Designing a decision service using DRL rules

The following procedure is a generic overview of creating data objects. It is not specific to a particular business process.

The when part of the rule contains the conditions that must be met to execute an action. For example, if a bank requires loan applicants to have over 21 years of age, then the when condition of an Underage rule would be Applicant( age < 21 ). Conditions consist of a series of stated patterns and constraints, with optional bindings and other supported DRL elements, based on the available data objects in the package.

The then part of the rule contains the actions to be performed when the conditional part of the rule has been met. For example, when a loan applicant is under 21 years old, the then action of an Underage rule would be setApproved( false ), declining the loan because the applicant is under age. Actions execute consequences based on the rule conditions and on available data objects in the package.

You can use Red Hat JBoss Developer Studio to create DRL files with rules and integrate the files with your Red Hat Decision Manager decision service. This method of creating DRL rules is helpful if you already use Red Hat Developer Studio for your decision service and want to continue with the same work flow. If you do not already use this method, then the Decision Central interface of Red Hat Decision Manager is recommended for creating DRL files and other rule assets.

To integrate the new rule assets with an existing project in Red Hat Decision Manager, you can compile the new project as a knowledge JAR (KJAR) and add it as a dependency in the pom.xml file of the project in Decision Central. To access the project pom.xml file in Decision Central, you can select any existing asset in the project and then in the Project Explorer menu on the left side of the screen, click the Customize View gear icon and select Repository View → pom.xml.

You can use Java objects to create DRL files with rules and integrate the objects with your Red Hat Decision Manager decision service. This method of creating DRL rules is helpful if you already use external Java objects for your decision service and want to continue with the same work flow. If you do not already use this method, then the Decision Central interface of Red Hat Decision Manager is recommended for creating DRL files and other rule assets.

To integrate the new rule assets with an existing project in Red Hat Decision Manager, you can compile the new Java project as a knowledge JAR (KJAR) and add it as a dependency in the pom.xml file of the project in Decision Central. To access the project pom.xml file in Decision Central, you can select any existing asset in the project and then in the Project Explorer menu on the left side of the screen, click the Customize View gear icon and select Repository View → pom.xml.

You can use Maven archetypes to create DRL files with rules and integrate the archetypes with your Red Hat Decision Manager decision service. This method of creating DRL rules is helpful if you already use external Maven archetypes for your decision service and want to continue with the same work flow. If you do not already use this method, then the Decision Central interface of Red Hat Decision Manager is recommended for creating DRL files and other rule assets.

To integrate the new rule assets with an existing project in Red Hat Decision Manager, you can compile the new Maven project as a knowledge JAR (KJAR) and add it as a dependency in the pom.xml file of the project in Decision Central. To access the project pom.xml file in Decision Central, you can select any existing asset in the project and then in the Project Explorer menu on the left side of the screen, click the Customize View gear icon and select Repository View → pom.xml.

Executable rule models are embeddable models that provide a Java-based representation of a rule set for execution at build time. The executable model is a more efficient alternative to the standard asset packaging in Red Hat Decision Manager and enables KIE containers and KIE bases to be created more quickly, especially when you have large lists of DRL (Drools Rule Language) files and other Red Hat Decision Manager assets. The model is low level and enables you to provide all necessary execution information, such as the lambda expressions for the index evaluation.

Executable rule models provide the following specific advantages for your projects:

...
then
  $input.setNo13Count(functions.sumOf(new Object[]{$no1Count_1, $no2Count_1, $no3Count_1, ..., $no13Count_1}).intValue());
  $input.getFirings().add("fired");
  update($input);

You can import Red Hat Decision Manager example decisions into your integrated development environment (IDE) and execute them to explore how the rules and code function. You can use these examples to better understand Red Hat Decision Manager decision engine capabilities or use them as a reference for the decisions that you define in your own Red Hat Decision Manager projects.

The Hello World example decision set demonstrates how to insert objects into the Red Hat Decision Manager decision engine working memory, how to match the objects using rules, and how to configure logging to trace the internal activity of the engine.

The following is an overview of the Hello World example:

In the Hello World example, a KIE session is generated to enable rule execution. All rules require a KIE session for execution.

KieServices ks = KieServices.Factory.get(); 1
KieContainer kc = ks.getKieClasspathContainer(); 2
KieSession ksession = kc.newKieSession("HelloWorldKS"); 3

Red Hat Decision Manager has an event model that exposes internal engine activity. Two default debug listeners, DebugAgendaEventListener and DebugWorkingMemoryEventListener, print debug event information to the System.err output. The KieRuntimeLogger provides execution auditing, the result of which you can view in a graphical viewer.

// Set up listeners.
ksession.addEventListener( new DebugAgendaEventListener() );
ksession.addEventListener( new DebugRuleRuntimeEventListener() );

// Set up a file-based audit logger.
KieRuntimeLogger logger = KieServices.get().getLoggers().newFileLogger( ksession, "./target/helloworld" );

// Set up a ThreadedFileLogger so that the audit view reflects events while debugging.
KieRuntimeLogger logger = ks.getLoggers().newThreadedFileLogger( ksession, "./target/helloworld", 1000 );

The logger is a specialized implementation built on the Agenda and RuleRuntime listeners. When the engine has finished executing, logger.close() is called.

The example creates a single Message object with the message "Hello World", inserts the status HELLO into the KieSession, executes rules with fireAllRules().

// Insert facts into the KIE session.
final Message message = new Message();
message.setMessage( "Hello World" );
message.setStatus( Message.HELLO );
ksession.insert( message );

// Fire the rules.
ksession.fireAllRules();

Rule execution uses a data model to pass data as inputs and outputs to the KieSession. The data model in this example has two fields: the message, which is a String, and the status, which can be HELLO or GOODBYE.

public static class Message {
    public static final int HELLO   = 0;
    public static final int GOODBYE = 1;

    private String          message;
    private int             status;
    ...
}

The two rules are located in the file src/main/resources/org/drools/examples/helloworld/HelloWorld.drl.

The when condition of the "Hello World" rule states that the rule is activated for each Message object inserted into the KIE session that has the status Message.HELLO. Additionally, two variable bindings are created: the variable message is bound to the message attribute and the variable m is bound to the matched Message object itself.

The then action of the rule is written using the MVEL expression language, as declared by the rule dialect attribute. After printing the content of the bound variable message to System.out, the rule changes the values of the message and status attributes of the Message object bound to m. The rule uses the MVEL modify statement to apply a block of assignments in one statement and to notify the engine of the changes at the end of the block.

rule "Hello World"
      dialect "mvel"
  when
    m : Message( status == Message.HELLO, message : message )
  then
    System.out.println( message );
    modify ( m ) { message = "Goodbye cruel world",
                   status = Message.GOODBYE };
end

The "Good Bye" rule, which specifies the java dialect, is similar to the "Hello World" rule except that it matches Message objects that have the status Message.GOODBYE.

rule "Good Bye"
      dialect "java"
  when
    Message( status == Message.GOODBYE, message : message )
  then
    System.out.println( message );
end

To execute the example, run the org.drools.examples.helloworld.HelloWorldExample class as a Java application in your IDE. The rule writes to System.out, the debug listener writes to System.err, and the audit logger creates a log file in target/helloworld.log.

Hello World
Goodbye cruel world

==>[ActivationCreated(0): rule=Hello World;
                   tuple=[fid:1:1:org.drools.examples.helloworld.HelloWorldExample$Message@17cec96]]
[ObjectInserted: handle=[fid:1:1:org.drools.examples.helloworld.HelloWorldExample$Message@17cec96];
                 object=org.drools.examples.helloworld.HelloWorldExample$Message@17cec96]
[BeforeActivationFired: rule=Hello World;
                   tuple=[fid:1:1:org.drools.examples.helloworld.HelloWorldExample$Message@17cec96]]
==>[ActivationCreated(4): rule=Good Bye;
                   tuple=[fid:1:2:org.drools.examples.helloworld.HelloWorldExample$Message@17cec96]]
[ObjectUpdated: handle=[fid:1:2:org.drools.examples.helloworld.HelloWorldExample$Message@17cec96];
                old_object=org.drools.examples.helloworld.HelloWorldExample$Message@17cec96;
                new_object=org.drools.examples.helloworld.HelloWorldExample$Message@17cec96]
[AfterActivationFired(0): rule=Hello World]
[BeforeActivationFired: rule=Good Bye;
                   tuple=[fid:1:2:org.drools.examples.helloworld.HelloWorldExample$Message@17cec96]]
[AfterActivationFired(4): rule=Good Bye]

To better understand the execution flow of this example, you can load the audit log file from target/helloworld.log into your IDE debug view or Audit View, if available (for example, in Window → Show View in some IDEs).

In this example, the Audit view shows that the object is inserted, which creates an activation for the "Hello World" rule. The activation is then executed, which updates the Message object and causes the "Good Bye" rule to activate. Finally, the "Good Bye" rule is executed. When you select an event in the Audit View, the origin event, which is the "Activation created" event in this example, is highlighted in green.

Figure 7.3. Hello World example Audit View

The State example decision set demonstrates how the decision engine uses forward chaining and any changes to facts in the working memory to resolve execution conflicts for rules in a sequence. The example focuses on resolving conflicts through salience values or through agenda groups that you can define in rules.

The following is an overview of the State example:

A forward-chaining rule system is a data-driven system that starts with a fact in the working memory of the decision engine and reacts to changes to that fact. When objects are inserted into working memory, any rule conditions that become true as a result of the change are scheduled for execution by the agenda.

In contrast, a backward-chaining rule system is a goal-driven system that starts with a conclusion that the decision engine attempts to satisfy, often using recursion. If the system cannot reach the conclusion or goal, it searches for subgoals, which are conclusions that complete part of the current goal. The system continues this process until either the initial conclusion is satisfied or all subgoals are satisfied.

