Getting Started With db4o

By Stefan Edlich and Eric Falsken

db4o is the open source object database that enables developers to store and retrieve any application object with one line of code.

The Object Database (ODB) arrived in the software industry with the advent of object oriented languages. The ODB is primarily used as an application specific database in either extreme scale systems or embedded systems where typical DBA activities are automated.

If you are familiar with object-relational mapping (ORM) tools, then you are already an object database expert because many of the APIs and query languages are comparable. You will even find db4o both familiar and yet simpler to use as ORM tools as there is no requirement for XML mapping, annotations or IDs.

db4o was introduced in 2000 with a focus on performance, compactness, zero administration, simplicity and (the most important of all) native object persistence. Developers were finally able to combine the power of a full database engine with plain undecorated objects.

db4o comes distributed as a few native .NET assemblies. Only Db4objects.Db4o.dll is required for basic db4o operation. Use the Visual Studio "Add Reference" command to add the necessary assemblies. Then use the Solution Explorer to locate the new reference, right-click and open the properties window to ensure that "Copy Local" is set to true for each db4o assembly. This will copy the necessary db4o libraries to your application's bin folder automatically when compiling your project.

Required Environment

Developing with db4o requires only the .NET SDK version 2.0 or better. (3.5 suggested)

Visual Studio 2008 or better is suggested for the best experience, but any .NET IDE will do. Microsoft Visual Studio 2008 Express editions are available for free download from Microsoft.

Running db4o requires only the .NET Framework to be installed. Some hosting environments, such as shared website hosting providers, do not allow code to run with full trust. When the environment is configured to run with reduced trust, all basic database operations require at least ReflectPermission (MemberAccess and ReflectionEmit) for the classes and types being persisted.

Required Libraries per Database Feature

Depending on the features your application requires, reference and distribute these assemblies when you distribute your application:

ObjectManager for Debugging

Included in the db4o distribution you'll find the installer for ObjectManager Enterprise (OME) which, once installed, will integrate into your Visual Studio Environment and allow you to open and inspect, query, and edit (value types only ) object instances stored in your database file.

A Complete Example

​

IObjectContainer db =

Db4oFactory.OpenFile([filename]), ([config]);

try{

        // Store a few Person objects

        db.Store(new Person("Petra"));

        db.Store(new Person("Gallad"));

        // Retrieve the Person

        var results = db.Query<Person>(x => x.Name == "Petra");

        Person p = result.First();

        // Update the Person

        p.Name = "Peter";

        db.Store(p);

        // Delete the person

        db.Delete(p);

        // Don't forget to commit!

        db.Commit();

}

catch{

        db.Rollback();

}

finally{

        // Close the db cleanly

        db.Close();

}

​

Create or Open a Database File

​

IObjectContainer container = Db4oFactory.OpenFile([filename]);

​

While a db4o database file is open, it is locked and cannot be accessed by another application at the same time.

In traditional relational databases, it's common to open a connection, get/update data, close connection, and then perform your operation. Because db4o uses the native object (or referential) identity, it's better to open the database or connection when your application begins, do all your work, then close the database when your program is shutting down. You'll see why when we get to updating an object with our changes.

Starting a db4o Server

By default, the db4o server runs in-process within your application. To start a db4o server, place this code into your application:

​

IObjectServer server =

        Db4oFactory.OpenServer([filename], [port]);

server.GrantAccess([user], [password]);

​

To shut down the server:

​

server.Close();

​

The port parameter specifies the network port number. Acceptable values are any number above 1024 which are not already in use.

The GrantAccess() method must be called for each username/ password combination. It is not required at all for embedded servers.

Connecting to a db4o Server

After starting a db4o server instance, use either of the commands below to open a db4o client connection:

​

// In-Process mode (embedded server)

IObjectContainer client = server.OpenClient();

// Client/Server mode (remote server)

IObjectContainer client =

        Db4oFactory.OpenClient([serverAddress], [port], [user],

[password]);

​

To close the client connection to the server:

​

client.Close();

​

Storing an Object

​

db.Store(anObject);

db.Commit()

​

Just one line of code is all it takes. All of an object's properties and child objects will be stored.

Updating an Object

​

db.Store(anObject);

db.Commit();

​

Looks familiar? You can use the same Store(object) command to update an object. One difference, however, is that db4o will (for performance reasons) not automatically check child objects for changes.

There is one more thing to be aware of: db4o uses an object's native identity (referential identity) to identify an object and map it to the stored instance of the object. This means that there is only ever 1 instance of an object in memory for each stored instance of the object. (per ObjectContainer or connection) This is important when dealing with class instances that may come from ASP.NET ViewState, Web Services, Interop, or any other serialized source of object data.

Deleting an Object

​

db.Delete(anObject);

db.Commit();

​

You didn't think it was any harder than that, did you? Like updates, db4o will not automatically delete child objects unless you configure cascading deletes for your object will remain in memory until the objects are refreshed or garbage collected.

