A model driven approach for bridging ILOG Rule Language and RIF

Abstract

Nowadays many companies run their business using Business Rule Management Systems (BRMS), that o er: a clear separation between decision logic and procedural structure; the ability to modify a rulebase set rather than processes and the reusability of rules across applications. All these factors allow a company to quickly react and align its policies to the ever-changing market needs. Despite these advantages, di erent BRMSs can be found on the market, each of them implementing a proprietary business rule language (ex.: JBoss uses Drools, IBM uses Ilog Rule Language, etc.). Rule Interchange Format(RIF) is a W3C open standard aiming at reducing the heterogeneity among business rule languages, which makes rules less reusable and interchangeable. Our work is focused on providing an implementation based on a Model Driven approach for bridging Ilog Rule Language to RIF.

-

The eld of BRMS is characterized by a number of normative, open source, or proprietary systems and languages (Ilog JRules, JBoss Drools, etc.), allowing the expression of various solutions to business problems at a high abstraction level, but with heterogeneous sets of capabilities and languages. Rule Interchange Format (RIF [ 1 ]) and Production Rules Representation (PRR [ 2 ]) are two standard proposals respectively developed at the W3C and at the OMG aiming at providing a level of standardization to the domain. An important di erence between RIF and PRR is that RIF is a standard for distribution of shared rules at runtime, whereas PRR is a standard for interchanging design-time rules.

This work presents a case study of bridging a business rule language: the Ilog Rule Language (IRL) to a standard format: the Rule Interchange Format (RIF) and vice versa. We show how to make it possible with the help of Model Driven Engineering (MDE [ 3 ]).

RIF is the W3C Rule Interchange Format and it is part of the infrastructure composing the semantic web. It is an XML language for expressing rules that computers can execute. One of the main goals of RIF is to promote a standard format to interchange rules between existing rule languages.

Because of the serious trade o s in the design of rule language, RIF provides multiple versions, called dialects: { Core: it is the fundamental RIF language. It is designed to be the common subset of most rule engines { BLD: it adds to Core dialect logic functions, equality in the then-part, and named arguments { PRD: adds a notion of forward-chaining rules, where a rule res and then performs some action (adding, updating or retracting some information) In this work we have implemented transformations for the Core and PRD dialects. An example of a RIF example is shown in (Fig. 1).

IRL is a formal rule language used inside WebSphere ILOG JRules BRMS. It supports the two di erent levels of a full- edged BRMS: the technical level targeted at software developers and the business action language targeted at business users. The technical rules of JRules are written in IRL. The complete speci cation of IRL can be found in the JRules documentation [ 4 ] an example of an IRL rule is provided in (Fig. 2).

Model Driven Engineering

The primary software artifacts of the MDE approach are models, which are considered as rst class entities. Every model de nes a concrete syntax and conforms to a meta-model (or grammar, or schema), which de nes its abstract syntax. One of the most common operations applied on models is transformation, which consists in the creation of a set of target models from a set of source models according to speci c rules.

In the work presented in this paper three main tools have been used: { Ecore allows to handle and create meta-models in Eclipse Modeling Framework (EMF) [ 5 ] { TCS (Textual Concrete Syntax) ([ 6 ],[ 7 ]) is a bidirectional mapping tool between meta-models and grammars. It is able to perform both text-to-model (injections) and model-to-text (extractions) translations from a single specication. TCS is used to map context-free concrete syntaxes to meta-models. { ATL (Atlas Transformation Language) [ 9 ] is a model transformation language speci ed as both a meta-model and a textual concrete syntax. It allows developers to produce a number of target models from a set of source models by writing rules that de ne how to create target models from source model elements. 3

Transformations

The RIF2IRL transformation takes as input a RIF le, a BOM le (Business Object Model [ 4 ]) and a Dictionary model. The output generates an IRL model and the related IRL le, generated by a TCS parser.

From the rst input, using a TCS parser, a model conforming to an implementation of the RIF meta-model is extracted. A BOM le is passed to the transformation to resolve the domain information declared inside the RIF le/model. The related BOM model is extracted out of the BOM le by a TCS parser and it conforms to the BOM meta-model implemented in [ 10 ]. A Dictionary is added as input parameter in the transformation in order to provide a translation from the element names declared in the RIF le to the corresponding ones appearing in the IRL le. The transformation has been implemented in ATL [ 8 ] and it de nes how to convert RIF to IRL.

The IRL2RIF transformation takes as input a IRL le and two BOMs and returns a RIF le. From the rst input, a model conforming to the IRL metamodel is extracted by means of a TCS parser. The rst BOM contains the information of the domain model, while the second one is used to map IRL functions to the built-in RIF functions for handling dates, numbers, strings, etc. The output is a RIF model, that is translated into a RIF le by a TCS parser. 4

The demonstration

The IRL2RIF and RIF2IRL transformations will be rst demonstrated using a set of simple examples (Appendix A) that illustrate the main functionalities of the transformation. We will then show a real world application of the transformation using data from the Arcelor use-case studied in the ONTORULE project [ 11 ]. In the context of the Rule Xchanger application depicted in the (Fig. 3), we will focus on the transformation of IRL rule edited using JRules into RIF rules that will be executed using the Tight engine. The transformation we intend to demonstrate is a key component in the interoperability between two rule languages that has been used in some real world applications such us the Arcelor use-case. it shows that rule application developed using commercial BRMS such as JRules can be exported to other platforms that support the RIF standard.

Annex

Fig. 5. Example 2

References

1. Rule Interchange Format http://www.w3.org/TR/rif-overview/

2. Production Rules Representation http://www.omg.org/spec/PRR/1.0/

3. Schmidt , D.C. Model-Driven Engineering . IEEE Computer 39 ( 2 ), Feb. 2006 .

4. JRules documentation http://pic.dhe.ibm.com/infocenter/dmanager/v7r5/index.jsp

5. Eclipse Modeling Framework http://www.eclipse.org/modeling/emf/

6. Textual Concrete Syntax http://wiki.eclipse.org/TCS

7. Jouault F. , Bezivin J. , Kurtev I. , TCS: a DSL for the Speci cation of Textual Concrete Syntaxes in Model Engineering . In proc. of GPCE'06 , Portland , Oregon, USA, pp 249 - 254

8. Atlas Transformation Language http://www.eclipse.org/atl

9. Jouault F , Allilaire A , Bezivin J , Kurtev I. ATL : a Model Transformation Tool . Science of Computer Programming 72 ( 3 , Special Issue on Second issue of experimental software and toolkits EST): 31 - 39 , 2008

10. Didonet Del Fabro M. , Albert P. , Bezivin J. , Jouault F. , Industrial-strength Rule Interoperability using Model Driven Engineering , Technical report 6747 , INRIA, Nov. 2008 . http://hal.inria.fr/docs/00/34/40/13/PDF/RR6747.pdf

11. Gonzalez-Moriyon G. , Final steel industry public demonstrators , ONTORULE Deliverable D5 .5, Jan . 2012