The decision engine in Red Hat Decision Manager uses both forward and backward chaining to evaluate rules.

The following diagram illustrates how the decision engine evaluates rules using forward chaining overall with a backward-chaining segment in the logic flow:

Figure 7.4. Rule evaluation logic using forward and backward chaining

In the State example, each State class has fields for its name and its current state (see the class org.drools.examples.state.State). The following states are the two possible states for each object:

public class State {
    public static final int NOTRUN   = 0;
    public static final int FINISHED = 1;

    private final PropertyChangeSupport changes =
        new PropertyChangeSupport( this );

    private String name;
    private int    state;

    ... setters and getters go here...
}

The State example contains two versions of the same example to resolve rule execution conflicts:

Both versions of the state example involve four State objects: A, B, C, and D. Initially, their states are set to NOTRUN, which is the default value for the constructor that the example uses.

State example using salience

The StateExampleUsingSalience version of the State example uses salience values in rules to resolve rule execution conflicts. Rules with a higher salience value are given higher priority when ordered in the activation queue.

The example inserts each State instance into the KIE session and then calls fireAllRules().

final State a = new State( "A" );
final State b = new State( "B" );
final State c = new State( "C" );
final State d = new State( "D" );

ksession.insert( a );
ksession.insert( b );
ksession.insert( c );
ksession.insert( d );

ksession.fireAllRules();

// Dispose KIE session if stateful (not required if stateless).
ksession.dispose();

To execute the example, run the org.drools.examples.state.StateExampleUsingSalience class as a Java application in your IDE.

After the execution, the following output appears in the IDE console window:

A finished
B finished
C finished
D finished

Four rules are present.

First, the "Bootstrap" rule fires, setting A to state FINISHED, which then causes B to change its state to FINISHED. Objects C and D are both dependent on B, causing a conflict that is resolved by the salience values.

To better understand the execution flow of this example, you can load the audit log file from target/state.log into your IDE debug view or Audit View, if available (for example, in Window → Show View in some IDEs).

In this example, the Audit View shows that the assertion of the object A in the state NOTRUN activates the "Bootstrap" rule, while the assertions of the other objects have no immediate effect.

Figure 7.5. Salience State example Audit View

rule "Bootstrap"
  when
    a : State(name == "A", state == State.NOTRUN )
  then
    System.out.println(a.getName() + " finished" );
    a.setState( State.FINISHED );
end

The execution of the "Bootstrap" rule changes the state of A to FINISHED, which activates rule "A to B".

rule "A to B"
  when
    State(name == "A", state == State.FINISHED )
    b : State(name == "B", state == State.NOTRUN )
  then
    System.out.println(b.getName() + " finished" );
    b.setState( State.FINISHED );
end

The execution of rule "A to B" changes the state of B to FINISHED, which activates both rules "B to C" and "B to D", placing their activations onto the engine agenda.

rule "B to C"
    salience 10
  when
    State(name == "B", state == State.FINISHED )
    c : State(name == "C", state == State.NOTRUN )
  then
    System.out.println(c.getName() + " finished" );
    c.setState( State.FINISHED );
end

rule "B to D"
  when
    State(name == "B", state == State.FINISHED )
    d : State(name == "D", state == State.NOTRUN )
  then
    System.out.println(d.getName() + " finished" );
    d.setState( State.FINISHED );
end

From this point on, both rules may fire and, therefore, the rules are in conflict. The conflict resolution strategy enables the engine agenda to decide which rule to fire. Rule "B to C" has the higher salience value (10 versus the default salience value of 0), so it fires first, modifying object C to state FINISHED.

The Audit View in your IDE shows the modification of the State object in the rule "A to B", which results in two activations being in conflict.

You can also use the Agenda View in your IDE to investigate the state of the engine agenda. In this example, the Agenda View shows the breakpoint in the rule "A to B" and the state of the agenda with the two conflicting rules. Rule "B to D" fires last, modifying object D to state FINISHED.

Figure 7.6. Salience State example Agenda View

State example using agenda groups

The StateExampleUsingAgendaGroups version of the State example uses agenda groups in rules to resolve rule execution conflicts. Agenda groups enable you to partition the engine agenda to provide more execution control over groups of rules. By default, all rules are in the agenda group MAIN. You can use the agenda-group attribute to specify a different agenda group for the rule.

Initially, a working memory has its focus on the agenda group MAIN. Rules in an agenda group only fire when the group receives the focus. You can set the focus either by using the method setFocus() or the rule attribute auto-focus. The auto-focus attribute enables the rule to be given a focus automatically for its agenda group when the rule is matched and activated.

In this example, the auto-focus attribute enables rule "B to C" to fire before "B to D".

rule "B to C"
    agenda-group "B to C"
    auto-focus true
  when
    State(name == "B", state == State.FINISHED )
    c : State(name == "C", state == State.NOTRUN )
  then
    System.out.println(c.getName() + " finished" );
    c.setState( State.FINISHED );
    kcontext.getKnowledgeRuntime().getAgenda().getAgendaGroup( "B to D" ).setFocus();
end

The rule "B to C" calls setFocus() on the agenda group "B to D", enabling its active rules to fire, which then enables the rule "B to D" to fire.

rule "B to D"
    agenda-group "B to D"
  when
    State(name == "B", state == State.FINISHED )
    d : State(name == "D", state == State.NOTRUN )
  then
    System.out.println(d.getName() + " finished" );
    d.setState( State.FINISHED );
end

To execute the example, run the org.drools.examples.state.StateExampleUsingAgendaGroups class as a Java application in your IDE.

After the execution, the following output appears in the IDE console window (same as the salience version of the State example):

A finished
B finished
C finished
D finished

Dynamic facts in the State example

Another notable concept in this State example is the use of dynamic facts, based on objects that implement a PropertyChangeListener object. In order for the engine to see and react to changes of fact properties, the application must notify the engine that changes occurred. You can configure this communication explicitly in the rules by using the modify statement, or implicitly by specifying that the facts implement the PropertyChangeSupport interface as defined by the JavaBeans specification.

This example demonstrates how to use the PropertyChangeSupport interface to avoid the need for explicit modify statements in the rules. To make use of this interface, ensure that your facts implement PropertyChangeSupport in the same way that the class org.drools.example.State implements it, and then use the following code in the DRL rule file to configure the engine to listen for property changes on those facts:

declare type State
  @propertyChangeSupport
end

When you use PropertyChangeListener objects, each setter must implement additional code for the notification. For example, the following setter for state is in the class org.drools.examples:

public void setState(final int newState) {
    int oldState = this.state;
    this.state = newState;
    this.changes.firePropertyChange( "state",
                                     oldState,
                                     newState );
}

The Fibonacci example decision set demonstrates how the decision engine uses recursion to resolve execution conflicts for rules in a sequence. The example focuses on resolving conflicts through salience values that you can define in rules.

The following is an overview of the Fibonacci example:

The Fibonacci Numbers form a sequence starting with 0 and 1. The next Fibonacci number is obtained by adding the two preceding Fibonacci numbers: 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, 233, 377, 610, 987, 1597, 2584, 4181, 6765, 10946, and so on.

The Fibonacci example uses the single fact class Fibonacci with the following two fields:

The sequence field indicates the position of the object in the Fibonacci number sequence. The value field shows the value of that Fibonacci object for that sequence position, where -1 indicates a value that still needs to be computed.

public static class Fibonacci {
    private int  sequence;
    private long value;

    public Fibonacci( final int sequence ) {
        this.sequence = sequence;
        this.value = -1;
    }

    ... setters and getters go here...
}

To execute the example, run the org.drools.examples.fibonacci.FibonacciExample class as a Java application in your IDE.

After the execution, the following output appears in the IDE console window:

recurse for 50
recurse for 49
recurse for 48
recurse for 47
...
recurse for 5
recurse for 4
recurse for 3
recurse for 2
1 == 1
2 == 1
3 == 2
4 == 3
5 == 5
6 == 8
...
47 == 2971215073
48 == 4807526976
49 == 7778742049
50 == 12586269025

To achieve this behavior in Java, the example inserts a single Fibonacci object with a sequence field of 50. The example then uses a recursive rule to insert the other 49 Fibonacci objects.

Instead of implementing the PropertyChangeSupport interface to use dynamic facts, this example uses the MVEL dialect modify keyword to enable a block setter action and notify the engine of changes.

ksession.insert( new Fibonacci( 50 ) );
ksession.fireAllRules();

This example uses the following three rules:

The rule "Recurse" matches each asserted Fibonacci object with a value of -1, creating and asserting a new Fibonacci object with a sequence of one less than the currently matched object. Each time a Fibonacci object is added while the one with a sequence field equal to 1 does not exist, the rule re-matches and fires again. The not conditional element is used to stop the rule matching once you have all 50 Fibonacci objects in memory. The rule also has a salience value because you need to have all 50 Fibonacci objects asserted before you execute the "Bootstrap" rule.

rule "Recurse"
    salience 10
  when
    f : Fibonacci ( value == -1 )
    not ( Fibonacci ( sequence == 1 ) )
  then
    insert( new Fibonacci( f.sequence - 1 ) );
    System.out.println( "recurse for " + f.sequence );
end

To better understand the execution flow of this example, you can load the audit log file from target/fibonacci.log into your IDE debug view or Audit View, if available (for example, in Window → Show View in some IDEs).

In this example, the Audit View shows the original assertion of the Fibonacci object with a sequence field of 50, done from Java code. From there on, the Audit View shows the continual recursion of the rule, where each asserted Fibonacci object causes the "Recurse" rule to become activated and to fire again.

