This example shows how to declare and use bidirectional one-to-one, one-to- many, and many-to-many relationships between persistent objects. It also shows how to work with lazy pointers. All the relationships presented in this example declare one side as inverse in order to produce canonical database schema. The example uses the shared_ptr and weak_ptr smart pointers from TR1 and requires a C++ compiler with TR1 support or an external TR1 implementation, such as the one provided by Boost. The example consists of the following files: employee.hxx Header file defining the 'employee', 'employer', 'position', and 'project' persistent classes as well as the employer-employee (one-to-many), employee-position (one-to-one), and employee-project (many-to-many) bidirectional relationships between them. employee-odb.hxx employee-odb.ixx employee-odb.cxx employee.sql The first three files contain the database support code and the last file contains the database schema for the employee.hxx header. These files are generated by the ODB compiler from employee.hxx using the following command line: odb -d <database> --generate-schema --generate-query \ --default-pointer std::tr1::shared_ptr employee.hxx Where <database> stands for the database system we are using, for example, 'mysql'. The --default-pointer option is used to make TR1 shared_ptr the default object pointer. database.hxx Contains the create_database() function which instantiates the concrete database class corresponding to the database system we are using. driver.cxx Driver for the example. It includes the employee.hxx and employee-odb.hxx headers to gain access to the persistent classes and their database support code. It also includes database.hxx for the create_database() function declaration. In main() the driver first calls create_database() to obtain the database instance. It then creates a number of 'employee', 'employer', 'position', and 'project' objects, sets the relationships between them, and persists them in the database. In the next few transactions the driver loads various objects, then accesses and modifies the relationships between them. Finally, the driver performs a database query which uses a data member from a related object in its criterion. To run the example we may first need to create the database schema (for some database systems, such as SQLite, the schema is embedded into the generated code which makes this step unnecessary). Using MySQL as an example, this can be achieved with the following command: mysql --user=odb_test --database=odb_test < employee.sql Here we use 'odb_test' as the database login and also 'odb_test' as the database name. Once the database schema is ready, we can run the example (using MySQL as the database): ./driver --user odb_test --database odb_test