The SAVEPOINT statement sets a savepoint within a unit of work to identify a point in time within the unit of work to which relational database changes can be rolled back.
This statement can be embedded in an application program or issued interactively. It is an executable statement that can be dynamically prepared.
None required.
>>-SAVEPOINT--savepoint-name--+--------+------------------------>
'-UNIQUE-'
(1)
.-ON ROLLBACK RETAIN LOCKS--------.
>--ON ROLLBACK RETAIN CURSORS--+---------------------------------+-><
Omitting UNIQUE indicates that the application can reuse the savepoint name within the unit of work. If savepoint-name identifies a savepoint that already exists within the unit of work and the savepoint was not created with the UNIQUE option, the existing savepoint is destroyed and a new savepoint is created. Destroying a savepoint to reuse its name for another savepoint is not the same as releasing the savepoint. Reusing a savepoint name destroys only one savepoint. Releasing a savepoint with the RELEASE SAVEPOINT statement releases the savepoint and all savepoints that have been subsequently set.
Effect on INSERT: In an application, inserts may be buffered. The buffer will be flushed when SAVEPOINT, ROLLBACK, or RELEASE TO SAVEPOINT statements are issued.
SAVEPOINT Restriction: A SAVEPOINT statement is not allowed if commitment control is not active for the activation group. For information on determining which commitment definition is used, see Notes.
Assume that you want to set three savepoints at various points in a unit of work. Name the first savepoint A and allow the savepoint name to be reused. Name the second savepoint B and do not allow the name to be reused. Because you no longer need savepoint A when you are ready to set the third savepoint, reuse A as the name of the savepoint.
SAVEPOINT A ON ROLLBACK RETAIN CURSORS; . . . SAVEPOINT B UNIQUE ON ROLLBACK RETAIN CURSORS; . . . SAVEPOINT A ON ROLLBACK RETAIN CURSORS;