Using the Domain Objects Persistence Pattern in .NET

Domain objects in an application represent the core data and business validation rules relating to it. And, domain objects are usually central to the entire application and used by most subsystems. Therefore, their good design is critical for a good application design that is robust, high performing, and yet flexible.

When it comes to developing object oriented applications that use relational databases, the domain object design has to be consistent with the database design. This make them easier to understand because they represent real-life "entities" and their relationships with each other. Therefore, in many situations, the domain objects are "mapped" to the relational database tables and relationships between tables. However, it is very easy to get this mapping wrong and end up with an undesirable domain object design. A good design for domain objects requires a solid understanding of object oriented and relational fundamentals on the part of developers.

Domain Objects Persistence Pattern attempts to provide a solution for domain object mapping to the relational databases that decouples the domain objects from the persistence logic. The domain objects in this pattern are unaware of the objects that persist them because the dependency is only one-way (from persistence objects to domain objects). This makes the domain objects design much simpler and easier to understand. It also hides the persistence objects from other subsystems in the application that are using the domain objects. This also works in distributed systems where only the domain objects are passed around. In this context, an attempt is made to incorporate the Factory Pattern into this pattern to help decouple domain objects and persistence logic.



Problem Definition

Domain objects form the backbone of any application. They capture the core data model from the database and also the business rules that apply to this data. It is very typical for most subsystems of an application to rely on these common domain objects. This means that the closer the domain objects map to the data model in the database, the easier it is for the application developers to understand and use them because they mimic real-life "entities" and "relationships" as represented in the database.

If domain objects are not separated from the rest of the application, we end up with duplication of code everywhere. Similarly, if domain objects are not separated from the persistence code, we face situations where any subsystem using the domain objects also knows and depends on the persistence objects. And, any change in persistence objects affects the entire application, hence a bad design.


Solution
One way to achieve the above mentioned goals is to separate the domain objects into a separate subsystem and let the entire application use them wherever it needs domain data. Additionally, we should separate domain objects from the persistence code. This double-decoupling allows us on one hand to avoid code duplication and on the other to hide the persistence details from the domain objects and make it more flexible in case it needs to change. The domain objects and the rest of the application is totally unaffected whether the data is coming from a relational database or any other source (e.g. XML, flat files, or Active Directory/LDAP).

In separating the persistence logic from domain objects, we ensure that the domain objects have no dependency on the persistence code. This allows the domain objects to become available in environments where we don't even want to expose our persistence code.


Sample Code

In this sample, we will look at a Customer object from Northwind database mapped to the "Customers" table in the database.

public class Customer {
// Private data members
String _CustomerID;
String _companyName;
String _contactName;
String _contactTitle;

public Customer() {}

// Properties for Customer object
public String CustomerId {
get { return _customerId; } set { _customerId = value;}
}

public String CompanyName {
get { return _companyName; } set { _companyName = value;}
}

public String ContactName {
get { return _contactName; } set { _contactName = value;}
}

public String ContactTitle {
get { return _contactTitle; } set { _contactTitle = value;}
}
}

public interface ICustomerFactory
{
// Standard transactional methods for single-row operations
void Load(Customer cust);
void Insert(Customer cust);
void Update(Customer cust);
void Delete(Customer cust);

// Query method to return a collection
ArrayList FindCustomersByState(String state);

}

public class CustomerFactory : ICustomerFactory
{
// Standard transactional methods for single-row operations
void Load(Customer cust) { /* Implement here */ }
void Insert(Customer cust) { /* Implement here */ }
void Update(Customer cust) { /* Implement here */ }
void Delete(Customer cust) { /* Implement here */ }

// Query method to return a collection
ArrayList FindCustomersByState(String state) { /* Implement here */ }
}
Below is an example of how a client application will use this code.

public class NorthwindApp
{
static void Main (string[] args) {
Customer cust = new Customer();
CustomerFactory custFactory = new CustomerFactory();

// Let's load a customer from Northwind database.
cust.CustomerId = "ALFKI";
custFactory.load(cust);

// Pass on the Customer object
FooBar(cust);

// custList is a collection of Customer objects
ArrayList custList = custFactory.FindCustomersByState("CA");
}
}

As you can see above, the "load" method loads the Customer object from the database based on the CustomerId. Once the Customer is loaded, then it can be passed on to any subsystem in the application without exposing the persistence code. Similarly, if you get an ArrayList of Customer objects, you can pass on the ArrayList which has no persistence code dependency