Using Internal data (data in excel) along with passing data from Java to the openrules engine

[indranil nanda]

Hi Team,

Is there any way to use internal data (data in excel) along with passing data from Java to the openrules engine.

Looks like in UserManual Page 34 there is a reference for this. But I am unable to locate standard example "DecisionLoan".

Please see the example below.

So my issue is, if business user next day want to add new "Eligibility Z" (in a new DT), I do not want to change in Java side as from Java side the concern is only with "Recommended Prevent Strategies" which is just a string (as of now).

And business user can add new item under PreventStrategy in glossary and Datatype and the java code should not be impacted.

DT DeciosnForRecommendation1
If	If	Conclusion
Var1	Var2	Eligibility for X
A	B	Eligible for X1, X2
DT DeciosnForRecommendation2
If	If	Conclusion
Var3	Var4	Eligibility for Y
C	D	Eligible for Y1, Y2
DecisionTable CustomerPreventStrategyDecisionTable
Action
Recommended Prevent Strategies
::= ${Eligibility for X} + "__ "+$Eligibility for Y}

 

 

Glossary glossary
Variable	Business Concept	Attribute
Var1	Customer	var1
Var2		var2
Var3		var3
Var4		var4
Var5		var5
Var6		var6
Eligibility for X	PreventStrategy	eligibilityForX
Eligibility for Y		eligibilityForY
Recommended Prevent Strategies	Response	recommendedPreventStrategies
Datatype PreventStrategy
String	eligibilityForX
String	eligibilityForY
Data PreventStrategy preventStrategy
eligibilityForX	eligibilityForY
Eligibility for X	Eligibility for Y
?	?

 

DecisionObject decisionObjects
Business Concept	Business Object
Customer	:=  decision.get("Customer")
Response	:=  decision.get("PreventResponse")
PreventStrategy	:= preventStrategy[0]

 

 

 

 

 In Java side:

                Response preventResponse = new Response();

decision.put("Customer", customer);

decision.put("PreventResponse", preventResponse);

decision.execute();

It is deployed using springBoot.

Now the scenario is

1st call: Customer attributes does not satisfy any Decision Table  - it is returning-- ?__? which is correct.

2nd call: Customer attribute satisfy 1st DT, response: Eligible for X1, X2__? which is correct.

3rd call: Customer attributes does not satisfy any Decision Table  - it is returning-- Eligible for X1, X2__? which is wrong.

So for all consecutive calls it is returning same.

Another scenario:

1st call: Customer attribute satisfy 1st DT, response: Eligible for X1, X2__? which is correct.

2nd call: Customer attributes does not satisfy any Decision Table  - it is returning-- Eligible for X1, X2__? which is wrong.

I tried removing table  Data PreventStrategy and passing preventStrategy (instead of preventStrategy[0]) in BusinessObject table but then getting

Caused by: org.openl.syntax.SyntaxErrorException: Error: Field not found: preventStrategy

Invalid Code Fragment:

=========================================

 preventStrategy

 ^^^^^^^^^^^^^^^

It will be really helpful if you can point the mistake.

Regards

Indranil

[OpenRules Support]

The example was renamed to DecisionLoanPreQualification - see "openrules.dmn". 

Excel table of types Datatype and Data should be used only for testing purposes and never in a production environment. The life-cycle of Data objects defined in Excel should start and end with ONE decision's run. When you switch from a test-mode to a production, the Glossary table and all decision tables will stay the same, but you will change the table "decisionObjects" to feed your already tested model with real data coming as Java objects. You may read the chapter 3 of the new book  about an integration with IT. So, if you build a real-world application, you probably keep your actual data somewhere in a data warehouse and want to pass it to an OpenRules-based decision service as Java objects. The service will make changes in these objects in accordance with the business logic presented in Excel, and will save the results back to the objects. 

[indranil nanda]

Hi,

I understand the data we will get in real time. But that is for "deciding" variable. In this case Customer data.

But my question was what about the interim data that decision engine might use.

Cause Prevent Strategy should be used in Excel only.

From Java side we will get only the concatenated string.

From Java side I am not concerned about if business user add one new strategy or remove one.

[OpenRules Support]

Indranil,

You are making a valid point, and you actually can define and use intermediate variables in Excel. It will work fine in a single user environment. However, if you want to deploy your rule service at a server and give allow multiple concurrent users to access this service, you will have two choices:

1) use the same instance of OpenRulesEngine to serve multiple concurrent requests

2) use different instances of OpenRulesEngine for different concurrent requests.

In the first mode you cannot use the intermediate decision variables defined in Excel Data tables as they all will be shared by concurrent requests and it could lead to a mess in your production environment. But you can use them in the second mode. Obviously, the first mode will work much faster to compare with the second one. So, it is your decision where to keep your intermediate variables: in Java or in Excel. Usually, Excel is sufficient for a test mode. However, we recommend our customers after a decision model is tested to move all intermediate variables into a simple Java bean. This solution will always be thread-safe even if today you don't plan to use your service in a multi-threaded environment.

Support 

[indranil nanda]

Great. I understand now.

However I have one doubt regarding different instances of OpenRulesEngine.

My understanding is when we are instantiating Decision 

Decision decision = new Decision(decisionName,fileName);

It will always return same instance unless main Decision.xls file is changed (last saved timestamp changed).

So to implement your next point

               "2) use different instances of OpenRulesEngine for different concurrent requests."

I have to do decision.clear() after decision.execute()

Please correct me if I am wrong.

Thanks.

[OpenRules Support]

To implement 2) you may re-create a new decision for each http-request by using 

Decision decision = new Decision(decisionName,fileName);

...

decision.execute();

But keep in mind the this is an expensive operation as all related rules will be downloaded from excel and interpreted again.

Contrary, to implement 1) you create an instance"engine" of OpenRulesEngine only once when you start your server:

OpenRulesEngine engine = new OpenRulesEngine(fileName);

Then for every http-request you may call a very quick Decision's constructor 

Decision decision = new Decision(decisionName,engine);

Hope it helps to clarify your questions and available options.

Support

[indranil nanda]

Thanks again.

But in this case will I loose the functionality of changing decisions on the fly, (without restarting the server)

I understand it may not be a good idea to change decisions on the fly in Prod environment.

Regards

Indranil

[indranil nanda]

Also as per:

http://openrules.com/pdf/OpenRulesUserManual.WebService.pdf  P5-6

only one Decision instance is created (Singleton pattern)

Not one for each request.

Got little confused here.

Thanks

Indranil

[OpenRules Support]

If you choose the mode 2), you will create a new instance of OpenRulesEngine and download the latest rules for every request, so you will get "hot deployment" automatically. But if you have too many rules, the mode 2) can become too expensive. So, we strongly recommend using the first mode.

[indranil nanda]

Thanks for your help. Really appreciate this.

Just to iterate to make my understanding clear:

In PROD env where we do not need "hot deployment" of rules we can use one Decision instance (singleton) in the lifetime of the application as mentioned in 

http://openrules.com/pdf/OpenRulesUserManual.WebService.pdf (Page5-6)

And for each web service request just pass the new Request(Customer) and new Response object to the Decision instance. Page-9

And the Decision engine will take care for concurrent requuets.

[OpenRules Support]

Correct!