, but is not regenerated on subsequent updates. Things like created-date
would fall into this category. Note that even though Section 5.1.7, “version (optional)”
and Section 5.1.8, “timestamp (optional)”
properties can be marked as generated, this option is not available there…

always – states that the property value is generated both
on insert and on update.

5.6. Auxiliary Database Objects

Allows CREATE and DROP of arbitrary database objects, in conjunction with NHibernate’s
schema evolution tools, to provide the ability to fully define a user schema within
the NHibernate mapping files. Although designed specifically for creating and dropping
things like triggers or stored procedures, really any SQL command that can be run
via a IDbCommand.ExecuteNonQuery() method is valid here
(ALTERs, INSERTS, etc). There are essentially two modes for defining auxiliary database

The first mode is to explicitly list the CREATE and DROP commands out in the mapping

        <create>CREATE TRIGGER my_trigger ...</create>
        <drop>DROP TRIGGER my_trigger</drop>

The second mode is to supply a custom class which knows how to construct the CREATE
and DROP commands. This custom class must implement the NHibernate.Mapping.IAuxiliaryDatabaseObject

        <definition class="MyTriggerDefinition, MyAssembly"/>

You may also specify parameters to be passed to the database object:

        <definition class="MyTriggerDefinition, MyAssembly">
            <param name="parameterName">parameterValue</param>

NHibernate will call IAuxiliaryDatabaseObject.SetParameterValues
passing it a dictionary of parameter names and values.

Additionally, these database objects can be optionally scoped such that they only
apply when certain dialects are used.

        <definition class="MyTriggerDefinition"/>
        <dialect-scope name="NHibernate.Dialect.Oracle9Dialect"/>
        <dialect-scope name="NHibernate.Dialect.OracleDialect"/>

Chapter 6. Collection Mapping

6.1. Persistent Collections

NHibernate requires that persistent collection-valued fields be declared as an interface
type, for example:

public class Product
    private string serialNumber;
    private ISet parts = new HashedSet();
    public ISet Parts
        get { return parts; }
        set { parts = value; }

    public string SerialNumber
        get { return serialNumber; }
        set { serialNumber = value; }

The actual interface might be Iesi.Collections.ISet, System.Collections.ICollection, System.Collections.IList,
System.Collections.IDictionary, System.Collections.Generic.ICollection<T>,
System.Collections.Generic.IDictionary<K, V>
, Iesi.Collections.Generic.ISet<T>
or … anything you like! (Where “anything you like” means you will have to write
an implementation of NHibernate.UserType.IUserCollectionType.)

Notice how we initialized the instance variable with an instance of
. This is the best way to initialize collection valued properties
of newly instantiated (non-persistent) instances. When you make the instance persistent
– by calling Save(), for example – NHibernate will actually
replace the HashedSet with an instance of NHibernate’s
own implementation of ISet. Watch out for errors like this:

Cat cat = new DomesticCat();
Cat kitten = new DomesticCat();
ISet kittens = new HashedSet();
cat.Kittens = kittens;
kittens = cat.Kittens; //Okay, kittens collection is an ISet
HashedSet hs = (HashedSet) cat.Kittens; //Error!

Collection instances have the usual behavior of value types. They are automatically
persisted when referenced by a persistent object and automatically deleted when
unreferenced. If a collection is passed from one persistent object to another, its
elements might be moved from one table to another. Two entities may not share a
reference to the same collection instance. Due to the underlying relational model,
collection-valued properties do not support null value semantics; NHibernate does
not distinguish between a null collection reference and an empty collection.

You shouldn’t have to worry much about any of this. Just use NHibernate’s collections
the same way you use ordinary .NET collections, but make sure you understand the
semantics of bidirectional associations (discussed later) before using them.

Collection instances are distinguished in the database by a foreign key to the owning
entity. This foreign key is referred to as the collection
. The collection key is mapped by the <key>

Collections may contain almost any other NHibernate type, including all basic types,
custom types, entity types and components. This is an important definition: An object
in a collection can either be handled with “pass by value” semantics (it therefore
fully depends on the collection owner) or it can be a reference to another entity
with an own lifecycle. Collections may not contain other collections. The contained
type is referred to as the collection element type.
Collection elements are mapped by <element>,
, <one-to-many>,
<many-to-many> or <many-to-any>.
The first two map elements with value semantics, the other three are used to map
entity associations.

All collection types except ISet and bag have an
column – a column that maps to an array or
index or IDictionary key. The index
of an IDictionary may be of any basic type, an entity type
or even a composite type (it may not be a collection). The index of an array or
list is always of type Int32. Indexes are mapped using
<index>, <index-many-to-many>,
<composite-index> or <index-many-to-any>.

There are quite a range of mappings that can be generated for collections, covering
many common relational models. We suggest you experiment with the schema generation
tool to get a feeling for how various mapping declarations translate to database

6.2. Mapping a Collection

Collections are declared by the <set>,
, <map>, <bag>,
<array> and <primitive-array>
elements. <map> is representative:

    table="table_name"                                          (2)
    schema="schema_name"                                        (3)
    lazy="true|false"                                           (4)
    inverse="true|false"                                        (5)
    cascade="all|none|save-update|delete|all-delete-orphan"     (6)
    sort="unsorted|natural|comparatorClass"                     (7)
    order-by="column_name asc|desc"                             (8)
    where="arbitrary sql where condition"                       (9)
    fetch="select|join"                                         (10)
    batch-size="N"                                              (11)
    access="field|property|ClassName"                           (12)
    optimistic-lock="true|false"                                (13)
    generic="true|false"                                        (14)

    <key .... />
    <index .... />
    <element .... />

name the collection property name


table (optional – defaults to property name) the name of
the collection table (not used for one-to-many associations)


schema (optional) the name of a table schema to override
the schema declared on the root element


lazy (optional – defaults to true)
may be used to disable lazy fetching and specify that the association is always
eagerly fetched.


inverse (optional – defaults to false)
mark this collection as the “inverse” end of a bidirectional association


cascade (optional – defaults to none)
enable operations to cascade to child entities


sort (optional) specify a sorted collection with
sort order, or a given comparator class


order-by (optional) specify a table column (or columns)
that define the iteration order of the IDictionary, ISet or bag, together with an optional asc
or desc


where (optional) specify an arbitrary SQL
condition to be used when retrieving or removing the collection (useful
if the collection should contain only a subset of the available data)


fetch (optional) Choose between outer-join fetching and
fetching by sequential select.


batch-size (optional, defaults to 1)
specify a “batch size” for lazily fetching instances of this collection.


access (optional – defaults to property):
The strategy NHibernate should use for accessing the property value.


optimistic-lock (optional – defaults to
): Species that changes to the state of the collection results in increment
of the owning entity’s version. (For one to many associations, it is often reasonable
to disable this setting.)


generic (optional): Choose between generic and non-generic
collection interface. If this option is not specified, NHibernate will use reflection
to choose the interface.

The mapping of an IList or array requires a seperate table
column holding the array or list index (the i in
). If your relational model doesn’t have an index column, e.g. if
you’re working with legacy data, use an unordered ISet
instead. This seems to put people off who assume that IList
should just be a more convenient way of accessing an unordered collection. NHibernate
collections strictly obey the actual semantics attached to the ISet,
IList and IDictionary interfaces.
IList elements don’t just spontaneously rearrange themselves!

On the other hand, people who planned to use the IList
to emulate bag semantics have a legitimate
grievance here. A bag is an unordered, unindexed collection which may contain the
same element multiple times. The .NET collections framework lacks an
interface, hence you have to emulate it with an IList.
NHibernate lets you map properties of type IList or ICollection with the <bag>
element. Note that bag semantics are not really part of the ICollection
contract and they actually conflict with the semantics of the IList
contract (however, you can sort the bag arbitrarily, discussed later in this chapter).

Note: Large NHibernate bags mapped with inverse="false"
are inefficient and should be avoided; NHibernate can’t create, delete or update
rows individually, because there is no key that may be used to identify an individual

6.3. Collections of Values and Many-To-Many

A collection table is required for any collection of values and any collection of
references to other entities mapped as a many-to-many association (the natural semantics
for a .NET collection). The table requires (foreign) key column(s), element column(s)
and possibly index column(s).

The foreign key from the collection table to the table of the owning class is declared
using a <key> element.

<key column="column_name"/>

column (required): The name of the foreign key column.

For indexed collections like maps and lists, we require an <index>
element. For lists, this column contains sequential integers numbered from zero.
Make sure that your index really starts from zero if you have to deal with legacy
data. For maps, the column may contain any values of any NHibernate type.

        column="column_name"                (1)
        type="typename"                     (2)

column (required): The name of the column holding the collection
index values.


type (optional, defaults to Int32):
The type of the collection index.

Alternatively, a map may be indexed by objects of entity type. We use the

        class="ClassName"                   (2)

column (required): The name of the foreign key column for
the collection index values.


class (required): The entity class used as the collection

For a collection of values, we use the <element>

        type="typename"                     (2)

column (required): The name of the column holding the collection
element values.


type (required): The type of the collection element.

A collection of entities with its own table corresponds to the relational notion
of many-to-many association. A many to many
association is the most natural mapping of a .NET collection but is not usually
the best relational model.

        class="ClassName"                                  (2)
        fetch="join|select"                                (3)
        not-found="ignore|exception"                       (4)

column (required): The name of the element foreign key


class (required): The name of the associated class.


fetch (optional, defaults to join):
enables outer-join or sequential select fetching for this association. This is a
special case; for full eager fetching (in a single SELECT) of an entity and its
many-to-many relationships to other entities, you would enable join fetching not
only of the collection itself, but also with this attribute on the
nested element.


not-found (optional – defaults to exception):
Specifies how foreign keys that reference missing rows will be handled:
will treat a missing row as a null association.

Some examples, first, a set of strings:

<set name="Names" table="NAMES">
    <key column="GROUPID"/>
    <element column="NAME" type="String"/>

A bag containing integers (with an iteration order determined by the

<bag name="Sizes" table="SIZES" order-by="SIZE ASC">
    <key column="OWNER"/>
    <element column="SIZE" type="Int32"/>

An array of entities – in this case, a many to many association (note that the entities
are lifecycle objects, cascade="all"):

<array name="Foos" table="BAR_FOOS" cascade="all">
    <key column="BAR_ID"/>
    <index column="I"/>
    <many-to-many column="FOO_ID" class="Eg.Foo, Eg"/>

A map from string indices to dates:

<map name="Holidays" table="holidays" 
     schema="dbo" order-by="hol_name asc">
    <key column="id"/>
    <index column="hol_name" type="String"/>
    <element column="hol_date" type="Date"/>

A list of components (discussed in the next chapter):

<list name="CarComponents" table="car_components">
    <key column="car_id"/>
    <index column="posn"/>
    <composite-element class="Eg.Car.CarComponent">
            <property name="Price" type="float"/>
            <property name="Type" type="Eg.Car.ComponentType, Eg"/>
            <property name="SerialNumber" 
                column="serial_no" type="String"/>

6.4. One-To-Many Associations

A one to many association links the tables
of two classes directly, with no intervening
collection table. (This implements a one-to-many
relational model.) This relational model loses some of the semantics of .NET collections:

  • No null values may be contained in a dictionary, set or list

  • An instance of the contained entity class may not belong to more than one instance
    of the collection

  • An instance of the contained entity class may not appear at more than one value
    of the collection index

An association from Foo to Bar
requires the addition of a key column and possibly an index column to the table
of the contained entity class, Bar. These columns are mapped
using the <key> and <index>
elements described above.

The <one-to-many> tag indicates a one to many association.

        not-found="ignore|exception"                       (2)

class (required): The name of the associated class.


not-found (optional – defaults to exception):
Specifies how foreign keys that reference missing rows will be handled:
will treat a missing row as a null association.


<set name="Bars">
    <key column="foo_id"/>
    <one-to-many class="Eg.Bar, Eg"/>

Notice that the <one-to-many> element does not need
to declare any columns. Nor is it necessary to specify the table
name anywhere.

Very Important Note: If the
column of a <one-to-many> association
is declared NOT NULL, NHibernate may cause constraint violations
when it creates or updates the association. To prevent this problem,
you must use a bidirectional association
with the many valued
end (the set or bag) marked as inverse="true". See the
discussion of bidirectional associations later in this chapter.

6.5. Lazy Initialization

Collections (other than arrays) may be lazily initialized, meaning they load their
state from the database only when the application needs to access it. Initialization
happens transparently to the user so the application would not normally need to
worry about this (in fact, transparent lazy initialization is the main reason why
NHibernate needs its own collection implementations). However, if the application
tries something like this:

s = sessions.OpenSession();
ITransaction tx = sessions.BeginTransaction();
User u = (User) s.Find("from User u where u.Name=?", 
                        userName, NHibernateUtil.String)[0];
IDictionary permissions = u.Permissions;

int accessLevel = (int) permissions["accounts"];  // Error!

It could be in for a nasty surprise. Since the permissions collection was not initialized
when the ISession was committed, the collection will never
be able to load its state. The fix is to move the line that reads from the collection
to just before the commit. (There are other more advanced ways to solve this problem,

Alternatively, use a non-lazy collection. Since lazy initialization can lead to
bugs like that above, non-laziness is the default. However, it is intended that
lazy initialization be used for almost all collections, especially for collections
of entities (for reasons of efficiency).

Exceptions that occur while lazily initializing a collection are wrapped in a LazyInitializationException.

Declare a lazy collection using the optional lazy attribute:

<set name="Names" table="NAMES" lazy="true">
    <key column="group_id"/>
    <element column="NAME" type="String"/>

In some application architectures, particularly where the code that accesses data
using NHibernate, and the code that uses it are in different application layers,
it can be a problem to ensure that the ISession is open
when a collection is initialized. There are two basic ways to deal with this issue:

  • In a web-based application, an event handler can be used to close the
    only at the very end of a user request, once the rendering of
    the view is complete. Of course, this places heavy demands upon the correctness
    of the exception handling of your application infrastructure. It is vitally important
    that the ISession is closed and the transaction ended before
    returning to the user, even when an exception occurs during rendering of the view.
    The event handler has to be able to access the ISession
    for this approach. We recommend that the current ISession
    is stored in the HttpContext.Items collection (see chapter
    Section 1.4, “Playing with cats”
    , for an example implementation).

  • In an application with a seperate business tier, the business logic must “prepare”
    all collections that will be needed by the web tier before returning. This means
    that the business tier should load all the data and return all the data already
    initialized to the presentation/web tier that is required for a particular use case.
    Usually, the application calls NHibernateUtil.Initialize()
    for each collection that will be needed in the web tier (this call must occur before
    the session is closed) or retrieves the collection eagerly using a NHibernate query
    with a FETCH clause.

  • You may also attach a previously loaded object to a new ISession
    with Update() or Lock() before
    accessing unitialized collections (or other proxies). NHibernate can not do this
    automatically, as it would introduce ad hoc transaction semantics!

You can use the Filter() method of the NHibernate ISession
API to get the size of a collection without initializing it:

ICollection countColl = s.Filter( collection, "select count(*)" );
IEnumerator countEn = countColl.GetEnumerator();
int count = (int) countEn.Current;

Filter() or CreateFilter() are
also used to efficiently retrieve subsets of a collection without needing to initialize
the whole collection.

6.6. Sorted Collections

NHibernate supports collections implemented by System.Collections.SortedList
and Iesi.Collections.SortedSet. You must specify a comparer
in the mapping file:

<set name="Aliases" table="person_aliases" sort="natural">
    <key column="person"/>
    <element column="name" type="String"/>

<map name="Holidays" 
     sort="My.Custom.HolidayComparer, MyAssembly" 
    <key column="year_id"/>
    <index column="hol_name" type="String"/>
    <element column="hol_date" type="Date"/>

Allowed values of the sort attribute are
, natural and the name of a class implementing

If you want the database itself to order the collection elements use the
attribute of set, bag
or map mappings. This performs the ordering in the SQL
query, not in memory.

Setting the order-by attribute tells NHibernate to use
ListDictionary or ListSet class
internally for dictionaries and sets, maintaining the order of the elements. Note that lookup operations on these collections are very slow
if they contain more than a few elements.

<set name="Aliases" table="person_aliases" order-by="name asc">
    <key column="person"/>
    <element column="name" type="String"/>

<map name="Holidays" order-by="hol_date, hol_name" lazy="true">
    <key column="year_id"/>
    <index column="hol_name" type="String"/>
    <element column="hol_date type="Date"/>

Note that the value of the order-by attribute is an SQL
ordering, not a HQL ordering!

Associations may even be sorted by some arbitrary criteria at runtime using a Filter().

sortedUsers = s.Filter( group.Users, "order by this.Name" );

6.7. Using an <idbag>

If you’ve fully embraced our view that composite keys are a bad thing and that entities
should have synthetic identifiers (surrogate keys), then you might find it a bit
odd that the many to many associations and collections of values that we’ve shown
so far all map to tables with composite keys! Now, this point is quite arguable;
a pure association table doesn’t seem to benefit much from a surrogate key (though
a collection of composite values might).
Nevertheless, NHibernate provides a feature that allows you to map many to many
associations and collections of values to a table with a surrogate key.

The <idbag> element lets you map a
(or Collection) with bag semantics.

<idbag name="Lovers" table="LOVERS" lazy="true">
    <collection-id column="ID" type="Int64">
        <generator class="hilo"/>
    <key column="PERSON1"/>
    <many-to-many column="PERSON2" class="Eg.Person" fetch="join"/>

As you can see, an <idbag> has a synthetic id generator,
just like an entity class! A different surrogate key is assigned to each collection
row. NHibernate does not provide any mechanism to discover the surrogate key value
of a particular row, however.

Note that the update performance of an <idbag> is
much better than a regular
! NHibernate can locate individual rows efficiently and update
or delete them individually, just like a list, map or set.

As of version 2.0, the native identifier generation strategy
is supported for <idbag> collection identifiers.

6.8. Bidirectional Associations

A bidirectional association allows navigation
from both “ends” of the association. Two kinds of bidirectional association are


set or bag valued at one end, single-valued at the other


set or bag valued at both ends

Please note that NHibernate does not support bidirectional one-to-many associations
with an indexed collection (list, map or array) as the “many” end, you have to use
a set or bag mapping.

You may specify a bidirectional many-to-many association simply by mapping two many-to-many
associations to the same database table and declaring one end as
(which one is your choice). Here’s an example of a bidirectional
many-to-many association from a class back to itself
(each category can have many items and each item can be in many categories):

<class name="NHibernate.Auction.Category, NHibernate.Auction">
    <id name="Id" column="ID"/>
    <bag name="Items" table="CATEGORY_ITEM" lazy="true">
        <key column="CATEGORY_ID"/>
            class="NHibernate.Auction.Item, NHibernate.Auction" 

<class name="NHibernate.Auction.Item, NHibernate.Auction">
    <id name="id" column="ID"/>

    <!-- inverse end -->
    <bag name="categories" table="CATEGORY_ITEM" inverse="true" 
        <key column="ITEM_ID"/>
            class="NHibernate.Auction.Category, NHibernate.Auction" 

Changes made only to the inverse end of the association are
persisted. This means that NHibernate has two representations
in memory for every bidirectional association, one link from A to B and another
link from B to A. This is easier to understand if you think about the .NET object
model and how we create a many-to-many relationship in C#:

// The category now "knows" about 
//the relationship
item.Categories.Add(category);     // The item now "knows" about 
                                   //  the relationship

session.Update(item);                     // No effect, nothing 
                                          // will be saved!
session.Update(category);                 // The relationship will 
                                          //be saved

The non-inverse side is used to save the in-memory representation to the database.
We would get an unneccessary INSERT/UPDATE and probably even a foreign key violation
if both would trigger changes! The same is of course also true for bidirectional
one-to-many associations.

You may map a bidirectional one-to-many association by mapping a one-to-many association
to the same table column(s) as a many-to-one association and declaring the many-valued
end inverse="true".

<class name="Eg.Parent, Eg">
    <id name="Id" column="id"/>
    <set name="Children" inverse="true" lazy="true">
        <key column="parent_id"/>
        <one-to-many class="Eg.Child, Eg"/>

<class name="Eg.Child, Eg">
    <id name="Id" column="id"/>
    <many-to-one name="Parent" class="Eg.Parent, Eg" 

Mapping one end of an association with inverse="true" doesn’t
affect the operation of cascades, both are different concepts!

6.9. Ternary Associations

There are two possible approaches to mapping a ternary association. One approach
is to use composite elements (discussed below). Another is to use an
with an association as its index:

<map name="Contracts" lazy="true">
    <key column="employer_id"/>
    <index-many-to-many column="employee_id" class="Employee"/>
    <one-to-many class="Contract"/>
<map name="Connections" lazy="true">
    <key column="node1_id"/>
    <index-many-to-many column="node2_id" class="Node"/>
    <many-to-many column="connection_id" class="Connection"/>

6.10. Heterogeneous Associations

The <many-to-any> and <index-many-to-any>
elements provide for true heterogeneous associations. These mapping elements work
in the same way as the <any> element – and should
also be used rarely, if ever.

6.11. Collection examples

The previous sections are pretty confusing. So lets look at an example. This class:

using System;
using System.Collections;

namespace Eg
    public class Parent
        private long id;
        private ISet children;
        public long Id
            get { return id; }
            set { id = value; }
        private ISet Children
            get { return children; }
            set { children = value; }

has a collection of Eg.Child instances. If each child has
at most one parent, the most natural mapping is a one-to-many association:

<hibernate-mapping xmlns="urn:nhibernate-mapping-2.2"
    assembly="Eg" namespace="Eg">

    <class name="Parent">
        <id name="Id">
            <generator class="sequence"/>
        <set name="Children" lazy="true">
            <key column="parent_id"/>
            <one-to-many class="Child"/>

    <class name="Child">
        <id name="Id">
            <generator class="sequence"/>
        <property name="Name"/>


This maps to the following table definitions:

create table parent ( Id bigint not null primary key )
create table child ( Id bigint not null primary key, 
                     Name varchar(255), parent_id bigint )
alter table child add constraint childfk0 (parent_id) 
    references parent

If the parent is required, use a bidirectional
one-to-many association:

<hibernate-mapping xmlns="urn:nhibernate-mapping-2.2"
    assembly="Eg" namespace="Eg">

    <class name="Parent">
        <id name="Id">
            <generator class="sequence"/>
        <set name="Children" inverse="true" lazy="true">
            <key column="parent_id"/>
            <one-to-many class="Child"/>

    <class name="Child">
        <id name="Id">
            <generator class="sequence"/>
        <property name="Name"/>
        <many-to-one name="parent" class="Parent" column="parent_id" 


Notice the NOT NULL constraint:

create table parent ( Id bigint not null primary key )
create table child ( Id bigint not null
                     primary key,
                     Name varchar(255),
                     parent_id bigint not null )
alter table child add constraint childfk0 (parent_id) 
 references parent

On the other hand, if a child might have multiple parents, a many-to-many association
is appropriate:

<hibernate-mapping xmlns="urn:nhibernate-mapping-2.2"
    assembly="Eg" namespace="Eg">

    <class name="Parent">
        <id name="Id">
            <generator class="sequence"/>
        <set name="Children" lazy="true" table="childset">
            <key column="parent_id"/>
            <many-to-many class="Child" column="child_id"/>

    <class name="eg.Child">
        <id name="Id">
            <generator class="sequence"/>
        <property name="Name"/>


Table definitions:

create table parent ( Id bigint not null primary key )
create table child ( Id bigint not null primary key, 
                     name varchar(255) )
create table childset ( parent_id bigint not null,
                        child_id bigint not null,
                        primary key ( parent_id, child_id ) )
alter table childset add constraint childsetfk0 (parent_id) 
references parent
alter table childset add constraint childsetfk1 (child_id) 
references child

Chapter 7. Component Mapping

The notion of a component is re-used in several
different contexts, for different purposes, throughout NHibernate.

7.1. Dependent objects

A component is a contained object that is persisted as a value type, not an entity.
The term “component” refers to the object-oriented notion of composition (not to
architecture-level components). For example, you might model a person like this:

public class Person
    private DateTime birthday;
    private Name name;
    private string key;
    public string Key
        get { return key; }
        set { key = value; }
    public DateTime Birthday
        get { return birthday; }
        set { birthday = value; }

    public Name Name
        get { return name; }
        set { name = value; }
public class Name
    char initial;
    string first;
    string last;
    public string First
        get { return first; }
        set { first = value; }
    public string Last
        get { return last; }
        set { last = value; }

    public char Initial
        get { return initial; }
        set { initial = value; }

Now Name may be persisted as a component of
. Notice that Name defines getter and setter
methods for its persistent properties, but doesn’t need to declare any interfaces
or identifier properties.

Our NHibernate mapping would look like:

<class name="Eg.Person, Eg" table="person">
    <id name="Key" column="pid" type="string">
        <generator class="uuid.hex"/>
    <property name="Birthday" type="date"/>
<!-- class attribute optional -->
    <component name="Name" class="Eg.Name, Eg"> 
        <property name="Initial"/>
        <property name="First"/>
        <property name="Last"/>

The person table would have the columns pid,
, Initial, First
and Last.

Like all value types, components do not support shared references. The null value
semantics of a component are ad hoc. When
reloading the containing object, NHibernate will assume that if all component columns
are null, then the entire component is null. This should be okay for most purposes.

The properties of a component may be of any NHibernate type (collections, many-to-one
associations, other components, etc). Nested components should
be considered an exotic usage. NHibernate is intended to
support a very fine-grained object model.

The <component> element allows a
subelement that maps a property of the component class as
a reference back to the containing entity.

<class name="Eg.Person, Eg" table="person">
    <id name="Key" column="pid" type="string">
        <generator class="uuid.hex"/>
    <property name="Birthday" type="date"/>
    <component name="Name" class="Eg.Name, Eg">
<!-- reference back to the Person -->
        <parent name="NamedPerson"/> 
        <property name="Initial"/>
        <property name="First"/>
        <property name="Last"/>

7.2. Collections of dependent objects

Collections of components are supported (eg. an array of type Name).
Declare your component collection by replacing the <element>
tag with a <composite-element> tag.

<set name="SomeNames" table="some_names" lazy="true">
    <key column="id"/>
<!-- class attribute required -->
    <composite-element class="Eg.Name, Eg"> 
        <property name="Initial"/>
        <property name="First"/>
        <property name="Last"/>

Note: if you define an ISet of composite elements, it is
very important to implement Equals() and

Composite elements may contain components but not collections. If your composite
element itself contains components, use the <nested-composite-element>
tag. This is a pretty exotic case – a collection of components which themselves
have components. By this stage you should be asking yourself if a one-to-many association
is more appropriate. Try remodelling the composite element as an entity – but note
that even though the object model is the same, the relational model and persistence
semantics are still slightly different.

Please note that a composite element mapping doesn’t support null-able properties
if you’re using a <set>. NHibernate has to use each
columns value to identify a record when deleting objects (there is no separate primary
key column in the composite element table), which is not possible with null values.
You have to either use only not-null properties in a composite-element or choose
a <list>, <map>, <bag> or <idbag>.

A special case of a composite element is a composite element with a nested
element. A mapping like this allows you to map extra
columns of a many-to-many association table to the composite element class. The
following is a many-to-many association from Order to Item where PurchaseDate,
and Quantity are properties of the association:

<class name="Order" .... >
    <set name="PurchasedItems" table="purchase_items" lazy="true">
        <key column="order_id">
        <composite-element class="Purchase">
            <property name="PurchaseDate"/>
            <property name="Price"/>
            <property name="Quantity"/>
            <!-- class attribute is optional -->
            <many-to-one name="Item" class="Item"/> 

Even ternary (or quaternary, etc) associations are possible:

<class name="Order" .... >
    <set name="PurchasedItems" table="purchase_items" lazy="true">
        <key column="order_id">
        <composite-element class="OrderLine">
            <many-to-one name="PurchaseDetails class="Purchase"/>
            <many-to-one name="Item" class="Item"/>

Composite elements may appear in queries using the same syntax as associations to
other entities.

7.3. Components as IDictionary indices

The <composite-index> element lets you map a component
class as the key of an IDictionary. Make sure you override
GetHashCode() and Equals() correctly
on the component class.

7.4. Components as composite identifiers

You may use a component as an identifier of an entity class. Your component class
must satisfy certain requirements:

  • It must be Serializable.

  • It must re-implement Equals() and GetHashCode(),
    consistently with the database’s notion of composite key equality.

You can’t use an IIdentifierGenerator to generate composite
keys. Instead the application must assign its own identifiers.

Since a composite identifier must be assigned to the object before saving it, we
can’t use unsaved-value of the identifier to distinguish
between newly instantiated instances and instances saved in a previous session.

You may instead implement IInterceptor.IsTransient() if
you wish to use SaveOrUpdate() or cascading save / update.
As an alternative, you may also set the unsaved-value attribute
on a <version> (or <timestamp>)
element to specify a value that indicates a new transient instance. In this case,
the version of the entity is used instead of the (assigned) identifier and you don’t
have to implement IInterceptor.IsTransient() yourself.