Figure 7.7. Rule "Recurse" in Audit View

When a Fibonacci object with a sequence field of 2 is asserted, the "Bootstrap" rule is matched and activated along with the "Recurse" rule. Notice the multiple restrictions on field sequence that test for equality with 1 or 2:

rule "Bootstrap"
  when
    f : Fibonacci( sequence == 1 || == 2, value == -1 ) // multi-restriction
  then
    modify ( f ){ value = 1 };
    System.out.println( f.sequence + " == " + f.value );
end

You can also use the Agenda View in your IDE to investigate the state of the engine agenda. The "Bootstrap" rule does not fire yet because the "Recurse" rule has a higher salience value.

Figure 7.8. Rules "Recurse" and "Bootstrap" in Agenda View 1

When a Fibonacci object with a sequence of 1 is asserted, the "Bootstrap" rule is matched again, causing two activations for this rule. The "Recurse" rule does not match and activate because the not conditional element stops the rule matching as soon as a Fibonacci object with a sequence of 1 exists.

Figure 7.9. Rules "Recurse" and "Bootstrap" in Agenda View 2

The "Bootstrap" rule sets the objects with a sequence of 1 and 2 to a value of 1. Now that you have two Fibonacci objects with values not equal to -1, the "Calculate" rule is able to match.

At this point in the example, nearly 50 Fibonacci objects exist in the working memory. You need to select a suitable triple to calculate each of their values in turn. If you use three Fibonacci patterns in a rule without field constraints to confine the possible cross products, the result would be 50x49x48 possible combinations, leading to about 125,000 possible rule firings, most of them incorrect.

The "Calculate" rule uses field constraints to evaluate the three Fibonacci patterns in the correct order. This technique is called cross-product matching.

The first pattern finds any Fibonacci object with a value != -1 and binds both the pattern and the field. The second Fibonacci object does the same thing, but adds an additional field constraint to ensure that its sequence is greater by one than the Fibonacci object bound to f1. When this rule fires for the first time, you know that only sequences 1 and 2 have values of 1, and the two constraints ensure that f1 references sequence 1 and that f2 references sequence 2.

The final pattern finds the Fibonacci object with a value equal to -1 and with a sequence one greater than f2.

At this point in the example, three Fibonacci objects are correctly selected from the available cross products, and you can calculate the value for the third Fibonacci object that is bound to f3.

rule "Calculate"
  when
    // Bind f1 and s1.
    f1 : Fibonacci( s1 : sequence, value != -1 )
    // Bind f2 and v2, refer to bound variable s1.
    f2 : Fibonacci( sequence == (s1 + 1), v2 : value != -1 )
    // Bind f3 and s3, alternative reference of f2.sequence.
    f3 : Fibonacci( s3 : sequence == (f2.sequence + 1 ), value == -1 )
  then
    // Note the various referencing techniques.
    modify ( f3 ) { value = f1.value + v2 };
    System.out.println( s3 + " == " + f3.value );
end

The modify statement updates the value of the Fibonacci object bound to f3. This means that you now have another new Fibonacci object with a value not equal to -1, which allows the "Calculate" rule to re-match and calculate the next Fibonacci number.

The debug view or Audit View of your IDE shows how the firing of the last "Bootstrap" rule modifies the Fibonacci object, enabling the "Calculate" rule to match, which then modifies another Fibonacci object that enables the "Calculate" rule to match again. This process continues until the value is set for all Fibonacci objects.

Figure 7.10. Rules in Audit View

The Pet Store example decision set demonstrates how to use agenda groups and global variables in rules and how to integrate Red Hat Decision Manager rules with a graphical user interface (GUI), in this case a Swing-based desktop application. The example also demonstrates how to use callbacks to interact with a running decision engine to update the GUI based on changes in the working memory at run time.

The following is an overview of the Pet Store example:

In the Pet Store example, the sample PetStoreExample.java class defines the following principal classes (in addition to several classes to handle Swing events):

Much of the Java code in this example is either plain JavaBean or Swing based. For more information about Swing components, see the Java tutorial on Creating a GUI with JFC/Swing.

Rule execution behavior in the Pet Store example

Unlike other example decision sets where the facts are asserted and fired immediately, the Pet Store example does not execute the rules until more facts are gathered based on user interaction. The example executes rules through a PetStoreUI object, created by a constructor, that accepts the Vector object stock for collecting the products. The example then uses an instance of the CheckoutCallback class containing the rule base that was previously loaded.

// KieServices is the factory for all KIE services.
KieServices ks = KieServices.Factory.get();

// Create a KIE container on the class path.
KieContainer kc = ks.getKieClasspathContainer();

// Create the stock.
Vector<Product> stock = new Vector<Product>();
stock.add( new Product( "Gold Fish", 5 ) );
stock.add( new Product( "Fish Tank", 25 ) );
stock.add( new Product( "Fish Food", 2 ) );

// A callback is responsible for populating the working memory and for firing all rules.
PetStoreUI ui = new PetStoreUI( stock,
                                new CheckoutCallback( kc ) );
ui.createAndShowGUI();

The Java code that fires the rules is in the CheckoutCallBack.checkout() method. This method is triggered when the user clicks Checkout in the UI.

public String checkout(JFrame frame, List<Product> items) {
    Order order = new Order();

    // Iterate through list and add to cart.
    for ( Product p: items ) {
        order.addItem( new Purchase( order, p ) );
    }

    // Add the JFrame to the ApplicationData to allow for user interaction.

    // From the KIE container, a KIE session is created based on
    // its definition and configuration in the META-INF/kmodule.xml file.
    KieSession ksession = kcontainer.newKieSession("PetStoreKS");

    ksession.setGlobal( "frame", frame );
    ksession.setGlobal( "textArea", this.output );

    ksession.insert( new Product( "Gold Fish", 5 ) );
    ksession.insert( new Product( "Fish Tank", 25 ) );
    ksession.insert( new Product( "Fish Food", 2 ) );

    ksession.insert( new Product( "Fish Food Sample", 0 ) );

    ksession.insert( order );

    // Execute rules.
    ksession.fireAllRules();

    // Return the state of the cart
    return order.toString();
}

The example code passes two elements into the CheckoutCallBack.checkout() method. One element is the handle for the JFrame Swing component surrounding the output text frame, found at the bottom of the GUI. The second element is a list of order items, which comes from the TableModel that stores the information from the Table area at the upper-right section of the GUI.

The for loop transforms the list of order items coming from the GUI into the Order JavaBean, also contained in the file PetStoreExample.java.

In this case, the rule is firing in a stateless KIE session because all of the data is stored in Swing components and is not executed until the user clicks Checkout in the UI. Each time the user clicks Checkout, the content of the list is moved from the Swing TableModel into the KIE session working memory and is then executed with the ksession.fireAllRules() method.

Within this code, there are nine calls to KieSession. The first of these creates a new KieSession from the KieContainer (the example passed in this KieContainer from the CheckoutCallBack class in the main() method). The next two calls pass in the two objects that hold the global variables in the rules: the Swing text area and the Swing frame used for writing messages. More inserts put information on products into the KieSession, as well as the order list. The final call is the standard fireAllRules().

Pet Store rule file imports, global variables, and Java functions

The PetStore.drl file contains the standard package and import statements to make various Java classes available to the rules. The rule file also includes global variables to be used within the rules, defined as frame and textArea. The global variables hold references to the Swing components JFrame and JTextArea components that were previously passed on by the Java code that called the setGlobal() method. Unlike standard variables in rules, which expire as soon as the rule has fired, global variables retain their value for the lifetime of the KIE session. This means the contents of these global variables are available for evaluation on all subsequent rules.

package org.drools.examples;

import org.kie.api.runtime.KieRuntime;
import org.drools.examples.petstore.PetStoreExample.Order;
import org.drools.examples.petstore.PetStoreExample.Purchase;
import org.drools.examples.petstore.PetStoreExample.Product;
import java.util.ArrayList;
import javax.swing.JOptionPane;

import javax.swing.JFrame;

global JFrame frame
global javax.swing.JTextArea textArea

The PetStore.drl file also contains two functions that the rules in the file use:

function void doCheckout(JFrame frame, KieRuntime krt) {
        Object[] options = {"Yes",
                            "No"};

        int n = JOptionPane.showOptionDialog(frame,
                                             "Would you like to checkout?",
                                             "",
                                             JOptionPane.YES_NO_OPTION,
                                             JOptionPane.QUESTION_MESSAGE,
                                             null,
                                             options,
                                             options[0]);

       if (n == 0) {
            krt.getAgenda().getAgendaGroup( "checkout" ).setFocus();
       }
}

function boolean requireTank(JFrame frame, KieRuntime krt, Order order, Product fishTank, int total) {
        Object[] options = {"Yes",
                            "No"};

        int n = JOptionPane.showOptionDialog(frame,
                                             "Would you like to buy a tank for your " + total + " fish?",
                                             "Purchase Suggestion",
                                             JOptionPane.YES_NO_OPTION,
                                             JOptionPane.QUESTION_MESSAGE,
                                             null,
                                             options,
                                             options[0]);

       System.out.print( "SUGGESTION: Would you like to buy a tank for your "
                           + total + " fish? - " );

       if (n == 0) {
             Purchase purchase = new Purchase( order, fishTank );
             krt.insert( purchase );
             order.addItem( purchase );
             System.out.println( "Yes" );
       } else {
            System.out.println( "No" );
       }
       return true;
}

The two functions perform the following actions:

Pet Store rules with agenda groups

Most of the rules in the Pet Store example use agenda groups to control rule execution. Agenda groups allow you to partition the engine agenda to provide more execution control over groups of rules. By default, all rules are in the agenda group MAIN. You can use the agenda-group attribute to specify a different agenda group for the rule.

Initially, a working memory has its focus on the agenda group MAIN. Rules in an agenda group only fire when the group receives the focus. You can set the focus either by using the method setFocus() or the rule attribute auto-focus. The auto-focus attribute enables the rule to be given a focus automatically for its agenda group when the rule is matched and activated.