Database Transactions

Whenever you start making changes to your database (using the Store() and Delete() commands) you are automatically in an open transaction. To close the transaction, use the Commit() method:

​

db.Commit();

​

Your changes will be permanently saved. If you wish to cancel or roll back any uncommitted changes, use the Rollback() method:

​

db.Rollback();

​

Closing a database cleanly will automatically call Commit() for you, so any uncommitted transactions are committed automatically.

If the database is not closed cleanly, or if the application crashes at any time and uncommitted (or incomplete) transactions are discarded.

Query by Example (QBE)

QBE lets you pass db4o an example object. db4o will search and return all objects which match the object you specify. To do this, db4o will reflect all of the properties of the object and assemble all non-default property values into a single query expression.

Here's an example QBE query that will contain all Customer objects with "Lee" as their Surname:

​

Customer proto = new Customer ()

        {Surname = "Lee"};

IObjectSet result = db.Get(proto);

​

Native Queries (NQ)

Like any query language, Native Queries are capable of expressing complex parameterized constraints, however NQ also have the benefit of being 100% compiler checked and can be refactored using common code refactoring tools. NQ can do all this because they are expressed as native .NET code rather than as strings (like SQL statements)

​

// NQ Lambda Expression (.NET 3.5 syntax)

var result = db.Query<type>(

        o => [boolean expression]);

// NQ Anonymous Method (.NET 2.0 syntax)

IList<type> result = db.Query<type>(

        delegate(type o){

                return [boolean expression];

        });

​

Native Query Examples

​

// Query all instances of a type.

IList<User> result = db.Query<User>();

// Query Users by Name.

IList<User> result = db.Query<User>(u => u.Name == "Joe");

// Query Users with at least 10 orders

IList<User> result = db.Query<user>(u => u.Orders.Count >= 10);

// Complex Query

IList<User> result = db.Query<User>(

        u => u.Name.StartsWith("Bob")

                && (u.Country == "Canada" | |

                        u.Country == "USA" );

​

Sorting Native Queries

Native Query results can be sorted by using an IComparer or a comparison delegate. Here is the query syntax:

​

IList<User> result = db.Query<User>([predicate],[comparer]);

​

And here's an example:

​

// All users with "Smith" in their name

// sorted by name

IList<User> result = db.Query<User>(

        u => u.Name.Contains("Smith"),

        (u1, u2) => u1.Name.CompareTo(u2.Name)

        );

​

LINQ Queries

LINQ was introduced by Microsoft .NET Framework version 3.5 (also called C# 3.0) in November 2007. To enable the use of LINQ queries, you'll need to add a reference to the Db4objects.Db4o.LINQ.dll assembly and import the Db4objects.Db4o.LINQ namespace. (with a using statement)

A full description of LINQ syntax will not fit in this document. You can find the URL to Microsoft's LINQ reference in the Resources section at the end.

LINQ Queries have all the benefit of compiler checking and automated code refactorability that Native Queries have, but are expressed in syntax more familiar to SQL developers. Here's one quick and easy example that gets all of the Customers with "Smith" in their name, and sorts the results by name.

​

var results =

        from Customer c in db

        where c.Name.Contains("Smith")

        orderby c.Name descending

        select c;

​

With LINQ queries you can:

• Use ORDERBY to sort the results.
• Use JOIN expressions to filter one dataset based on the contents of another.
• Use Aggregate Expressions to get the sum, min, max, or average values and GROUPBY to group the results.
• Use Anonymous Types to get back only the fields you wish.
• Use an expression to process, filter, or format the data as it is returned from the database.
• Use LINQ extensions like First(), Any(), All(), Single(), and ElementAt() to access or constrain the result set.

SODA Queries

SODA query expressions are a standard that was present in all but the earliest versions of db4o. Using combinations of SODA query and constraint keywords, you can build up what is called a query "graph". A graph is a network of objects which represent a segment of data.

In this example, we're querying for all of the customers named "Smith".

​

IQuery query = db.Query();

query.Constrain(typeof(Customer));

query.Descend("Name").Constrain("Smith");

IObjectSet results = query.Execute();

​

And in this more complicated example, we'll use a compound constraint:

​

IQuery query = db.Query();

query.Constrain(typof(Customer));

IConstraint c1 = query

        .Descend("Name")

        .Constrain("Smith");

IConstraint c2 = query

        .Descend("Country")

        .Constrain("USA");

c1.And(c2);

IObjectSet results = query.Execute();

​

Notice how each of the calls to Constrain() will return an IConstraint? You can keep references to those constraints and then use constraint keywords like And(), Or(), and Not() to relate the constraints together, as we did at the end of that example.