Use the <composite-id> tag (same attributes and elements
as <component>) in place of <id>
for the declaration of a composite identifier class:

<class name="Foo" table="FOOS">
    <composite-id name="CompId" class="FooCompositeID">
        <key-property name="String"/>
        <key-property name="Short"/>
        <key-property name="Date" column="date_" type="Date"/>
    <property name="Name"/>

Now, any foreign keys into the table FOOS are also composite.
You must declare this in your mappings for other classes. An association to
would be declared like this:

<many-to-one name="Foo" class="Foo">
<!-- the "class" attribute is optional, as usual -->
    <column name="foo_string"/>
    <column name="foo_short"/>
    <column name="foo_date"/>

This new <column> tag is also used by multi-column
custom types. Actually it is an alternative to the column
attribute everywhere. A collection with elements of type Foo
would use:

<set name="Foos">
    <key column="owner_id"/>
    <many-to-many class="Foo">
        <column name="foo_string"/>
        <column name="foo_short"/>
        <column name="foo_date"/>

On the other hand, <one-to-many>, as usual, declares
no columns.

If Foo itself contains collections, they will also need
a composite foreign key.

<class name="Foo">
    <set name="Dates" lazy="true">
        <key>   <!-- a collection inherits the composite key type -->
            <column name="foo_string"/>
            <column name="foo_short"/>
            <column name="foo_date"/>
        <element column="foo_date" type="Date"/>

7.5. Dynamic components

You may even map a property of type IDictionary:

<dynamic-component name="UserAttributes">
    <property name="Foo" column="FOO"/>
    <property name="Bar" column="BAR"/>
    <many-to-one name="Baz" class="Baz" column="BAZ"/>

The semantics of a <dynamic-component> mapping are
identical to <component>. The advantage of this kind
of mapping is the ability to determine the actual properties of the component at
deployment time, just by editing the mapping document. (Runtime manipulation of
the mapping document is also possible, using a DOM parser.)

Chapter 8. Inheritance Mapping

8.1. The Three Strategies

NHibernate supports the three basic inheritance mapping strategies.

  • table per class hierarchy

  • table per subclass

  • table per concrete class

In addition, NHibernate supports a fourth, slightly different kind of polymorphism:

  • implicit polymorphism

It is possible to use different mapping strategies for different branches of the
same inheritance hierarchy, and then make use of implicit polymorphism to achieve
polymorphism across the whole hierarchy. However, NHibernate does not support mixing
<subclass>, and <joined-subclass>
and <union-subclass> mappings under the same root
<class> element. It is possible to mix together the
table per hierarchy and table per subclass strategies, under the the same
element, by combining the <subclass>
and <join> elements (see below).

It is possible to define subclass, union-subclass,
and joined-subclass mappings in separate mapping documents,
directly beneath hibernate-mapping. This allows you to
extend a class hierachy just by adding a new mapping file. You must specify an extends attribute in the subclass mapping, naming a previously
mapped superclass.

     <subclass name="DomesticCat" extends="Cat" 
          <property name="name" type="string"/>

8.1.1. Table per class hierarchy

Suppose we have an interface IPayment, with implementors
CreditCardPayment, CashPayment,
ChequePayment. The table-per-hierarchy mapping would look

<class name="IPayment" table="PAYMENT">
    <id name="Id" type="Int64" column="PAYMENT_ID">
        <generator class="native"/>
    <discriminator column="PAYMENT_TYPE" type="String"/>
    <property name="Amount" column="AMOUNT"/>
    <subclass name="CreditCardPayment" discriminator-value="CREDIT">
    <subclass name="CashPayment" discriminator-value="CASH">
    <subclass name="ChequePayment" discriminator-value="CHEQUE">

Exactly one table is required. There is one big limitation of this mapping strategy:
columns declared by the subclasses may not have NOT NULL

8.1.2. Table per subclass

A table-per-subclass mapping would look like:

<class name="IPayment" table="PAYMENT">
    <id name="Id" type="Int64" column="PAYMENT_ID">
        <generator class="native"/>
    <property name="Amount" column="AMOUNT"/>
    <joined-subclass name="CreditCardPayment" table="CREDIT_PAYMENT">
        <key column="PAYMENT_ID"/>
    <joined-subclass name="CashPayment" table="CASH_PAYMENT">
        <key column="PAYMENT_ID"/>
    <joined-subclass name="ChequePayment" table="CHEQUE_PAYMENT">
        <key column="PAYMENT_ID"/>

Four tables are required. The three subclass tables have primary key associations
to the superclass table (so the relational model is actually a one-to-one association).

8.1.3. Table per subclass,
using a discriminator

Note that NHibernate’s implementation of table-per-subclass requires no discriminator
column. Other object/relational mappers use a different implementation of table-per-subclass
which requires a type discriminator column in the superclass table. The approach
taken by NHibernate is much more difficult to implement but arguably more correct
from a relational point of view. If you would like to use a discriminator column
with the table per subclass strategy, you may combine the use of
and <join>, as follow:

<class name="Payment" table="PAYMENT">
    <id name="Id" type="Int64" column="PAYMENT_ID">
        <generator class="native"/>
    <discriminator column="PAYMENT_TYPE" type="string"/>
    <property name="Amount" column="AMOUNT"/>
    <subclass name="CreditCardPayment" discriminator-value="CREDIT">
        <join table="CREDIT_PAYMENT">
            <key column="PAYMENT_ID"/>
            <property name="CreditCardType" column="CCTYPE"/>
    <subclass name="CashPayment" discriminator-value="CASH">
        <join table="CASH_PAYMENT">
            <key column="PAYMENT_ID"/>
    <subclass name="ChequePayment" discriminator-value="CHEQUE">
        <join table="CHEQUE_PAYMENT" fetch="select">
            <key column="PAYMENT_ID"/>

The optional fetch="select" declaration tells NHibernate
not to fetch the ChequePayment subclass data using an outer
join when querying the superclass.

8.1.4. Mixing
table per class hierarchy with table per subclass

You may even mix the table per hierarchy and table per subclass strategies using
this approach:

<class name="Payment" table="PAYMENT">
    <id name="Id" type="Int64" column="PAYMENT_ID">
        <generator class="native"/>
    <discriminator column="PAYMENT_TYPE" type="string"/>
    <property name="Amount" column="AMOUNT"/>
    <subclass name="CreditCardPayment" discriminator-value="CREDIT">
        <join table="CREDIT_PAYMENT">
            <property name="CreditCardType" column="CCTYPE"/>
    <subclass name="CashPayment" discriminator-value="CASH">
    <subclass name="ChequePayment" discriminator-value="CHEQUE">

For any of these mapping strategies, a polymorphic association to
is mapped using <many-to-one>.

<many-to-one name="Payment" column="PAYMENT" class="IPayment"/>

8.1.5. Table per concrete class

There are two ways we could go about mapping the table per concrete class strategy.
The first is to use <union-subclass>.

<class name="Payment">
    <id name="Id" type="Int64" column="PAYMENT_ID">
        <generator class="sequence"/>
    <property name="Amount" column="AMOUNT"/>
    <union-subclass name="CreditCardPayment" table="CREDIT_PAYMENT">
        <property name="CreditCardType" column="CCTYPE"/>
    <union-subclass name="CashPayment" table="CASH_PAYMENT">
    <union-subclass name="ChequePayment" table="CHEQUE_PAYMENT">

Three tables are involved for the subclasses. Each table defines columns for all
properties of the class, including inherited properties.

The limitation of this approach is that if a property is mapped on the superclass,
the column name must be the same on all subclass tables. (We might relax this in
a future release of NHibernate.) The identity generator strategy is not allowed
in union subclass inheritance, indeed the primary key seed has to be shared accross
all unioned subclasses of a hierarchy.

If your superclass is abstract, map it with abstract="true".
Of course, if it is not abstract, an additional table (defaults to
in the example above) is needed to hold instances of the superclass.

8.1.6. Table per concrete
class, using implicit polymorphism

An alternative approach is to make use of implicit polymorphism:

<class name="CreditCardPayment" table="CREDIT_PAYMENT">
    <id name="Id" type="Int64" column="CREDIT_PAYMENT_ID">
        <generator class="native"/>
    <property name="Amount" column="CREDIT_AMOUNT"/>

<class name="CashPayment" table="CASH_PAYMENT">
    <id name="Id" type="Int64" column="CASH_PAYMENT_ID">
        <generator class="native"/>
    <property name="Amount" column="CASH_AMOUNT"/>

<class name="ChequePayment" table="CHEQUE_PAYMENT">
    <id name="Id" type="Int64" column="CHEQUE_PAYMENT_ID">
        <generator class="native"/>
    <property name="Amount" column="CHEQUE_AMOUNT"/>

Notice that nowhere do we mention the IPayment interface
explicitly. Also notice that properties of IPayment are
mapped in each of the subclasses. If you want to avoid duplication, consider using
XML entities (e.g. [ <!ENTITY allproperties SYSTEM "allproperties.xml">
in the DOCTYPE declartion and &allproperties;
in the mapping).

The disadvantage of this approach is that NHibernate does not generate SQL
s when performing polymorphic queries.

For this mapping strategy, a polymorphic association to IPayment
is usually mapped using <any>.

<any name="Payment" meta-type="string" id-type="Int64">
    <meta-value value="CREDIT" class="CreditCardPayment"/>
    <meta-value value="CASH" class="CashPayment"/>
    <meta-value value="CHEQUE" class="ChequePayment"/>
    <column name="PAYMENT_CLASS"/>
    <column name="PAYMENT_ID"/>

8.1.7. Mixing implicit polymorphism
with other inheritance mappings

There is one further thing to notice about this mapping. Since the subclasses are
each mapped in their own <class> element (and since
IPayment is just an interface), each of the subclasses
could easily be part of another table-per-class or table-per-subclass inheritance
hierarchy! (And you can still use polymorphic queries against the

<class name="CreditCardPayment" table="CREDIT_PAYMENT">
    <id name="Id" type="Int64" column="CREDIT_PAYMENT_ID">
        <generator class="native"/>
    <discriminator column="CREDIT_CARD" type="String"/>
    <property name="Amount" column="CREDIT_AMOUNT"/>
    <subclass name="MasterCardPayment" discriminator-value="MDC"/>
    <subclass name="VisaPayment" discriminator-value="VISA"/>

<class name="NonelectronicTransaction" table="NONELECTRONIC_TXN">
    <id name="Id" type="Int64" column="TXN_ID">
        <generator class="native"/>
    <joined-subclass name="CashPayment" table="CASH_PAYMENT">
        <key column="PAYMENT_ID"/>
        <property name="Amount" column="CASH_AMOUNT"/>
    <joined-subclass name="ChequePayment" table="CHEQUE_PAYMENT">
        <key column="PAYMENT_ID"/>
        <property name="Amount" column="CHEQUE_AMOUNT"/>

Once again, we don’t mention IPayment explicitly. If we
execute a query against the IPayment interface – for example,
from IPayment – NHibernate automatically returns instances
of CreditCardPayment (and its subclasses, since they also
implement IPayment), CashPayment
and ChequePayment but not instances of

8.2. Limitations

There are certain limitations to the “implicit polymorphism” approach to the table
per concrete-class mapping strategy. There are somewhat less restrictive limitations
to <union-subclass> mappings.

The following table shows the limitations of table per concrete-class mappings,
and of implicit polymorphism, in NHibernate.

Table 8.1. Features of inheritance mappings

Outer join
table per
<many-to-one> <one-to-one> <one-to-many> <many-to-many> s.Get(typeof(IPayment), id) from IPayment p from
o join
o.Payment p
table per
<many-to-one> <one-to-one> <one-to-many> <many-to-many> s.
Order o
table per
<many-to-one> <one-to-one> <one-to-many>

<many-to-many> s
IPayment p
Order o
table per
<any> not
<many-to-any> use a query from IPayment p not

Chapter 9. Manipulating Persistent Data

9.1. Creating a persistent object

An object (entity instance) is either transient
or persistent with respect to a particular
ISession. Newly instantiated objects are, of course, transient.
The session offers services for saving (ie. persisting) transient instances:

DomesticCat fritz = new DomesticCat();
fritz.Color = Color.Ginger;
fritz.Sex = 'M';
fritz.Name = "Fritz";
long generatedId = (long) sess.Save(fritz);
DomesticCat pk = new DomesticCat();
pk.Color = Color.Tabby;
pk.Sex = 'F';
pk.Name = "PK";
pk.Kittens = new HashSet();
sess.Save( pk, 1234L );

The single-argument Save() generates and assigns a unique
identifier to fritz. The two-argument form attempts to
persist pk using the given identifier. We generally discourage
the use of the two-argument form since it may be used to create primary keys with
business meaning.

Associated objects may be made persistent in any order you like unless you have
a NOT NULL constraint upon a foreign key column. There
is never a risk of violating foreign key constraints. However, you might violate
a NOT NULL constraint if you Save()
the objects in the wrong order.

9.2. Loading an object

The Load() methods of ISession
give you a way to retrieve a persistent instance if you already know its identifier.
One version takes a class object and will load the state into a newly instantiated
object. The second version allows you to supply an instance into which the state
will be loaded. The form which takes an instance is only useful in special circumstances
(DIY instance pooling etc.)

Cat fritz = (Cat) sess.Load(typeof(Cat), generatedId);
long pkId = 1234;
DomesticCat pk = (DomesticCat) sess.Load( typeof(Cat), pkId );
Cat cat = new DomesticCat();
// load pk's state into cat
sess.Load( cat, pkId );
ISet kittens = cat.Kittens;

Note that Load() will throw an unrecoverable exception
if there is no matching database row. If the class is mapped with a proxy,
returns an object that is an uninitialized proxy and does not actually
hit the database until you invoke a method of the object. This behaviour is very
useful if you wish to create an association to an object without actually loading
it from the database.

If you are not certain that a matching row exists, you should use the
method, which hits the database immediately and returns null if there
is no matching row.

Cat cat = (Cat) sess.Get(typeof(Cat), id);
if (cat==null) {
    cat = new Cat();
    sess.Save(cat, id);
return cat;

You may also load an objects using an SQL SELECT ... FOR UPDATE.
See the next section for a discussion of NHibernate LockModes.

Cat cat = (Cat) sess.Get(typeof(Cat), id, LockMode.Upgrade);

Note that any associated instances or contained collections are
selected FOR UPDATE.

It is possible to re-load an object and all its collections at any time, using the
Refresh() method. This is useful when database triggers
are used to initialize some of the properties of the object.

sess.Flush(); //force the SQL INSERT
sess.Refresh(cat); //re-read the state (after the trigger executes)

An important question usually appears at this point: How much does NHibernate load
from the database and how many SQL SELECTs will it use?
This depends on the fetching strategy and
is explained in Section 19.1, “Fetching strategies”.

9.3. Querying

If you don’t know the identifier(s) of the object(s) you are looking for, use the
Find() methods of ISession. NHibernate
supports a simple but powerful object oriented query language.

IList cats = sess.Find(
    "from Cat as cat where cat.Birthdate = ?",

IList mates = sess.Find(
    "select mate from Cat as cat join cat.Mate as mate " +
    "where cat.name = ?",

IList cats = sess.Find( "from Cat as cat where cat.Mate.Birthdate is null" );

IList moreCats = sess.Find(
    "from Cat as cat where " + 
    "cat.Name = 'Fritz' or cat.id = ? or cat.id = ?",
    new object[] { id1, id2 },
    new IType[] { NHibernateUtil.Int64, NHibernateUtil.Int64 }

IList mates = sess.Find(
    "from Cat as cat where cat.Mate = ?",

IList problems = sess.Find(
    "from GoldFish as fish " +
    "where fish.Birthday > fish.Deceased or fish.Birthday is null"

The second argument to Find() accepts an object or array
of objects. The third argument accepts a NHibernate type or array of NHibernate
types. These given types are used to bind the given objects to the
query placeholders (which map to input parameters of an ADO.NET
). Just as in ADO.NET, you should use this binding mechanism
in preference to string manipulation.

The NHibernateUtil class defines a number of static methods
and constants, providing access to most of the built-in types, as instances of NHibernate.Type.IType.

If you expect your query to return a very large number of objects, but you don’t
expect to use them all, you might get better performance from the
methods, which return a System.Collections.IEnumerable.
The iterator will load objects on demand, using the identifiers returned by an initial
SQL query (n+1 selects total).

// fetch ids
IEnumerable en = sess.Enumerable("from eg.Qux q order by q.Likeliness"); 
foreach ( Qux qux in en )
    // something we couldnt express in the query
    if ( qux.CalculateComplicatedAlgorithm() ) {
        // dont need to process the rest

The Enumerable() method also performs better if you expect
that many of the objects are already loaded and cached by the session, or if the
query results contain the same objects many times. (When no data is cached or repeated,
Find() is almost always faster.) Heres an example of a
query that should be called using Enumerable():

IEnumerable en = sess.Enumerable(
    "select customer, product " + 
    "from Customer customer, " +
    "Product product " +
    "join customer.Purchases purchase " +
    "where product = purchase.Product"

Calling the previous query using Find() would return a
very large ADO.NET result set containing the same data many times.

NHibernate queries sometimes return tuples of objects, in which case each tuple
is returned as an array:

IEnumerable foosAndBars = sess.Enumerable(
    "select foo, bar from Foo foo, Bar bar " +
    "where bar.Date = foo.Date"
foreach (object[] tuple in foosAndBars)
    Foo foo = tuple[0]; Bar bar = tuple[1];

9.3.1. Scalar queries

Queries may specify a property of a class in the select
clause. They may even call SQL aggregate functions. Properties or aggregates are
considered “scalar” results.

IEnumerable results = sess.Enumerable(
        "select cat.Color, min(cat.Birthdate), count(cat) from Cat cat " +
        "group by cat.Color"
foreach ( object[] row in results )
    Color type = (Color) row[0];
    DateTime oldest = (DateTime) row[1];
    int count = (int) row[2];
IEnumerable en = sess.Enumerable(
    "select cat.Type, cat.Birthdate, cat.Name from DomesticCat cat"
IList list = sess.Find(
    "select cat, cat.Mate.Name from DomesticCat cat"

9.3.2. The IQuery interface

If you need to specify bounds upon your result set (the maximum number of rows you
want to retrieve and / or the first row you want to retrieve) you should obtain
an instance of NHibernate.IQuery:

IQuery q = sess.CreateQuery("from DomesticCat cat");
IList cats = q.List();

You may even define a named query in the mapping document. (Remember to use a CDATA section if your query contains characters that could
be interpreted as markup.)

<query name="Eg.DomesticCat.by.name.and.minimum.weight"><![CDATA[
    from Eg.DomesticCat as cat
        where cat.Name = ?
        and cat.Weight > ?
] ]></query>
IQuery q = sess.GetNamedQuery("Eg.DomesticCat.by.name.and.minimum.weight");
q.SetString(0, name);
q.SetInt32(1, minWeight);
IList cats = q.List();

The query interface supports the use of named parameters. Named parameters are identifiers
of the form :name in the query string. There are methods
on IQuery for binding values to named or positional parameters.
NHibernate numbers parameters from zero. The advantages of named parameters are:

  • named parameters are insensitive to the order they occur in the query string

  • they may occur multiple times in the same query

  • they are self-documenting

//named parameter (preferred)
IQuery q = sess.CreateQuery("from DomesticCat cat where cat.Name = :name");
q.SetString("name", "Fritz");
IEnumerable cats = q.Enumerable();
//positional parameter
IQuery q = sess.CreateQuery("from DomesticCat cat where cat.Name = ?");
q.SetString(0, "Izi");
IEnumerable cats = q.Enumerable();
//named parameter list
IList names = new ArrayList();
IQuery q = sess.CreateQuery("from DomesticCat cat where cat.Name in (:namesList)");
q.SetParameterList("namesList", names);
IList cats = q.List();

9.3.3. Filtering collections

A collection filter is a special type of
query that may be applied to a persistent collection or array. The query string
may refer to this, meaning the current collection element.

ICollection blackKittens = session.Filter(
    pk.Kittens, "where this.Color = ?", Color.Black, NHibernateUtil.Enum(typeof(Color))

The returned collection is considered a bag.

Observe that filters do not require a from clause (though
they may have one if required). Filters are not limited to returning the collection
elements themselves.

ICollection blackKittenMates = session.Filter(
    pk.Kittens, "select this.Mate where this.Color = Eg.Color.Black"

9.3.4. Criteria queries

HQL is extremely powerful but some people prefer to build queries dynamically, using
an object oriented API, rather than embedding strings in their .NET code. For these
people, NHibernate provides an intuitive ICriteria query

ICriteria crit = session.CreateCriteria(typeof(Cat));
crit.Add( Expression.Eq("color", Eg.Color.Black) );
IList cats = crit.List();

If you are uncomfortable with SQL-like syntax, this is perhaps the easiest way to
get started with NHibernate. This API is also more extensible than HQL. Applications
might provide their own implementations of the ICriterion

9.3.5. Queries in native SQL

You may express a query in SQL, using CreateSQLQuery().
You must enclose SQL aliases in braces.

IList cats = session.CreateSQLQuery(
    "SELECT {cat.*} FROM CAT {cat} WHERE ROWNUM<10",
IList cats = session.CreateSQLQuery(
    "SELECT {cat}.ID AS {cat.Id}, {cat}.SEX AS {cat.Sex}, " +
           "{cat}.MATE AS {cat.Mate}, {cat}.SUBCLASS AS {cat.class}, ... " +
    "FROM CAT {cat} WHERE ROWNUM<10",

SQL queries may contain named and positional parameters, just like NHibernate queries.

9.4. Updating objects

9.4.1. Updating in the same

Transactional persistent instances (ie. objects
loaded, saved, created or queried by the ISession) may
be manipulated by the application and any changes to persistent state will be persisted
when the ISession is flushed
(discussed later in this chapter). So the most straightforward way to update the
state of an object is to Load() it, and then manipulate
it directly, while the ISession is open:

DomesticCat cat = (DomesticCat) sess.Load( typeof(Cat), 69L );
cat.Name = "PK";
sess.Flush();  // changes to cat are automatically detected and persisted

Sometimes this programming model is inefficient since it would require both an SQL
SELECT (to load an object) and an SQL UPDATE
(to persist its updated state) in the same session. Therefore NHibernate offers
an alternate approach.

9.4.2. Updating detached objects

Many applications need to retrieve an object in one transaction, send it to the
UI layer for manipulation, then save the changes in a new transaction. (Applications
that use this kind of approach in a high-concurrency environment usually use versioned
data to ensure transaction isolation.) This approach requires a slightly different
programming model to the one described in the last section. NHibernate supports
this model by providing the method Session.Update().

// in the first session
Cat cat = (Cat) firstSession.Load(typeof(Cat), catId);
Cat potentialMate = new Cat();

// in a higher tier of the application
cat.Mate = potentialMate;

// later, in a new session
secondSession.Update(cat);  // update cat
secondSession.Update(mate); // update mate

If the Cat with identifier catId
had already been loaded by secondSession when the application
tried to update it, an exception would have been thrown.

The application should individually Update() transient
instances reachable from the given transient instance if and
if it wants their state also updated. (Except for lifecycle
objects, discussed later.)

NHibernate users have requested a general purpose method that either saves a transient
instance by generating a new identifier or update the persistent state associated
with its current identifier. The SaveOrUpdate() method
now implements this functionality.

NHibernate distinguishes “new” (unsaved) instances from “existing” (saved or loaded
in a previous session) instances by the value of their identifier (or version, or
timestamp) property. The unsaved-value attribute of the
<id> (or <version>,
or <timestamp>) mapping specifies which values should
be interpreted as representing a “new” instance.

<id name="Id" type="Int64" column="uid" unsaved-value="0">
    <generator class="hilo"/>

The allowed values of unsaved-value are:

  • any – always save

  • none – always update

  • null – save when identifier is null

  • valid identifier value – save when identifier is null or the given value

  • undefined – if set for version
    or timestamp, then identifier check is used

If unsaved-value is not specified for a class, NHibernate
will attempt to guess it by creating an instance of the class using the no-argument
constructor and reading the property value from the instance.

// in the first session
Cat cat = (Cat) firstSession.Load(typeof(Cat), catID);

// in a higher tier of the application
Cat mate = new Cat();
cat.Mate = mate;

// later, in a new session
secondSession.SaveOrUpdate(cat);   // update existing state (cat has a non-null id)
secondSession.SaveOrUpdate(mate);  // save the new instance (mate has a null id)

The usage and semantics of SaveOrUpdate() seems to be confusing
for new users. Firstly, so long as you are not trying to use instances from one
session in another new session, you should not need to use Update()
or SaveOrUpdate(). Some whole applications will never use
either of these methods.

Usually Update() or SaveOrUpdate()
are used in the following scenario:

  • the application loads an object in the first session

  • the object is passed up to the UI tier

  • some modifications are made to the object

  • the object is passed back down to the business logic tier

  • the application persists these modifications by calling Update()
    in a second session

SaveOrUpdate() does the following:

  • if the object is already persistent in this session, do nothing

  • if the object has no identifier property, Save() it

  • if the object’s identifier matches the criteria specified by unsaved-value,
    Save() it

  • if the object is versioned (version or
    ), then the version will take precedence to identifier check,
    unless the versions unsaved-value="undefined" (default

  • if another object associated with the session has the same identifier, throw an

The last case can be avoided by using Merge(Object o).
This method copies the state of the given object onto the persistent object with
the same identifier. If there is no persistent instance currently associated with
the session, it will be loaded. The method returns the persistent instance. If the
given instance is unsaved or does not exist in the database, NHibernate will save
it and return it as a newly persistent instance. Otherwise, the given instance does
not become associated with the session. In most applications with detached objects,
you need both methods, SaveOrUpdate() and

9.4.3. Reattaching detached objects

The Lock() method allows the application to reassociate
an unmodified object with a new session.

//just reassociate:
sess.Lock(fritz, LockMode.None);
//do a version check, then reassociate:
sess.Lock(izi, LockMode.Read);
//do a version check, using SELECT ... FOR UPDATE, then reassociate:
sess.Lock(pk, LockMode.Upgrade);

9.5. Deleting persistent objects

ISession.Delete() will remove an object’s state from the
database. Of course, your application might still hold a reference to it. So it’s
best to think of Delete() as making a persistent instance


You may also delete many objects at once by passing a NHibernate query string to

You may now delete objects in any order you like, without risk of foreign key constraint
violations. Of course, it is still possible to violate a NOT NULL
constraint on a foreign key column by deleting objects in the wrong order.

9.6. Flush

From time to time the ISession will execute the SQL statements
needed to synchronize the ADO.NET connection’s state with the state of objects held
in memory. This process, flush, occurs by
default at the following points

  • from some invocations of Find() or Enumerable()

  • from NHibernate.ITransaction.Commit()

  • from ISession.Flush()

The SQL statements are issued in the following order

  1. all entity insertions, in the same order the corresponding objects were saved using

  2. all entity updates

  3. all collection deletions

  4. all collection element deletions, updates and insertions

  5. all collection insertions

  6. all entity deletions, in the same order the corresponding objects were deleted using

(An exception is that objects using native ID generation
are inserted when they are saved.)

Except when you explicity Flush(), there are absolutely
no guarantees about when the
executes the ADO.NET calls, only the order
in which they are executed. However, NHibernate does guarantee that the
methods will never return stale data; nor will they return
the wrong data.

It is possible to change the default behavior so that flush occurs less frequently.
The FlushMode class defines three different modes: only
flush at commit time (and only when the NHibernate ITransaction
API is used), flush automatically using the explained routine (will only work inside
an explicit NHibernate ITransaction), or never flush unless
Flush() is called explicitly. The last mode is useful for
long running units of work, where an ISession is kept open and disconnected for
a long time (see Section 11.4, “Optimistic concurrency

sess = sf.OpenSession();
ITransaction tx = sess.BeginTransaction();
sess.FlushMode = FlushMode.Commit; //allow queries to return stale state
Cat izi = (Cat) sess.Load(typeof(Cat), id);
izi.Name = "iznizi";
// execute some queries....
sess.Find("from Cat as cat left outer join cat.Kittens kitten");
//change to izi is not flushed!
tx.Commit(); //flush occurs

9.7. Ending a Session

Ending a session involves four distinct phases:

  • flush the session

  • commit the transaction

  • close the session

  • handle exceptions

9.7.1. Flushing the Session

If you happen to be using the ITransaction API, you don’t
need to worry about this step. It will be performed implicitly when the transaction
is committed. Otherwise you should call ISession.Flush()
to ensure that all changes are synchronized with the database.

9.7.2. Committing the database

If you are using the NHibernate ITransaction API, this
looks like:

tx.Commit(); // flush the session and commit the transaction

If you are managing ADO.NET transactions yourself you should manually
the ADO.NET transaction.


If you decide not to commit your changes:

tx.Rollback();  // rollback the transaction



If you rollback the transaction you should immediately close and discard the current
session to ensure that NHibernate’s internal state is consistent.

9.7.3. Closing the ISession

A call to ISession.Close() marks the end of a session.
The main implication of Close() is that the ADO.NET connection
will be relinquished by the session.


If you provided your own connection, Close() returns a
reference to it, so you can manually close it or return it to the pool. Otherwise
Close() returns it to the pool.

9.8. Exception handling

NHibernate use might lead to exceptions, usually HibernateException.
This exception can have a nested inner exception (the root cause), use the
property to access it.

If the ISession throws an exception you should immediately
rollback the transaction, call ISession.Close() and discard
the ISession instance. Certain methods of
will not leave the session
in a consistent state.

