如何在 Drools 中编写有状态和基于时间的规则？

Question

I'm trying to write a rule based engine using Drools. The rules go like this:

Rule "Alarm": If the status is "Alarm" then send notification right away.

Rule "Warning": If the status is "Warning" then save the device ID in memory and wait for 5 minutes. If another message with the same device ID and status "Resolved" is received within 5 minutes, then cancel this rule. Otherwise, upgrade the status to Alarm and trigger the Rule "Alarm".

Rule "Resolved": If status is "Resolved" and device ID is already in memory, then clear the "Warning" for that device ID ie remove the device ID from memory.

Rule Alarm is easy and is working. But rules Warning and Resolved are obviously not working in my code. How to write such rules in Drools?

Here's the code I've written so far:

DroolsTest.java

package com.sample;

import org.drools.KnowledgeBase;
import org.drools.KnowledgeBaseFactory;
import org.drools.builder.KnowledgeBuilder;
import org.drools.builder.KnowledgeBuilderError;
import org.drools.builder.KnowledgeBuilderErrors;
import org.drools.builder.KnowledgeBuilderFactory;
import org.drools.builder.ResourceType;
import org.drools.io.ResourceFactory;
import org.drools.logger.KnowledgeRuntimeLogger;
import org.drools.logger.KnowledgeRuntimeLoggerFactory;
import org.drools.runtime.StatefulKnowledgeSession;

public class DroolsTest {

    public static final void main(String[] args) {
        try {
            // load up the knowledge base
            KnowledgeBase kbase = readKnowledgeBase();
            StatefulKnowledgeSession ksession = kbase.newStatefulKnowledgeSession();
            KnowledgeRuntimeLogger logger = KnowledgeRuntimeLoggerFactory.newFileLogger(ksession, "test");
            // go !
            MachineMessage[] messages = new MachineMessage[4]; 
            messages[0] = new MachineMessage("1", "Warning");
            messages[1] = new MachineMessage("2", "Alarm");
            messages[2] = new MachineMessage("3", "Alarm");
            messages[3] = new MachineMessage("1", "Resolved");

            for(int i = 0; i < messages.length; i++)
            {
                ksession.insert(messages[i]);
                ksession.fireAllRules();
                Thread.sleep(9000);
            }
            logger.close();
        } catch (Throwable t) {
            t.printStackTrace();
        }
    }

    private static KnowledgeBase readKnowledgeBase() throws Exception {
        KnowledgeBuilder kbuilder = KnowledgeBuilderFactory.newKnowledgeBuilder();
        kbuilder.add(ResourceFactory.newClassPathResource("Sample.drl"), ResourceType.DRL);
        KnowledgeBuilderErrors errors = kbuilder.getErrors();
        if (errors.size() > 0) {
            for (KnowledgeBuilderError error: errors) {
                System.err.println(error);
            }
            throw new IllegalArgumentException("Could not parse knowledge.");
        }
        KnowledgeBase kbase = KnowledgeBaseFactory.newKnowledgeBase();
        kbase.addKnowledgePackages(kbuilder.getKnowledgePackages());
        return kbase;
    }

    public static class MachineMessage {

        private String deviceID;
        private String status;

        public MachineMessage() { }
        public MachineMessage(String deviceID, String status) {
            this.deviceID = deviceID;
            this.status = status;           
        }

        public String getDeviceID() {
            return this.deviceID;
        }

        public void setDeviceID(String deviceID) {
            this.deviceID = deviceID;
        }

        public String getStatus() {
            return this.status;
        }

        public void setStatus(String status) {
            this.status = status;
        }
    }
}

Sample.drl

package com.sample

import com.sample.DroolsTest.MachineMessage;

rule "Alarm"
when
    m : MachineMessage( status == "Alarm", myDeviceID : deviceID )
then
    System.out.println("Send notification to technician, device in Alarm state. Device ID: " +  myDeviceID );
end

rule "Warning"
when
    m : MachineMessage( status == "Warning", myDeviceID : deviceID )
then
    System.out.println("Wait for 5 minutes, device in Warning state. Device ID: " + myDeviceID);        
    // if we get Machine Message with the same device ID and Resolved status in 5 minutes then 
    // we'll not trigger the following line
    m.setStatus("Alarm");
    update(m);
end

rule "Resolved"
when
    m : MachineMessage( status == "Resolved", myDeviceID : deviceID )
then
    System.out.println("Device has resolved its warning state. Device ID: " + myDeviceID);

end

The actual output I'm getting is this:

    Wait for 5 minutes, device in Warning state. Device ID: 1
    Send notification to technician, device in Alarm state. Device ID: 1
    Send notification to technician, device in Alarm state. Device ID: 2
    Send notification to technician, device in Alarm state. Device ID: 3
    Device has resolved its warning state. Device ID: 1

Whereas I wanted:

    Wait for 5 minutes, device in Warning state. Device ID: 1
    Send notification to technician, device in Alarm state. Device ID: 2
    Send notification to technician, device in Alarm state. Device ID: 3
    Device has resolved its warning state. Device ID: 1

Answer 1

As per the Drools documentation on CEP Events are supposed to be immutable:

As a best practice, the application is allowed to populate un-populated event attributes (to enrich the event with inferred data), but already populated attributes should never be changed.

So clearly rule "Warning" is already failing the 1st requirement of CEP by the way it's proposed by OP (mutating the field "status").

A potential suggestion would be along the lines of:

rule "Warning"
when
    m : MachineMessage( status == "Warning", myDeviceID : deviceID )
    not( MachineMessage( status == "Resolved", deviceID == m.deviceID, this after[0s, 30s] m) )
then
    System.out.println("I waited for 5 minutes, device in Warning state not followed by Resolved. Device ID: " + myDeviceID);        
    insert(new MachineMessage(m.deviceID, "Alarm"));
end

which would NOT mutate any valorized fields of existing Events, and would actually make CEP knowledge inference trying to match the business requirements described by OP.

In general the CEP example in the Drools manual of sprinkler/fire alarms seems to align with the business requirements described by OP.