The Pet Store example uses the following agenda groups for rules:

For example, the sample rule "Explode Cart" uses the "init" agenda group to ensure that it has the option to fire and insert shopping cart items into the KIE session working memory:

// Insert each item in the shopping cart into the working memory.
rule "Explode Cart"
    agenda-group "init"
    auto-focus true
    salience 10
    dialect "java"
  when
    $order : Order( grossTotal == -1 )
    $item : Purchase() from $order.items
  then
    insert( $item );
    kcontext.getKnowledgeRuntime().getAgenda().getAgendaGroup( "show items" ).setFocus();
    kcontext.getKnowledgeRuntime().getAgenda().getAgendaGroup( "evaluate" ).setFocus();
end

This rule matches against all orders that do not yet have their grossTotal calculated. The execution loops for each purchase item in that order.

The rule uses the following features related to its agenda group:

The "show items" agenda group contains only one rule, "Show Items". For each purchase in the order currently in the KIE session working memory, the rule logs details to the text area at the bottom of the GUI, based on the textArea variable defined in the rule file.

rule "Show Items"
    agenda-group "show items"
    dialect "mvel"
  when
    $order : Order( )
    $p : Purchase( order == $order )
  then
   textArea.append( $p.product + "\n");
end

The "evaluate" agenda group also gains focus from the "Explode Cart" rule. This agenda group contains two rules, "Free Fish Food Sample" and "Suggest Tank", which are executed in that order.

// Free fish food sample when users buy a goldfish if they did not already buy
// fish food and do not already have a fish food sample.
rule "Free Fish Food Sample"
    agenda-group "evaluate" 1
    dialect "mvel"
  when
    $order : Order()
    not ( $p : Product( name == "Fish Food") && Purchase( product == $p ) ) 2
    not ( $p : Product( name == "Fish Food Sample") && Purchase( product == $p ) ) 3
    exists ( $p : Product( name == "Gold Fish") && Purchase( product == $p ) ) 4
    $fishFoodSample : Product( name == "Fish Food Sample" );
  then
    System.out.println( "Adding free Fish Food Sample to cart" );
    purchase = new Purchase($order, $fishFoodSample);
    insert( purchase );
    $order.addItem( purchase );
end

The rule "Free Fish Food Sample" fires only if all of the following conditions are true:

If the order facts meet all of these requirements, then a new product is created (Fish Food Sample) and is added to the order in working memory.

// Suggest a fish tank if users buy more than five goldfish and
// do not already have a tank.
rule "Suggest Tank"
    agenda-group "evaluate"
    dialect "java"
  when
    $order : Order()
    not ( $p : Product( name == "Fish Tank") && Purchase( product == $p ) ) 1
    ArrayList( $total : size > 5 ) from collect( Purchase( product.name == "Gold Fish" ) ) 2
    $fishTank : Product( name == "Fish Tank" )
  then
    requireTank(frame, kcontext.getKieRuntime(), $order, $fishTank, $total);
end

The rule "Suggest Tank" fires only if the following conditions are true:

When the rule fires, it calls the requireTank() function defined in the rule file. This function displays a dialog that asks the user if she or he wants to buy a fish tank. If the user does, a new fish tank Product is added to the order list in the working memory. When the rule calls the requireTank() function, the rule passes the frame global variable so that the function has a handle for the Swing GUI.

The "do checkout" rule in the Pet Store example has no agenda group and no when conditions, so the rule is always executed and considered part of the default MAIN agenda group.

rule "do checkout"
    dialect "java"
  when
  then
    doCheckout(frame, kcontext.getKieRuntime());
end

When the rule fires, it calls the doCheckout() function defined in the rule file. This function displays a dialog that asks the user if she or he wants to check out. If the user does, the focus is set to the checkout agenda group, enabling rules in that group to (potentially) fire. When the rule calls the doCheckout() function, the rule passes the frame global variable so that the function has a handle for the Swing GUI.

The "checkout" agenda group contains three rules for processing the order checkout and applying any discounts: "Gross Total", "Apply 5% Discount", and "Apply 10% Discount".

rule "Gross Total"
    agenda-group "checkout"
    dialect "mvel"
  when
    $order : Order( grossTotal == -1)
    Number( total : doubleValue ) from accumulate( Purchase( $price : product.price ),
                                                              sum( $price ) )
  then
    modify( $order ) { grossTotal = total }
    textArea.append( "\ngross total=" + total + "\n" );
end

rule "Apply 5% Discount"
    agenda-group "checkout"
    dialect "mvel"
  when
    $order : Order( grossTotal >= 10 && < 20 )
  then
    $order.discountedTotal = $order.grossTotal * 0.95;
    textArea.append( "discountedTotal total=" + $order.discountedTotal + "\n" );
end

rule "Apply 10% Discount"
    agenda-group "checkout"
    dialect "mvel"
  when
    $order : Order( grossTotal >= 20 )
  then
    $order.discountedTotal = $order.grossTotal * 0.90;
    textArea.append( "discountedTotal total=" + $order.discountedTotal + "\n" );
end

If the user has not already calculated the gross total, the Gross Total accumulates the product prices into a total, puts this total into the KIE session, and displays it through the Swing JTextArea using the textArea global variable.

If the gross total is between 10 and 20 (currency units), the "Apply 5% Discount" rule calculates the discounted total, adds it to the KIE session, and displays it in the text area.

If the gross total is not less than 20, the "Apply 10% Discount" rule calculates the discounted total, adds it to the KIE session, and displays it in the text area.

Pet Store example execution

Similar to other Red Hat Decision Manager decision examples, you execute the Pet Store example by running the org.drools.examples.petstore.PetStoreExample class as a Java application in your IDE.

When you execute the Pet Store example, the Pet Store Demo GUI window appears. This window displays a list of available products (upper left), an empty list of selected products (upper right), Checkout and Reset buttons (middle), and an empty system messages area (bottom).

Figure 7.11. Pet Store example GUI after launch

The following events occurred in this example to establish this execution behavior:

You can click on various products from the list to explore the UI setup:

Figure 7.12. Explore the Pet Store example GUI

No rules code has been fired yet. The UI uses Swing code to detect user mouse clicks and add selected products to the TableModel object for display in the upper-right corner of the UI. This example illustrates the Model-View-Controller design pattern.

When you click Checkout, the rules are then fired in the following way:

You can add more System.out calls to demonstrate this flow of events in your IDE console:

Adding free Fish Food Sample to cart
SUGGESTION: Would you like to buy a tank for your 6 fish? - Yes

The Honest Politician example decision set demonstrates the concept of truth maintenance with logical insertions and the use of salience in rules.

The following is an overview of the Honest Politician example:

The basic premise of the Honest Politician example is that an object can only exist while a statement is true. A rule consequence can logically insert an object with the insertLogical() method. This means the object remains in the KIE session working memory as long as the rule that logically inserted it remains true. When the rule is no longer true, the object is automatically retracted.

In this example, rule execution causes a group of politicians to change from being honest to being dishonest as a result of a corrupt corporation. As each politician is evaluated, they start out with their honesty attribute being set to true, but a rule fires that makes the politicians no longer honest. As they switch their state from being honest to dishonest, they are then removed from the working memory. The rule salience notifies the engine how to prioritize any rules that have a salience defined for them, otherwise utilizing the default salience value of 0. Rules with a higher salience value are given higher priority when ordered in the activation queue.

Politician and Hope classes

The sample class Politician in the example is configured for an honest politician. The Politician class is made up of a String item name and a Boolean item honest:

public class Politician {
    private String name;
    private boolean honest;
    ...
}

The Hope class determines if a Hope object exists. This class has no meaningful members, but is present in the working memory as long as society has hope.

public class Hope {

    public Hope() {

    }
  }

Rule definitions for politician honesty

In the Honest Politician example, when at least one honest politician exists in the working memory, the "We have an honest Politician" rule logically inserts a new Hope object. As soon as all politicians become dishonest, the Hope object is automatically retracted. This rule has a salience attribute with a value of 10 to ensure that it fires before any other rule, because at that stage the "Hope is Dead" rule is true.

rule "We have an honest Politician"
    salience 10
  when
    exists( Politician( honest == true ) )
  then
    insertLogical( new Hope() );
end

As soon as a Hope object exists, the "Hope Lives" rule matches and fires. This rule also has a salience value of 10 so that it takes priority over the "Corrupt the Honest" rule.

rule "Hope Lives"
    salience 10
  when
    exists( Hope() )
  then
    System.out.println("Hurrah!!! Democracy Lives");
end

Initially, four honest politicians exist so this rule has four activations, all in conflict. Each rule fires in turn, corrupting each politician so that they are no longer honest. When all four politicians have been corrupted, no politicians have the property honest == true. The rule "We have an honest Politician" is no longer true and the object it logically inserted (due to the last execution of new Hope()) is automatically retracted.

rule "Corrupt the Honest"
  when
    politician : Politician( honest == true )
    exists( Hope() )
  then
    System.out.println( "I'm an evil corporation and I have corrupted " + politician.getName() );
    modify ( politician ) { honest = false };
end

With the Hope object automatically retracted through the truth maintenance system, the conditional element not applied to Hope is no longer true so that the "Hope is Dead" rule matches and fires.

rule "Hope is Dead"
  when
    not( Hope() )
  then
    System.out.println( "We are all Doomed!!! Democracy is Dead" );
end

Example execution and audit trail

In the HonestPoliticianExample.java class, the four politicians with the honest state set to true are inserted for evaluation against the defined business rules:

public static void execute( KieContainer kc ) {
        KieSession ksession = kc.newKieSession("HonestPoliticianKS");

        final Politician p1 = new Politician( "President of Umpa Lumpa", true );
        final Politician p2 = new Politician( "Prime Minster of Cheeseland", true );
        final Politician p3 = new Politician( "Tsar of Pringapopaloo", true );
        final Politician p4 = new Politician( "Omnipotence Om", true );

        ksession.insert( p1 );
        ksession.insert( p2 );
        ksession.insert( p3 );
        ksession.insert( p4 );

        ksession.fireAllRules();

        ksession.dispose();
    }

To execute the example, run the org.drools.examples.honestpolitician.HonestPoliticianExample class as a Java application in your IDE.