For exceptions thrown by the data provider while interacting with the database,
NHibernate will wrap the error in an instance of ADOException.
The underlying exception is accessible by calling ADOException.InnerException.
NHibernate converts the DbException into an appropriate ADOException subclass using
the ISQLExceptionConverter attached to the SessionFactory. By default, the ISQLExceptionConverter
is defined by the configured dialect; however, it is also possible to plug in a
custom implementation (see the api-docs for the ISQLExceptionConverter class for

The following exception handling idiom shows the typical case in NHibernate applications:

using (ISession sess = factory.OpenSession())
using (ITransaction tx = sess.BeginTransaction())
    // do some work

Or, when manually managing ADO.NET transactions:

ISession sess = factory.openSession();
    // do some work
catch (Exception e)

9.9. Lifecyles and object graphs

To save or update all objects in a graph of associated objects, you must either

  • Save(), SaveOrUpdate() or
    each individual object OR

  • map associated objects using cascade="all" or

Likewise, to delete all objects in a graph, either

  • Delete() each individual object OR

  • map associated objects using cascade="all",
    or cascade="delete".


  • If the child object’s lifespan is bounded by the lifespan of the of the parent object
    make it a lifecycle object by specifying

  • Otherwise, Save() and Delete()
    it explicitly from application code. If you really want to save yourself some extra
    typing, use cascade="save-update" and explicit

Mapping an association (many-to-one, or collection) with cascade="all"
marks the association as a parent/child style
relationship where save/update/deletion of the parent results in save/update/deletion
of the child(ren). Futhermore, a mere reference to a child from a persistent parent
will result in save / update of the child. The metaphor is incomplete, however.
A child which becomes unreferenced by its parent is not
automatically deleted, except in the case of a <one-to-many>
association mapped with cascade="all-delete-orphan". The
precise semantics of cascading operations are as follows:

  • If a parent is saved, all children are passed to SaveOrUpdate()

  • If a parent is passed to Update() or SaveOrUpdate(),
    all children are passed to SaveOrUpdate()

  • If a transient child becomes referenced by a persistent parent, it is passed to

  • If a parent is deleted, all children are passed to Delete()

  • If a transient child is dereferenced by a persistent parent,
    nothing special happens
    (the application should explicitly delete
    the child if necessary) unless cascade="all-delete-orphan",
    in which case the “orphaned” child is deleted.

NHibernate does not fully implement “persistence by reachability”, which would imply
(inefficient) persistent garbage collection. However, due to popular demand, NHibernate
does support the notion of entities becoming persistent when referenced by another
persistent object. Associations marked cascade="save-update"
behave in this way. If you wish to use this approach throughout your application,
it’s easier to specify the default-cascade attribute of
the <hibernate-mapping> element.

9.10. Interceptors

The IInterceptor interface provides callbacks from the
session to the application allowing the application to inspect and / or manipulate
properties of a persistent object before it is saved, updated, deleted or loaded.
One possible use for this is to track auditing information. For example, the following
IInterceptor automatically sets the CreateTimestamp
when an IAuditable is created and updates the
property when an IAuditable
is updated.

using System;
using NHibernate.Type;

namespace NHibernate.Test
    public class AuditInterceptor : IInterceptor
        private int updates;
        private int creates;
        public void OnDelete(object entity,
                             object id,
                             object[] state,
                             string[] propertyNames,
                             IType[] types)
            // do nothing
        public boolean OnFlushDirty(object entity, 
                                    object id, 
                                    object[] currentState,
                                    object[] previousState,
                                    string[] propertyNames,
                                    IType[] types) {
            if ( entity is IAuditable )
                for ( int i=0; i < propertyNames.Length; i++ )
                    if ( "LastUpdateTimestamp" == propertyNames[i] )
                        currentState[i] = DateTime.Now;
                        return true;
            return false;
        public boolean OnLoad(object entity, 
                              object id,
                              object[] state,
                              string[] propertyNames,
                              IType[] types)
            return false;
        public boolean OnSave(object entity,
                              object id,
                              object[] state,
                              string[] propertyNames,
                              IType[] types)
            if ( entity is IAuditable )
                for ( int i=0; i<propertyNames.Length; i++ )
                    if ( "CreateTimestamp" == propertyNames[i] )
                        state[i] = DateTime.Now;
                        return true;
            return false;
        public void PostFlush(ICollection entities)
            Console.Out.WriteLine("Creations: {0}, Updates: {1}", creates, updates);
        public void PreFlush(ICollection entities) {

The interceptor would be specified when a session is created.

ISession session = sf.OpenSession( new AuditInterceptor() );

You may also set an interceptor on a global level, using the Configuration:

new Configuration().SetInterceptor( new AuditInterceptor() );

9.11. Metadata API

NHibernate requires a very rich meta-level model of all entity and value types.
From time to time, this model is very useful to the application itself. For example,
the application might use NHibernate’s metadata to implement a “smart” deep-copy
algorithm that understands which objects should be copied (eg. mutable value types)
and which should not (eg. immutable value types and, possibly, associated entities).

NHibernate exposes metadata via the IClassMetadata and
ICollectionMetadata interfaces and the
hierarchy. Instances of the metadata interfaces may be obtained from
the ISessionFactory.

Cat fritz = ......;
IClassMetadata catMeta = sessionfactory.GetClassMetadata(typeof(Cat));
long id = (long) catMeta.GetIdentifier(fritz);
object[] propertyValues = catMeta.GetPropertyValues(fritz);
string[] propertyNames = catMeta.PropertyNames;
IType[] propertyTypes = catMeta.PropertyTypes;

// get an IDictionary of all properties which are not collections or associations
// TODO: what about components?

IDictionary namedValues = new Hashtable();
for ( int i=0; i<propertyNames.Length; i++ )
    if ( !propertyTypes[i].IsEntityType && !propertyTypes[i].IsCollectionType )
        namedValues[ propertyNames[i] ] = propertyValues[i];

Chapter 10. Read-only entities


NHibernate’s treatment of read-only entities
may differ from what you may have encountered elsewhere. Incorrect usage may cause
unexpected results.

When an entity is read-only:

  • NHibernate does not dirty-check the entity’s simple properties or single-ended associations;

  • NHibernate will not update simple properties or updatable single-ended associations;

  • NHibernate will not update the version of the read-only entity if only simple properties
    or single-ended updatable associations are changed;

In some ways, NHibernate treats read-only entities the same as entities that are
not read-only:

  • NHibernate cascades operations to associations as defined in the entity mapping.

  • NHibernate updates the version if the entity has a collection with changes that
    dirties the entity;

  • A read-only entity can be deleted.

Even if an entity is not read-only, its collection association can be affected if
it contains a read-only entity.

For details about the affect of read-only entities on different property and association
types, see
Section 10.2, “Read-only affect on property type”

For details about how to make entities read-only, see Section 10.1, “Making persistent
entities read-only”

NHibernate does some optimizing for read-only entities:

  • It saves execution time by not dirty-checking simple properties or single-ended

  • It saves memory by deleting database snapshots.

10.1. Making persistent entities read-only

Only persistent entities can be made read-only. Transient and detached entities
must be put in persistent state before they can be made read-only.

NHibernate provides the following ways to make persistent entities read-only:

10.1.1. Entities of immutable classes

When an entity instance of an immutable class is made persistent, NHibernate automatically
makes it read-only.

An entity of an immutable class can created and deleted the same as an entity of
a mutable class.

NHibernate treats a persistent entity of an immutable class the same way as a read-only
persistent entity of a mutable class. The only exception is that NHibernate will
not allow an entity of an immutable class to be changed so it is not read-only.

10.1.2. Loading persistent entities
as read-only


Entities of immutable classes are automatically loaded as read-only.

To change the default behavior so NHibernate loads entity instances of mutable classes
into the session and automatically makes them read-only, call:

Session.DefaultReadOnly = true;

To change the default back so entities loaded by NHibernate are not made read-only,

Session.DefaultReadOnly = false;

You can determine the current setting by using the property:


If Session.DefaultReadOnly property returns true, entities loaded by the following
are automatically made read-only:

Changing this default has no effect on:

  • persistent entities already in the session when the default was changed

  • persistent entities that are refreshed via Session.Refresh(); a refreshed persistent
    entity will only be read-only if it was read-only before refreshing

  • persistent entities added by the application via Session.Persist(), Session.Save(),
    and Session.Update() Session.SaveOrUpdate()

10.1.3. Loading read-only entities
from an HQL query/criteria


Entities of immutable classes are automatically loaded as read-only.

If Session.DefaultReadOnly returns false (the default) when an HQL query or criteria
executes, then entities and proxies of mutable classes loaded by the query will
not be read-only.

You can override this behavior so that entities and proxies loaded by an HQL query
or criteria are automatically made read-only.

For an HQL query, call:


Query.SetReadOnly(true) must be called before
, Query.UniqueResult(), or

For an HQL criteria, call:


Criteria.SetReadOnly(true) must be called before
, or Criteria.UniqueResult()

Entities and proxies that exist in the session before being returned by an HQL query
or criteria are not affected.

Uninitialized persistent collections returned by the query are not affected. Later,
when the collection is initialized, entities loaded into the session will be read-only
if Session.DefaultReadOnly returns true.

Using Query.SetReadOnly(true) or Criteria.SetReadOnly(true)
works well when a single HQL query or criteria loads all the entities and intializes
all the proxies and collections that the application needs to be read-only.

When it is not possible to load and initialize all necessary entities in a single
query or criteria, you can temporarily change the session default to load entities
as read-only before the query is executed. Then you can explicitly initialize proxies
and collections before restoring the session default.

ISession session = factory.OpenSession();
ITransaction tx = session.BeginTransaction();
session.DefaultReadOnly = true;
Contract contract = session.CreateQuery("from Contract where CustomerName = 'Sherman'").UniqueResult<Contract>();
session.DefaultReadOnly = false;

If Session.DefaultReadOnly returns true, then you can use Query.SetReadOnly(false)
and Criteria.SetReadOnly(false) to override this session setting and load entities
that are not read-only.

10.1.4. Making a persistent entity read-only


Persistent entities of immutable classes are automatically made read-only.

To make a persistent entity or proxy read-only, call:

Session.SetReadOnly(entityOrProxy, true)

To change a read-only entity or proxy of a mutable class so it is no longer read-only,

Session.SetReadOnly(entityOrProxy, false)


When a read-only entity or proxy is changed so it is no longer read-only, NHibernate
assumes that the current state of the read-only entity is consistent with its database
representation. If this is not true, then any non-flushed changes made before or
while the entity was read-only, will be ignored.

To throw away non-flushed changes and make the persistent entity consistent with
its database representation, call:


To flush changes made before or while the entity was read-only and make the database
representation consistent with the current state of the persistent entity:

// evict the read-only entity so it is detached

// make the detached entity (with the non-flushed changes) persistent

// now entity is no longer read-only and its changes can be flushed

10.2. Read-only affect on property type

The following table summarizes how different property types are affected by making
an entity read-only.

Table 10.1. Affect of read-only entity on property types

Property/Association Type Changes flushed to DB?

(Section 10.2.1,
“Simple properties”


Unidirectional one-to-one

Unidirectional many-to-one

“Unidirectional one-to-one and many-to-one”



Unidirectional one-to-many

Unidirectional many-to-many

“Unidirectional one-to-many and many-to-many”



Bidirectional one-to-one

(Section, “Bidirectional

only if the owning entity is not read-only*

Bidirectional one-to-many/many-to-one

inverse collection

non-inverse collection

(Section, “Bidirectional

only added/removed entities that are not read-only*


Bidirectional many-to-many

(Section, “Bidirectional


* Behavior is different when the entity having the property/association is read-only,
compared to when it is not read-only.

10.2.1. Simple properties

When a persistent object is read-only, NHibernate does not dirty-check simple properties.

NHibernate will not synchronize simple property state changes to the database. If
you have automatic versioning, NHibernate will not increment the version if any
simple properties change.

ISession session = factory.OpenSession();
ITransaction tx = session.BeginTransaction();

// get a contract and make it read-only
Contract contract = session.Get<Contract>(contractId);
session.SetReadOnly(contract, true);

// contract.CustomerName is "Sherman"
contract.CustomerName = "Yogi";

tx = session.BeginTransaction();

contract = session.Get<Contract>(contractId);
// contract.CustomerName is still "Sherman"

10.2.2. Unidirectional associations Unidirectional
one-to-one and many-to-one

NHibernate treats unidirectional one-to-one and many-to-one associations in the
same way when the owning entity is read-only.

We use the term unidirectional single-ended association
when referring to functionality that is common to unidirectional one-to-one and
many-to-one associations.

NHibernate does not dirty-check unidirectional single-ended associations when the
owning entity is read-only.

If you change a read-only entity’s reference to a unidirectional single-ended association
to null, or to refer to a different entity, that change will not be flushed to the


If an entity is of an immutable class, then its references to unidirectional single-ended
associations must be assigned when that entity is first created. Because the entity
is automatically made read-only, these references can not be updated.

If automatic versioning is used, NHibernate will not increment the version due to
local changes to unidirectional single-ended associations.

In the following examples, Contract has a unidirectional many-to-one association
with Plan. Contract cascades save and update operations to the association.

The following shows that changing a read-only entity’s many-to-one association reference
to null has no effect on the entity’s database representation.

// get a contract with an existing plan;
// make the contract read-only and set its plan to null 
tx = session.BeginTransaction();
Contract contract = session.Get<Contract>(contractId);
session.SetReadOnly(contract, true);
contract.Plan = null;

// get the same contract
tx = session.BeginTransaction();
Contract contract = session.Get<Contract>(contractId);

// contract.Plan still refers to the original plan;


The following shows that, even though an update to a read-only entity’s many-to-one
association has no affect on the entity’s database representation, flush still cascades
the save-update operation to the locally changed association.

// get a contract with an existing plan;
// make the contract read-only and change to a new plan
tx = session.BeginTransaction();
Contract contract = session.Get<Contract>(contractId);
session.SetReadOnly(contract, true);
Plan newPlan = new Plan("new plan");
contract.Plan = newPlan;

// get the same contract
tx = session.BeginTransaction();
contract = session.Get<Contract>(contractId);
newPlan = session.Get<Plan>(newPlan.Id);

// contract.Plan still refers to the original plan;
// newPlan is non-null because it was persisted when 
// the previous transaction was committed; 

session.Close(); Unidirectional
one-to-many and many-to-many

NHibernate treats unidirectional one-to-many and many-to-many associations owned
by a read-only entity the same as when owned by an entity that is not read-only.

NHibernate dirty-checks unidirectional one-to-many and many-to-many associations;

The collection can contain entities that are read-only, as well as entities that
are not read-only.

Entities can be added and removed from the collection; changes are flushed to the

If automatic versioning is used, NHibernate will update the version due to changes
in the collection if they dirty the owning entity.

10.2.3. Bidirectional associations Bidirectional one-to-one

If a read-only entity owns a bidirectional one-to-one association:

  • NHibernate does not dirty-check the association.

  • updates that change the association reference to null or to refer to a different
    entity will not be flushed to the database.

  • If automatic versioning is used, NHibernate will not increment the version due to
    local changes to the association.


If an entity is of an immutable class, and it owns a bidirectional one-to-one association,
then its reference must be assigned when that entity is first created. Because the
entity is automatically made read-only, these references cannot be updated.

When the owner is not read-only, NHibernate treats an association with a read-only
entity the same as when the association is with an entity that is not read-only. Bidirectional

A read-only entity has no impact on a bidirectional one-to-many/many-to-one association

  • the read-only entity is on the one-to-many side using an inverse collection;

  • the read-only entity is on the one-to-many side using a non-inverse collection;

  • the one-to-many side uses a non-inverse collection that contains the read-only entity

When the one-to-many side uses an inverse collection:

  • a read-only entity can only be added to the collection when it is created;

  • a read-only entity can only be removed from the collection by an orphan delete or
    by explicitly deleting the entity. Bidirectional many-to-many

NHibernate treats bidirectional many-to-many associations owned by a read-only entity
the same as when owned by an entity that is not read-only.

NHibernate dirty-checks bidirectional many-to-many associations.

The collection on either side of the association can contain entities that are read-only,
as well as entities that are not read-only.

Entities are added and removed from both sides of the collection; changes are flushed
to the database.

If automatic versioning is used, NHibernate will update the version due to changes
in both sides of the collection if they dirty the entity owning the respective collections.

Chapter 11. Transactions And Concurrency

NHibernate is not itself a database. It is a lightweight object-relational mapping
tool. Transaction management is delegated to the underlying database connection.
If the connection is enlisted with a distributed transaction, operations performed
by the ISession are atomically part of the wider distributed
transaction. NHibernate can be seen as a thin adapter to ADO.NET, adding object-oriented

11.1. Configurations, Sessions and Factories

An ISessionFactory is an expensive-to-create, threadsafe
object intended to be shared by all application threads. An ISession
is an inexpensive, non-threadsafe object that should be used once, for a single
business process, and then discarded. For example, when using NHibernate in an ASP.NET
application, pages could obtain an ISessionFactory using:

ISessionFactory sf = Global.SessionFactory;

Each call to a service method could create a new ISession,
Flush() it, Commit() its transaction,
Close() it and finally discard it. (The
may also be kept in a static Singleton
helper variable.)

We use the NHibernate ITransaction API as discussed previously,
a single Commit() of a NHibernate ITransaction
flushes the state and commits any underlying database connection (with special handling
of distributed transactions).

Ensure you understand the semantics of Flush(). Flushing
synchronizes the persistent store with in-memory changes but
vice-versa. Note that for all NHibernate ADO.NET connections/transactions,
the transaction isolation level for that connection applies to all operations executed
by NHibernate!

The next few sections will discuss alternative approaches that utilize versioning
to ensure transaction atomicity. These are considered “advanced” approaches to be
used with care.

11.2. Threads and connections

You should observe the following practices when creating NHibernate Sessions:

  • Never create more than one concurrent ISession or
    instance per database connection.

  • Be extremely careful when creating more than one ISession
    per database per transaction. The ISession itself keeps
    track of updates made to loaded objects, so a different ISession
    might see stale data.

  • The ISession is not
    threadsafe! Never access the same ISession in two concurrent
    threads. An ISession is usually only a single unit-of-work!

11.3. Considering object identity

The application may concurrently access the same persistent state in two different
units-of-work. However, an instance of a persistent class is never shared between
two ISession instances. Hence there are two different notions
of identity:

Database Identity

foo.Id.Equals( bar.Id )

CLR Identity

foo == bar

Then for objects attached to a particular
Session, the two notions are equivalent. However, while
the application might concurrently access the “same” (persistent identity) business
object in two different sessions, the two instances will actually be “different”
(CLR identity).

This approach leaves NHibernate and the database to worry about concurrency. The
application never needs to synchronize on any business object, as long as it sticks
to a single thread per ISession or object identity (within
an ISession the application may safely use
to compare objects).

11.4. Optimistic concurrency control

Many business processes require a whole series of interactions with the user interleaved
with database accesses. In web and enterprise applications it is not acceptable
for a database transaction to span a user interaction.

Maintaining isolation of business processes becomes the partial responsibility of
the application tier, hence we call this process a long running
application transaction
. A single application transaction usually
spans several database transactions. It will be atomic if only one of these database
transactions (the last one) stores the updated data, all others simply read data.

The only approach that is consistent with high concurrency and high scalability
is optimistic concurrency control with versioning. NHibernate provides for three
possible approaches to writing application code that uses optimistic concurrency.

11.4.1. Long session with
automatic versioning

A single ISession instance and its persistent instances
are used for the whole application transaction.

The ISession uses optimistic locking with versioning to
ensure that many database transactions appear to the application as a single logical
application transaction. The ISession is disconnected from
any underlying ADO.NET connection when waiting for user interaction. This approach
is the most efficient in terms of database access. The application need not concern
itself with version checking or with reattaching detached instances.

// foo is an instance loaded earlier by the Session
transaction = session.BeginTransaction();
foo.Property = "bar";

The foo object still knows which ISession
it was loaded it. As soon as the ISession has an ADO.NET
connection, we commit the changes to the object.

This pattern is problematic if our ISession is too big
to be stored during user think time, e.g. an HttpSession
should be kept as small as possible. As the ISession is
also the (mandatory) first-level cache and contains all loaded objects, we can propably
use this strategy only for a few request/response cycles. This is indeed recommended,
as the ISession will soon also have stale data.

11.4.2. Many sessions with automatic

Each interaction with the persistent store occurs in a new ISession.
However, the same persistent instances are reused for each interaction with the
database. The application manipulates the state of detached instances originally
loaded in another ISession and then “reassociates” them
using ISession.Update() or ISession.SaveOrUpdate().

// foo is an instance loaded by a previous Session
foo.Property = "bar";
session = factory.OpenSession();
transaction = session.BeginTransaction();

You may also call Lock() instead of Update()
and use LockMode.Read (performing a version check, bypassing
all caches) if you are sure that the object has not been modified.

11.4.3. Customizing automatic

You may disable NHibernate’s automatic version increment for particular properties
and collections by setting the optimistic-lock mapping
attribute to false. NHibernate will then no longer increment
versions if the property is dirty.

Legacy database schemas are often static and can’t be modified. Or, other applications
might also access the same database and don’t know how to handle version numbers
or even timestamps. In both cases, versioning can’t rely on a particular column
in a table. To force a version check without a version or timestamp property mapping,
with a comparison of the state of all fields in a row, turn on optimistic-lock="all"
in the <class> mapping. Note that this concepetually
only works if NHibernate can compare the old and new state, i.e. if you use a single
long ISession and not session-per-request-with-detached-objects.

Sometimes concurrent modification can be permitted as long as the changes that have
been made don’t overlap. If you set optimistic-lock="dirty"
when mapping the <class>, NHibernate will only compare
dirty fields during flush.

In both cases, with dedicated version/timestamp columns or with full/dirty field
comparison, NHibernate uses a single UPDATE statement (with
an appropriate WHERE clause) per entity to execute the
version check and update the information. If you use transitive persistence to cascade
reattachment to associated entities, NHibernate might execute uneccessary updates.
This is usually not a problem, but on update
triggers in the database might be executed even when no changes have been made to
detached instances. You can customize this behavior by setting select-before-update="true"
in the <class> mapping, forcing NHibernate to SELECT the instance to ensure that changes did actually
occur, before updating the row.

11.4.4. Application version checking

Each interaction with the database occurs in a new ISession
that reloads all persistent instances from the database before manipulating them.
This approach forces the application to carry out its own version checking to ensure
application transaction isolation. (Of course, NHibernate will still
version numbers for you.) This approach is the least
efficient in terms of database access.

// foo is an instance loaded by a previous Session
session = factory.OpenSession();
transaction = session.BeginTransaction();
int oldVersion = foo.Version;
session.Load( foo, foo.Key );
if ( oldVersion != foo.Version ) throw new StaleObjectStateException();
foo.Property = "bar";

Of course, if you are operating in a low-data-concurrency environment and don’t
require version checking, you may use this approach and just skip the version check.

11.5. Session disconnection

The first approach described above is to maintain a single ISession
for a whole business process thats spans user think time. (For example, a servlet
might keep an ISession in the user’s HttpSession.)
For performance reasons you should

  1. commit the ITransaction and then

  2. disconnect the ISession from the ADO.NET connection

before waiting for user activity. The method ISession.Disconnect()
will disconnect the session from the ADO.NET connection and return the connection
to the pool (unless you provided the connection).

ISession.Reconnect() obtains a new connection (or you may
supply one) and restarts the session. After reconnection, to force a version check
on data you aren’t updating, you may call ISession.Lock()
on any objects that might have been updated by another transaction. You don’t need
to lock any data that you are updating.

Heres an example:

ISessionFactory sessions;
IList fooList;
Bar bar;
ISession s = sessions.OpenSession();
ITransaction tx = null;

    tx = s.BeginTransaction())

    fooList = s.Find(
    	"select foo from Eg.Foo foo where foo.Date = current date"
        // uses db2 date function

    bar = new Bar();

catch (Exception)
    if (tx != null) tx.Rollback();

Later on:


    tx = s.BeginTransaction();

    bar.FooTable = new HashMap();
    foreach (Foo foo in fooList)
        s.Lock(foo, LockMode.Read);    //check that foo isn't stale
        bar.FooTable.Put( foo.Name, foo );

catch (Exception)
    if (tx != null) tx.Rollback();

You can see from this how the relationship between ITransactions
and ISessions is many-to-one, An ISession
represents a conversation between the application and the database. The
breaks that conversation up into atomic units of work at the
database level.

11.6. Pessimistic Locking

It is not intended that users spend much time worring about locking strategies.
It’s usually enough to specify an isolation level for the ADO.NET connections and
then simply let the database do all the work. However, advanced users may sometimes
wish to obtain exclusive pessimistic locks, or re-obtain locks at the start of a
new transaction.

NHibernate will always use the locking mechanism of the database, never lock objects
in memory!

The LockMode class defines the different lock levels that
may be acquired by NHibernate. A lock is obtained by the following mechanisms:

  • LockMode.Write is acquired automatically when NHibernate
    updates or inserts a row.

  • LockMode.Upgrade may be acquired upon explicit user request
    using SELECT ... FOR UPDATE on databases which support
    that syntax.

  • LockMode.UpgradeNoWait may be acquired upon explicit user
    request using a SELECT ... FOR UPDATE NOWAIT under Oracle.

  • LockMode.Read is acquired automatically when NHibernate
    reads data under Repeatable Read or Serializable isolation level. May be re-acquired
    by explicit user request.

  • LockMode.None represents the absence of a lock. All objects
    switch to this lock mode at the end of an ITransaction.
    Objects associated with the session via a call to Update()
    or SaveOrUpdate() also start out in this lock mode.

The “explicit user request” is expressed in one of the following ways:

  • A call to ISession.Load(), specifying a

  • A call to ISession.Lock().

  • A call to IQuery.SetLockMode().

If ISession.Load() is called with Upgrade
or UpgradeNoWait, and the requested object was not yet
loaded by the session, the object is loaded using SELECT ... FOR
. If Load() is called for an object that
is already loaded with a less restrictive lock than the one requested, NHibernate
calls Lock() for that object.

ISession.Lock() performs a version number check if the
specified lock mode is Read, Upgrade
or UpgradeNoWait. (In the case of Upgrade
or UpgradeNoWait, SELECT ... FOR UPDATE
is used.)

If the database does not support the requested lock mode, NHibernate will use an
appropriate alternate mode (instead of throwing an exception). This ensures that
applications will be portable.

11.7. Connection Release Modes

The legacy (1.0.x) behavior of NHibernate in regards to ADO.NET connection management
was that a ISession would obtain a connection when it was
first needed and then hold unto that connection until the session was closed. NHibernate
introduced the notion of connection release modes to tell a session how to handle
its ADO.NET connections. Note that the following discussion is pertinent only to
connections provided through a configured IConnectionProvider;
user-supplied connections are outside the breadth of this discussion. The different
release modes are identified by the enumerated values of NHibernate.ConnectionReleaseMode:

  • OnClose – is essentially the legacy behavior described
    above. The NHibernate session obtains a connection when it first needs to perform
    some database access and holds unto that connection until the session is closed.

  • AfterTransaction – says to release connections after a
    NHibernate.ITransaction has completed.

The configuration parameter hibernate.connection.release_mode
is used to specify which release mode to use. The possible values:

  • auto (the default) – equivalent to after_transaction
    in the current release. It is rarely a good idea to change this default behavior
    as failures due to the value of this setting tend to indicate bugs and/or invalid
    assumptions in user code.

  • on_close – says to use ConnectionReleaseMode.OnClose.
    This setting is left for backwards compatibility, but its use is highly discouraged.

  • after_transaction – says to use ConnectionReleaseMode.AfterTransaction.
    Note that with ConnectionReleaseMode.AfterTransaction,
    if a session is considered to be in auto-commit mode (i.e. no transaction was started)
    connections will be released after every operation.

As of NHibernate, if your application manages transactions through .NET APIs such
as System.Transactions library, ConnectionReleaseMode.AfterTransaction
may cause NHibernate to open and close several connections during one transaction,
leading to unnecessary overhead and transaction promotion from local to distributed.
Specifying ConnectionReleaseMode.OnClose will revert to
the legacy behavior and prevent this problem from occuring.

Chapter 12. Interceptors and events

It is often useful for the application to react to certain events that occur inside
NHibernate. This allows implementation of certain kinds of generic functionality,
and extension of NHibernate functionality.

12.1. Interceptors

The IInterceptor interface provides callbacks from the
session to the application allowing the application to inspect and/or manipulate
properties of a persistent object before it is saved, updated, deleted or loaded.
One possible use for this is to track auditing information. For example, the following
IInterceptor automatically sets the createTimestamp
when an IAuditable is created and updates the
property when an IAuditable
is updated.