Note that the Descend() method returns an IQuery too. So you could Descend() into an object, and then Execute() at a deeper level to return only the matching child objects, like in this example:

​

IQuery q1 = db.Query();

q1.Constrain(typeof(Customer)); // start with Customers

q1.Descend("Name")

        .Constrain("Smith");

IQuery q2 = q1.Descend("Orders"); // constrain Orders

// Order totals greater than $100.

q2.Descend("Total")

        .Constrain(100)

        .Greater();

// Since we want to return Orders, execute q2 instead of q1

IObjectSet results = q2.Execute();

​

Query Performance

• If you are experiencing poor NQ performance, then you probably forgot to either enable run-time optimization of Native Queries (by including a reference to the assemblies listed in the "Required Libraries" section) or running Db4oTool.exe on your compiled assembly.
• You can change the Query EvaluationMode configuration to control how and when a query should be evaluated. (See below in the Configuration section)
• You can index fields to aid query evaluation. Indexing fields causes db4o to store the values of the field in a separate index lookup table in the db. As a result, when evaluating the query, db4o does not have to seek through all of the object data to resolve the query results. (See the Class-Specific Configuration Options below)

When dealing with objects that may have relations to other objects quite deep (think of the path of data from Customer to Order to OrderItem to Product with relations to Address objects for billing and shipping and then PO and payment transactions) it would be quite expensive to have to pull all of that data into memory from the DB if all you wanted was the Customer object. Modern object databases use the idea of activation to control the depth to which objects are instantiated and populated when retrieved from the database.

With an ActivationDepth of 5, objects will be populated up to 5 levels deep. Properties of the 5th descendant object will have their values left as default or null.

If you encounter an object that is not yet activated, you can pass it to db4o for manual (late) activation:

​

db.Activate([unactivatedObject], [depth]);

​

Objects can also be manually de-activated:

​

db.Deactivate([activatedObject], [depth]);

​

Fine-grained activation depth can be configured per class. (see ActivationDepth in the Configuration section below) Activation can also be managed transparently using Transparent Activation. (see below)

For all but the simplest db4o use cases, you'll probably want to specify one or more configuration settings when opening your db4o database:

​

IConfiguration config =

Db4oFactory.NewConfiguration();

// Set configuration properties here

IObjectContainer db= Db4oFactory.OpenFile([config], [filename]);

​

The IConfiguration object must be passed in the call to open the db4o file, server, or client connection.

Query Evaluation Mode

This property controls when and how much of a query is executed.

​

IConfiguration config =

        Db4oFactory.NewConfiguration();

config.Queries().EvaluationMode([mode]);

​

Global Update Depth

We said earlier that when calling Store() to update an object graph, that db4o will not (by default) descend into child objects to detect changes to the graph. If you know that you'll be often changing child properties, or when changing a parent object often results in changes to child objects, then you may want to change the UpdateDepth.

​

IConfiguration config =

Db4oFactory.NewConfiguration();

config.UpdateDepth([depth]);

​

The default value 1 means db4o will not descend into child objects when updating stored object instances.

Setting the UpdateDepth to int.MaxValue will cause db4o to descend as deeply as possible to look for changes.

Setting the UpdateDepth to 0 will prevent any changes from being saved to the database.

Global Activation Depth

As explained in the section on Dealing with Object Activation, the ActivationDepth controls how much data is loaded when an object is retrieved from the database.

​

Iconfiguration config =

        Db4oFactory.NewConfiguration();

config.ActivationDepth([depth]);

​

The default value of 5 is a good balance for most applications, but developers should balance this against the weight of their classes and their access patterns.

Setting the ActivationDepth to int.MaxValue will cause all related objects to be instantiated as deeply as possible, restoring the entire object graph to memory.

Setting the ActivationDepth to 0 will cause nothing to be activated. The object returned will have none of its values loaded. You can then call the Activate(object, depth) method to manually activate the object as described above.

Transparent Activation (TA)

The Db4oTool.exe tool found in the distribution bin folder can be used to instrument your classes to perform activation on-demand transparently as you navigate between object references. When accessing a child object, db4o can automatically activate the child object from the database if it is not yet already activated.

Transparent Activation must be enabled in the configuration when opening a db4o database and your compiled assembly must be instrumented by Db4oTool.exe as part of the build (as an MSBuild task) or post-build by running the tool manually.

Using Db4oTool.exe is not complicated, but is beyond the scope of this DZone Refcard. You can find complete instructions in the db4o reference documentation.

The global UpdateDepth and ActivationDepth configurations are good for general testing. But oftentimes, you will want to configure specific behaviors per-class-type, or per-field. (e.g. field indexing)

​

IConfiguration config =

Db4oFactory.NewConfiguration();

//Get the class-specific canfiguration

IObjectClass objectClassConfig =

        config.ObjectClass([typeName]);

//Get a field-specific configuration

IObjectField objectFieldConfig =

        objectClassConfig.ObjectField([fieldName]);

// Usually shortened to one line. e.g. config.

objectClass(typeof(Customer)).ObjectField ("Name").Indexed(true);

​

A full description of Class-specific and field-specific configuration settings can be found in the db4o reference documentation.

db4o also allows you to specify event handlers when an object is retrieved, activated, deleted, etc. These events can be handled globally using the EventRegistryFactory or individually within your classes using any of the IObjectCallbacks methods like objectOnActivate, objectOnDelete, and objectOnNew. These handlers are great for selectively activating, refreshing, or cascading db operations.

Lastly, setting the Transient attribute on your class's field members will prevent db4o from storing the values of those fields. When read from the DB, these members will always be left at their default (null) values, useful for remote connections, non- native data (interop) and temporary state.