Inferencing Overview › Discretely Reactive Rules

Discretely Reactive Rules

For this type of rule, a rule's threads react to field-value modifications only while threads are pended and, even then, only in a restricted manner.

Discretely reactive rules are inherently more efficient, more stable and safer than continually reactive rules. For example, it should be impossible for a rule author to write discretely-reactive rules that loop endlessly. Such looping would be possible for continually reactive rules.

Thread States

The Engine processes rule threads independently of one another.

All threads are initialized in a READY state.

As the Engine visits rule threads, thread states change:

• If the Engine can find a non-empty statement-action to perform and can complete execution of all the action statements, the Engine sets the thread's state to FIRED and retires the thread from the agenda.
• If the Engine cannot find a statement-action to perform - or only finds an empty statement-action - the Engine sets the thread's state to FAILED and retires the thread from the agenda.
• If the Engine is unable to evaluate premise expressions or to complete statement-actions due to references to unresolved fields, the Engine sets the thread's state to PENDED but retains the thread on the agenda. Other threads (of the same or different rules) may subsequently modify field values - thereby unpending the pended thread and return it to a READY state.

The Engine visits only READY rule threads. When the Engine visits a thread, the thread's state always changes to a non-READY state.

A thread may remain indefinitely in a PENDED state because relevant fields were never resolved.