You may either implement IInterceptor directly or (better)
extend EmptyInterceptor.

using System;
using NHibernate;
using NHibernate.Type;

public class AuditInterceptor : EmptyInterceptor {

    private int updates;
    private int creates;
    private int loads;

    public override void OnDelete(object entity,
                                  object id,
                                  object[] state,
                                  string[] propertyNames,
                                  IType[] types)
        // do nothing

    public override bool OnFlushDirty(object entity, 
                                      object id, 
				      object[] currentState,
				      object[] previousState, 
				      string[] propertyNames,
				      IType[] types) 
        if ( entity is IAuditable ) {
            for ( int i=0; i < propertyNames.Length; i++ ) {
                if ( "lastUpdateTimestamp".Equals( propertyNames[i] ) ) {
                    currentState[i] = new DateTime();
                    return true;
        return false;

    public override bool OnLoad(object entity, 
                                object id, 
				object[] state, 
				string[] propertyNames, 
				IType[] types)
        if ( entity is IAuditable ) {
        return false;

    public override bool OnSave(object entity, 
                                object id, 
				object[] state, 
				string[] propertyNames, 
				IType[] types)
        if ( entity is IAuditable ) {
            for ( int i=0; i<propertyNames.Length; i++ ) {
                if ( "createTimestamp".Equals( propertyNames[i] ) ) {
                    state[i] = new DateTime();
                    return true;
        return false;

    public override void AfterTransactionCompletion(ITransaction tx)
        if ( tx.WasCommitted ) {
            System.Console.WriteLine("Creations: " + creates + ", Updates: " + updates, "Loads: " + loads);


Interceptors come in two flavors: ISession-scoped and ISessionFactory-scoped.

An ISession-scoped interceptor is specified when a session
is opened using one of the overloaded ISessionFactory.OpenSession() methods accepting
an IInterceptor.

ISession session = sf.OpenSession( new AuditInterceptor() );

An ISessionFactory-scoped interceptor is registered with
the Configuration object prior to building the
. In this case, the supplied interceptor will be applied
to all sessions opened from that ISessionFactory; this
is true unless a session is opened explicitly specifying the interceptor to use.
ISessionFactory-scoped interceptors must be thread safe,
taking care to not store session-specific state since multiple sessions will use
this interceptor (potentially) concurrently.

new Configuration().SetInterceptor( new AuditInterceptor() );

12.2. Event system

If you have to react to particular events in your persistence layer, you may also
use the NHibernate2 event architecture. The
event system can be used in addition or as a replacement for interceptors.

Essentially all of the methods of the ISession interface
correlate to an event. You have a LoadEvent, a
, etc (consult the XML configuration-file XSD or the
namespace for the full list of defined event types).
When a request is made of one of these methods, the ISession
generates an appropriate event and passes it to the configured event listeners for
that type. Out-of-the-box, these listeners implement the same processing in which
those methods always resulted. However, you are free to implement a customization
of one of the listener interfaces (i.e., the LoadEvent
is processed by the registered implemenation of the ILoadEventListener
interface), in which case their implementation would be responsible for processing
any Load() requests made of the ISession.

The listeners should be considered effectively singletons; meaning, they are shared
between requests, and thus should not save any state as instance variables.

A custom listener should implement the appropriate interface for the event it wants
to process and/or extend one of the convenience base classes (or even the default
event listeners used by NHibernate out-of-the-box as their methods are declared
virtual for this purpose). Custom listeners can either be registered programmatically
through the Configuration object, or specified in the NHibernate
configuration XML. Here’s an example of a custom load event listener:

public class MyLoadListener : ILoadEventListener 
    // this is the single method defined by the LoadEventListener interface
    public void OnLoad(LoadEvent theEvent, LoadType loadType)
        if ( !MySecurity.IsAuthorized( theEvent.EntityClassName, theEvent.EntityId ) ) {
            throw new MySecurityException("Unauthorized access");

You also need a configuration entry telling NHibernate to use the listener in addition
to the default listener:

        <event type="load">
            <listener class="MyLoadListener"/>
            <listener class="NHibernate.Event.Default.DefaultLoadEventListener"/>

Instead, you may register it programmatically:

Configuration cfg = new Configuration();
ILoadEventListener[] stack = new ILoadEventListener[] { new MyLoadListener(), new DefaultLoadEventListener() };
cfg.EventListeners.LoadEventListeners = stack;

Listeners registered declaratively cannot share instances. If the same class name
is used in multiple <listener/> elements, each reference
will result in a separate instance of that class. If you need the capability to
share listener instances between listener types you must use the programmatic registration

Why implement an interface and define the specific type during configuration? Well,
a listener implementation could implement multiple event listener interfaces. Having
the type additionally defined during registration makes it easier to turn custom
listeners on or off during configuration.

Chapter 13. Batch processing

A naive approach to inserting 100 000 rows in the database using NHibernate might
look like this:

ISession session = sessionFactory.OpenSession();
ITransaction tx = session.BeginTransaction();
for ( int i=0; i<100000; i++ ) {
    Customer customer = new Customer(.....);

This would fall over with an OutOfMemoryException somewhere
around the 50 000th row. That’s because NHibernate caches all the newly inserted
Customer instances in the session-level cache.

In this chapter we’ll show you how to avoid this problem. First, however, if you
are doing batch processing, it is absolutely critical that you enable the use of
ADO batching, if you intend to achieve reasonable performance. Set the ADO batch
size to a reasonable number (say, 10-50):

adonet.batch_size 20

Note that NHibernate disables insert batching at the
ADO level transparently if you use an identiy identifier

You also might like to do this kind of work in a process where interaction with
the second-level cache is completely disabled:

cache.use_second_level_cache false

However, this is not absolutely necessary, since we can explicitly set the
to disable interaction with the second-level cache.

13.1. Batch inserts

When making new objects persistent, you must Flush() and
then Clear() the session regularly, to control the size
of the first-level cache.

ISession session = sessionFactory.openSession();
ITransaction tx = session.BeginTransaction();
for ( int i=0; i<100000; i++ ) {
    Customer customer = new Customer(.....);
    if ( i % 20 == 0 ) { //20, same as the ADO batch size
        //flush a batch of inserts and release memory:

13.2. The StatelessSession interface

Alternatively, NHibernate provides a command-oriented API that may be used for streaming
data to and from the database in the form of detached objects. A
has no persistence context associated with it and does
not provide many of the higher-level life cycle semantics. In particular, a stateless
session does not implement a first-level cache nor interact with any second-level
or query cache. It does not implement transactional write-behind or automatic dirty
checking. Operations performed using a stateless session do not ever cascade to
associated instances. Collections are ignored by a stateless session. Operations
performed via a stateless session bypass NHibernate’s event model and interceptors.
Stateless sessions are vulnerable to data aliasing effects, due to the lack of a
first-level cache. A stateless session is a lower-level abstraction, much closer
to the underlying ADO.

IStatelessSession session = sessionFactory.OpenStatelessSession();
ITransaction tx = session.BeginTransaction();
var customers = session.GetNamedQuery("GetCustomers")
while ( customers.MoveNext() ) {
    Customer customer = customers.Current;

Note that in this code example, the Customer instances
returned by the query are immediately detached. They are never associated with any
persistence context.

The insert(), update() and delete()
operations defined by the StatelessSession interface are
considered to be direct database row-level operations, which result in immediate
execution of a SQL INSERT, UPDATE or DELETE
respectively. Thus, they have very different semantics to the Save(),
and Delete() operations defined
by the ISession interface.

13.3. DML-style operations

As already discussed, automatic and transparent object/relational mapping is concerned
with the management of object state. This implies that the object state is available
in memory, hence manipulating (using the SQL Data Manipulation Language
(DML) statements: INSERT, UPDATE,
DELETE) data directly in the database will not affect in-memory
state. However, NHibernate provides methods for bulk SQL-style DML statement execution
which are performed through the Hibernate Query Language (HQL).

The pseudo-syntax for UPDATE and DELETE
statements is: ( UPDATE | DELETE ) FROM? EntityName (WHERE where_conditions)?.
Some points to note:

  • In the from-clause, the FROM keyword is optional

  • There can only be a single entity named in the from-clause; it can optionally be
    aliased. If the entity name is aliased, then any property references must be qualified
    using that alias; if the entity name is not aliased, then it is illegal for any
    property references to be qualified.

  • No joins (either implicit
    or explicit) can be specified in a bulk HQL query. Sub-queries may be used in the
    where-clause; the subqueries, themselves, may contain joins.

  • The where-clause is also optional.

As an example, to execute an HQL UPDATE, use the

ISession session = sessionFactory.OpenSession();
ITransaction tx = session.BeginTransaction();

string hqlUpdate = "update Customer c set c.name = :newName where c.name = :oldName";
// or string hqlUpdate = "update Customer set name = :newName where name = :oldName";
int updatedEntities = s.CreateQuery( hqlUpdate )
        .SetString( "newName", newName )
        .SetString( "oldName", oldName )

HQL UPDATE statements, by default do not effect the version or the timestamp property values for the
affected entities. However, you can force NHibernate to properly reset the
or timestamp property values through the
use of a versioned update. This is achieved by adding the
VERSIONED keyword after the UPDATE

ISession session = sessionFactory.OpenSession();
ITransaction tx = session.BeginTransaction();
string hqlVersionedUpdate = "update versioned Customer set name = :newName where name = :oldName";
int updatedEntities = s.CreateQuery( hqlUpdate )
        .SetString( "newName", newName )
        .SetString( "oldName", oldName )

Note that custom version types (NHibernate.Usertype.IUserVersionType)
are not allowed in conjunction with a update versioned

To execute an HQL DELETE, use the same

ISession session = sessionFactory.OpenSession();
ITransaction tx = session.BeginTransaction();

String hqlDelete = "delete Customer c where c.name = :oldName";
// or String hqlDelete = "delete Customer where name = :oldName";
int deletedEntities = s.CreateQuery( hqlDelete )
        .SetString( "oldName", oldName )

The int value returned by the IQuery.ExecuteUpdate()
method indicate the number of entities effected by the operation. Consider this
may or may not correlate to the number of rows effected in the database. An HQL
bulk operation might result in multiple actual SQL statements being executed, for
joined-subclass, for example. The returned number indicates the number of actual
entities affected by the statement. Going back to the example of joined-subclass,
a delete against one of the subclasses may actually result in deletes against not
just the table to which that subclass is mapped, but also the “root” table and potentially
joined-subclass tables further down the inheritence hierarchy.

The pseudo-syntax for INSERT statements is:
INSERT INTO EntityName properties_list select_statement
. Some points to

  • Only the INSERT INTO … SELECT … form is supported; not the INSERT INTO … VALUES
    … form.

    The properties_list is analogous to the column speficiation
    in the SQL INSERT statement. For entities involved in mapped
    inheritence, only properties directly defined on that given class-level can be used
    in the properties_list. Superclass properties are not allowed; and subclass properties
    do not make sense. In other words, INSERT statements are
    inherently non-polymorphic.

  • select_statement can be any valid HQL select query, with the caveat that the return
    types must match the types expected by the insert. Currently, this is checked during
    query compilation rather than allowing the check to relegate to the database. Note
    however that this might cause problems between NHibernate Types
    which are equivalent as opposed to
    . This might cause issues with mismatches between a property
    defined as a NHibernate.Type.DateType and a property defined
    as a NHibernate.Type.TimestampType, even though the database
    might not make a distinction or might be able to handle the conversion.

  • For the id property, the insert statement gives you two options. You can either
    explicitly specify the id property in the properties_list (in which case its value
    is taken from the corresponding select expression) or omit it from the properties_list
    (in which case a generated value is used). This later option is only available when
    using id generators that operate in the database; attempting to use this option
    with any “in memory” type generators will cause an exception during parsing. Note
    that for the purposes of this discussion, in-database generators are considered
    to be NHibernate.Id.SequenceGenerator (and its subclasses)
    and any implementors of NHibernate.Id.IPostInsertIdentifierGenerator.
    The most notable exception here is NHibernate.Id.TableHiLoGenerator,
    which cannot be used because it does not expose a selectable way to get its values.

  • For properties mapped as either version or
    , the insert statement gives you two options. You can either specify
    the property in the properties_list (in which case its value is taken from the corresponding
    select expressions) or omit it from the properties_list (in which case the
    seed value
    defined by the NHibernate.Type.IVersionType
    is used).

An example HQL INSERT statement execution:

ISession session = sessionFactory.OpenSession();
ITransaction tx = session.BeginTransaction();

var hqlInsert = "insert into DelinquentAccount (id, name) select c.id, c.name from Customer c where ...";
int createdEntities = s.CreateQuery( hqlInsert )

Chapter 14. HQL: The Hibernate Query Language

NHibernate is equiped with an extremely powerful query language that (quite intentionally)
looks very much like SQL. But don’t be fooled by the syntax; HQL is fully object-oriented,
understanding notions like inheritence, polymorphism and association.

14.1. Case Sensitivity

Queries are case-insensitive, except for names of .NET classes and properties. So
SeLeCT is the same as sELEct is
the same as SELECT but Eg.FOO
is not Eg.Foo and foo.barSet is
not foo.BARSET.

This manual uses lowercase HQL keywords. Some users find queries with uppercase
keywords more readable, but we find this convention ugly when embedded in C# code.

14.2. The from clause

The simplest possible NHibernate query is of the form:

from Eg.Cat

which simply returns all instances of the class Eg.Cat.

Most of the time, you will need to assign an alias,
since you will want to refer to the Cat in other parts
of the query.

from Eg.Cat as cat

This query assigns the alias cat to Cat
instances, so we could use that alias later in the query. The as
keyword is optional; we could also write:

from Eg.Cat cat

Multiple classes may appear, resulting in a cartesian product or “cross” join.

from Formula, Parameter
from Formula as form, Parameter as param

It is considered good practice to name query aliases using an initial lowercase,
consistent with naming standards for local variables (eg. domesticCat).

14.3. Associations and joins

We may also assign aliases to associated entities, or even to elements of a collection
of values, using a join.

from Eg.Cat as cat 
    inner join cat.Mate as mate
    left outer join cat.Kittens as kitten

from Eg.Cat as cat left join cat.Mate.Kittens as kittens

from Formula form full join form.Parameter param

The supported join types are borrowed from ANSI SQL

  • inner join

  • left outer join

  • right outer join

  • full join (not usually useful)

The inner join, left outer join
and right outer join constructs may be abbreviated.

from Eg.Cat as cat 
    join cat.Mate as mate
    left join cat.Kittens as kitten

In addition, a “fetch” join allows associations or collections of values to be initialized
along with their parent objects, using a single select. This is particularly useful
in the case of a collection. It effectively overrides the outer join and lazy declarations
of the mapping file for associations and collections. See Section 19.1, “Fetching strategies”
for more information.

from Eg.Cat as cat 
    inner join fetch cat.Mate
    left join fetch cat.Kittens

The associated objects are not returned directly in the query results. Instead,
they may be accessed via the parent object.

It is possible to create a cartesian product by join fetching more than one collection
in a query, so take care in this case. Join fetching multiple collection roles is
also disabled for bag mappings. Note also that the fetch
construct may not be used in queries called using Enumerable().
Finally, note that full join fetch and
right join fetch
are not meaningful.

14.4. The select clause

The select clause picks which objects and properties to
return in the query result set. Consider:

select mate 
from Eg.Cat as cat 
    inner join cat.Mate as mate

The query will select Mates of other Cats.
Actually, you may express this query more compactly as:

select cat.Mate from Eg.Cat cat

You may even select collection elements, using the special elements
function. The following query returns all kittens of any cat.

select elements(cat.Kittens) from Eg.Cat cat

Queries may return properties of any value type including properties of component

select cat.Name from Eg.DomesticCat cat
where cat.Name like 'fri%'

select cust.Name.FirstName from Customer as cust

Queries may return multiple objects and/or properties as an array of type

select mother, offspr, mate.Name 
from Eg.DomesticCat as mother
    inner join mother.Mate as mate
    left outer join mother.Kittens as offspr

or as an actual typesafe object

select new Family(mother, mate, offspr)
from Eg.DomesticCat as mother
    join mother.Mate as mate
    left join mother.Kittens as offspr

assuming that the class Family has an appropriate constructor.

14.5. Aggregate functions

HQL queries may even return the results of aggregate functions on properties:

select avg(cat.Weight), sum(cat.Weight), max(cat.Weight), count(cat)
from Eg.Cat cat

Collections may also appear inside aggregate functions in the select

select cat, count( elements(cat.Kittens) ) 
from Eg.Cat cat group by cat.Id, cat.Weight, ...

The supported aggregate functions are

  • avg(...), sum(...), min(...), max(...)

  • count(*)

  • count(...), count(distinct ...), count(all...)

The distinct and all keywords
may be used and have the same semantics as in SQL.

select distinct cat.Name from Eg.Cat cat

select count(distinct cat.Name), count(cat) from Eg.Cat cat

14.6. Polymorphic queries

A query like:

from Eg.Cat as cat

returns instances not only of Cat, but also of subclasses
like DomesticCat. NHibernate queries may name
.NET class or interface in the from
clause. The query will return instances of all persistent classes that extend that
class or implement the interface. The following query would return all persistent

from System.Object o

The interface INamed might be implemented by various persistent

from Eg.Named n, Eg.Named m where n.Name = m.Name

Note that these last two queries will require more than one SQL
. This means that the order by clause does
not correctly order the whole result set.

In order to use non-mapped base classes or interfaces in HQL queries, they have
to be imported. See Section 5.1.20, “import” for more information.

14.7. The where clause

The where clause allows you to narrow the list of instances

from Eg.Cat as cat where cat.Name='Fritz'

returns instances of Cat named ‘Fritz’.

select foo 
from Eg.Foo foo, Eg.Bar bar
where foo.StartDate = bar.Date

will return all instances of Foo for which there exists
an instance of Bar with a Date
property equal to the StartDate property of the
. Compound path expressions make the where
clause extremely powerful. Consider:

from Eg.Cat cat where cat.Mate.Name is not null

This query translates to an SQL query with a table (inner) join. If you were to
write something like

from Eg.Foo foo  
where foo.Bar.Baz.Customer.Address.City is not null

you would end up with a query that would require four table joins in SQL.

The = operator may be used to compare not only properties,
but also instances:

from Eg.Cat cat, Eg.Cat rival where cat.Mate = rival.Mate

select cat, mate 
from Eg.Cat cat, Eg.Cat mate
where cat.Mate = mate

The special property (lowercase) id may be used to reference
the unique identifier of an object. (You may also use its property name.)

from Eg.Cat as cat where cat.id = 123

from Eg.Cat as cat where cat.Mate.id = 69

The second query is efficient. No table join is required!

Properties of composite identifiers may also be used. Suppose Person
has a composite identifier consisting of Country and MedicareNumber.

from Bank.Person person
where person.id.Country = 'AU' 
    and person.id.MedicareNumber = 123456

from Bank.Account account
where account.Owner.id.Country = 'AU' 
    and account.Owner.id.MedicareNumber = 123456

Once again, the second query requires no table join.

Likewise, the special property class accesses the discriminator
value of an instance in the case of polymorphic persistence. A .Net class name embedded
in the where clause will be translated to its discriminator value.

from Eg.Cat cat where cat.class = Eg.DomesticCat

You may also specify properties of components or composite user types (and of components
of components, etc). Never try to use a path-expression that ends in a property
of component type (as opposed to a property of a component). For example, if store.Owner is an entity with a component Address

store.Owner.Address.City    // okay
store.Owner.Address         // error!

An “any” type has the special properties id and
, allowing us to express a join in the following way (where
is a property mapped with <any>).

from Eg.AuditLog log, Eg.Payment payment 
where log.Item.class = 'Eg.Payment, Eg, Version=...' and log.Item.id = payment.id

Notice that log.Item.class and payment.class
would refer to the values of completely different database columns in the above

14.8. Expressions

Expressions allowed in the where clause include most of
the kind of things you could write in SQL:

  • mathematical operators +, -, *, /

  • binary comparison operators =, >=, <=, <>, !=, like

  • logical operations and, or, not

  • string concatenation ||

  • SQL scalar functions like upper() and lower()

  • Parentheses ( ) indicate grouping

  • in, between, is

  • positional parameters ?

  • named parameters :name, :start_date,

  • SQL literals 'foo', 69,
    '1970-01-01 10:00:01.0'

  • Enumeration values and constants Eg.Color.Tabby

in and between may be used as

from Eg.DomesticCat cat where cat.Name between 'A' and 'B'

from Eg.DomesticCat cat where cat.Name in ( 'Foo', 'Bar', 'Baz' )

and the negated forms may be written

from Eg.DomesticCat cat where cat.Name not between 'A' and 'B'

from Eg.DomesticCat cat where cat.Name not in ( 'Foo', 'Bar', 'Baz' )

Likewise, is null and is not null
may be used to test for null values.

Booleans may be easily used in expressions by declaring HQL query substitutions
in NHibernate configuration:

<property name="hibernate.query.substitutions">true 1, false 0</property>

This will replace the keywords true and
with the literals 1 and 0
in the translated SQL from this HQL:

from Eg.Cat cat where cat.Alive = true

You may test the size of a collection with the special property
, or the special size() function.

from Eg.Cat cat where cat.Kittens.size > 0

from Eg.Cat cat where size(cat.Kittens) > 0

For indexed collections, you may refer to the minimum and maximum indices using
minIndex and maxIndex. Similarly,
you may refer to the minimum and maximum elements of a collection of basic type
using minElement and maxElement.

from Calendar cal where cal.Holidays.maxElement > current date

There are also functional forms (which, unlike the constructs above, are not case

from Order order where maxindex(order.Items) > 100

from Order order where minelement(order.Items) > 10000

The SQL functions any, some, all, exists, in are supported
when passed the element or index set of a collection (elements
and indices functions) or the result of a subquery (see

select mother from Eg.Cat as mother, Eg.Cat as kit
where kit in elements(mother.Kittens)

select p from Eg.NameList list, Eg.Person p
where p.Name = some elements(list.Names)

from Eg.Cat cat where exists elements(cat.Kittens)

from Eg.Player p where 3 > all elements(p.Scores)

from Eg.Show show where 'fizard' in indices(show.Acts)

Note that these constructs – size, elements,
indices, minIndex,
, minElement, maxElement
– have certain usage restrictions:

  • in a where clause: only for databases with subselects

  • in a select clause: only elements
    and indices make sense

Elements of indexed collections (arrays, lists, maps) may be referred to by index
(in a where clause only):

from Order order where order.Items[0].id = 1234

select person from Person person, Calendar calendar
where calendar.Holidays['national day'] = person.BirthDay
    and person.Nationality.Calendar = calendar

select item from Item item, Order order
where order.Items[ order.DeliveredItemIndices[0] ] = item and order.id = 11

select item from Item item, Order order
where order.Items[ maxindex(order.items) ] = item and order.id = 11

The expression inside [] may even be an arithmetic expression.

select item from Item item, Order order
where order.Items[ size(order.Items) - 1 ] = item

HQL also provides the built-in index() function, for elements
of a one-to-many association or collection of values.

select item, index(item) from Order order 
    join order.Items item
where index(item) < 5

Scalar SQL functions supported by the underlying database may be used

from Eg.DomesticCat cat where upper(cat.Name) like 'FRI%'

If you are not yet convinced by all this, think how much longer and less readable
the following query would be in SQL:

select cust
from Product prod,
    Store store
    inner join store.Customers cust
where prod.Name = 'widget'
    and store.Location.Name in ( 'Melbourne', 'Sydney' )
    and prod = all elements(cust.CurrentOrder.LineItems)

Hint: something like

SELECT cust.name, cust.address, cust.phone, cust.id, cust.current_order
FROM customers cust,
    stores store,
    locations loc,
    store_customers sc,
    product prod
WHERE prod.name = 'widget'
    AND store.loc_id = loc.id
    AND loc.name IN ( 'Melbourne', 'Sydney' )
    AND sc.store_id = store.id
    AND sc.cust_id = cust.id
    AND prod.id = ALL(
        SELECT item.prod_id
        FROM line_items item, orders o
        WHERE item.order_id = o.id
            AND cust.current_order = o.id

14.9. The order by clause

The list returned by a query may be ordered by any property of a returned class
or components:

from Eg.DomesticCat cat
order by cat.Name asc, cat.Weight desc, cat.Birthdate

The optional asc or desc indicate
ascending or descending order respectively.

14.10. The group by clause

A query that returns aggregate values may be grouped by any property of a returned
class or components:

select cat.Color, sum(cat.Weight), count(cat) 
from Eg.Cat cat
group by cat.Color

select foo.id, avg( elements(foo.Names) ), max( indices(foo.Names) ) 
from Eg.Foo foo
group by foo.id

Note: You may use the elements and indices
constructs inside a select clause, even on databases with no subselects.

A having clause is also allowed.

select cat.color, sum(cat.Weight), count(cat) 
from Eg.Cat cat
group by cat.Color 
having cat.Color in (Eg.Color.Tabby, Eg.Color.Black)

SQL functions and aggregate functions are allowed in the having
and order by clauses, if supported by the underlying database
(ie. not in MySQL).

select cat
from Eg.Cat cat
    join cat.Kittens kitten
group by cat.Id, cat.Name, cat.Other, cat.Properties
having avg(kitten.Weight) > 100
order by count(kitten) asc, sum(kitten.Weight) desc

Note that neither the group by clause nor the
order by
clause may contain arithmetic expressions. Also note that NHibernate
currently does not expand a grouped entity, so you can’t write group
by cat
if all properties of cat are non-aggregated.
You have to list all non-aggregated properties explicitly.

14.11. Subqueries

For databases that support subselects, NHibernate supports subqueries within queries.
A subquery must be surrounded by parentheses (often by an SQL aggregate function
call). Even correlated subqueries (subqueries that refer to an alias in the outer
query) are allowed.

from Eg.Cat as fatcat 
where fatcat.Weight > ( 
    select avg(cat.Weight) from Eg.DomesticCat cat 

from Eg.DomesticCat as cat 
where cat.Name = some ( 
    select name.NickName from Eg.Name as name 
from Eg.Cat as cat 
where not exists ( 
    from eg.Cat as mate where mate.Mate = cat 

from Eg.DomesticCat as cat 
where cat.Name not in ( 
    select name.NickName from Eg.Name as name 

14.12. HQL examples

NHibernate queries can be quite powerful and complex. In fact, the power of the
query language is one of NHibernate’s main selling points. Here are some example
queries very similar to queries that I used on a recent project. Note that most
queries you will write are much simpler than these!

The following query returns the order id, number of items and total value of the
order for all unpaid orders for a particular customer and given minimum total value,
ordering the results by total value. In determining the prices, it uses the current
catalog. The resulting SQL query, against the ORDER, ORDER_LINE, PRODUCT,
and PRICE tables has four inner joins
and an (uncorrelated) subselect.

select order.id, sum(price.Amount), count(item)
from Order as order
    join order.LineItems as item
    join item.Product as product,
    Catalog as catalog
    join catalog.Prices as price
where order.Paid = false
    and order.Customer = :customer
    and price.Product = product
    and catalog.EffectiveDate < sysdate
    and catalog.EffectiveDate >= all (
        select cat.EffectiveDate 
        from Catalog as cat
        where cat.EffectiveDate < sysdate
group by order
having sum(price.Amount) > :minAmount
order by sum(price.Amount) desc

What a monster! Actually, in real life, I’m not very keen on subqueries, so my query
was really more like this:

select order.id, sum(price.amount), count(item)
from Order as order
    join order.LineItems as item
    join item.Product as product,
    Catalog as catalog
    join catalog.Prices as price
where order.Paid = false
    and order.Customer = :customer
    and price.Product = product
    and catalog = :currentCatalog
group by order
having sum(price.Amount) > :minAmount
order by sum(price.Amount) desc

The next query counts the number of payments in each status, excluding all payments
in the AwaitingApproval status where the most recent status
change was made by the current user. It translates to an SQL query with two inner
joins and a correlated subselect against the PAYMENT, PAYMENT_STATUS and PAYMENT_STATUS_CHANGE

select count(payment), status.Name 
from Payment as payment 
    join payment.CurrentStatus as status
    join payment.StatusChanges as statusChange
where payment.Status.Name <> PaymentStatus.AwaitingApproval
    or (
        statusChange.TimeStamp = ( 
            select max(change.TimeStamp) 
            from PaymentStatusChange change 
            where change.Payment = payment
        and statusChange.User <> :currentUser
group by status.Name, status.SortOrder
order by status.SortOrder

If I would have mapped the StatusChanges collection as
a list, instead of a set, the query would have been much simpler to write.

select count(payment), status.Name 
from Payment as payment
    join payment.CurrentStatus as status
where payment.Status.Name <> PaymentStatus.AwaitingApproval
    or payment.StatusChanges[ maxIndex(payment.StatusChanges) ].User <> :currentUser
group by status.Name, status.SortOrder
order by status.SortOrder

The next query uses the MS SQL Server isNull() function
to return all the accounts and unpaid payments for the organization to which the
current user belongs. It translates to an SQL query with three inner joins, an outer
join and a subselect against the ACCOUNT,

select account, payment
from Account as account
    left outer join account.Payments as payment
where :currentUser in elements(account.Holder.Users)
    and PaymentStatus.Unpaid = isNull(payment.CurrentStatus.Name, PaymentStatus.Unpaid)
order by account.Type.SortOrder, account.AccountNumber, payment.DueDate

For some databases, we would need to do away with the (correlated) subselect.

select account, payment
from Account as account
    join account.Holder.Users as user
    left outer join account.Payments as payment
where :currentUser = user
    and PaymentStatus.Unpaid = isNull(payment.CurrentStatus.Name, PaymentStatus.Unpaid)
order by account.Type.SortOrder, account.AccountNumber, payment.DueDate

14.13. Tips & Tricks

You can count the number of query results without actually returning them:

int count = (int) session.CreateQuery("select count(*) from ....").UniqueResult();

To order a result by the size of a collection, use the following query:

select usr.id, usr.Name
from User as usr 
    left join usr.Messages as msg
group by usr.id, usr.Name
order by count(msg)

If your database supports subselects, you can place a condition upon selection size
in the where clause of your query:

from User usr where size(usr.Messages) >= 1

If your database doesn’t support subselects, use the following query:

select usr.id, usr.Name
from User usr
    join usr.Messages msg
group by usr.id, usr.Name
having count(msg) >= 1

As this solution can’t return a User with zero messages
because of the inner join, the following form is also useful:

select usr.id, usr.Name
from User as usr
    left join usr.Messages as msg
group by usr.id, usr.Name
having count(msg) = 0

Properties of an object can be bound to named query parameters:

IQuery q = s.CreateQuery("from foo in class Foo where foo.Name=:Name and foo.Size=:Size");
q.SetProperties(fooBean); // fooBean has properties Name and Size
IList foos = q.List();

Collections are pageable by using the IQuery interface
with a filter:

IQuery q = s.CreateFilter( collection, "" ); // the trivial filter
q.setFirstResult(PageSize * pageNumber);
IList page = q.List();

Collection elements may be ordered or grouped using a query filter:

ICollection orderedCollection = s.Filter( collection, "order by this.Amount" );
ICollection counts = s.Filter( collection, "select this.Type, count(this) group by this.Type" );

Chapter 15. Criteria Queries

NHibernate features an intuitive, extensible criteria query API.