After the execution, the following output appears in the IDE console window:

Hurrah!!! Democracy Lives
I'm an evil corporation and I have corrupted President of Umpa Lumpa
I'm an evil corporation and I have corrupted Prime Minster of Cheeseland
I'm an evil corporation and I have corrupted Tsar of Pringapopaloo
I'm an evil corporation and I have corrupted Omnipotence Om
We are all Doomed!!! Democracy is Dead

The output shows that, while there is at least one honest politician, democracy lives. However, as each politician is corrupted by some corporation, all politicians become dishonest, and democracy is dead.

To better understand the execution flow of this example, you can modify the HonestPoliticianExample.java class to include a RuleRuntime listener and an audit logger to view execution details:

package org.drools.examples.honestpolitician;

import org.kie.api.KieServices;
import org.kie.api.event.rule.DebugAgendaEventListener; 1
import org.kie.api.event.rule.DebugRuleRuntimeEventListener;
import org.kie.api.runtime.KieContainer;
import org.kie.api.runtime.KieSession;

public class HonestPoliticianExample {

    /**
     * @param args
     */
    public static void main(final String[] args) {
    	KieServices ks = KieServices.Factory.get(); 2
    	//ks = KieServices.Factory.get();
        KieContainer kc = KieServices.Factory.get().getKieClasspathContainer();
        System.out.println(kc.verify().getMessages().toString());
        //execute( kc );
        execute( ks, kc); 3
    }

    public static void execute( KieServices ks, KieContainer kc ) { 4
        KieSession ksession = kc.newKieSession("HonestPoliticianKS");

        final Politician p1 = new Politician( "President of Umpa Lumpa", true );
        final Politician p2 = new Politician( "Prime Minster of Cheeseland", true );
        final Politician p3 = new Politician( "Tsar of Pringapopaloo", true );
        final Politician p4 = new Politician( "Omnipotence Om", true );

        ksession.insert( p1 );
        ksession.insert( p2 );
        ksession.insert( p3 );
        ksession.insert( p4 );

        // The application can also setup listeners 5
        ksession.addEventListener( new DebugAgendaEventListener() );
        ksession.addEventListener( new DebugRuleRuntimeEventListener() );

        // Set up a file-based audit logger.
        ks.getLoggers().newFileLogger( ksession, "./target/honestpolitician" ); 6

        ksession.fireAllRules();

        ksession.dispose();
    }

}

When you run the Honest Politician with this modified logging capability, you can load the audit log file from target/honestpolitician.log into your IDE debug view or Audit View, if available (for example, in Window → Show View in some IDEs).

In this example, the Audit View shows the flow of executions, insertions, and retractions as defined in the example classes and rules:

Figure 7.15. Honest Politician example Audit View

When the first politician is inserted, two activations occur. The rule "We have an honest Politician" is activated only one time for the first inserted politician because it uses an exists conditional element, which matches when at least one politician is inserted. The rule "Hope is Dead" is also activated at this stage because the Hope object is not yet inserted. The rule "We have an honest Politician" fires first because it has a higher salience value than the rule "Hope is Dead", and inserts the Hope object (highlighted in green). The insertion of the Hope object activates the rule "Hope Lives" and deactivates the rule "Hope is Dead". The insertion also activates the rule "Corrupt the Honest" for each inserted honest politician. The rule "Hope Lives" is executed and prints "Hurrah!!! Democracy Lives".

Next, for each politician, the rule "Corrupt the Honest" fires, printing "I’m an evil corporation and I have corrupted X", where X is the name of the politician, and modifies the politician honesty value to false. When the last honest politician is corrupted, Hope is automatically retracted by the truth maintenance system (highlighted in blue). The green highlighted area shows the origin of the currently selected blue highlighted area. After the Hope fact is retracted, the rule "Hope is dead" fires, printing "We are all Doomed!!! Democracy is Dead".

The Sudoku example decision set, based on the popular number puzzle Sudoku, demonstrates how to use rules in Red Hat Decision Manager to find a solution in a large potential solution space based on various constraints. This example also shows how to integrate Red Hat Decision Manager rules into a graphical user interface (GUI), in this case a Swing-based desktop application, and how to use callbacks to interact with a running decision engine to update the GUI based on changes in the working memory at run time.

The following is an overview of the Sudoku example:

Sudoku is a logic-based number placement puzzle. The objective is to fill a 9x9 grid so that each column, each row, and each of the nine 3x3 zones contains the digits from 1 to 9 only one time. The puzzle setter provides a partially completed grid and the puzzle solver’s task is to complete the grid with these constraints.

The general strategy to solve the problem is to ensure that when you insert a new number, it must be unique in its particular 3x3 zone, row, and column. This Sudoku example decision set uses Red Hat Decision Manager rules to solve Sudoku puzzles from a range of difficulty levels, and to attempt to resolve flawed puzzles that contain invalid entries.

Sudoku example execution and interaction

Similar to other Red Hat Decision Manager decision examples, you execute the Sudoku example by running the org.drools.examples.sudoku.SudokuExample class as a Java application in your IDE.

When you execute the Sudoku example, the Drools Sudoku Example GUI window appears. This window contains an empty grid, but the program comes with various grids stored internally that you can load and solve.

Click File → Samples → Simple to load one of the examples. Notice that all buttons are disabled until a grid is loaded.

Figure 7.16. Sudoku example GUI after launch

When you load the Simple example, the grid is filled according to the puzzle’s initial state.

Figure 7.17. Sudoku example GUI after loading Simple sample

Choose from the following options:

The Sudoku example includes a deliberately broken sample file that the rules defined in the example can resolve.

Click File → Samples → !DELIBERATELY BROKEN! to load the broken sample. The grid starts with some issues, for example, the value 5 appears two times in the first row, which is not allowed.

Figure 7.19. Broken Sudoku example initial state

Click Solve to apply the solving rules to this invalid grid. The associated solving rules in the Sudoku example detect the issues in the sample and attempts to solve the puzzle as far as possible. This process does not complete and leaves some cells empty.

The solving rule activity is displayed in the IDE console window:

cell [0,8]: 5 has a duplicate in row 0
cell [0,0]: 5 has a duplicate in row 0
cell [6,0]: 8 has a duplicate in col 0
cell [4,0]: 8 has a duplicate in col 0
Validation complete.

Figure 7.20. Broken sample solution attempt

The sample Sudoku files labeled Hard are more complex and the solving rules might not be able to solve them. The unsuccessful solution attempt is displayed in the IDE console window:

Validation complete.
...
Sorry - can't solve this grid.

The rules that work to solve the broken sample implement standard solving techniques based on the sets of values that are still candidates for a cell. For example, if a set contains a single value, then this is the value for the cell. For a single occurrence of a value in one of the groups of nine cells, the rules insert a fact of type Setting with the solution value for some specific cell. This fact causes the elimination of this value from all other cells in any of the groups the cell belongs to and the value is retracted.

Other rules in the example reduce the permissible values for some cells. The rules "naked pair", "hidden pair in row", "hidden pair in column", and "hidden pair in square" eliminate possibilities but do not establish solutions. The rules "X-wings in rows", "`X-wings in columns"`, "intersection removal row", and "intersection removal column" perform more sophisticated eliminations.

Sudoku example classes

The package org.drools.examples.sudoku.swing contains the following core set of classes that implement a framework for Sudoku puzzles:

The package org.drools.examples.sudoku contains the following core set of classes that implement the elementary Cell object and its various aggregations:

Sudoku validation rules (validate.drl)

The validate.drl file in the Sudoku example contains validation rules that detect duplicate numbers in cell groups. They are combined in a "validate" agenda group that enables the rules to be explicitly activated after a user loads the puzzle.

The when conditions of the three rules "duplicate in cell …​" all function in the following ways:

rule "duplicate in cell row"
  when
    $c: Cell( $v: value != null )
    $cr: CellRow( cells contains $c )
    exists Cell( this != $c, value == $v, cellRow == $cr )
  then
    System.out.println( "cell " + $c.toString() + " has a duplicate in row " + $cr.getNumber() );
end

rule "duplicate in cell col"
  when
    $c: Cell( $v: value != null )
    $cc: CellCol( cells contains $c )
    exists Cell( this != $c, value == $v, cellCol == $cc )
  then
    System.out.println( "cell " + $c.toString() + " has a duplicate in col " + $cc.getNumber() );
end

rule "duplicate in cell sqr"
  when
    $c: Cell( $v: value != null )
    $cs: CellSqr( cells contains $c )
    exists Cell( this != $c, value == $v, cellSqr == $cs )
  then
    System.out.println( "cell " + $c.toString() + " has duplicate in its square of nine." );
end

The rule "terminate group" is the last to fire. This rule prints a message and stops the sequence.

rule "terminate group"
    salience -100
  when
  then
    System.out.println( "Validation complete." );
    drools.halt();
end

Sudoku solving rules (sudoku.drl)

The sudoku.drl file in the Sudoku example contains three types of rules: one group handles the allocation of a number to a cell, another group detects feasible allocations, and the third group eliminates values from candidate sets.

The rules "set a value", "eliminate a value from Cell", and "retract setting" depend on the presence of a Setting object. The first rule handles the assignment to the cell and the operations for removing the value from the free sets of the three groups of the cell. This group also reduces a counter that, when zero, returns control to the Java application that has called fireUntilHalt().

