Decision model services: Removal from IBM Operational Decision Manager on Cloud

For assistance in converting your decision model services, contact your IBM representative or create a help desk ticket: Let’s troubleshoot. The process includes the Decision Model Object (DMO) converter, a tool created by IBM specifically for converting decision model services.

Note: You cannot convert a decision service back into a decision modeling service.

Provided by IBM, the converter is delivered as a compressed archive, named dmo-rfl-converter.zip. The archive contains the following files:

• run.sh

• run.bat (For Windows. For other operating systems, contact support.)

• README  (Contains the instructions for using the converter.)

• dmo-rfl-converter.jar

To apply the converter:

• Ensure that JAVA_HOME points to Java 17.
• Use run.bat or run.sh with the following parameters:

./run.sh -i <path_to_decision_model_project> -o <path_to_decision_service_project> -m <mode>

• path_to_decision_model_project: The path to the decision model project (exported as archive ) that is to be converted.
• path_to_decision_service_project: The path to the resultant decision service project. The converted decision service can be imported by Operational Decision Manager on Cloud 2023.04 Rule Designer, which uses Java 8. 
• mode: Its values can be Flat or Layer.

Converter limitations

The DMO converter does not handle the following features. You must convert these features manually, or you can reach out to support for assistance.

Modeling capabilities

• Hit policies: If multiple rules determine values for the same decision variable, a hit policy specifies how to determine the final value of the decision variable. The policy First simply chooses the first value. The policies Sum, Min, Max, and Collect aggregate the values. The policy Sequential permits each applicable rule to determine a new value for the decision variable that is based on its previous value. It corresponds to the way rules are evaluated in rule tasks. When converting a decision model, only use the Sequential hit policy.

• Default rules: If no rule determines a value for a decision variable, a default rule provides a fallback value. In some cases, default rules might be replaced by an initialization of the decision variable. However, an initial value might impact the results of the rules, whereas a default rule has no impact on those results.

Object model

• Time periods, calendar durations, and time zones in the boot business object model (BOM): These parameters are not supported in standard decision services.

• Mapping of temporal boot BOM types to java.time types: Temporal types such as dates are mapped to java.time types in decision models, and java.util.Date in standard decision services.

• BOM property de.generated: Java classes are automatically generated for types that are defined in a data model. In standard decision models, these Java classes must be written manually.

• Temporal BAL operators: Rules in decision models might use a rich set of temporal operators such as adding a calendar duration to a time point (see Dates and times). Among these operations only the comparison operators “is after“, “is after or the same as“, “is before“, “is before or the same as“, “is after … and before”, “is after … and on or before”, “is on or after … and before”, “is on or after … and on or before” (see Tables 1, 6, 9, 10, 11, 12 of Section “Time point functions”) and the testing functions “is on <day of week> at <time>”, “is in <month> <year>“, “is on <date & time>“, “is at <time>“, “is in <month>“, “is on <day of week>“, and “is in <year>“ (see Tables 3, 7 of Section “Time point functions”) are supported in standard decision projects. The other operations require the development of custom Java methods.

• Rules that have guards protecting them against undefined input variables: Rules in decision models might use an implicit condition of the form “<variable> is not null”. If required, those conditions need to be added manually in standard decision services.

• Implicit construction of owner object in attribute setters (automatic instantiation for object navigation): When a rule in a decision model assigns a value to an attribute of an object and this object has not been created, it is automatically created. In standard decision projects, executing such a rule encounters a Java exception. Therefore, it is necessary to add actions to such a rule in order to ensure that setter methods are only called for objects that have been created.

• Implicit deep copy to avoid side effects on input variables: Rules in decision models must not modify values in decision variables. This might happen because of side effects. For example, the input variable and decision variables mighty both have the type “Address”. If the address value of the input variable is assigned to the decision variable, and then the value of the street attribute is modified, the value of the street attribute of the input variable is modified as well. Decision models use a sophisticated method to add copy instructions in order to avoid these side effects. This method is not needed in standard decision projects as they permit modifications of all variables. However, it should be noted that the converted project might behave differently than the decision model.