15.1. Creating an ICriteria

The interface NHibernate.ICriteria represents a query against
a particular persistent class. The ISession is a factory
for ICriteria instances.

ICriteria crit = sess.CreateCriteria(typeof(Cat));
List cats = crit.List();

15.2. Narrowing the result set

An individual query criterion is an instance of the interface NHibernate.Expression.ICriterion.
The class NHibernate.Expression.Expression defines factory
methods for obtaining certain built-in ICriterion types.

IList cats = sess.CreateCriteria(typeof(Cat))
    .Add( Expression.Like("Name", "Fritz%") )
    .Add( Expression.Between("Weight", minWeight, maxWeight) )

Expressions may be grouped logically.

IList cats = sess.CreateCriteria(typeof(Cat))
    .Add( Expression.Like("Name", "Fritz%") )
    .Add( Expression.Or(
        Expression.Eq( "Age", 0 ),
    ) )
IList cats = sess.CreateCriteria(typeof(Cat))
    .Add( Expression.In( "Name", new String[] { "Fritz", "Izi", "Pk" } ) )
    .Add( Expression.Disjunction()
        .Add( Expression.IsNull("Age") )
    	.Add( Expression.Eq("Age", 0 ) )
    	.Add( Expression.Eq("Age", 1 ) )
    	.Add( Expression.Eq("Age", 2 ) )
    ) )

There are quite a range of built-in criterion types (Expression
subclasses), but one that is especially useful lets you specify SQL directly.

        // Create a string parameter for the SqlString below
        IList cats = sess.CreateCriteria(typeof(Cat))
            .Add( Expression.Sql("lower({alias}.Name) like lower(?)", "Fritz%", NHibernateUtil.String )

The {alias} placeholder with be replaced by the row alias
of the queried entity.

15.3. Ordering the results

You may order the results using NHibernate.Expression.Order.

IList cats = sess.CreateCriteria(typeof(Cat))
    .Add( Expression.Like("Name", "F%")
    .AddOrder( Order.Asc("Name") )
    .AddOrder( Order.Desc("Age") )

15.4. Associations

You may easily specify constraints upon related entities by navigating associations
using CreateCriteria().

IList cats = sess.CreateCriteria(typeof(Cat))
    .Add( Expression.Like("Name", "F%")
        .Add( Expression.Like("Name", "F%") )

note that the second CreateCriteria() returns a new instance
of ICriteria, which refers to the elements of the

The following, alternate form is useful in certain circumstances.

IList cats = sess.CreateCriteria(typeof(Cat))
    .CreateAlias("Kittens", "kt")
    .CreateAlias("Mate", "mt")
    .Add( Expression.EqProperty("kt.Name", "mt.Name") )

(CreateAlias() does not create a new instance of

Note that the kittens collections held by the Cat instances
returned by the previous two queries are not
pre-filtered by the criteria! If you wish to retrieve just the kittens that match
the criteria, you must use SetResultTransformer(Transformers.AliasToEntityMap).

IList cats = sess.CreateCriteria(typeof(Cat))
    .CreateCriteria("Kittens", "kt")
        .Add( Expression.Eq("Name", "F%") )
foreach ( IDictionary map in cats )
    Cat cat = (Cat) map[CriteriaUtil.RootAlias];
    Cat kitten = (Cat) map["kt"];

15.5. Dynamic association fetching

You may specify association fetching semantics at runtime using

IList cats = sess.CreateCriteria(typeof(Cat))
    .Add( Expression.Like("Name", "Fritz%") )
    .SetFetchMode("Mate", FetchMode.Eager)
    .SetFetchMode("Kittens", FetchMode.Eager)

This query will fetch both Mate and Kittens
by outer join. See Section 19.1, “Fetching strategies”
for more information.

15.6. Example queries

The class NHibernate.Expression.Example allows you to construct
a query criterion from a given instance.

Cat cat = new Cat();
cat.Sex = 'F';
cat.Color = Color.Black;
List results = session.CreateCriteria(typeof(Cat))
    .Add( Example.Create(cat) )

Version properties, identifiers and associations are ignored. By default, null-valued
properties and properties which return an empty string from the call to
are excluded.

You can adjust how the Example is applied.

Example example = Example.Create(cat)
    .ExcludeZeroes()           //exclude null- or zero-valued properties
    .ExcludeProperty("Color")  //exclude the property named "color"
    .IgnoreCase()              //perform case insensitive string comparisons
    .EnableLike();             //use like for string comparisons
IList results = session.CreateCriteria(typeof(Cat))

You can even use examples to place criteria upon associated objects.

IList results = session.CreateCriteria(typeof(Cat))
    .Add( Example.Create(cat) )
        .Add( Example.Create( cat.Mate ) )

15.7. Projections, aggregation and grouping

The class NHibernate.Expression.Projections is a factory
for IProjection instances. We apply a projection to a query
by calling SetProjection().

IList results = session.CreateCriteria(typeof(Cat))
    .SetProjection( Projections.RowCount() )
    .Add( Expression.Eq("Color", Color.BLACK) )
List results = session.CreateCriteria(typeof(Cat))
    .SetProjection( Projections.ProjectionList()
        .Add( Projections.RowCount() )
        .Add( Projections.Avg("Weight") )
        .Add( Projections.Max("Weight") )
        .Add( Projections.GroupProperty("Color") )

There is no explicit “group by” necessary in a criteria query. Certain projection
types are defined to be grouping projections,
which also appear in the SQL group by clause.

An alias may optionally be assigned to a projection, so that the projected value
may be referred to in restrictions or orderings. Here are two different ways to
do this:

IList results = session.CreateCriteria(typeof(Cat))
    .SetProjection( Projections.Alias( Projections.GroupProperty("Color"), "colr" ) )
    .AddOrder( Order.Asc("colr") )
IList results = session.CreateCriteria(typeof(Cat))
    .SetProjection( Projections.GroupProperty("Color").As("colr") )
    .AddOrder( Order.Asc("colr") )

The Alias() and As() methods simply
wrap a projection instance in another, aliased, instance of IProjection.
As a shortcut, you can assign an alias when you add the projection to a projection

IList results = session.CreateCriteria(typeof(Cat))
    .SetProjection( Projections.ProjectionList()
        .Add( Projections.RowCount(), "catCountByColor" )
        .Add( Projections.Avg("Weight"), "avgWeight" )
        .Add( Projections.Max("Weight"), "maxWeight" )
        .Add( Projections.GroupProperty("Color"), "color" )
    .AddOrder( Order.Desc("catCountByColor") )
    .AddOrder( Order.Desc("avgWeight") )
IList results = session.CreateCriteria(typeof(DomesticCat), "cat")
    .CreateAlias("kittens", "kit")
    .SetProjection( Projections.ProjectionList()
        .Add( Projections.Property("cat.Name"), "catName" )
        .Add( Projections.Property("kit.Name"), "kitName" )
    .AddOrder( Order.Asc("catName") )
    .AddOrder( Order.Asc("kitName") )

15.8. Detached queries and subqueries

The DetachedCriteria class lets you create a query outside
the scope of a session, and then later execute it using some arbitrary

DetachedCriteria query = DetachedCriteria.For(typeof(Cat))
    .Add( Expression.Eq("sex", 'F') );
ISession session = ....;
ITransaction txn = session.BeginTransaction();
IList results = query.GetExecutableCriteria(session).SetMaxResults(100).List();

A DetachedCriteria may also be used to express a subquery.
ICriterion instances involving subqueries may be obtained via Subqueries

DetachedCriteria avgWeight = DetachedCriteria.For(typeof(Cat))
    .SetProjection( Projections.Avg("Weight") );
    .Add( Subqueries.Gt("Weight", avgWeight) )
DetachedCriteria weights = DetachedCriteria.For(typeof(Cat))
    .SetProjection( Projections.Property("Weight") );
    .add( Subqueries.GeAll("Weight", weights) )

Even correlated subqueries are possible:

DetachedCriteria avgWeightForSex = DetachedCriteria.For(typeof(Cat), "cat2")
    .SetProjection( Projections.Avg("Weight") )
    .Add( Expression.EqProperty("cat2.Sex", "cat.Sex") );
session.CreateCriteria(typeof(Cat), "cat")
    .Add( Subqueries.Gt("weight", avgWeightForSex) )

Chapter 16. QueryOver Queries

The ICriteria API is NHibernate’s implementation of Query Object. NHibernate 3.0
introduces the QueryOver api, which combines the use of Extension Methods and Lambda
Expressions (both new in .Net 3.5) to provide a statically typesafe wrapper round
the ICriteria API.

QueryOver uses Lambda Expressions to provide some extra syntax to remove the ‘magic
strings’ from your ICriteria queries.

So, for example:

.Add(Expression.Eq("Name", "Smith"))


.Where<Person>(p => p.Name == "Smith")

With this kind of syntax there are no ‘magic strings’, and refactoring tools like
‘Find All References’, and ‘Refactor->Rename’ work perfectly.

Note: QueryOver is intended to remove the references to ‘magic strings’ from the
ICriteria API while maintaining it’s opaqueness. It is not
a LINQ provider; NHibernate has a built-in Linq provider for this.

16.1. Structure of a Query

Queries are created from an ISession using the syntax:

IList<Cat> cats =
        .Where(c => c.Name == "Max")


Detached QueryOver (analagous to DetachedCriteria) can be created, and then used
with an ISession using:

QueryOver<Cat> query =
        .Where(c => c.Name == "Paddy");
IList<Cat> cats =

Queries can be built up to use restrictions, projections, and ordering using a fluent
inline syntax:

var catNames =
        .WhereRestrictionOn(c => c.Age).IsBetween(2).And(8)
        .Select(c => c.Name)
        .OrderBy(c => c.Name).Asc

16.2. Simple Expressions

The Restrictions class (used by ICriteria) has been extended to include overloads
that allow Lambda Expression syntax. The Where() method works for simple expressions
(<, <=, ==, !=, >, >=) so instead of:

ICriterion equalCriterion = Restrictions.Eq("Name", "Max")

You can write:

ICriterion equalCriterion = Restrictions.Where<Cat>(c => c.Name == "Max")


Since the QueryOver class (and IQueryOver interface) is generic and knows the type
of the query, there is an inline syntax for restrictions that does not require the
additional qualification of class name. So you can also write:

var cats =
        .Where(c => c.Name == "Max")
        .And(c => c.Age > 4)

Note, the methods Where() and And() are semantically identical; the And() method
is purely to allow QueryOver to look similar to HQL/SQL.


Boolean comparisons can be made directly instead of comparing to true/false:

        .Where(p => p.IsParent)
        .And(p => !p.IsRetired)


Simple expressions can also be combined using the || and && operators. So
ICriteria like:

                Restrictions.Eq("Name", "test name"),
                    Restrictions.Gt("Age", 21),
                    Restrictions.Eq("HasCar", true))))

Can be written in QueryOver as:

        .Where(p => p.Name == "test name" && (p.Age > 21 || p.HasCar))


Each of the corresponding overloads in the QueryOver API allows the use of regular
ICriterion to allow access to private properties.

        .Where(Restrictions.Eq("Name", "Max"))


It is worth noting that the QueryOver API is built on top of the ICriteria API.
Internally the structures are the same, so at runtime the statement below, and the
statement above, are stored as exactly the same ICriterion. The actual Lambda Expression
is not stored in the query.

        .Where(c => c.Name == "Max")

16.3. Additional Restrictions

Some SQL operators/functions do not have a direct equivalent in C#. (e.g., the SQL
where name like '%anna%'). These operators have overloads
for QueryOver in the Restrictions class, so you can write:

        .Where(Restrictions.On<Cat>(c => c.Name).IsLike("%anna%"))

There is also an inline syntax to avoid the qualification of the type:

        .WhereRestrictionOn(c => c.Name).IsLike("%anna%")


While simple expressions (see above) can be combined using the || and &&
operators, this is not possible with the other restrictions. So this ICriteria:

            Restrictions.Gt("Age", 5)
            Restrictions.In("Name", new string[] { "Max", "Paddy" })))

Would have to be written as:

            Restrictions.Where<Cat>(c => c.Age > 5)
            Restrictions.On<Cat>(c => c.Name).IsIn(new string[] { "Max", "Paddy" })))

However, in addition to the additional restrictions factory methods, there are extension
methods to allow a more concise inline syntax for some of the operators. So this:

        .WhereRestrictionOn(c => c.Name).IsLike("%anna%")

May also be written as:

        .Where(c => c..Name.IsLike("%anna%"))

16.4. Associations

QueryOver can navigate association paths using JoinQueryOver() (analagous to ICriteria.CreateCriteria()
to create sub-criteria).

The factory method QuerOver<T>() on ISession returns an IQueryOver<T>.
More accurately, it returns an IQueryOver<T,T> (which inherits from IQueryOver<T>).

An IQueryOver has two types of interest; the root type (the type of entity that
the query returns), and the type of the ‘current’ entity being queried. For example,
the following query uses a join to create a sub-QueryOver (analagous to creating
sub-criteria in the ICriteria API):

IQueryOver<Cat,Kitten> catQuery =
        .JoinQueryOver(c => c.Kittens)
            .Where(k => k.Name == "Tiddles");

The JoinQueryOver returns a new instance of the IQueryOver than has its root at
the Kittens collection. The default type for restrictions is now Kitten (restricting
on the name ‘Tiddles’ in the above example), while calling .List() will return an
IList<Cat>. The type IQueryOver<Cat,Kitten> inherits from IQueryOver<Cat>.

Note, the overload for JoinQueryOver takes an IEnumerable<T>, and the C# compiler
infers the type from that. If your collection type is not IEnumerable<T>,
then you need to qualify the type of the sub-criteria:

IQueryOver<Cat,Kitten> catQuery =
Kitten>(c => c.Kittens)
            .Where(k => k.Name == "Tiddles");


The default join is an inner-join. Each of the additional join types can be specified
using the methods .Inner, .Left, .Right, or .Full.
For example, to left outer-join on Kittens use:

IQueryOver<Cat,Kitten> catQuery =
        .Left.JoinQueryOver(c => c.Kittens)
            .Where(k => k.Name == "Tiddles");

16.5. Aliases

In the traditional ICriteria interface aliases are assigned using ‘magic strings’,
however their value does not correspond to a name in the object domain. For example,
when an alias is assigned using .CreateAlias("Kitten", "kittenAlias"),
the string “kittenAlias” does not correspond to a property or class in the domain.

In QueryOver, aliases are assigned using an empty variable. The variable can be
declared anywhere (but should be null at runtime). The compiler
can then check the syntax against the variable is used correctly, but at runtime
the variable is not evaluated (it’s just used as a placeholder for the alias).

Each Lambda Expression function in QueryOver has a corresponding overload to allow
use of aliases, and a .JoinAlias function to traverse associations using aliases
without creating a sub-QueryOver.

Cat catAlias = null;
Kitten kittenAlias = null;

IQueryOver<Cat,Cat> catQuery =
    session.QueryOver<Cat>(() => catAlias)
        .JoinAlias(() => catAlias.Kittens, () => kittenAlias)
        .Where(() => catAlias.Age > 5)
        .And(() => kittenAlias.Name == "Tiddles");

16.6. Projections

Simple projections of the properties of the root type can be added using the .Select method which can take multiple Lambda Expression arguments:

IList selection =
            c => c.Name,
            c => c.Age)

Because this query no longer returns a Cat, the return type must be explicitly specified.
If a single property is projected, the return type can be specified using:

IList<int> ages =
        .Select(c => c.Age)

However, if multiple properties are projected, then the returned list will contain
object arrays, as per a projection in ICriteria. This could be fed into an anonymous
type using:

var catDetails =
            c => c.Name,
            c => c.Age)
        .Select(properties => new {
            CatName = (string)properties[0],
            CatAge = (int)properties[1],

Note that the second .Select call in this example is an extension
method on IEnumerable<T> supplied in System.Linq; it is not part of NHibernate.


QueryOver allows arbitrary IProjection to be added (allowing private properties
to be projected). The Projections factory class also has overloads to allow Lambda
Expressions to be used:

IList selection =
            .Add(Projections.Property<Cat>(c => c.Name))
            .Add(Projections.Avg<Cat>(c => c.Age)))


In addition there is an inline syntax for creating projection lists that does not
require the explicit class qualification:

IList selection =
        .SelectList(list => list
            .Select(c => c.Name)
            .SelectAvg(c => c.Age))


Projections can also have arbitrary aliases assigned to them to allow result transformation.
If there is a CatSummary DTO class defined as:

public class CatSummary
    public string Name { get; set; }
    public int AverageAge { get; set; }

… then aliased projections can be used with the AliasToBean<T> transformer:

CatSummary summaryDto = null;
IList<CatSummary> catReport =
        .SelectList(list => list
            .SelectGroup(c => c.Name).WithAlias(() => summaryDto.Name)
            .SelectAvg(c => c.Age).WithAlias(() => summaryDto.AverageAge))

16.7. Projection Functions

In addition to projecting properties, there are extension methods to allow certain
common dialect-registered functions to be applied. For example you can write the
following to extract just the year part of a date:

        .Where(p => p.BirthDate.YearPart() == 1971)

The functions can also be used inside projections:

            p => Projections.Concat(p.LastName, ", ", p.FirstName),
            p => p.Height.Abs())

16.8. Subqueries

The Subqueries factory class has overloads to allow Lambda Expressions to express
sub-query restrictions. For example:

QueryOver<Cat> maximumAge =
        .SelectList(p => p.SelectMax(c => c.Age));

IList<Cat> oldestCats =
        .Where(Subqueries.WhereProperty<Cat>(c => c.Age).Eq(maximumAge))


The inline syntax allows you to use subqueries without requalifying the type:

IList<Cat> oldestCats =
        .WithSubquery.WhereProperty(c => c.Age).Eq(maximumAge)


There is an extension method As() on (a detached) QueryOver
that allows you to cast it to any type. This is used in conjunction with the overloads
Where(), WhereAll(), and WhereSome()
to allow use of the built-in C# operators for comparison, so the above query can
be written as:

IList<Cat> oldestCats =
        .WithSubquery.Where(c => c.Age == maximumAge.As<int>())

Chapter 17. Native SQL

You may also express queries in the native SQL dialect of your database. This is
useful if you want to utilize database specific features such as query hints or
the CONNECT keyword in Oracle. It also provides a clean
migration path from a direct SQL/ADO.NET based application to NHibernate.

NHibernate allows you to specify handwritten SQL (including stored procedures) for
all create, update, delete, and load operations.

17.1. Using an ISQLQuery

Execution of native SQL queries is controlled via the ISQLQuery
interface, which is obtained by calling ISession.CreateSQLQuery().
The following describes how to use this API for querying.

17.1.1. Scalar queries

The most basic SQL query is to get a list of scalars (values).

sess.CreateSQLQuery("SELECT * FROM CATS")
 .AddScalar("ID", NHibernateUtil.Int64)
 .AddScalar("NAME", NHibernateUtil.String)
 .AddScalar("BIRTHDATE", NHibernateUtil.Date)

This query specified:

  • the SQL query string

  • the columns and types to return

This will return an IList of Object arrays (object[]) with scalar values for each
column in the CATS table. Only these three columns will be returned, even though
the query is using * and could return more than the three
listed columns.

17.1.2. Entity queries

The above query was about returning scalar values, basically returning the “raw”
values from the result set. The following shows how to get entity objects from a
native SQL query via AddEntity().

sess.CreateSQLQuery("SELECT * FROM CATS").AddEntity(typeof(Cat));
sess.CreateSQLQuery("SELECT ID, NAME, BIRTHDATE FROM CATS").AddEntity(typeof(Cat));

This query specified:

  • the SQL query string

  • the entity returned by the query

Assuming that Cat is mapped as a class with the columns ID, NAME and BIRTHDATE the
above queries will both return an IList where each element is a Cat entity.

If the entity is mapped with a many-to-one to another entity
it is required to also return its identifier when performing the native query, otherwise
a database specific “column not found” error will occur. The additional columns
will automatically be returned when using the * notation, but we prefer to be explicit
as in the following example for a many-to-one to a

sess.CreateSQLQuery("SELECT ID, NAME, BIRTHDATE, DOG_ID FROM CATS").AddEntity(typeof(Cat));

This will allow cat.Dog property access to function properly.

17.1.3. Handling associations and collections

It is possible to eagerly join in the Dog to avoid the
possible extra roundtrip for initializing the proxy. This is done via the
method, which allows you to join in an association or collection.

 .AddEntity("cat", typeof(Cat))

In this example the returned Cat‘s will have their
property fully initialized without any extra roundtrip to the database.
Notice that we added a alias name (“cat”) to be able to specify the target property
path of the join. It is possible to do the same eager joining for collections, e.g.
if the Cat had a one-to-many to Dog

 .AddEntity("cat", typeof(Cat))

At this stage we are reaching the limits of what is possible with native queries
without starting to enhance the SQL queries to make them usable in NHibernate; the
problems start to arise when returning multiple entities of the same type or when
the default alias/column names are not enough.

17.1.4. Returning multiple entities

Until now the result set column names are assumed to be the same as the column names
specified in the mapping document. This can be problematic for SQL queries which
join multiple tables, since the same column names may appear in more than one table.

Column alias injection is needed in the following query (which most likely will

sess.CreateSQLQuery("SELECT c.*, m.*  FROM CATS c, CATS m WHERE c.MOTHER_ID = c.ID")
 .AddEntity("cat", typeof(Cat))
 .AddEntity("mother", typeof(Cat))

The intention for this query is to return two Cat instances per row, a cat and its
mother. This will fail since there is a conflict of names since they are mapped
to the same column names and on some databases the returned column aliases will
most likely be on the form “c.ID”, “c.NAME”, etc. which are not equal to the columns
specificed in the mappings (“ID” and “NAME”).

The following form is not vulnerable to column name duplication:

sess.CreateSQLQuery("SELECT {cat.*}, {mother.*}  FROM CATS c, CATS m WHERE c.MOTHER_ID = c.ID")
 .AddEntity("cat", typeof(Cat))
 .AddEntity("mother", typeof(Cat))

This query specified:

  • the SQL query string, with placeholders for NHibernate to inject column aliases

  • the entities returned by the query

The {cat.*} and {mother.*} notation used above is a shorthand for “all properties”.
Alternatively, you may list the columns explicity, but even in this case we let
NHibernate inject the SQL column aliases for each property. The placeholder for
a column alias is just the property name qualified by the table alias. In the following
example, we retrieve Cats and their mothers from a different table (cat_log) to
the one declared in the mapping metadata. Notice that we may even use the property
aliases in the where clause if we like.

String sql = "SELECT ID as {c.Id}, NAME as {c.Name}, " + 
         "BIRTHDATE as {c.BirthDate}, MOTHER_ID as {c.Mother}, {mother.*} " +
         "FROM CAT_LOG c, CAT_LOG m WHERE {c.Mother} = c.ID";

IList loggedCats = sess.CreateSQLQuery(sql)
        .AddEntity("cat", typeof(Cat))
        .AddEntity("mother", typeof(Cat)).List(); Alias and property references

For most cases the above alias injection is needed, but for queries relating to
more complex mappings like composite properties, inheritance discriminators, collections
etc. there are some specific aliases to use to allow NHibernate to inject the proper

The following table shows the different possibilities of using the alias injection.
Note: the alias names in the result are examples, each alias will have a unique
and probably different name when used.

Table 17.1. Alias injection names

Description Syntax Example
A simple property {[aliasname].[propertyname]} A_NAME as {item.Name}
A composite property {[aliasname].[componentname].[propertyname]} CURRENCY as {item.Amount.Currency}, VALUE as {item.Amount.Value}
Discriminator of an entity {[aliasname].class} DISC as {item.class}
All properties of an entity {[aliasname].*} {item.*}
A collection key {[aliasname].key} ORGID as {coll.key}
The id of an collection {[aliasname].id} EMPID as {coll.id}
The element of an collection {[aliasname].element} XID as {coll.element}
property of the element in the collection {[aliasname].element.[propertyname]} NAME as {coll.element.Name}
All properties of the element in the collection {[aliasname].element.*} {coll.element.*}
All properties of the the collection {[aliasname].*} {coll.*}

17.1.5. Returning non-managed entities

It is possible to apply an IResultTransformer to native
sql queries. Allowing it to e.g. return non-managed entities.


This query specified:

  • the SQL query string

  • a result transformer

The above query will return a list of CatDTO which has
been instantiated and injected the values of NAME and BIRTHNAME into its corresponding
properties or fields.

IMPORTANT: The custom IResultTransformer should override
Equals and GetHashCode, otherwise
the query translation won’t be cached. This also will result in memory leak.

17.1.6. Handling inheritance

Native SQL queries which query for entities that are mapped as part of an inheritance
hierarchy must include all properties for the base class and all its subclasses.

17.1.7. Parameters

Native SQL queries support positional as well as named parameters:

Query query = sess.CreateSQLQuery("SELECT * FROM CATS WHERE NAME like ?").AddEntity(typeof(Cat));
IList pusList = query.SetString(0, "Pus%").List();
query = sess.createSQLQuery("SELECT * FROM CATS WHERE NAME like :name").AddEntity(typeof(Cat));
IList pusList = query.SetString("name", "Pus%").List();          

17.2. Named SQL queries

Named SQL queries may be defined in the mapping document and called in exactly the
same way as a named HQL query. In this case, we do not
need to call AddEntity().

<sql-query name="persons">
    <return alias="person" class="eg.Person"/>
    SELECT person.NAME AS {person.Name},
           person.AGE AS {person.Age},
           person.SEX AS {person.Sex}
    FROM PERSON person
    WHERE person.NAME LIKE :namePattern
IList people = sess.GetNamedQuery("persons")
    .SetString("namePattern", namePattern)

The <return-join> and <load-collection>
elements are used to join associations and define queries which initialize collections,

<sql-query name="personsWith">
    <return alias="person" class="eg.Person"/>
    <return-join alias="address" property="person.MailingAddress"/>
    SELECT person.NAME AS {person.Name},
           person.AGE AS {person.Age},
           person.SEX AS {person.Sex},
           adddress.STREET AS {address.Street},
           adddress.CITY AS {address.City},
           adddress.STATE AS {address.State},
           adddress.ZIP AS {address.Zip}
    FROM PERSON person
    JOIN ADDRESS adddress
        ON person.ID = address.PERSON_ID AND address.TYPE='MAILING'
    WHERE person.NAME LIKE :namePattern

A named SQL query may return a scalar value. You must declare the column alias and
NHibernate type using the <return-scalar> element:

<sql-query name="mySqlQuery">
    <return-scalar column="name" type="String"/>
    <return-scalar column="age" type="Int64"/>
    SELECT p.NAME AS name,
           p.AGE AS age,

You can externalize the resultset mapping informations in a <resultset>
element to either reuse them accross several named queries or through the

<resultset name="personAddress">
    <return alias="person" class="eg.Person"/>
    <return-join alias="address" property="person.MailingAddress"/>

<sql-query name="personsWith" resultset-ref="personAddress">
    SELECT person.NAME AS {person.Name},
           person.AGE AS {person.Age},
           person.SEX AS {person.Sex},
           adddress.STREET AS {address.Street},
           adddress.CITY AS {address.City},
           adddress.STATE AS {address.State},
           adddress.ZIP AS {address.Zip}
    FROM PERSON person
    JOIN ADDRESS adddress
        ON person.ID = address.PERSON_ID AND address.TYPE='MAILING'
    WHERE person.NAME LIKE :namePattern

You can alternatively use the resultset mapping information in your .hbm.xml files
directly in code.