The purpose of the rule "eliminate a value from Cell" is to reduce the candidate lists of all cells that are related to the newly assigned cell. Finally, when all eliminations have been made, the rule "retract setting" retracts the triggering Setting fact.

// A Setting object is inserted to define the value of a Cell.
// Rule for updating the cell and all cell groups that contain it
rule "set a value"
  when
    // A Setting with row and column number, and a value
    $s: Setting( $rn: rowNo, $cn: colNo, $v: value )

    // A matching Cell, with no value set
    $c: Cell( rowNo == $rn, colNo == $cn, value == null,
              $cr: cellRow, $cc: cellCol, $cs: cellSqr )

    // Count down
    $ctr: Counter( $count: count )
  then
    // Modify the Cell by setting its value.
    modify( $c ){ setValue( $v ) }
    // System.out.println( "set cell " + $c.toString() );
    modify( $cr ){ blockValue( $v ) }
    modify( $cc ){ blockValue( $v ) }
    modify( $cs ){ blockValue( $v ) }
    modify( $ctr ){ setCount( $count - 1 ) }
end

// Rule for removing a value from all cells that are siblings
// in one of the three cell groups
rule "eliminate a value from Cell"
  when
    // A Setting with row and column number, and a value
    $s: Setting( $rn: rowNo, $cn: colNo, $v: value )

    // The matching Cell, with the value already set
    Cell( rowNo == $rn, colNo == $cn, value == $v, $exCells: exCells )

    // For all Cells that are associated with the updated cell
    $c: Cell( free contains $v ) from $exCells
  then
    // System.out.println( "clear " + $v + " from cell " + $c.posAsString()  );
    // Modify a related Cell by blocking the assigned value.
    modify( $c ){ blockValue( $v ) }
end

// Rule for eliminating the Setting fact
rule "retract setting"
  when
    // A Setting with row and column number, and a value
    $s: Setting( $rn: rowNo, $cn: colNo, $v: value )

    // The matching Cell, with the value already set
    $c: Cell( rowNo == $rn, colNo == $cn, value == $v )

    // This is the negation of the last pattern in the previous rule.
    // Now the Setting fact can be safely retracted.
    not( $x: Cell( free contains $v )
         and
         Cell( this == $c, exCells contains $x ) )
  then
    // System.out.println( "done setting cell " + $c.toString() );
    // Discard the Setter fact.
    delete( $s );
    // Sudoku.sudoku.consistencyCheck();
end

Two solving rules detect a situation where an allocation of a number to a cell is possible. The rule "single" fires for a Cell with a candidate set containing a single number. The rule "hidden single" fires when no cell exists with a single candidate, but when a cell exists containing a candidate, this candidate is absent from all other cells in one of the three groups to which the cell belongs. Both rules create and insert a Setting fact.

// Detect a set of candidate values with cardinality 1 for some Cell.
// This is the value to be set.
rule "single"
  when
    // Currently no setting underway
    not Setting()

    // One element in the "free" set
    $c: Cell( $rn: rowNo, $cn: colNo, freeCount == 1 )
  then
    Integer i = $c.getFreeValue();
    if (explain) System.out.println( "single " + i + " at " + $c.posAsString() );
    // Insert another Setter fact.
    insert( new Setting( $rn, $cn, i ) );
end

// Detect a set of candidate values with a value that is the only one
// in one of its groups. This is the value to be set.
rule "hidden single"
  when
    // Currently no setting underway
    not Setting()
    not Cell( freeCount == 1 )

    // Some integer
    $i: Integer()

    // The "free" set contains this number
    $c: Cell( $rn: rowNo, $cn: colNo, freeCount > 1, free contains $i )

    // A cell group contains this cell $c.
    $cg: CellGroup( cells contains $c )
    // No other cell from that group contains $i.
    not ( Cell( this != $c, free contains $i ) from $cg.getCells() )
  then
    if (explain) System.out.println( "hidden single " + $i + " at " + $c.posAsString() );
    // Insert another Setter fact.
    insert( new Setting( $rn, $cn, $i ) );
end

Rules from the largest group, either individually or in groups of two or three, implement various solving techniques used for solving Sudoku puzzles manually.

The rule "naked pair" detects identical candidate sets of size 2 in two cells of a group. These two values may be removed from all other candidate sets of that group.

// A "naked pair" is two cells in some cell group with their sets of
// permissible values being equal with cardinality 2. These two values
// can be removed from all other candidate lists in the group.
rule "naked pair"
  when
    // Currently no setting underway
    not Setting()
    not Cell( freeCount == 1 )

    // One cell with two candidates
    $c1: Cell( freeCount == 2, $f1: free, $r1: cellRow, $rn1: rowNo, $cn1: colNo, $b1: cellSqr )

    // The containing cell group
    $cg: CellGroup( freeCount > 2, cells contains $c1 )

    // Another cell with two candidates, not the one we already have
    $c2: Cell( this != $c1, free == $f1 /*** , rowNo >= $rn1, colNo >= $cn1 ***/ ) from $cg.cells

    // Get one of the "naked pair".
    Integer( $v: intValue ) from $c1.getFree()

    // Get some other cell with a candidate equal to one from the pair.
    $c3: Cell( this != $c1 && != $c2, freeCount > 1, free contains $v ) from $cg.cells
  then
    if (explain) System.out.println( "remove " + $v + " from " + $c3.posAsString() + " due to naked pair at " + $c1.posAsString() + " and " + $c2.posAsString() );
    // Remove the value.
    modify( $c3 ){ blockValue( $v ) }
end

The three rules "hidden pair in …​" functions similarly to the rule "naked pair". These rules detect a subset of two numbers in exactly two cells of a group, with neither value occurring in any of the other cells of the group. This means that all other candidates can be eliminated from the two cells harboring the hidden pair.

// If two cells within the same cell group contain candidate sets with more than
// two values, with two values being in both of them but in none of the other
// cells, then we have a "hidden pair". We can remove all other candidates from
// these two cells.
rule "hidden pair in row"
  when
    // Currently no setting underway
    not Setting()
    not Cell( freeCount == 1 )

    // Establish a pair of Integer facts.
    $i1: Integer()
    $i2: Integer( this > $i1 )

    // Look for a Cell with these two among its candidates. (The upper bound on
    // the number of candidates avoids a lot of useless work during startup.)
    $c1: Cell( $rn1: rowNo, $cn1: colNo, freeCount > 2 && < 9, free contains $i1 && contains $i2, $cellRow: cellRow )

    // Get another one from the same row, with the same pair among its candidates.
    $c2: Cell( this != $c1, cellRow == $cellRow, freeCount > 2, free contains $i1 && contains $i2 )

    // Ascertain that no other cell in the group has one of these two values.
    not( Cell( this != $c1 && != $c2, free contains $i1 || contains $i2 ) from $cellRow.getCells() )
  then
    if( explain) System.out.println( "hidden pair in row at " + $c1.posAsString() + " and " + $c2.posAsString() );
    // Set the candidate lists of these two Cells to the "hidden pair".
    modify( $c1 ){ blockExcept( $i1, $i2 ) }
    modify( $c2 ){ blockExcept( $i1, $i2 ) }
end

rule "hidden pair in column"
  when
    not Setting()
    not Cell( freeCount == 1 )

    $i1: Integer()
    $i2: Integer( this > $i1 )
    $c1: Cell( $rn1: rowNo, $cn1: colNo, freeCount > 2 && < 9, free contains $i1 && contains $i2, $cellCol: cellCol )
    $c2: Cell( this != $c1, cellCol == $cellCol, freeCount > 2, free contains $i1 && contains $i2 )
    not( Cell( this != $c1 && != $c2, free contains $i1 || contains $i2 ) from $cellCol.getCells() )
  then
    if (explain) System.out.println( "hidden pair in column at " + $c1.posAsString() + " and " + $c2.posAsString() );
    modify( $c1 ){ blockExcept( $i1, $i2 ) }
    modify( $c2 ){ blockExcept( $i1, $i2 ) }
end

rule "hidden pair in square"
  when
    not Setting()
    not Cell( freeCount == 1 )

    $i1: Integer()
    $i2: Integer( this > $i1 )
    $c1: Cell( $rn1: rowNo, $cn1: colNo, freeCount > 2 && < 9, free contains $i1 && contains $i2,
               $cellSqr: cellSqr )
    $c2: Cell( this != $c1, cellSqr == $cellSqr, freeCount > 2, free contains $i1 && contains $i2 )
    not( Cell( this != $c1 && != $c2, free contains $i1 || contains $i2 ) from $cellSqr.getCells() )
  then
    if (explain) System.out.println( "hidden pair in square " + $c1.posAsString() + " and " + $c2.posAsString() );
    modify( $c1 ){ blockExcept( $i1, $i2 ) }
    modify( $c2 ){ blockExcept( $i1, $i2 ) }
end

Two rules deal with "X-wings" in rows and columns. When only two possible cells for a value exist in each of two different rows (or columns) and these candidates lie also in the same columns (or rows), then all other candidates for this value in the columns (or rows) can be eliminated. When you follow the pattern sequence in one of these rules, notice how the conditions that are conveniently expressed by words such as same or only result in patterns with suitable constraints or that are prefixed with not.

rule "X-wings in rows"
  when
    not Setting()
    not Cell( freeCount == 1 )

    $i: Integer()
    $ca1: Cell( freeCount > 1, free contains $i,
                $ra: cellRow, $rano: rowNo,         $c1: cellCol,        $c1no: colNo )
    $cb1: Cell( freeCount > 1, free contains $i,
                $rb: cellRow, $rbno: rowNo > $rano,      cellCol == $c1 )
    not( Cell( this != $ca1 && != $cb1, free contains $i ) from $c1.getCells() )

