The ideas of success and failure are used to express algorithms with choices. Indeed, a goal can be defined as a choice between different goals. Such a goal is called a choice point.

A choice point is created by the execution of the goal IlcOr. A choice point can be labeled when it is created so that you can direct CP Optimizer to return to it explicitly.

Here's how a choice point is executed in a depth first search:

• The state of CP Optimizer is saved, including the state of the goal stack.
• The first subgoal is added to the top of the goal stack.
• The other subgoals are saved as untried subgoals for the choice point.
• Then the first subgoal is popped from the goal stack and executed. If this subgoal fails, the state of CP Optimizer is restored, and the first untried choice is pushed onto the goal stack. This activity is called backtracking. Backtracking is done as long as no subgoal succeeds. If no subgoal succeeds, the choice point fails.

Thus a CP Optimizer user can define an algorithm without knowing in advance which subgoal will succeed. This kind of programming is often called non-deterministic programming.

See also

IlcAnd, IlcGoal, IlcGoalI, IlcOr