IList cats = sess.CreateSQLQuery(
        "select {cat.*}, {kitten.*} from cats cat, cats kitten where kitten.mother = cat.id"

17.2.1. Using return-property to explicitly specify
column/alias names

With <return-property> you can explicitly tell NHibernate
what column aliases to use, instead of using the {}-syntax
to let NHibernate inject its own aliases.

<sql-query name="mySqlQuery">
    <return alias="person" class="eg.Person">
        <return-property name="Name" column="myName"/>
        <return-property name="Age" column="myAge"/>
        <return-property name="Sex" column="mySex"/>
    SELECT person.NAME AS myName,
           person.AGE AS myAge,
           person.SEX AS mySex,
    FROM PERSON person WHERE person.NAME LIKE :name

<return-property> also works with multiple columns.
This solves a limitation with the {}-syntax which can not
allow fine grained control of multi-column properties.

<sql-query name="organizationCurrentEmployments">
    <return alias="emp" class="Employment">
        <return-property name="Salary">
            <return-column name="VALUE"/>
            <return-column name="CURRENCY"/>
        <return-property name="EndDate" column="myEndDate"/>
        SELECT EMPLOYEE AS {emp.Employee}, EMPLOYER AS {emp.Employer},
        STARTDATE AS {emp.StartDate}, ENDDATE AS {emp.EndDate},
        REGIONCODE as {emp.RegionCode}, EID AS {emp.Id}, VALUE, CURRENCY

Notice that in this example we used <return-property>
in combination with the {}-syntax for injection, allowing
users to choose how they want to refer column and properties.

If your mapping has a discriminator you must use <return-discriminator>
to specify the discriminator column.

17.2.2. Using stored procedures for querying

NHibernate introduces support for queries via stored procedures and functions. Most
of the following documentation is equivalent for both. The stored procedure/function
must return a resultset to be able to work with NHibernate. An example of such a
stored function in MS SQL Server 2000 and higher is as follows:

CREATE PROCEDURE selectAllEmployments AS

To use this query in NHibernate you need to map it via a named query.

<sql-query name="selectAllEmployments_SP">
    <return alias="emp" class="Employment">
        <return-property name="employee" column="EMPLOYEE"/>
        <return-property name="employer" column="EMPLOYER"/>
        <return-property name="startDate" column="STARTDATE"/>
        <return-property name="endDate" column="ENDDATE"/>
        <return-property name="regionCode" column="REGIONCODE"/>
        <return-property name="id" column="EID"/>
        <return-property name="salary">
            <return-column name="VALUE"/>
            <return-column name="CURRENCY"/>
    exec selectAllEmployments

Notice that stored procedures currently only return scalars and entities.
and <load-collection>
are not supported. Rules/limitations
for using stored procedures

To use stored procedures with NHibernate the procedures/functions have to follow
some rules. If they do not follow those rules they are not usable with NHibernate.
If you still want to use these procedures you have to execute them via
. The rules are different for each database, since database
vendors have different stored procedure semantics/syntax.

Stored procedure queries can’t be paged with SetFirstResult()/SetMaxResults().

Recommended call form is dependent on your database. For MS SQL Server use
exec functionName <parameters>

For Oracle the following rules apply:

  • A function must return a result set. The first parameter of a procedure must be
    an OUT that returns a result set. This is done by using
    a SYS_REFCURSOR type in Oracle 9 or 10. In Oracle you need
    to define a REF CURSOR type, see Oracle literature.

For MS SQL server the following rules apply:

  • The procedure must return a result set. NHibernate will use IDbCommand.ExecuteReader()
    to obtain the results.

  • If you can enable SET NOCOUNT ON in your procedure it will
    probably be more efficient, but this is not a requirement.

17.3. Custom SQL for create, update and delete

NHibernate can use custom SQL statements for create, update, and delete operations.
The class and collection persisters in NHibernate already contain a set of configuration
time generated strings (insertsql, deletesql, updatesql etc.). The mapping tags
<sql-insert>, <sql-delete>,
and <sql-update> override these strings:

<class name="Person">
    <id name="id">
        <generator class="increment"/>
    <property name="name" not-null="true"/>
    <sql-insert>INSERT INTO PERSON (NAME, ID) VALUES ( UPPER(?), ? )</sql-insert>
    <sql-update>UPDATE PERSON SET NAME=UPPER(?) WHERE ID=?</sql-update>
    <sql-delete>DELETE FROM PERSON WHERE ID=?</sql-delete>

Note that the custom sql-insert will not be used if you
use identity to generate identifier values for the class.

The SQL is directly executed in your database, so you are free to use any dialect
you like. This will of course reduce the portability of your mapping if you use
database specific SQL.

Stored procedures are supported if the database-native syntax is used:

<class name="Person">
    <id name="id">
        <generator class="increment"/>
    <property name="name" not-null="true"/>
    <sql-insert>exec createPerson ?, ?</sql-insert>
    <sql-delete>exec deletePerson ?</sql-delete>
    <sql-update>exec updatePerson ?, ?</sql-update>

The order of the positional parameters is currently vital, as they must be in the
same sequence as NHibernate expects them.

You can see the expected order by enabling debug logging for the
level. With this level enabled NHibernate will
print out the static SQL that is used to create, update, delete etc. entities. (To
see the expected sequence, remember to not include your custom SQL in the mapping
files as that will override the NHibernate generated static sql.)

The stored procedures are by default required to affect the same number of rows
as NHibernate-generated SQL would. NHibernate uses IDbCommand.ExecuteNonQuery
to retrieve the number of rows affected. This check can be disabled by using check="none" attribute in sql-insert

17.4. Custom SQL for loading

You may also declare your own SQL (or HQL) queries for entity loading:

<sql-query name="person">
    <return alias="pers" class="Person" lock-mode="upgrade"/>
    SELECT NAME AS {pers.Name}, ID AS {pers.Id}
    WHERE ID=?

This is just a named query declaration, as discussed earlier. You may reference
this named query in a class mapping:

<class name="Person">
    <id name="Id">
        <generator class="increment"/>
    <property name="Name" not-null="true"/>
    <loader query-ref="person"/>

This even works with stored procedures.

You may even define a query for collection loading:

<set name="Employments" inverse="true">
    <one-to-many class="Employment"/>
    <loader query-ref="employments"/>
<sql-query name="employments">
    <load-collection alias="emp" role="Person.Employments"/>
    SELECT {emp.*}

You could even define an entity loader that loads a collection by join fetching:

<sql-query name="person">
    <return alias="pers" class="Person"/>
    <return-join alias="emp" property="pers.Employments"/>
    SELECT NAME AS {pers.*}, {emp.*}
    FROM PERSON pers
        ON pers.ID = emp.PERSON_ID
    WHERE ID=?

Chapter 18. Filtering data

NHibernate provides an innovative new approach to handling data with “visibility”
rules. A NHibernate filter is a global, named,
parameterized filter that may be enabled or disabled for a particular NHibernate

18.1. NHibernate filters

NHibernate adds the ability to pre-define filter criteria and attach those filters
at both a class and a collection level. A filter criteria is the ability to define
a restriction clause very similiar to the existing “where” attribute available on
the class and various collection elements. Except these filter conditions can be
parameterized. The application can then make the decision at runtime whether given
filters should be enabled and what their parameter values should be. Filters can
be used like database views, but parameterized inside the application.

In order to use filters, they must first be defined and then attached to the appropriate
mapping elements. To define a filter, use the <filter-def/>
element within a <hibernate-mapping/> element:

<filter-def name="myFilter">
    <filter-param name="myFilterParam" type="String"/>

Then, this filter can be attached to a class:

<class name="MyClass" ...>
    <filter name="myFilter" condition=":myFilterParam = MY_FILTERED_COLUMN"/>

or, to a collection:

<set ...>
    <filter name="myFilter" condition=":myFilterParam = MY_FILTERED_COLUMN"/>

or, even to both (or multiples of each) at the same time.

The methods on ISession are: EnableFilter(string
, GetEnabledFilter(string filterName),
and DisableFilter(string filterName). By default, filters
are not enabled for a given session; they
must be explcitly enabled through use of the ISession.EnableFilter()
method, which returns an instance of the IFilter interface.
Using the simple filter defined above, this would look like:

session.EnableFilter("myFilter").SetParameter("myFilterParam", "some-value");

Note that methods on the NHibernate.IFilter interface do
allow the method-chaining common to much of NHibernate.

A full example, using temporal data with an effective record date pattern:

<filter-def name="effectiveDate">
    <filter-param name="asOfDate" type="date"/>

<class name="Employee" ...>
    <many-to-one name="Department" column="dept_id" class="Department"/>
    <property name="EffectiveStartDate" type="date" column="eff_start_dt"/>
    <property name="EffectiveEndDate" type="date" column="eff_end_dt"/>
        Note that this assumes non-terminal records have an eff_end_dt set to
        a max db date for simplicity-sake
    <filter name="effectiveDate"
            condition=":asOfDate BETWEEN eff_start_dt and eff_end_dt"/>

<class name="Department" ...>
    <set name="Employees" lazy="true">
        <key column="dept_id"/>
        <one-to-many class="Employee"/>
        <filter name="effectiveDate"
                condition=":asOfDate BETWEEN eff_start_dt and eff_end_dt"/>

Then, in order to ensure that you always get back currently effective records, simply
enable the filter on the session prior to retrieving employee data:

ISession session = ...;
session.EnableFilter("effectiveDate").SetParameter("asOfDate", DateTime.Today);
IList results = session.CreateQuery("from Employee as e where e.Salary > :targetSalary")
         .SetInt64("targetSalary", 1000000L)

In the HQL above, even though we only explicitly mentioned a salary constraint on
the results, because of the enabled filter the query will return only currently
active employees who have a salary greater than a million dollars.

Note: if you plan on using filters with outer joining (either through HQL or load
fetching) be careful of the direction of the condition expression. It’s safest to
set this up for left outer joining; in general, place the parameter first followed
by the column name(s) after the operator.

Default all filter definitions are applied to <many-to-one/>
and <one-to-one/> elements. You can turn off this
behaviour by using use-many-to-one attribute on

<filter-def name="effectiveDate" use-many-to-one="false"/>

Chapter 19. Improving performance

19.1. Fetching strategies

A fetching strategy is the strategy NHibernate
will use for retrieving associated objects if the application needs to navigate
the association. Fetch strategies may be declared in the O/R mapping metadata, or
overridden by a particular HQL or Criteria query.

NHibernate defines the following fetching strategies:

  • Join fetching – NHibernate retrieves the
    associated instance or collection in the same SELECT, using
    an OUTER JOIN.

  • Select fetching – a second
    is used to retrieve the associated entity or collection. Unless
    you explicitly disable lazy fetching by specifying lazy="false",
    this second select will only be executed when you actually access the association.

  • Subselect fetching – a second
    is used to retrieve the associated collections for all entities
    retrieved in a previous query or fetch. Unless you explicitly disable lazy fetching
    by specifying lazy="false", this second select will only
    be executed when you actually access the association.

  • “Extra-lazy” collection fetching – individual
    elements of the collection are accessed from the database as needed. NHibernate
    tries not to fetch the whole collection into memory unless absolutely needed (suitable
    for very large collections)

  • Batch fetching – an optimization strategy
    for select fetching – NHibernate retrieves a batch of entity instances or collections
    in a single SELECT, by specifying a list of primary keys
    or foreign keys.

NHibernate also distinguishes between:

  • Immediate fetching – an association, collection
    or attribute is fetched immediately, when the owner is loaded.

  • Lazy collection fetching – a collection is
    fetched when the application invokes an operation upon that collection. (This is
    the default for collections.)

  • Proxy fetching – a single-valued association
    is fetched when a method other than the identifier getter is invoked upon the associated

We have two orthogonal notions here: when
is the association fetched, and how is it
fetched (what SQL is used). Don’t confuse them! We use fetch
to tune performance. We may use lazy to define a contract
for what data is always available in any detached instance of a particular class.

19.1.1. Working with lazy associations

By default, NHibernate uses lazy select fetching for collections and lazy proxy
fetching for single-valued associations. These defaults make sense for almost all
associations in almost all applications.

However, lazy fetching poses one problem that you must be aware of. Access to a
lazy association outside of the context of an open NHibernate session will result
in an exception. For example:

s = sessions.OpenSession();
Transaction tx = s.BeginTransaction();
User u = (User) s.CreateQuery("from User u where u.Name=:userName")
    .SetString("userName", userName).UniqueResult();
IDictionary permissions = u.Permissions;


int accessLevel = (int) permissions["accounts"];  // Error!

Since the permissions collection was not initialized when
the ISession was closed, the collection will not be able
to load its state. NHibernate does not support lazy initialization
for detached objects
. The fix is to move the code that reads from
the collection to just before the transaction is committed.

Alternatively, we could use a non-lazy collection or association, by specifying
lazy="false" for the association mapping. However, it is
intended that lazy initialization be used for almost all collections and associations.
If you define too many non-lazy associations in your object model, NHibernate will
end up needing to fetch the entire database into memory in every transaction!

On the other hand, we often want to choose join fetching (which is non-lazy by nature)
instead of select fetching in a particular transaction. We’ll now see how to customize
the fetching strategy. In NHibernate, the mechanisms for choosing a fetch strategy
are identical for single-valued associations and collections.

19.1.2. Tuning fetch strategies

Select fetching (the default) is extremely vulnerable to N+1 selects problems, so
we might want to enable join fetching in the mapping document:

<set name="Permissions" 
    <key column="userId"/>
    <one-to-many class="Permission"/>
<many-to-one name="Mother" class="Cat" fetch="join"/>

The fetch strategy defined in the mapping document affects:

  • retrieval via Get() or Load()

  • retrieval that happens implicitly when an association is navigated

  • ICriteria queries

  • HQL queries if subselect fetching is used

No matter what fetching strategy you use, the defined non-lazy graph is guaranteed
to be loaded into memory. Note that this might result in several immediate selects
being used to execute a particular HQL query.

Usually, we don’t use the mapping document to customize fetching. Instead, we keep
the default behavior, and override it for a particular transaction, using
left join fetch
in HQL. This tells NHibernate to fetch the association
eagerly in the first select, using an outer join. In the ICriteria
query API, you would use SetFetchMode(FetchMode.Join).

If you ever feel like you wish you could change the fetching strategy used by Get() or Load(), simply use a
ICriteria query, for example:

User user = (User) session.CreateCriteria(typeof(User))
                .SetFetchMode("Permissions", FetchMode.Join)
                .Add( Expression.Eq("Id", userId) )

(This is NHibernate’s equivalent of what some ORM solutions call a “fetch plan”.)

A completely different way to avoid problems with N+1 selects is to use the second-level

19.1.3. Single-ended association

Lazy fetching for collections is implemented using NHibernate’s own implementation
of persistent collections. However, a different mechanism is needed for lazy behavior
in single-ended associations. The target entity of the association must be proxied.
NHibernate implements lazy initializing proxies for persistent objects using runtime
bytecode enhancement (via the excellent Castle.DynamicProxy library).

By default, NHibernate generates proxies (at startup) for all persistent classes
and uses them to enable lazy fetching of many-to-one and
one-to-one associations.

The mapping file may declare an interface to use as the proxy interface for that
class, with the proxy attribute. By default, NHibernate
uses a subclass of the class. Note that the proxied class
must implement a non-private default constructor. We recommend this constructor
for all persistent classes!

There are some gotchas to be aware of when extending this approach to polymorphic
classes, eg.

<class name="Cat" proxy="Cat">
    <subclass name="DomesticCat">

Firstly, instances of Cat will never be castable to DomesticCat, even if the underlying instance is an instance
of DomesticCat:

Cat cat = (Cat) session.Load(typeof(Cat), id);  // instantiate a proxy (does not hit the db)
if ( cat.IsDomesticCat ) {                  // hit the db to initialize the proxy
    DomesticCat dc = (DomesticCat) cat;       // Error!

Secondly, it is possible to break proxy ==.

Cat cat = (Cat) session.Load(typeof(Cat), id);            // instantiate a Cat proxy
DomesticCat dc = 
        (DomesticCat) session.Load(typeof(DomesticCat), id);  // acquire new DomesticCat proxy!
System.out.println(cat==dc);                            // false

However, the situation is not quite as bad as it looks. Even though we now have
two references to different proxy objects, the underlying instance will still be
the same object:

cat.Weight = 11.0;  // hit the db to initialize the proxy
Console.WriteLine( dc.Weight );  // 11.0

Third, you may not use a proxy for a sealed class or a
class with any non-overridable public members.

Finally, if your persistent object acquires any resources upon instantiation (eg.
in initializers or default constructor), then those resources will also be acquired
by the proxy. The proxy class is an actual subclass of the persistent class.

These problems are all due to fundamental limitations in .NET’s single inheritance
model. If you wish to avoid these problems your persistent classes must each implement
an interface that declares its business methods. You should specify these interfaces
in the mapping file. eg.

<class name="CatImpl" proxy="ICat">
    <subclass name="DomesticCatImpl" proxy="IDomesticCat">

where CatImpl implements the interface
and DomesticCatImpl implements the interface
IDomesticCat. Then proxies for instances of
and IDomesticCat may be returned by
or Enumerable(). (Note that
does not usually return proxies.)

ICat cat = (ICat) session.Load(typeof(CatImpl), catid);
IEnumerator iter = session.Enumerable("from CatImpl as cat where cat.Name='fritz'").GetEnumerator();
ICat fritz = (ICat) iter.Current;

Relationships are also lazily initialized. This means you must declare any properties
to be of type ICat, not CatImpl.

Certain operations do not require proxy initialization

  • Equals(), if the persistent class does not override Equals()

  • GetHashCode(), if the persistent class does not override

  • The identifier getter method

NHibernate will detect persistent classes that override Equals()
or GetHashCode().

19.1.4. Initializing collections
and proxies

A LazyInitializationException will be thrown by NHibernate
if an uninitialized collection or proxy is accessed outside of the scope of the
ISession, ie. when the entity owning the collection or
having the reference to the proxy is in the detached state.

Sometimes we need to ensure that a proxy or collection is initialized before closing
the ISession. Of course, we can alway force initialization
by calling cat.Sex or cat.Kittens.Count,
for example. But that is confusing to readers of the code and is not convenient
for generic code.

The static methods NHibernateUtil.Initialize() and
provide the application with a convenient
way of working with lazily initialized collections or proxies. NHibernateUtil.Initialize(cat)
will force the initialization of a proxy, cat, as long
as its ISession is still open. NHibernateUtil.Initialize(
cat.Kittens )
has a similar effect for the collection of kittens.

Another option is to keep the ISession open until all needed
collections and proxies have been loaded. In some application architectures, particularly
where the code that accesses data using NHibernate, and the code that uses it are
in different application layers or different physical processes, it can be a problem
to ensure that the ISession is open when a collection is
initialized. There are two basic ways to deal with this issue:

  • In a web-based application, a HttpModule can be used to
    close the ISession only at the very end of a user request,
    once the rendering of the view is complete (the Open Session
    in View
    pattern). Of course, this places heavy demands on the correctness
    of the exception handling of your application infrastructure. It is vitally important
    that the ISession is closed and the transaction ended before
    returning to the user, even when an exception occurs during rendering of the view.
    See the NHibernate Wiki for examples of this “Open Session in View” pattern.

  • In an application with a separate business tier, the business logic must “prepare”
    all collections that will be needed by the web tier before returning. This means
    that the business tier should load all the data and return all the data already
    initialized to the presentation/web tier that is required for a particular use case.
    Usually, the application calls NHibernateUtil.Initialize()
    for each collection that will be needed in the web tier (this call must occur before
    the session is closed) or retrieves the collection eagerly using a NHibernate query
    with a FETCH clause or a FetchMode.Join
    in ICriteria. This is usually easier if you adopt the Command pattern instead of a
    Session Facade

  • You may also attach a previously loaded object to a new ISession
    with Merge() or Lock() before
    accessing uninitialized collections (or other proxies). No, NHibernate does not,
    and certainly should not do this automatically,
    since it would introduce ad hoc transaction semantics!

Sometimes you don’t want to initialize a large collection, but still need some information
about it (like its size) or a subset of the data.

You can use a collection filter to get the size of a collection without initializing

(int) s.CreateFilter( collection, "select count(*)" ).List()[0]

The CreateFilter() method is also used to efficiently retrieve
subsets of a collection without needing to initialize the whole collection:

s.CreateFilter( lazyCollection, "").SetFirstResult(0).SetMaxResults(10).List();

19.1.5. Using batch fetching

NHibernate can make efficient use of batch fetching, that is, NHibernate can load
several uninitialized proxies if one proxy is accessed (or collections. Batch fetching
is an optimization of the lazy select fetching strategy. There are two ways you
can tune batch fetching: on the class and the collection level.

Batch fetching for classes/entities is easier to understand. Imagine you have the
following situation at runtime: You have 25 Cat instances
loaded in an ISession, each Cat
has a reference to its Owner, a Person.
The Person class is mapped with a proxy,
. If you now iterate through all cats and call cat.Owner
on each, NHibernate will by default execute 25 SELECT statements,
to retrieve the proxied owners. You can tune this behavior by specifying a
in the mapping of Person:

<class name="Person" batch-size="10">...</class>

NHibernate will now execute only three queries, the pattern is 10, 10, 5.

You may also enable batch fetching of collections. For example, if each
has a lazy collection of Cats, and 10 persons
are currently loaded in the ISesssion, iterating through
all persons will generate 10 SELECTs, one for every call
to person.Cats. If you enable batch fetching for the Cats collection in the mapping of Person,
NHibernate can pre-fetch collections:

<class name="Person">
    <set name="Cats" batch-size="3">

With a batch-size of 3, NHibernate will load 3, 3, 3, 1
collections in four SELECTs. Again, the value of the attribute
depends on the expected number of uninitialized collections in a particular

Batch fetching of collections is particularly useful if you have a nested tree of
items, ie. the typical bill-of-materials pattern. (Although a
nested set
or a materialized path
might be a better option for read-mostly trees.)

19.1.6. Using subselect fetching

If one lazy collection or single-valued proxy has to be fetched, NHibernate loads
all of them, re-running the original query in a subselect. This works in the same
way as batch-fetching, without the piecemeal loading.

19.2. The Second Level Cache

A NHibernate ISession is a transaction-level cache of persistent
data. It is possible to configure a cluster or process-level (ISessionFactory-level)
cache on a class-by-class and collection-by-collection basis. You may even plug
in a clustered cache. Be careful. Caches are never aware of changes made to the
persistent store by another application (though they may be configured to regularly
expire cached data). In NHibernate 1.x the second level cache
does not work correctly in combination with distributed transactions.

By default, NHibernate uses HashtableCache for process-level caching. You may choose
a different implementation by specifying the name of a class that implements NHibernate.Cache.ICacheProvider using the property

Table 19.1. Cache Providers

Cache Provider class Type Cluster Safe Query Cache Supported
Hashtable (not intended for production use) NHibernate.Cache.HashtableCacheProvider memory   yes
ASP.NET Cache (System.Web.Cache) NHibernate.Caches.SysCache.SysCacheProvider, NHibernate.Caches.SysCache memory   yes
Prevalence Cache NHibernate.Caches.Prevalence.PrevalenceCacheProvider, NHibernate.Caches.Prevalence memory, disk   yes

19.2.1. Cache mappings

The <cache> element of a class or collection mapping
has the following form:

    region="RegionName"                                              (2)

usage specifies the caching strategy: read-write,
nonstrict-read-write or read-only


region (optional, defaults to the class or collection role
name) specifies the name of the second level cache region

Alternatively (preferrably?), you may specify <class-cache>
and <collection-cache> elements in

The usage attribute specifies a
cache concurrency strategy

19.2.2. Strategy: read only

If your application needs to read but never modify instances of a persistent class,
a read-only cache may be used. This is the simplest and
best performing strategy. Its even perfectly safe for use in a cluster.

<class name="Eg.Immutable" mutable="false">
    <cache usage="read-only"/>

19.2.3. Strategy: read/write

If the application needs to update data, a read-write cache
might be appropriate. This cache strategy should never be used if serializable transaction
isolation level is required. You should ensure that the transaction is completed
when ISession.Close() or ISession.Disconnect()
is called. If you wish to use this strategy in a cluster, you should ensure that
the underlying cache implementation supports locking. The built-in cache providers
do not.

<class name="eg.Cat" .... >
    <cache usage="read-write"/>
    <set name="Kittens" ... >
        <cache usage="read-write"/>

19.2.4. Strategy: nonstrict read/write

If the application only occasionally needs to update data (ie. if it is extremely
unlikely that two transactions would try to update the same item simultaneously)
and strict transaction isolation is not required, a nonstrict-read-write
cache might be appropriate. When using this strategy you should ensure that the
transaction is completed when ISession.Close() or
is called.

The following table shows which providers are compatible with which concurrency

Table 19.2. Cache Concurrency Strategy Support

Cache read-only nonstrict-read-write read-write
Hashtable (not intended for production use) yes yes yes
SysCache yes yes yes
PrevalenceCache yes yes yes

Refer to
Chapter 25, NHibernate.Caches
for more details.

19.3. Managing the caches

Whenever you pass an object to Save(),
or SaveOrUpdate() and whenever you retrieve
an object using Load(), Get(),
List(), or Enumerable(), that
object is added to the internal cache of the ISession.

When Flush() is subsequently called, the state of that
object will be synchronized with the database. If you do not want this synchronization
to occur or if you are processing a huge number of objects and need to manage memory
efficiently, the Evict() method may be used to remove the
object and its collections from the first-level cache.

IEnumerable cats = sess.Enumerable("from Eg.Cat as cat"); //a huge result set
foreach( Cat cat in cats )

NHibernate will evict associated entities automatically if the association is mapped
with cascade="all" or cascade="all-delete-orphan".

The ISession also provides a Contains()
method to determine if an instance belongs to the session cache.

To completely evict all objects from the session cache, call ISession.Clear()

For the second-level cache, there are methods defined on ISessionFactory
for evicting the cached state of an instance, entire class, collection instance
or entire collection role.

sessionFactory.Evict(typeof(Cat), catId); //evict a particular Cat
sessionFactory.Evict(typeof(Cat));  //evict all Cats
sessionFactory.EvictCollection("Eg.Cat.Kittens", catId); //evict a particular collection of kittens
sessionFactory.EvictCollection("Eg.Cat.Kittens"); //evict all kitten collections

19.4. The Query Cache

Query result sets may also be cached. This is only useful for queries that are run
frequently with the same parameters. To use the query cache you must first enable

<add key="hibernate.cache.use_query_cache" value="true" />

This setting causes the creation of two new cache regions – one holding cached query
result sets (NHibernate.Cache.StandardQueryCache), the
other holding timestamps of the most recent updates to queryable tables (NHibernate.Cache.UpdateTimestampsCache).
Note that the query cache does not cache the state of any entities in the result
set; it caches only identifier values and results of value type. So the query cache
should always be used in conjunction with the second-level cache.

Most queries do not benefit from caching, so by default queries are not cached.
To enable caching, call IQuery.SetCacheable(true). This
call allows the query to look for existing cache results or add its results to the
cache when it is executed.

If you require fine-grained control over query cache expiration policies, you may
specify a named cache region for a particular query by calling IQuery.SetCacheRegion().

IList blogs = sess.CreateQuery("from Blog blog where blog.Blogger = :blogger")
    .SetEntity("blogger", blogger)

If the query should force a refresh of its query cache region, you may call
to true. This is
particularly useful in cases where underlying data may have been updated via a seperate
process (i.e., not modified through NHibernate) and allows the application to selectively
refresh the query cache regions based on its knowledge of those events. This is
a more efficient alternative to eviction of a query cache region via

19.5. Understanding Collection performance

We’ve already spent quite some time talking about collections. In this section we
will highlight a couple more issues about how collections behave at runtime.

19.5.1. Taxonomy

NHibernate defines three basic kinds of collections:

  • collections of values

  • one to many associations

  • many to many associations

This classification distinguishes the various table and foreign key relationships
but does not tell us quite everything we need to know about the relational model.
To fully understand the relational structure and performance characteristics, we
must also consider the structure of the primary key that is used by NHibernate to
update or delete collection rows. This suggests the following classification:

  • indexed collections

  • sets

  • bags

All indexed collections (maps, lists, arrays) have a primary key consisting of the
<key> and <index>
columns. In this case collection updates are usually extremely efficient – the primary
key may be efficiently indexed and a particular row may be efficiently located when
NHibernate tries to update or delete it.

Sets have a primary key consisting of <key> and element
columns. This may be less efficient for some types of collection element, particularly
composite elements or large text or binary fields; the database may not be able
to index a complex primary key as efficently. On the other hand, for one to many
or many to many associations, particularly in the case of synthetic identifiers,
it is likely to be just as efficient. (Side-note: if you want SchemaExport
to actually create the primary key of a <set> for
you, you must declare all columns as not-null="true".)

<idbag> mappings define a surrogate key, so they
are always very efficient to update. In fact, they are the best case.

Bags are the worst case. Since a bag permits duplicate element values and has no
index column, no primary key may be defined. NHibernate has no way of distinguishing
between duplicate rows. NHibernate resolves this problem by completely removing
(in a single DELETE) and recreating the collection whenever
it changes. This might be very inefficient.

Note that for a one-to-many association, the “primary key” may not be the physical
primary key of the database table – but even in this case, the above classification
is still useful. (It still reflects how NHibernate “locates” individual rows of
the collection.)