    $ca2: Cell( freeCount > 1, free contains $i,
                cellRow == $ra, $c2: cellCol,       $c2no: colNo > $c1no )
    $cb2: Cell( freeCount > 1, free contains $i,
                cellRow == $rb,      cellCol == $c2 )
    not( Cell( this != $ca2 && != $cb2, free contains $i ) from $c2.getCells() )

    $cx: Cell( rowNo == $rano || == $rbno, colNo != $c1no && != $c2no,
               freeCount > 1, free contains $i )
  then
    if (explain) {
        System.out.println( "X-wing with " + $i + " in rows " +
            $ca1.posAsString() + " - " + $cb1.posAsString() +
            $ca2.posAsString() + " - " + $cb2.posAsString() + ", remove from " + $cx.posAsString() );
    }
    modify( $cx ){ blockValue( $i ) }
end

rule "X-wings in columns"
  when
    not Setting()
    not Cell( freeCount == 1 )

    $i: Integer()
    $ca1: Cell( freeCount > 1, free contains $i,
                $c1: cellCol, $c1no: colNo,         $ra: cellRow,        $rano: rowNo )
    $ca2: Cell( freeCount > 1, free contains $i,
                $c2: cellCol, $c2no: colNo > $c1no,      cellRow == $ra )
    not( Cell( this != $ca1 && != $ca2, free contains $i ) from $ra.getCells() )

    $cb1: Cell( freeCount > 1, free contains $i,
                cellCol == $c1, $rb: cellRow,  $rbno: rowNo > $rano )
    $cb2: Cell( freeCount > 1, free contains $i,
                cellCol == $c2,      cellRow == $rb )
    not( Cell( this != $cb1 && != $cb2, free contains $i ) from $rb.getCells() )

    $cx: Cell( colNo == $c1no || == $c2no, rowNo != $rano && != $rbno,
               freeCount > 1, free contains $i )
  then
    if (explain) {
        System.out.println( "X-wing with " + $i + " in columns " +
            $ca1.posAsString() + " - " + $ca2.posAsString() +
            $cb1.posAsString() + " - " + $cb2.posAsString() + ", remove from " + $cx.posAsString()  );
    }
    modify( $cx ){ blockValue( $i ) }
end

The two rules "intersection removal …​" are based on the restricted occurrence of some number within one square, either in a single row or in a single column. This means that this number must be in one of those two or three cells of the row or column and can be removed from the candidate sets of all other cells of the group. The pattern establishes the restricted occurrence and then fires for each cell outside of the square and within the same cell file.

rule "intersection removal column"
  when
    not Setting()
    not Cell( freeCount == 1 )

    $i: Integer()
    // Occurs in a Cell
    $c: Cell( free contains $i, $cs: cellSqr, $cc: cellCol )
    // Does not occur in another cell of the same square and a different column
    not Cell( this != $c, free contains $i, cellSqr == $cs, cellCol != $cc )

    // A cell exists in the same column and another square containing this value.
    $cx: Cell( freeCount > 1, free contains $i, cellCol == $cc, cellSqr != $cs )
  then
    // Remove the value from that other cell.
    if (explain) {
        System.out.println( "column elimination due to " + $c.posAsString() +
                            ": remove " + $i + " from " + $cx.posAsString() );
    }
    modify( $cx ){ blockValue( $i ) }
end

rule "intersection removal row"
  when
    not Setting()
    not Cell( freeCount == 1 )

    $i: Integer()
    // Occurs in a Cell
    $c: Cell( free contains $i, $cs: cellSqr, $cr: cellRow )
    // Does not occur in another cell of the same square and a different row.
    not Cell( this != $c, free contains $i, cellSqr == $cs, cellRow != $cr )

    // A cell exists in the same row and another square containing this value.
    $cx: Cell( freeCount > 1, free contains $i, cellRow == $cr, cellSqr != $cs )
  then
    // Remove the value from that other cell.
    if (explain) {
        System.out.println( "row elimination due to " + $c.posAsString() +
                            ": remove " + $i + " from " + $cx.posAsString() );
    }
    modify( $cx ){ blockValue( $i ) }
end

These rules are sufficient for many but not all Sudoku puzzles. To solve very difficult grids, the rule set requires more complex rules. (Ultimately, some puzzles can be solved only by trial and error.)

The Conway’s Game of Life example decision set, based on the famous cellular automaton by John Conway, demonstrates how to use ruleflow groups in rules to control rule execution. The example also demonstrates how to integrate Red Hat Decision Manager rules with a graphical user interface (GUI), in this case a Swing-based implementation of Conway’s Game of Life.

The following is an overview of the Conway’s Game of Life (Conway) example:

In Conway’s Game of Life, a user interacts with the game by creating an initial configuration or an advanced pattern with defined properties and then observing how the initial state evolves. The objective of the game is to show the development of a population, generation by generation. Each generation results from the preceding one, based on the simultaneous evaluation of all cells.

The following basic rules govern what the next generation looks like:

Any cell that does not meet any of those criteria is left as is for the next generation.

The Conway’s Game of Life example uses Red Hat Decision Manager rules with ruleflow-group attributes to define the pattern implemented in the game. The example also contains a version of the decision set that achieves the same behavior using agenda groups. Agenda groups enable you to partition the engine agenda to provide execution control over groups of rules. By default, all rules are in the agenda group MAIN. You can use the agenda-group attribute to specify a different agenda group for the rule.

This overview does not explore the version of the Conway example using agenda groups. For more information about agenda groups, see the Red Hat Decision Manager example decision sets that specifically address agenda groups.

Conway example execution and interaction

Similar to other Red Hat Decision Manager decision examples, you execute the Conway ruleflow example by running the org.drools.examples.conway.ConwayRuleFlowGroupRun class as a Java application in your IDE.

When you execute the Conway example, the Conway’s Game of Life GUI window appears. This window contains an empty grid, or "arena" where the life simulation takes place. Initially the grid is empty because no live cells are in the system yet.

Figure 7.21. Conway example GUI after launch

Select a predefined pattern from the Pattern drop-down menu and click Next Generation to click through each population generation. Each cell is either alive or dead, where live cells contain a green ball. As the population evolves from the initial pattern, cells live or die relative to neighboring cells, according to the rules of the game.

Figure 7.22. Generation evolution in Conway example

Neighbors include not only cells to the left, right, top, and bottom but also cells that are connected diagonally, so that each cell has a total of eight neighbors. Exceptions are the corner cells, which have only three neighbors, and the cells along the four borders, with five neighbors each.

You can manually intervene to create or kill cells by clicking the cell.

To run through an evolution automatically from the initial pattern, click Start.

Conway example rules with ruleflow groups

The rules in the ConwayRuleFlowGroupRun example use ruleflow groups to control rule execution. A ruleflow group is a group of rules associated by the ruleflow-group rule attribute. These rules can only fire when the group is activated. The group itself can only become active when the elaboration of the ruleflow diagram reaches the node representing the group.

The Conway example uses the following ruleflow groups for rules:

All of the Cell objects are inserted into the KIE session and the "register …​" rules in the ruleflow group "register neighbor" are allowed to execute by the ruleflow process. This group of four rules creates Neighbor relations between some cell and its northeastern, northern, northwestern, and western neighbors.

This relation is bidirectional and handles the other four directions. Border cells do not require any special treatment. These cells are not paired with neighboring cells where there is not any.

By the time all activations have fired for these rules, all cells are related to all their neighboring cells.

rule "register north east"
    ruleflow-group "register neighbor"
  when
    $cell: Cell( $row : row, $col : col )
    $northEast : Cell( row  == ($row - 1), col == ( $col + 1 ) )
  then
    insert( new Neighbor( $cell, $northEast ) );
    insert( new Neighbor( $northEast, $cell ) );
end

rule "register north"
    ruleflow-group "register neighbor"
  when
    $cell: Cell( $row : row, $col : col )
    $north : Cell( row  == ($row - 1), col == $col )
  then
    insert( new Neighbor( $cell, $north ) );
    insert( new Neighbor( $north, $cell ) );
end

rule "register north west"
    ruleflow-group "register neighbor"
  when
    $cell: Cell( $row : row, $col : col )
    $northWest : Cell( row  == ($row - 1), col == ( $col - 1 ) )
  then
    insert( new Neighbor( $cell, $northWest ) );
    insert( new Neighbor( $northWest, $cell ) );
end

rule "register west"
    ruleflow-group "register neighbor"
  when
    $cell: Cell( $row : row, $col : col )
    $west : Cell( row  == $row, col == ( $col - 1 ) )
  then
    insert( new Neighbor( $cell, $west ) );
    insert( new Neighbor( $west, $cell ) );
end

After all the cells are inserted, some Java code applies the pattern to the grid, setting certain cells to Live. Then, when the user clicks Start or Next Generation, the example executes the Generation ruleflow. This ruleflow manages all changes of cells in each generation cycle.

Figure 7.23. Generation ruleflow

The ruleflow process enters the "evaluate" ruleflow group and any active rules in the group can fire. The rules "Kill the …​" and "Give Birth" in this group apply the game rules to birth or kill cells. The example uses the phase attribute to drive the reasoning of the Cell object by specific groups of rules. Typically, the phase is tied to a ruleflow group in the ruleflow process definition.

