Multithreading in SQLJ

This section discusses SQLJ support and requirements for multithreading and the relation between multithreading and execution context instances.

You can use SQLJ in writing multithreaded applications; however, any use of multithreading in your SQLJ application is subject to the limitations of your JDBC driver or proprietary database access vehicle.

You are required to use a different execution context instance for each thread. You can accomplish this in one of two ways:

• Specify connection context instances for your SQLJ statements such that a different connection context instance is used for each thread. Each connection context instance automatically has its own default execution context instance.

• If you are using the same connection context instance with multiple threads, then declare additional execution context instances and specify execution context instances for your SQLJ statements such that a different execution context instance is used for each thread.

For information about how to specify connection context instances and execution context instances for your SQLJ statements, see "Specifying Connection Context Instances and Execution Context Instances".

If you are using one of the Oracle JDBC drivers, multiple threads can use the same connection context instance if desired (as long as different execution context instances are specified), and there are no synchronization requirements directly visible to the user. Note, however, that database access is sequential--only one thread is accessing the database at any given time. (Synchronization refers to the control flow of the various stages of the SQL operations executing through your threads. Each statement, for example, may bind input parameters, then execute, then bind output parameters. With some JDBC drivers, special care must be taken not to intermingle these stages of different operations.)

If a thread attempts to execute a SQL operation that uses an execution context that is in use by another operation, then the thread is blocked until the current operation completes. If an execution context were shared between threads, the results of a SQL operation performed by one thread would be visible in the other thread. If both threads were executing SQL operations, a race condition might occur--the results of an execution in one thread might be overwritten by the results of an execution in the other thread before the first thread had processed the original results. This is why multiple threads are not allowed to share an execution context instance.

For a complete multithreading sample application, see "Multithreading--MultiThreadDemo.sqlj".

---