19.5.2. Lists, maps,
idbags and sets are the most efficient collections to update

From the discussion above, it should be clear that indexed collections and (usually)
sets allow the most efficient operation in terms of adding, removing and updating

There is, arguably, one more advantage that indexed collections have over sets for
many to many associations or collections of values. Because of the structure of
an ISet, NHibernate doesn’t ever UPDATE
a row when an element is “changed”. Changes to an ISet
always work via INSERT and DELETE
(of individual rows). Once again, this consideration does not apply to one to many

After observing that arrays cannot be lazy, we would conclude that lists, maps and
idbags are the most performant (non-inverse) collection types, with sets not far
behind. Sets are expected to be the most common kind of collection in NHibernate
applications. This is because the “set” semantics are most natural in the relational

However, in well-designed NHibernate domain models, we usually see that most collections
are in fact one-to-many associations with inverse="true".
For these associations, the update is handled by the many-to-one end of the association,
and so considerations of collection update performance simply do not apply.

19.5.3. Bags and
lists are the most efficient inverse collections

Just before you ditch bags forever, there is a particular case in which bags (and
also lists) are much more performant than sets. For a collection with
(the standard bidirectional one-to-many relationship idiom,
for example) we can add elements to a bag or list without needing to initialize
(fetch) the bag elements! This is because IList.Add() must
always succeed for a bag or IList (unlike an
). This can make the following common code much faster.

Parent p = (Parent) sess.Load(typeof(Parent), id);
    Child c = new Child();
    c.Parent = p;
    p.Children.Add(c);  //no need to fetch the collection!

19.5.4. One shot delete

Occasionally, deleting collection elements one by one can be extremely inefficient.
NHibernate isn’t completely stupid, so it knows not to do that in the case of an
newly-empty collection (if you called list.Clear(), for
example). In this case, NHibernate will issue a single DELETE
and we are done!

Suppose we add a single element to a collection of size twenty and then remove two
elements. NHibernate will issue one INSERT statement and
two DELETE statements (unless the collection is a bag).
This is certainly desirable.

However, suppose that we remove eighteen elements, leaving two and then add thee
new elements. There are two possible ways to proceed:

  • Delete eighteen rows one by one and then insert three rows

  • Remove the whole collection (in one SQL DELETE) and insert
    all five current elements (one by one)

NHibernate isn’t smart enough to know that the second option is probably quicker
in this case. (And it would probably be undesirable for NHibernate to be that smart;
such behaviour might confuse database triggers, etc.)

Fortunately, you can force this behaviour (ie. the second strategy) at any time
by discarding (ie. dereferencing) the original collection and returning a newly
instantiated collection with all the current elements. This can be very useful and
powerful from time to time.

one-shot-delete apply to collections mapped inverse="true".

19.6. Batch updates

NHibernate supports batching SQL update commands (INSERT,
UPDATE, DELETE) with the following

  • .NET Framework 2.0 or above is required,

  • the Nhibernate’s drive used for your RDBMS may not supports batching,

  • since the implementation uses reflection to access members and types in System.Data
    assembly which are not normally visible, it may not function in environments where
    necessary permissions are not granted,

  • optimistic concurrency checking may be impaired since ADO.NET 2.0 does not return
    the number of rows affected by each statement in the batch, only the total number
    of rows affected by the batch.

Update batching is enabled by setting adonet.batch_size
to a non-zero value.

19.7. Multi Query

This functionality allows you to execute several HQL queries in one round-trip against
the database server. A simple use case is executing a paged query while also getting
the total count of results, in a single round-trip. Here is a simple example:

IMultiQuery multiQuery = s.CreateMultiQuery()
    .Add(s.CreateQuery("from Item i where i.Id > ?")
            .SetInt32(0, 50).SetFirstResult(10))
    .Add(s.CreateQuery("select count(*) from Item i where i.Id > ?")
            .SetInt32(0, 50));
IList results = multiQuery.List();
IList items = (IList)results[0];
long count = (long)((IList)results[1])[0];

The result is a list of query results, ordered according to the order of queries
added to the multi query. Named parameters can be set on the multi query, and are
shared among all the queries contained in the multi query, like this:

IList results = s.CreateMultiQuery()
    .Add(s.CreateQuery("from Item i where i.Id > :id")
        .SetFirstResult(10) )
    .Add("select count(*) from Item i where i.Id > :id")
    .SetInt32("id", 50)
IList items = (IList)results[0];
long count = (long)((IList)results[1])[0];

Positional parameters are not supported on the multi query, only on the individual

As shown above, if you do not need to configure the query separately, you can simply
pass the HQL directly to the IMultiQuery.Add() method.

Multi query is executed by concatenating the queries and sending the query to the
database as a single string. This means that the database should support returning
several result sets in a single query. At the moment this functionality is only
enabled for Microsoft SQL Server and SQLite.

Note that the database server is likely to impose a limit on the maximum number
of parameters in a query, in which case the limit applies to the multi query as
a whole. Queries using in with a large number of arguments
passed as parameters may easily exceed this limit. For example, SQL Server has a
limit of 2,100 parameters per round-trip, and will throw an exception executing
this query:

IList allEmployeesId  = ...; //1,500 items
IMultiQuery multiQuery = s.CreateMultiQuery()
	.Add(s.CreateQuery("from Employee e where e.Id in :empIds")
		.SetParameter("empIds", allEmployeesId).SetFirstResult(10))
	.Add(s.CreateQuery("select count(*) from Employee e where e.Id in :empIds")
		.SetParameter("empIds", allEmployeesId));
	IList results = multiQuery.List(); // will throw an exception from SQL Server

An interesting usage of this feature is to load several collections of an object
in one round-trip, without an expensive cartesian product (blog * users * posts).

Blog blog = s.CreateMultiQuery()
    .Add("select b from Blog b left join fetch b.Users where b.Id = :id")
    .Add("select b from Blog b left join fetch b.Posts where b.Id = :id")
    .SetInt32("id", 123)

19.8. Multi Criteria

This is the counter-part to Multi Query, and allows you to perform several criteria
queries in a single round trip. A simple use case is executing a paged query while
also getting the total count of results, in a single round-trip. Here is a simple

IMultiCriteria multiCrit = s.CreateMultiCriteria()
            .Add(Expression.Gt("Id", 50))
            .Add(Expression.Gt("Id", 50))
IList results = multiCrit.List();
IList items = (IList)results[0];
long count = (long)((IList)results[1])[0];

The result is a list of query results, ordered according to the order of queries
added to the multi criteria.

You can add ICriteria or DetachedCriteria
to the Multi Criteria query. In fact, using DetachedCriteria in this fashion has
some interesting implications.

DetachedCriteria customersCriteria = AuthorizationService.GetAssociatedCustomersQuery();
IList results = session.CreateMultiCriteria()
		.Add( Subqueries.PropertyIn("id", CriteriaTransformer.Clone(customersCriteria)
                      ) )

ICollection<Customer> customers = CollectionHelper.ToArray<Customer>(results[0]);
ICollection<Policy> policies = CollectionHelper.ToArray<Policy>(results[1]);

As you see, we get a query that represnt the customers we can access, and then we
can utlize this query further in order to perform additional logic (getting the
policies of the customers we are associated with), all in a single database roundtrip.

Chapter 20. Toolset Guide

Roundtrip engineering with NHibernate is possible using a set of commandline tools
maintained as part of the NHibernate project, along with NHibernate support built
into various code generation tools (MyGeneration, CodeSmith, ObjectMapper, AndroMDA).

The NHibernate main package comes bundled with the most important tool (it can even
be used from “inside” NHibernate on-the-fly):

  • DDL schema generation from a mapping file (aka SchemaExport,

Other tools directly provided by the NHibernate project are delivered with a separate
package, NHibernateContrib. This package
includes tools for the following tasks:

  • C# source generation from a mapping file (aka hbm2net)

  • mapping file generation from .NET classes marked with attributes (NHibernate.Mapping.Attributes,
    or NHMA for short)

Third party tools with NHibernate support are:

  • CodeSmith, MyGeneration, and ObjectMapper (mapping file generation from an existing
    database schema)

  • AndroMDA (MDA (Model-Driven Architecture) approach generating code for persistent
    classes from UML diagrams and their XML/XMI representation)

These 3rd party tools are not documented in this reference. Please refer to the
NHibernate website for up-to-date information.

20.1. Schema Generation

The generated schema includes referential integrity constraints (primary and foreign
keys) for entity and collection tables. Tables and sequences are also created for
mapped identifier generators.

You must specify a SQL Dialect
via the hibernate.dialect property when using this tool.

20.1.1. Customizing the schema

Many NHibernate mapping elements define an optional attribute named
. You may set the length of a column with this attribute. (Or, for
numeric/decimal data types, the precision.)

Some tags also accept a not-null attribute (for generating
a NOT NULL constraint on table columns) and a
attribute (for generating UNIQUE constraint
on table columns).

Some tags accept an index attribute for specifying the
name of an index for that column. A unique-key attribute
can be used to group columns in a single unit key constraint. Currently, the specified
value of the unique-key attribute is
used to name the constraint, only to group the columns in
the mapping file.


<property name="Foo" type="String" length="64" not-null="true"/>

<many-to-one name="Bar" foreign-key="fk_foo_bar" not-null="true"/>

<element column="serial_number" type="Int64" not-null="true" unique="true"/>

Alternatively, these elements also accept a child <column>
element. This is particularly useful for multi-column types:

<property name="Foo" type="String">
    <column name="foo" length="64" not-null="true" sql-type="text"/>

<property name="Bar" type="My.CustomTypes.MultiColumnType, My.CustomTypes"/>
    <column name="fee" not-null="true" index="bar_idx"/>
    <column name="fi" not-null="true" index="bar_idx"/>
    <column name="fo" not-null="true" index="bar_idx"/>

The sql-type attribute allows the user to override the
default mapping of NHibernate type to SQL datatype.

The check attribute allows you to specify a check constraint.

<property name="Foo" type="Int32">
    <column name="foo" check="foo > 10"/>

<class name="Foo" table="foos" check="bar < 100.0">
    <property name="Bar" type="Single"/>

Table 20.1. Summary

Attribute Values Interpretation
length number column length/decimal precision
not-null true|false specfies that the column should be non-nullable
unique true|false specifies that the column should have a unique constraint
index index_name specifies the name of a (multi-column) index
unique-key unique_key_name specifies the name of a multi-column unique constraint
foreign-key foreign_key_name specifies the name of the foreign key constraint generated for an association, use
it on <one-to-one>, <many-to-one>, <key>, and <many-to-many>
mapping elements. Note that inverse="true" sides will not
be considered by SchemaExport.
sql-type column_type overrides the default column type (attribute of <column>
element only)
check SQL expression create an SQL check constraint on either column or table

20.1.2. Running the tool

The SchemaExport tool writes a DDL script to standard out
and/or executes the DDL statements.

You may embed SchemaExport in your application:

Configuration cfg = ....;
new SchemaExport(cfg).Create(false, true);

20.1.3. Properties

Database properties may be specified

  • as system properties with -D<property>

  • in hibernate.properties

  • in a named properties file with --properties

The needed properties are:

Table 20.2. SchemaExport Connection Properties

Property Name Description
hibernate.connection.driver_class jdbc driver class
hibernate.connection.url jdbc url
hibernate.connection.username database user
hibernate.connection.password user password
hibernate.dialect dialect

20.1.4. Using Ant

You can call SchemaExport from your Ant build script:

<target name="schemaexport">
    <taskdef name="schemaexport"
        <fileset dir="src">
            <include name="**/*.hbm.xml"/>

If you don’t specify properties or a config
file, the SchemaExportTask will try to use normal Ant project
properties instead. In other words, if you don’t want or need an external configuration
or properties file, you may put hibernate.* configuration
properties in your build.xml or build.properties.

20.2. Code Generation

The NHibernate code generator may be used to generate skeletal C# implementation
classes from a NHibernate mapping file. This tool is included in the NHibernate
Contrib package (a seperate download in http://sourceforge.net/projects/nhcontrib/).

hbm2net parses the mapping files and generates fully working
C# source files from these. Thus with hbm2net one could
“just” provide the .hbm files, and then don’t worry about
hand-writing/coding the C# files.

hbm2net options mapping_files

Table 20.3. Code Generator Command Line Options

Option Description
-output:output_dir root directory for generated code
-config:config_file optional file for configuring hbm2net

A more detailed guide of hbm2net is available in http://nhforge.org/blogs/nhibernate/archive/2009/12/12/t4-hbm2net-alpha-2.aspx

Chapter 21. Example: Parent/Child

One of the very first things that new users try to do with NHibernate is to model
a parent / child type relationship. There are two different approaches to this.
For various reasons the most convenient approach, especially for new users, is to
model both Parent and Child as
entity classes with a <one-to-many> association from
Parent to Child. (The alternative
approach is to declare the Child as a <composite-element>.)
Now, it turns out that default semantics of a one to many association (in NHibernate)
are much less close to the usual semantics of a parent / child relationship than
those of a composite element mapping. We will explain how to use a
bidirectional one to many association with cascades
to model
a parent / child relationship efficiently and elegantly. It’s not at all difficult!

21.1. A note about collections

NHibernate collections are considered to be a logical part of their owning entity;
never of the contained entities. This is a crucial distinction! It has the following

  • When we remove / add an object from / to a collection, the version number of the
    collection owner is incremented.

  • If an object that was removed from a collection is an instance of a value type (eg,
    a composite element), that object will cease to be persistent and its state will
    be completely removed from the database. Likewise, adding a value type instance
    to the collection will cause its state to be immediately persistent.

  • On the other hand, if an entity is removed from a collection (a one-to-many or many-to-many
    association), it will not be deleted, by default. This behavior is completely consistent
    – a change to the internal state of another entity should not cause the associated
    entity to vanish! Likewise, adding an entity to a collection does not cause that
    entity to become persistent, by default.

Instead, the default behavior is that adding an entity to a collection merely creates
a link between the two entities, while removing it removes the link. This is very
appropriate for all sorts of cases. Where it is not appropriate at all is the case
of a parent / child relationship, where the life of the child is bound to the lifecycle
of the parent.

21.2. Bidirectional one-to-many

Suppose we start with a simple <one-to-many> association
from Parent to Child.

<set name="Children">
    <key column="parent_id" />
    <one-to-many class="Child" />

If we were to execute the following code

Parent p = .....;
Child c = new Child();

NHibernate would issue two SQL statements:

  • an INSERT to create the record for c

  • an UPDATE to create the link from p
    to c

This is not only inefficient, but also violates any NOT NULL
constraint on the parent_id column.

The underlying cause is that the link (the foreign key parent_id)
from p to c is not considered
part of the state of the Child object and is therefore
not created in the INSERT. So the solution is to make the
link part of the Child mapping.

<many-to-one name="Parent" column="parent_id" not-null="true"/>

(We also need to add the Parent property to the

Now that the Child entity is managing the state of the
link, we tell the collection not to update the link. We use the

<set name="Children" inverse="true">
    <key column="parent_id"/>
    <one-to-many class="Child"/>

The following code would be used to add a new Child.

Parent p = (Parent) session.Load(typeof(Parent), pid);
Child c = new Child();
c.Parent = p;

And now, only one SQL INSERT would be issued!

To tighten things up a bit, we could create an AddChild()
method of Parent.

public void AddChild(Child c)
    c.Parent = this;

Now, the code to add a Child looks like

Parent p = (Parent) session.Load(typeof(Parent), pid);
Child c = new Child();

21.3. Cascading lifecycle

The explicit call to Save() is still annoying. We will
address this by using cascades.

<set name="Children" inverse="true" cascade="all">
    <key column="parent_id"/>
    <one-to-many class="Child"/>

This simplifies the code above to

Parent p = (Parent) session.Load(typeof(Parent), pid);
Child c = new Child();

Similarly, we don’t need to iterate over the children when saving or deleting a
Parent. The following removes p
and all its children from the database.

Parent p = (Parent) session.Load(typeof(Parent), pid);

However, this code

Parent p = (Parent) session.Load(typeof(Parent), pid);
// Get one child out of the set
IEnumerator childEnumerator = p.Children.GetEnumerator();
Child c = (Child) childEnumerator.Current;

c.Parent = null;

will not remove c from the database; it will only remove
the link to p (and cause a NOT NULL
constraint violation, in this case). You need to explicitly Delete()
the Child.

Parent p = (Parent) session.Load(typeof(Parent), pid);
// Get one child out of the set
IEnumerator childEnumerator = p.Children.GetEnumerator();
Child c = (Child) childEnumerator.Current;


Now, in our case, a Child can’t really exist without its
parent. So if we remove a Child from the collection, we
really do want it to be deleted. For this, we must use cascade="all-delete-orphan".

<set name="Children" inverse="true" cascade="all-delete-orphan">
    <key column="parent_id"/>
    <one-to-many class="Child"/>

Note: even though the collection mapping specifies inverse="true",
cascades are still processed by iterating the collection elements. So if you require
that an object be saved, deleted or updated by cascade, you must add it to the collection.
It is not enough to simply set its parent.

21.4. Using cascading

Suppose we loaded up a Parent in one ISession,
made some changes in a UI action and wish to persist these changes in a new ISession
(by calling Update()). The Parent
will contain a collection of children and, since cascading update is enabled, NHibernate
needs to know which children are newly instantiated and which represent existing
rows in the database. Let’s assume that both Parent and
Child have (synthetic) identifier properties of type long. NHibernate will use the identifier property value
to determine which of the children are new. (You may also use the version or timestamp
property, see Section 9.4.2, “Updating detached

The unsaved-value attribute is used to specify the identifier
value of a newly instantiated instance. In NHibernate it
is not necessary to specify unsaved-value explicitly.

The following code will update parent and
and insert newChild.

//parent and child were both loaded in a previous session
Child newChild = new Child();

Well, thats all very well for the case of a generated identifier, but what about
assigned identifiers and composite identifiers? This is more difficult, since unsaved-value can’t distinguish between a newly instantiated
object (with an identifier assigned by the user) and an object loaded in a previous
session. In these cases, you will probably need to give NHibernate a hint; either

  • define an unsaved-value on a <version>
    or <timestamp> property mapping for the class.

  • set unsaved-value="none" and explicitly
    newly instantiated children before calling Update(parent)

  • set unsaved-value="any" and explicitly
    previously persistent children before calling Update(parent)

null is the default unsaved-value
for assigned identifiers, none is the default
for composite identifiers.

There is one further possibility. There is a new IInterceptor
method named IsTransient() which lets the application implement
its own strategy for distinguishing newly instantiated objects. For example, you
could define a base class for your persistent classes.

public class Persistent
    private bool _saved = false;
    public void OnSave()
    public void OnLoad()
    public bool IsSaved
        get { return _saved; }

(The saved property is non-persistent.) Now implement IsTransient(), along with OnLoad()
and OnSave() as follows.

	public object IsTransient(object entity)
    if (entity is Persistent)
        return !( (Persistent) entity ).IsSaved;
        return null;

public bool OnLoad(object entity, 
    object id,
    object[] state,
    string[] propertyNames,
    IType[] types)
    if (entity is Persistent) ( (Persistent) entity ).OnLoad();
    return false;

public boolean OnSave(object entity,
    object id,
    object[] state,
    string[] propertyNames,
    IType[] types)
    if (entity is Persistent) ( (Persistent) entity ).OnSave();
    return false;

21.5. Conclusion

There is quite a bit to digest here and it might look confusing first time around.
However, in practice, it all works out quite nicely. Most NHibernate applications
use the parent / child pattern in many places.

We mentioned an alternative in the first paragraph. None of the above issues exist
in the case of <composite-element> mappings, which
have exactly the semantics of a parent / child relationship. Unfortunately, there
are two big limitations to composite element classes: composite elements may not
own collections, and they should not be the child of any entity other than the unique
parent. (However, they may have a surrogate
primary key, using an <idbag> mapping.)

Chapter 22. Example: Weblog Application

22.1. Persistent Classes

The persistent classes represent a weblog, and an item posted in a weblog. They
are to be modelled as a standard parent/child relationship, but we will use an ordered
bag, instead of a set.

using System;
using System.Collections;

namespace Eg
    public class Blog
        private long _id;
        private string _name;
        private IList _items;
        public virtual long Id
            get { return _id; }
            set { _id = value; }
        public virtual IList Items
            get { return _items; }
            set { _items = value; }
        public virtual string Name
            get { return _name; }
            set { _name = value; }
using System;

namespace Eg
    public class BlogItem
        private long _id;
        private DateTime _dateTime;
        private string _text;
        private string _title;
        private Blog _blog;

        public virtual Blog Blog
            get { return _blog; }
            set { _blog = value; }

        public virtual DateTime DateTime
            get { return _dateTime; }
            set { _dateTime = value; }

        public virtual long Id
            get { return _id; }
            set { _id = value; }

        public virtual string Text
            get { return _text; }
            set { _text = value; }

        public virtual string Title
            get { return _title; }
            set { _title = value; }

22.2. NHibernate Mappings

The XML mappings should now be quite straightforward.

<?xml version="1.0"?>
<hibernate-mapping xmlns="urn:nhibernate-mapping-2.2"
    assembly="Eg" namespace="Eg">

            <generator class="native"/>
            <key column="BLOG_ID"/>
            <one-to-many class="BlogItem"/>
<?xml version="1.0"?>
<hibernate-mapping xmlns="urn:nhibernate-mapping-2.2"
    assembly="Eg" namespace="Eg">
            <generator class="native"/>

22.3. NHibernate Code

The following class demonstrates some of the kinds of things we can do with these
classes, using NHibernate.

using System;
using System.Collections;

using NHibernate.Tool.hbm2ddl;

namespace Eg
    public class BlogMain
        private ISessionFactory _sessions;
        public void Configure()
            _sessions = new Configuration()
        public void ExportTables()
            Configuration cfg = new Configuration()
            new SchemaExport(cfg).create(true, true);
        public Blog CreateBlog(string name)
            Blog blog = new Blog();
            blog.Name = name;
            blog.Items = new ArrayList();
            using (ISession session = _sessions.OpenSession())
            using (ITransaction tx = session.BeginTransaction())

            return blog;
        public BlogItem CreateBlogItem(Blog blog, string title, string text)
            BlogItem item = new BlogItem();
            item.Title = title;
            item.Text = text;
            item.Blog = blog;
            item.DateTime = DateTime.Now;
            using (ISession session = _sessions.OpenSession())
            using (ITransaction tx = session.BeginTransaction())

            return item;
        public BlogItem CreateBlogItem(long blogId, string title, string text)
            BlogItem item = new BlogItem();
            item.Title = title;
            item.Text = text;
            item.DateTime = DateTime.Now;
            using (ISession session = _sessions.OpenSession())
            using (ITransaction tx = session.BeginTransaction())
                Blog blog = (Blog) session.Load(typeof(Blog), blogId);
                item.Blog = blog;

            return item;
        public void UpdateBlogItem(BlogItem item, string text)
            item.Text = text;

            using (ISession session = _sessions.OpenSession())
            using (ITransaction tx = session.BeginTransaction())
        public void UpdateBlogItem(long itemId, string text)
            using (ISession session = _sessions.OpenSession())
            using (ITransaction tx = session.BeginTransaction())
                BlogItem item = (BlogItem) session.Load(typeof(BlogItem), itemId);
                item.Text = text;
        public IList listAllBlogNamesAndItemCounts(int max)
            IList result = null;

            using (ISession session = _sessions.OpenSession())
            using (ITransaction tx = session.BeginTransaction())
                IQuery q = session.CreateQuery(
                    "select blog.id, blog.Name, count(blogItem) " +
                    "from Blog as blog " +
                    "left outer join blog.Items as blogItem " +
                    "group by blog.Name, blog.id " +
                    "order by max(blogItem.DateTime)"
                result = q.List();

            return result;
        public Blog GetBlogAndAllItems(long blogId)
            Blog blog = null;

            using (ISession session = _sessions.OpenSession())
            using (ITransaction tx = session.BeginTransaction())
                IQuery q = session.createQuery(
                    "from Blog as blog " +
                    "left outer join fetch blog.Items " +
                    "where blog.id = :blogId"
                q.SetParameter("blogId", blogId);
                blog  = (Blog) q.List()[0];

            return blog;
        public IList ListBlogsAndRecentItems()
            IList result = null;

            using (ISession session = _sessions.OpenSession())
            using (ITransaction tx = session.BeginTransaction())
                IQuery q = session.CreateQuery(
                    "from Blog as blog " +
                    "inner join blog.Items as blogItem " +
                    "where blogItem.DateTime > :minDate"
                DateTime date = DateTime.Now.AddMonths(-1);
                q.SetDateTime("minDate", date);
                result = q.List();

            return result;

Chapter 23. Example: Various Mappings

This chapter shows off some more complex association mappings.

23.1. Employer/Employee

The following model of the relationship between Employer
and Employee uses an actual entity class (Employment)
to represent the association. This is done because there might be more than one
period of employment for the same two parties. Components are used to model monetary
values and employee names.

Here’s a possible mapping document:

<hibernate-mapping xmlns="urn:nhibernate-mapping-2.2"
    assembly="..." namespace="...">

    <class name="Employer" table="employers">
        <id name="Id">
            <generator class="sequence">
                <param name="sequence">employer_id_seq</param>
        <property name="Name"/>

    <class name="Employment" table="employment_periods">

        <id name="Id">
            <generator class="sequence">
                <param name="sequence">employment_id_seq</param>
        <property name="StartDate" column="start_date"/>
        <property name="EndDate" column="end_date"/>

        <component name="HourlyRate" class="MonetaryAmount">
            <property name="Amount">
                <column name="hourly_rate" sql-type="NUMERIC(12, 2)"/>
            <property name="Currency" length="12"/>

        <many-to-one name="Employer" column="employer_id" not-null="true"/>
        <many-to-one name="Employee" column="employee_id" not-null="true"/>


    <class name="Employee" table="employees">
        <id name="Id">
            <generator class="sequence">
                <param name="sequence">employee_id_seq</param>
        <property name="TaxfileNumber"/>
        <component name="Name" class="Name">
            <property name="FirstName"/>
            <property name="Initial"/>
            <property name="LastName"/>


And here’s the table schema generated by SchemaExport.

create table employers (
    Id BIGINT not null, 
    Name VARCHAR(255), 
    primary key (Id)

create table employment_periods (
    Id BIGINT not null,
    hourly_rate NUMERIC(12, 2),
    Currency VARCHAR(12), 
    employee_id BIGINT not null, 
    employer_id BIGINT not null, 
    end_date TIMESTAMP, 
    start_date TIMESTAMP, 
    primary key (Id)

create table employees (
    Id BIGINT not null, 
    FirstName VARCHAR(255), 
    Initial CHAR(1), 
    LastName VARCHAR(255), 
    TaxfileNumber VARCHAR(255), 
    primary key (Id)

alter table employment_periods 
    add constraint employment_periodsFK0 foreign key (employer_id) references employers
alter table employment_periods 
    add constraint employment_periodsFK1 foreign key (employee_id) references employees
create sequence employee_id_seq
create sequence employment_id_seq
create sequence employer_id_seq

23.2. Author/Work

Consider the following model of the relationships between Work,
Author and Person. We represent
the relationship between Work and Author
as a many-to-many association. We choose to represent the relationship between Author and Person as one-to-one
association. Another possibility would be to have Author
extend Person.

The following mapping document correctly represents these relationships:

<hibernate-mapping xmlns="urn:nhibernate-mapping-2.2"
    assembly="..." namespace="...">

    <class name="Work" table="works" discriminator-value="W">

        <id name="Id" column="id" generator="native" />

        <discriminator column="type" type="character"/>

        <property name="Title"/>
        <set name="Authors" table="author_work" lazy="true">
                <column name="work_id" not-null="true"/>
            <many-to-many class="Author">
                <column name="author_id" not-null="true"/>

        <subclass name="Book" discriminator-value="B">
            <property name="Text" column="text" />

        <subclass name="Song" discriminator-value="S">
            <property name="Tempo" column="tempo" />
            <property name="Genre" column="genre" />


    <class name="Author" table="authors">

        <id name="Id" column="id">
            <!-- The Author must have the same identifier as the Person -->
            <generator class="assigned"/> 

        <property name="Alias" column="alias" />
        <one-to-one name="Person" constrained="true"/>

        <set name="Works" table="author_work" inverse="true" lazy="true">
            <key column="author_id"/>
            <many-to-many class="Work" column="work_id"/>


    <class name="Person" table="persons">
        <id name="Id" column="id">
            <generator class="native"/>
        <property name="Name" column="name" />


There are four tables in this mapping. works,
and persons hold work, author and person
data respectively. author_work is an association table
linking authors to works. Heres the table schema, as generated by

create table works (
    id BIGINT not null generated by default as identity, 
    tempo FLOAT, 
    genre VARCHAR(255), 
    text INTEGER, 
    title VARCHAR(255), 
    type CHAR(1) not null, 
    primary key (id)

create table author_work (
    author_id BIGINT not null, 
    work_id BIGINT not null, 
    primary key (work_id, author_id)

create table authors (
    id BIGINT not null generated by default as identity, 
    alias VARCHAR(255), 
    primary key (id)

create table persons (
    id BIGINT not null generated by default as identity, 
    name VARCHAR(255), 
    primary key (id)

alter table authors 
    add constraint authorsFK0 foreign key (id) references persons
alter table author_work 
    add constraint author_workFK0 foreign key (author_id) references authors
alter table author_work
    add constraint author_workFK1 foreign key (work_id) references works

23.3. Customer/Order/Product

Now consider a model of the relationships between Customer,
Order and LineItem and
. There is a one-to-many association between Customer
and Order, but how should we represent
/ LineItem / Product?
I’ve chosen to map LineItem as an association class representing
the many-to-many association between Order and
. In NHibernate, this is called a composite element.