Notice that the example does not change the state of any Cell objects at this point because it must complete the full evaluation before those changes can be applied. The example sets the cell to a phase that is either Phase.KILL or Phase.BIRTH, which is used later to control actions applied to the Cell object.

rule "Kill The Lonely"
    ruleflow-group "evaluate"
    no-loop
  when
    // A live cell has fewer than 2 live neighbors.
    theCell: Cell( liveNeighbors < 2, cellState == CellState.LIVE,
                   phase == Phase.EVALUATE )
  then
    modify( theCell ){
        setPhase( Phase.KILL );
    }
end

rule "Kill The Overcrowded"
    ruleflow-group "evaluate"
    no-loop
  when
    // A live cell has more than 3 live neighbors.
    theCell: Cell( liveNeighbors > 3, cellState == CellState.LIVE,
                   phase == Phase.EVALUATE )
  then
    modify( theCell ){
        setPhase( Phase.KILL );
    }
end

rule "Give Birth"
    ruleflow-group "evaluate"
    no-loop
  when
    // A dead cell has 3 live neighbors.
    theCell: Cell( liveNeighbors == 3, cellState == CellState.DEAD,
                   phase == Phase.EVALUATE )
  then
    modify( theCell ){
        theCell.setPhase( Phase.BIRTH );
    }
end

After all Cell objects in the grid have been evaluated, the example uses the "reset calculate" rule to clear any activations in the "calculate" ruleflow group. The example then enters a split in the ruleflow that enables the rules "kill" and "birth" to fire, if the ruleflow group is activated. These rules apply the state change.

rule "reset calculate"
    ruleflow-group "reset calculate"
  when
  then
    WorkingMemory wm = drools.getWorkingMemory();
    wm.clearRuleFlowGroup( "calculate" );
end

rule "kill"
    ruleflow-group "kill"
    no-loop
  when
    theCell: Cell( phase == Phase.KILL )
  then
    modify( theCell ){
        setCellState( CellState.DEAD ),
        setPhase( Phase.DONE );
    }
end

rule "birth"
    ruleflow-group "birth"
    no-loop
  when
    theCell: Cell( phase == Phase.BIRTH )
  then
    modify( theCell ){
        setCellState( CellState.LIVE ),
        setPhase( Phase.DONE );
    }
end

At this stage, several Cell objects have been modified with the state changed to either LIVE or DEAD. When a cell becomes live or dead, the example uses the Neighbor relation in the rules "Calculate …​" to iterate over all surrounding cells, increasing or decreasing the liveNeighbor count. Any cell that has its count changed is also set to to the EVALUATE phase to make sure it is included in the reasoning during the evaluation stage of the ruleflow process.

After the live count has been determined and set for all cells, the ruleflow process ends. If the user initially clicked Start, the engine restarts the ruleflow at that point. If the user initially clicked Next Generation, the user can request another generation.

rule "Calculate Live"
    ruleflow-group "calculate"
    lock-on-active
  when
    theCell: Cell( cellState == CellState.LIVE )
    Neighbor( cell == theCell, $neighbor : neighbor )
  then
    modify( $neighbor ){
        setLiveNeighbors( $neighbor.getLiveNeighbors() + 1 ),
        setPhase( Phase.EVALUATE );
    }
end

rule "Calculate Dead"
    ruleflow-group "calculate"
    lock-on-active
  when
    theCell: Cell( cellState == CellState.DEAD )
    Neighbor( cell == theCell, $neighbor : neighbor )
  then
    modify( $neighbor ){
        setLiveNeighbors( $neighbor.getLiveNeighbors() - 1 ),
        setPhase( Phase.EVALUATE );
    }
end

The House of Doom example decision set demonstrates how the decision engine uses backward chaining and recursion to reach defined goals or subgoals in a hierarchical system.

The following is an overview of the House of Doom example:

A backward-chaining rule system is a goal-driven system that starts with a conclusion that the decision engine attempts to satisfy, often using recursion. If the system cannot reach the conclusion or goal, it searches for subgoals, which are conclusions that complete part of the current goal. The system continues this process until either the initial conclusion is satisfied or all subgoals are satisfied.

In contrast, a forward-chaining rule system is a data-driven system that starts with a fact in the working memory of the decision engine and reacts to changes to that fact. When objects are inserted into working memory, any rule conditions that become true as a result of the change are scheduled for execution by the agenda.

The decision engine in Red Hat Decision Manager uses both forward and backward chaining to evaluate rules.

The following diagram illustrates how the decision engine evaluates rules using forward chaining overall with a backward-chaining segment in the logic flow:

Figure 7.24. Rule evaluation logic using forward and backward chaining

The House of Doom example uses rules with various types of queries to find the location of rooms and items within the house. The sample class Location.java contains the item and location elements used in the example. The sample class HouseOfDoomMain.java inserts the items or rooms in their respective locations in the house and executes the rules.

ksession.insert( new Location("Office", "House") );
ksession.insert( new Location("Kitchen", "House") );
ksession.insert( new Location("Knife", "Kitchen") );
ksession.insert( new Location("Cheese", "Kitchen") );
ksession.insert( new Location("Desk", "Office") );
ksession.insert( new Location("Chair", "Office") );
ksession.insert( new Location("Computer", "Desk") );
ksession.insert( new Location("Drawer", "Desk") );

The example rules rely on backward chaining and recursion to determine the location of all items and rooms in the house structure.

The following diagram illustrates the structure of the House of Doom and the items and rooms within it:

Figure 7.25. House of Doom structure

To execute the example, run the org.drools.examples.backwardchaining.HouseOfDoomMain class as a Java application in your IDE.

After the execution, the following output appears in the IDE console window:

go1
Office is in the House
---
go2
Drawer is in the House
---
go3
---
Key is in the Office
---
go4
Chair is in the Office
Desk is in the Office
Key is in the Office
Computer is in the Office
Drawer is in the Office
---
go5
Chair is in Office
Desk is in Office
Drawer is in Desk
Key is in Drawer
Kitchen is in House
Cheese is in Kitchen
Knife is in Kitchen
Computer is in Desk
Office is in House
Key is in Office
Drawer is in House
Computer is in House
Key is in House
Desk is in House
Chair is in House
Knife is in House
Cheese is in House
Computer is in Office
Drawer is in Office
Key is in Desk

All rules in the example have fired to detect the location of all items in the house and to print the location of each in the output.

Recursive query and related rules

A recursive query repeatedly searches through the hierarchy of a data structure for relationships between elements.

In the House of Doom example, the BC-Example.drl file contains an isContainedIn query that most of the rules in the example use to recursively evaluate the house data structure for data inserted into the decision engine:

query isContainedIn( String x, String y )
  Location( x, y; )
  or
  ( Location( z, y; ) and isContainedIn( x, z; ) )
end

The rule "go" prints every string inserted into the system to determine how items are implemented, and the rule "go1" calls the query isContainedIn:

rule "go" salience 10
  when
    $s : String( )
  then
    System.out.println( $s );
end

rule "go1"
  when
    String( this == "go1" )
    isContainedIn("Office", "House"; )
  then
    System.out.println( "Office is in the House" );
end

The example inserts the "go1" string into the engine and activates the "go1" rule to detect that item Office is in the location House:

ksession.insert( "go1" );
ksession.fireAllRules();

go1
Office is in the House

Transitive closure rule

Transitive closure is a relationship between an element contained in a parent element that is multiple levels higher in a hierarchical structure.

The rule "go2" identifies the transitive closure relationship of the Drawer and the House: The Drawer is in the Desk in the Office in the House.

rule "go2"
  when
    String( this == "go2" )
    isContainedIn("Drawer", "House"; )
  then
    System.out.println( "Drawer is in the House" );
end

The example inserts the "go2" string into the engine and activates the "go2" rule to detect that item Drawer is ultimately within the location House:

ksession.insert( "go2" );
ksession.fireAllRules();

go2
Drawer is in the House

The engine determines this outcome based on the following logic:

Reactive query rule

A reactive query searches through the hierarchy of a data structure for relationships between elements and is dynamically updated when elements in the structure are modified.

The rule "go3" functions as a reactive query that detects if a new item Key ever becomes present in the Office by transitive closure: A Key in the Drawer in the Office.

rule "go3"
  when
    String( this == "go3" )
    isContainedIn("Key", "Office"; )
  then
    System.out.println( "Key is in the Office" );
end

The example inserts the "go3" string into the engine and activates the "go3" rule. Initially, this rule is not satisfied because no item Key exists in the house structure, so the rule produces no output.

ksession.insert( "go3" );
ksession.fireAllRules();

go3

The example then inserts a new item Key in the location Drawer, which is in Office. This change satisfies the transitive closure in the "go3" rule and the output is populated accordingly.

ksession.insert( new Location("Key", "Drawer") );
ksession.fireAllRules();

Key is in the Office

This change also adds another level in the structure that the query includes in subsequent recursive searches.

Queries with unbound arguments in rules

A query with one or more unbound arguments returns all undefined (unbound) items within a defined (bound) argument of the query. If all arguments in a query are unbound, then the query returns all items within the scope of the query.

The rule "go4" uses an unbound argument thing to search for all items within the bound argument Office, instead of using a bound argument to search for a specific item in the Office:

rule "go4"
  when
    String( this == "go4" )
    isContainedIn(thing, "Office"; )
  then
    System.out.println( thing + "is in the Office" );
end

The example inserts the "go4" string into the engine and activates the "go4" rule to return all items in the Office:

ksession.insert( "go4" );
ksession.fireAllRules();

go4
Chair is in the Office
Desk is in the Office
Key is in the Office
Computer is in the Office
Drawer is in the Office

The rule "go5" uses both unbound arguments thing and location to search for all items and their locations in the entire House data structure:

rule "go5"
  when
    String( this == "go5" )
    isContainedIn(thing, location; )
  then
    System.out.println(thing + " is in " + location );
end

The example inserts the "go5" string into the engine and activates the "go5" rule to return all items and their locations in the House data structure:

ksession.insert( "go5" );
ksession.fireAllRules();

go5
Chair is in Office
Desk is in Office
Drawer is in Desk
Key is in Drawer
Kitchen is in House
Cheese is in Kitchen
Knife is in Kitchen
Computer is in Desk
Office is in House
Key is in Office
Drawer is in House
Computer is in House
Key is in House
Desk is in House
Chair is in House
Knife is in House
Cheese is in House
Computer is in Office
Drawer is in Office
Key is in Desk