The mapping document:

<hibernate-mapping xmlns="urn:nhibernate-mapping-2.2"
    assembly="..." namespace="...">

    <class name="Customer" table="customers">
        <id name="Id" column="id" generator="native" />
        <property name="Name" column="name"/>
        <set name="Orders" inverse="true" lazy="true">
            <key column="customer_id"/>
            <one-to-many class="Order"/>

    <class name="Order" table="orders">
        <id name="Id" column="id" generator="native" />
        <property name="Date" column="date"/>
        <many-to-one name="Customer" column="customer_id"/>
        <list name="LineItems" table="line_items" lazy="true">
            <key column="order_id"/>
            <index column="line_number"/>
            <composite-element class="LineItem">
                <property name="Quantity" column="quantity"/>
                <many-to-one name="Product" column="product_id"/>

    <class name="Product" table="products">
        <id name="Id" column="id">
            <generator class="native"/>
        <property name="SerialNumber" column="serial_number" />


customers, orders,
and products hold customer, order,
order line item and product data respectively. line_items
also acts as an association table linking orders with products.

create table customers (
    id BIGINT not null generated by default as identity, 
    name VARCHAR(255), 
    primary key (id)

create table orders (
    id BIGINT not null generated by default as identity, 
    customer_id BIGINT, 
    date TIMESTAMP, 
    primary key (id)

create table line_items (
    line_number INTEGER not null, 
    order_id BIGINT not null, 
    product_id BIGINT, 
    quantity INTEGER, 
    primary key (order_id, line_number)

create table products (
    id BIGINT not null generated by default as identity, 
    serial_number VARCHAR(255), 
    primary key (id)

alter table orders 
    add constraint ordersFK0 foreign key (customer_id) references customers
alter table line_items
    add constraint line_itemsFK0 foreign key (product_id) references products
alter table line_items
    add constraint line_itemsFK1 foreign key (order_id) references orders

Chapter 24. Best Practices

Write fine-grained classes and map them using

Use an Address class to encapsulate street,
suburb, state,
. This encourages code reuse and simplifies refactoring.

Declare identifier properties on persistent classes.

NHibernate makes identifier properties optional. There are all sorts of reasons
why you should use them. We recommend that identifiers be ‘synthetic’ (generated,
with no business meaning) and of a non-primitive type. For maximum flexibility,
use Int64 or String.

Place each class mapping in its own file.

Don’t use a single monolithic mapping document. Map Eg.Foo
in the file Eg/Foo.hbm.xml. This makes particularly good
sense in a team environment.

Embed mappings in assemblies.

Place mapping files along with the classes they map and declare them as
Embedded Resource
s in Visual Studio.

Consider externalising query strings.

This is a good practice if your queries call non-ANSI-standard SQL functions. Externalising
the query strings to mapping files will make the application more portable.

Use parameters.

As in ADO.NET, always replace non-constant values by “?”. Never use string manipulation
to bind a non-constant value in a query! Even better, consider using named parameters
in queries.

Don’t manage your own ADO.NET connections.

NHibernate lets the application manage ADO.NET connections. This approach should
be considered a last-resort. If you can’t use the built-in connections providers,
consider providing your own implementation of NHibernate.Connection.IConnectionProvider.

Consider using a custom type.

Suppose you have a type, say from some library, that needs to be persisted but doesn’t
provide the accessors needed to map it as a component. You should consider implementing
NHibernate.UserTypes.IUserType. This approach frees the
application code from implementing transformations to / from an NHibernate type.

Use hand-coded ADO.NET in bottlenecks.

In performance-critical areas of the system, some kinds of operations (eg. mass
update / delete) might benefit from direct ADO.NET. But please, wait until you know something is a bottleneck. And don’t assume
that direct ADO.NET is necessarily faster. If need to use direct ADO.NET, it might
be worth opening a NHibernate ISession and using that SQL
connection. That way you can still use the same transaction strategy and underlying
connection provider.

Understand ISession flushing.

From time to time the ISession synchronizes its persistent state with the database.
Performance will be affected if this process occurs too often. You may sometimes
minimize unnecessary flushing by disabling automatic flushing or even by changing
the order of queries and other operations within a particular transaction.

In a three tiered architecture, consider using

When using a distributed architecture, you could pass persistent objects loaded
in the middle tier to and from the user interface tier. Use a new session to service
each request. Use ISession.Update() or
to update the persistent state of an object.

In a two tiered architecture, consider using session disconnection.

Database Transactions have to be as short as possible for best scalability. However,
it is often neccessary to implement long running Application Transactions, a single
unit-of-work from the point of view of a user. This Application Transaction might
span several client requests and response cycles. Either use Detached Objects or,
in two tiered architectures, simply disconnect the NHibernate Session from the ADO.NET
connection and reconnect it for each subsequent request. Never use a single Session
for more than one Application Transaction usecase, otherwise, you will run into
stale data.

Don’t treat exceptions as recoverable.

This is more of a necessary practice than a “best” practice. When an exception occurs,
roll back the ITransaction and close the
. If you don’t, NHibernate can’t guarantee that in-memory state
accurately represents persistent state. As a special case of this, do not use ISession.Load() to determine if an instance with the given
identifier exists on the database; use Get() or a query

Prefer lazy fetching for associations.

Use eager (outer-join) fetching sparingly. Use
proxies and/or lazy collections for most associations to classes that are not cached
in the second-level cache. For associations to cached classes, where there is a
high probability of a cache hit, explicitly disable eager fetching using
. When an outer-join fetch is appropriate to a particular
use case, use a query with a left join fetch.

Consider abstracting your business logic from NHibernate.

Hide (NHibernate) data-access code behind an interface. Combine the
and Thread Local Session
patterns. You can even have some classes persisted by handcoded ADO.NET, associated
to NHibernate via an IUserType. (This advice is intended
for “sufficiently large” applications; it is not appropriate for an application
with five tables!)

Implement Equals() and
using a unique business key.

If you compare objects outside of the ISession scope, you have to implement
and GetHashCode(). Inside the ISession
scope, object identity is guaranteed. If you implement these methods, never ever
use the database identifier! A transient object doesn’t have an identifier value
and NHibernate would assign a value when the object is saved. If the object is in
an ISet while being saved, the hash code changes, breaking the contract. To implement
Equals() and GetHashCode(), use
a unique business key, that is, compare a unique combination of class properties.
Remember that this key has to be stable and unique only while the object is in an
ISet, not for the whole lifetime (not as stable as a database primary key). Never
use collections in the Equals() comparison (lazy loading)
and be careful with other associated classes that might be proxied.

Don’t use exotic association mappings.

Good usecases for a real many-to-many associations are rare. Most of the time you
need additional information stored in the “link table”. In this case, it is much
better to use two one-to-many associations to an intermediate link class. In fact,
we think that most associations are one-to-many and many-to-one, you should be careful
when using any other association style and ask yourself if it is really neccessary.

NHibernateContrib Documentation


The NHibernateContrib is various programs contributed to NHibernate by members of
the NHibernate Team or by the end users. The projects in here are not considered
core pieces of NHibernate but they extend it in a useful way.

Chapter 25. NHibernate.Caches

What is NHibernate.Caches?

NHibernate.Caches namespace contains several second-level cache providers for NHibernate. 
A cache is place where entities are kept after being loaded from the database; once
cached, they can be retrieved without going to the database. This means that they
are faster to (re)load.

An NHibernate session has an internal (first-level) cache where it keeps its entities.
There is no sharing between these caches – a first-level cache belongs to a given
session and is destroyed with it. NHibernate provides a second-level
system; it works at the session factory level. A second-level
cache is shared by all sessions created by the same session factory.

An important point is that the second-level cache does not
cache instances of the object type being cached; instead it caches the individual
values of the properties of that object. This provides two benefits. One, NHibernate
doesn’t have to worry that your client code will manipulate the objects in a way
that will disrupt the cache. Two, the relationships and associations do not become
stale, and are easy to keep up-to-date because they are simply identifiers. The
cache is not a tree of objects but rather a map of arrays.

With the session-per-request model, a high
number of sessions can concurrently access the same entity without hitting the database
each time; hence the performance gain.

Several cache providers have been contributed by NHibernate users:


Uses Bamboo.Prevalence as the cache provider. Open the
file Bamboo.Prevalence.license.txt for more information
about its license; you can also visit its website.


Uses System.Web.Caching.Cache as the cache provider.
This means that you can rely on ASP.NET caching feature to understand how it works.
For more information, read (on the MSDN): Caching Application Data.


Similar to NHibernate.Caches.SysCache, uses ASP.NET cache.
This provider also supports SQL dependency-based expiration, meaning that it is
possible to configure certain cache regions to automatically expire when the relevant
data in the database changes.

SysCache2 requires Microsoft SQL Server 2000 or higher and .NET Framework version
2.0 or higher.


Uses memcached. See memcached homepage for more information.

NCache provider for NHibernate

Uses NCache,NCache is a commercial distributed caching
system with a provider for NHibernate. The NCache Express version is free for use,
see NCache
Express homepage
for more information.

25.1. How to use a cache?

Here are the steps to follow to enable the second-level cache in NHibernate:

  • Choose the cache provider you want to use and copy its assembly in your assemblies
    directory (NHibernate.Caches.Prevalence.dll or

  • To tell NHibernate which cache provider to use, add in your NHibernate configuration
    file (can be YourAssembly.exe.config or
    or a .cfg.xml file, in the latter
    case the syntax will be different from what is shown below):

    <add key="hibernate.cache.provider_class" value="XXX" />(1)
    <add key="expiration" value="120" />(2)


    XXX” is the assembly-qualified class name of a class implementing
    ICacheProvider, eg. “NHibernate.Caches.SysCache.SysCacheProvider,


    The expiration value is the number of seconds you wish
    to cache each entry (here two minutes). This example applies to SysCache only.

  • Add <cache usage="read-write|nonstrict-read-write|read-only"/>
    (just after <class>) in the mapping of the entities
    you want to cache. It also works for collections (bag, list, map, set, …).

Be careful.  Caches are never aware of changes made to the persistent
store by another process (though they may be configured to regularly expire cached
data). As the caches are created at the session factory level, they are destroyed
with the SessionFactory instance; so you must keep them alive as long as you need

25.2. Prevalence Cache Configuration

There is only one configurable parameter: prevalenceBase.
This is the directory on the file system where the Prevalence engine will save data.
It can be relative to the current directory or a full path. If the directory doesn’t
exist, it will be created.

25.3. SysCache Configuration

As SysCache relies on System.Web.Caching.Cache for the
underlying implementation, the configuration is based on the available options that
make sense for NHibernate to utilize.

Number of seconds to wait before expiring each item.
A numeric cost of expiring each item, where 1 is a low cost, 5 is the highest, and
3 is normal. Only values 1 through 5 are valid.

SysCache has a config file section handler to allow configuring different expirations
and priorities for different regions. Here’s an example:

Example 25.1. 

<?xml version="1.0" encoding="utf-8" ?>
		<section name="syscache" type="NHibernate.Caches.SysCache.SysCacheSectionHandler,NHibernate.Caches.SysCache" />

		<cache region="foo" expiration="500" priority="4" />
		<cache region="bar" expiration="300" priority="3" />

25.4. SysCache2 Configuration

SysCache2 can use SqlCacheDependencies to invalidate cache regions when data in
an underlying SQL Server table or query changes. Query dependencies are only available
for SQL Server 2005. To use the cache provider, the application must be setup and
configured to support SQL notifications as described in the MSDN documentation.

To configure cache regions with SqlCacheDependencies a syscache2
config section must be defined in the application’s configuration file. See the
sample below.

Example 25.2. 

	<section name="syscache2" type="NHibernate.Caches.SysCache2.SysCacheSection, NHibernate.Caches.SysCache2"/>

25.4.1. Table-based Dependency

A table-based dependency will monitor the data in a database table for changes.
Table-based dependencies are generally used for a SQL Server 2000 database but will
work with SQL Server 2005 as well. Before you can use SQL Server cache invalidation
with table based dependencies, you need to enable notifications for the database.
This task is performed with the aspnet_regsql
command. With table-based notifications, the application will poll the database
for changes at a predefined interval. A cache region will not be invalidated immediately
when data in the table changes. The cache will be invalidated the next time the
application polls the database for changes.

To configure the data in a cache region to be invalidated when data in an underlying
table is changed, a cache region must be configured in the application’s configuration
file. See the sample below.

Example 25.3. 

	<cacheRegion name="Product">
				<add name="price"
					tableName="VideoTitle" />

Table-based Dependency Configuration Properties

Unique name for
the dependency
The name of the database table that the dependency is associated with. The table
must be enabled for notification support with the AspNet_SqlCacheRegisterTableStoredProcedure.
The name of a database defined in the databases element
for sqlCacheDependency for caching (ASP.NET Settings Schema)
element of the application’s Web.config file.

25.4.2. Command-Based Dependencies

A command-based dependency will use a SQL command to identify records to monitor
for data changes. Command-based dependencies work only with SQL Server 2005.

Before you can use SQL Server cache invalidation with command-based dependencies,
you need to enable the Service Broker for query notifications. The application must
also start the listener for receiving change notifications from SQL Server. With
command based notifications, SQL Server will notify the application when the data
of a record returned in the results of a SQL query has changed. Note that a change
will be indicated if the data in any of the columns of a record change, not just
the columns returned by a query. The query is a way to limit the number of records
monitored for changes, not the columns. As soon as data in one of the records is
modified, the data in the cache region will be invalidated immediately.

To configure the data in a cache region to be invalidated based on a SQL command,
a cache region must be configured in the application’s configuration file. See the
samples below.

Example 25.4. Stored Procedure

<cacheRegion name="Product" priority="High" >
			<add name="price"

Example 25.5. SELECT Statement

<cacheRegion name="Product" priority="High">
			<add name="price"
				command="Select VideoTitleId from dbo.VideoTitle where Active = 1"
                "NHibernate.Caches.SysCache2.ConfigConnectionStringProvider, NHibernate.Caches.SysCache2"/>

Command Configuration Properties

Unique name for
the dependency
command (required)
command that returns results which should be monitored for data changes
isStoredProcedure (optional)
Indicates if command is a stored procedure. The default is false.
connectionName (optional)
The name of the connection in the applications configuration file to use for registering
the cache dependency for change notifications. If no value is supplied for
or connectionStringProviderType,
the connection properties from the NHibernate configruation will be used.
connectionStringProviderType (optional)
IConnectionStringProvider to use for retrieving the connection
string to use for registering the cache dependency for change notifications. If
no value is supplied for connectionName, the unnamed connection
supplied by the provider will be used.

25.4.3. Aggregate Dependencies

Multiple cache dependencies can be specified. If any of the dependencies triggers
a change notification, the data in the cache region will be invalidated. See the
samples below.

Example 25.6. Multiple commands

<cacheRegion name="Product">
			<add name="price"
			<add name="quantity"
				command="Select quantityAvailable from dbo.VideoAvailability"/>


Example 25.7. Mixed

<cacheRegion name="Product">
			<add name="price"
			<add name="quantity"
				tableName=" VideoAvailability" />

25.4.4. Additional Settings

In addition to data dependencies for the cache regions, time based expiration policies
can be specified for each item added to the cache. Time based expiration policies
will not invalidate the data dependencies for the whole cache region, but serve
as a way to remove items from the cache after they have been in the cache for a
specified amount of time. See the samples below.

Example 25.8. Relative Expiration

<cacheRegion name="Product" relativeExpiration="300" priority="High" />

Example 25.9. Time of Day Expiration

<cacheRegion name="Product" timeOfDayExpiration="2:00:00" priority="High" />

Additional Configuration Properties


Number of seconds that an individual item will exist in the cache before being removed.


24 hour based time of day that an item will exist in the cache until. 12am is specified
as 00:00:00; 10pm is specified as 22:00:00. Only valid if relativeExpiration is
not specified. Time of Day Expiration is useful for scenarios where items should
be expired from the cache after a daily process completes.

that identifies the relative priority of items stored in the cache.

25.4.5. Patches

There is a known issue where some SQL Server 2005 notifications might not be received
when an application subscribes to query notifications by using ADO.NET 2.0. To fix
this problem install SQL hotfix for kb 913364.

Chapter 26. NHibernate.Mapping.Attributes

What is NHibernate.Mapping.Attributes?

NHibernate.Mapping.Attributes is an add-in for NHibernate contributed by Pierre Henri Kuaté (aka
); the former implementation was made by John Morris. 
require mapping streams to bind your domain model to your database. Usually, they
are written (and maintained) in separated hbm.xml files.

With NHibernate.Mapping.Attributes, you can use .NET attributes to decorate your
entities and these attributes will be used to generate these mapping .hbm.xml (as
files or streams). So you will no longer have to bother with these
xml files ;).

Content of this library. 

  1. NHibernate.Mapping.Attributes: That the only project
    you need (as end-user)

  2. Test: a working sample using attributes and HbmSerializer
    as NUnit TestFixture

  3. Generator: The program used to generate attributes and

  4. Refly
    : Thanks to Jonathan de Halleux
    for this library which make it so easy to generate code


This library is generated using the file /src/NHibernate.Mapping.Attributes/nhibernate-mapping.xsd
(which is embedded in the assembly to be able to validate generated XML streams).
As this file can change at each new release of NHibernate, you should regenerate
it before using it with a different version (open the Generator solution, compile
and run the Generator project). But, no test has been done with versions prior to

26.1. What’s new?

NHibernate. introduces many new features, improvements and changes:

  1. It is possible to import classes by simply decorating them with
    [Import] class ImportedClass1 {}
    . Note that you must use HbmSerializer.Serialize(assembly);
    The <import/> mapping will be added before the classes
    mapping. If you prefer to keep these imports in the class using them, you can specify
    them all on the class: [Import(ClassType=typeof(ImportedClass1))]
    class Query {}

  2. [RawXmlAttribute] is a new attribute allowing to insert
    xml as-is in the mapping. This feature can be very useful to do complex mapping
    (eg: components). It may also be used to quickly move the mapping from xml files
    to attributes. Usage: [RawXml(After=typeof(ComponentAttribute),
    Content="<component name="...">...</component>")]
    tells after which kind of mapping the xml should be inserted
    (generally, it is the type of the mapping you are inserting); it is optional (in
    which case the xml is inserted on the top of the mapping). Note: At the moment,
    all raw xmls are prefixed by a <!----> (in the generated
    stream); this is a known side-effect.

  3. [AttributeIdentifierAttribute] is a new attribute allowing
    to provide the value of a defined “place holder”. Eg:

    public class Base {
        [Id(..., Column="{{Id.Column}}")]
        [AttributeIdentifier(Name="Id.Column", Value="ID")] // Default value
        public int Id { ... }
    [AttributeIdentifier(Name="Id.Column", Value="SUB_ID")]
    [Class] public class MappedSubClass : Base { }

    The idea is that, when you have a mapping which is shared by many subclasses but
    which has minor differences (like different column names), you can put the mapping
    in the base class with place holders on these fields and give their values in subclasses.
    Note that this is possible for any mapping field taking a string (column, name,
    type, access, etc.). And, instead of Value, you can
    use ValueType or ValueObject
    (if you use an enum, you can control its formatting with ValueObject).

    The “place holder” is defined like this: {{XXX}}. If you
    don’t want to use these double curly brackets, you can change them using the properties
    StartQuote and EndQuote
    of the class HbmWriter.

  4. It is possible to register patterns (using Regular Expressions) to automatically
    transform fully qualified names of properties types into something else. Eg: HbmSerializer.Default.HbmWriter.Patterns.Add(@"Namespace.(\S+),
    Assembly", "$1");
    will map all properties with a not-qualified type name.

  5. Two methods have been added to allow writing: cfg.AddInputStream(
    HbmSerializer.Default.Serialize(typeof(XXX)) )
    and cfg.AddInputStream(
    HbmSerializer.Default.Serialize(typeof(XXX).Assembly) )
    . So it is no
    longer required to create a MemoryStream for these simple cases.

  6. Two WriteUserDefinedContent() methods have been added
    to HbmWriter. They improve the extensibility of this
    library; it is now very easy to create a .NET attribute and integrate it in the

  7. Attributes [(Jcs)Cache], [Discriminator]
    and [Key] can be specified at class-level.

  8. Interfaces can be mapped (just like classes and structs).

  9. A notable “bug” fix is the re-ordering of (joined-)subclasses; This operation may
    be required when a subclass extends another subclass. In this case, the extended
    class mapping must come before the extending class mapping. Note that the re-ordering
    takes place only for “top-level” classes (that is not nested in other mapped classes).
    Anyway, it is quite unusual to put a interdependent mapped subclasses in a mapped

  10. There are also many other little changes; refer to the release notes for more details.

26.2. How to use it?

The end-user class is
This class
your domain model to mapping streams. You can
either serialize classes one by one or an assembly. Look at NHibernate.Mapping.Attributes.Test
project for a working sample.

The first step is to decorate your entities with attributes; you can use:
, [Subclass], [JoinedSubclass]
or [Component]. Then, you decorate your members (fields/properties);
they can take as many attributes as required by your mapping. Eg:

    public class Example
        public string Name;

After this step, you use NHibernate.Mapping.Attributes.HbmSerializer:
(here, we use Default which is an instance you can use
if you don’t need/want to create it yourself).

    NHibernate.Cfg.Configuration cfg = new NHibernate.Cfg.Configuration();
    NHibernate.Mapping.Attributes.HbmSerializer.Default.Validate = true; // Enable validation (optional)
    // Here, we serialize all decorated classes (but you can also do it class by class)
    cfg.AddInputStream( NHibernate.Mapping.Attributes.HbmSerializer.Default.Serialize(
        System.Reflection.Assembly.GetExecutingAssembly() ); );
    // Now you can use this configuration to build your SessionFactory...


As you can see here: NHibernate.Mapping.Attributes is not
(really) intrusive. Setting attributes on your objects doesn’t force you to use
them with NHibernate and doesn’t break any constraint on your architecture. Attributes
are purely informative (like documentation)!

26.3. Tips

  1. In production, it is recommended to generate a XML mapping file from NHibernate.Mapping.Attributes
    and use this file each time the SessionFactory need to be built. Use:
    HbmSerializer.Default.Serialize(typeof(XXX).Assembly, "DomainModel.hbm.xml");

    It is slightly faster.

  2. Use HbmSerializer.Validate to enable/disable the validation
    of generated xml streams (against NHibernate mapping schema); this is useful to
    quickly find errors (they are written in StringBuilder HbmSerializer.Error).
    If the error is due to this library then see if it is a know issue and report it;
    you can contribute a solution if you solve the problem 🙂

  3. Your classes, fields and properties (members) can be private; just make sure that
    you have the permission to access private members using reflection (ReflectionPermissionFlag.MemberAccess).

  4. Members of a mapped classes are also seek in its base classes (until we reach mapped base class). So you can decorate some
    members of a (not mapped) base class and use it in its (mapped) sub class(es).

  5. For a Name taking a System.Type, set the type with Name="xxx" (as
    ) or NameType=typeof(xxx);
    (add “Type” to “Name“)

  6. By default, .NET attributes don’t keep the order of attributes; so you need to set
    it yourself when the order matter (using the first parameter of each attribute);
    it is highly recommended to set it when you
    have more than one attribute on the same member.

  7. As long as there is no ambiguity, you can decorate a member with many unrelated
    attributes. A good example is to put class-related attributes (like
    ) on the identifier member. But don’t forget that
    the order matters (the <discriminator> must be after
    the <id>). The order used comes from the order of
    elements in the NHibernate mapping schema. Personally, I prefer using negative numbers
    for these attributes (if they come before!).

  8. You can add [HibernateMapping] on your classes to specify
    <hibernate-mapping> attributes (used when serializing
    the class in its stream). You can also use HbmSerializer.Hbm*
    properties (used when serializing an assembly or a type that is not decorated with

  9. Instead of using a string for DiscriminatorValue (in
    [Class] and [Subclass]), you
    can use any object you want. Example:

    [Subclass(DiscriminatorValueEnumFormat="d", DiscriminatorValueObject=DiscEnum.Val1)]

    Here, the object is an Enum, and you can set the format you want (the default value
    is “g”). Note that you must put it before! For others
    types, It simply use the ToString() method of the object.

  10. If you are using members of the type Nullables.NullableXXX
    (from the library Nullables),
    then they will be mapped to Nullables.NHibernate.NullableXXXType
    automatically; don’t set Type="..." in
    (leave it null). This is also the case for SqlTypes
    (and you can add your own patterns). Thanks to Michael Third
    for the idea 🙂

  11. Each stream generated by NHibernate.Mapping.Attributes can contain a comment with
    the date of the generation; You may enable/disable this by using the property HbmSerializer.WriteDateComment.

  12. If you forget to provide a required xml attribute, it will obviously throw an exception
    while generating the mapping.

  13. The recommended and easiest way to map [Component] is
    to use [ComponentProperty]. The first step is to put
    [Component] on the component class and map its fields/properties.
    Note that you shouldn’t set the Name in
    . Then, on each member in your classes, add [ComponentProperty].
    But you can’t override Access, Update
    or Insert for each member.

    There is a working example in NHibernate.Mapping.Attributes.Test
    (look for the class CompAddress and its usage in others

  14. Another way to map [Component] is to use the way this
    library works: If a mapped class contains a mapped component, then this component
    will be include in the class. NHibernate.Mapping.Attributes.Test
    contains the classes JoinedBaz and
    which both use the component Address.

    Basically, it is done by adding

    [Component(Name = "MyComp")] private class SubComp : Comp {}

    in each class. One of the advantages is that you can override
    , Update or Insert
    for each member. But you have to add the component subclass in
    class (and it can not be inherited). Another advantage is that you
    can use [AttributeIdentifier].

  15. Finally, whenever you think that it is easier to write the mapping in XML (this
    is often the case for [Component]), you can use

  16. About customization. HbmSerializer uses HbmWriter to serialize each kind of attributes. Their
    methods are virtual; so you can create a subclass and override any method you want
    (to change its default behavior).

    Use the property HbmSerializer.HbmWriter to change the
    writer used (you may set a subclass of HbmWriter).

Example using some this tips: (0, 1 and 2 are position indexes)

    [NHibernate.Mapping.Attributes.Id(0, TypeType=typeof(int))] // Don't put it after [ManyToOne] !!!
        [NHibernate.Mapping.Attributes.Generator(1, Class="uuid.hex")]
    [NHibernate.Mapping.Attributes.ManyToOne(2, ClassType=typeof(Foo), OuterJoin=OuterJoinStrategy.True)]
    private Foo Entity;


    <id type="Int32">
        <generator class="uuid.hex" />
    <many-to-one name="Entity" class="Namespaces.Foo, SampleAssembly" outer-join="true" />

26.4. Known issues and TODOs

First, read TODOs in the source code 😉

A Position property has been added to all attributes
to order them. But there is still a problem:

When a parent element “p” has a child element “x” that is also the child element
of another child element “c” of “p” (preceding “x”) 😀 Illustration:

            <x />
        <x />

In this case, when writing:

    public MyType MyProperty;

X(3) will always belong to C(1) ! (as X(2)).

It is the case for <dynamic-component> and

Another bad news is that, currently, XML elements coming after this elements can
not be included in them. Eg: There is no way put a collection in
. The reason is that the file nhibernate-mapping.xsd
tells how elements are built and in which order, and NHibernate.Mapping.Attributes
use this order.

Anyway, the solution would be to add a int ParentNode
property to BaseAttribute so that you can create a real graph…

For now, you can fallback on [RawXml].

Actually, there is no other know issue nor planned modification. This library should
be stable and complete; but if you find a bug or think of an useful improvement,
contact us!

On side note, it would be nice to write a better TestFixture than

26.5. Developer Notes

Any change to the schema (nhibernate-mapping.xsd) implies:

  1. Checking if there is any change to do in the Generator (like updating KnowEnums
    / AllowMultipleValue / IsRoot / IsSystemType / IsSystemEnum / CanContainItself)

  2. Updating /src/NHibernate.Mapping.Attributes/nhibernate-mapping.xsd
    (copy/paste) and running the Generator again (even if it wasn’t modified)

  3. Running the Test project and make sure that no exception is thrown. A class/property
    should be modified/added in this project to be sure that any new breaking change
    will be caught (=> update the reference hbm.xml files and/or the project

This implementation is based on NHibernate mapping schema; so there is probably
lot of “standard schema features” that are not supported…

The version of NHibernate.Mapping.Attributes should be the version of the NHibernate
schema used to generate it (=> the version of NHibernate library).

In the design of this project, performance is a (very)
minor goal 🙂 Easier implementation and maintenance are far more important because
you can (and should) avoid to use this library in production (Cf. the first tip
Section 26.3, “Tips”


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: