Modeling Reactive Behavior in ORM Terry Halpin Gerd

Modeling Reactive Behavior in ORM Terry Halpin Gerd Wagner Northface University Eindhoven University of Technology Salt Lake City, USA. Eindhoven, The Netherlands. G. Wagner@tm. tue. nl terry. halpin@northface. edu © 2003, T. A. Halpin & Gerd Wagner 1

Contents v Object-Role Modeling (ORM) v Agent-Object-Relationship modeling (AOR) v Extending ORM with Agents and Actions v ORM and Reaction Rules v Conclusion and Further Reading 2

Object-Role Modeling (ORM) 3

ORM is fact-oriented • Grouping of facts into structures is considered sub-conceptual (hence attribute-free) • Semantically stable • Models can be fully verbalized in natural sentences • Validation by verbalization and population • Rich graphical coverage of business rules (static constraints and derivation rules) • Richer textual coverage of business rules • Ideal for conceptual analysis of data • Formal and Executable But weak or no support for • Reactive behavior (e. g. dynamic constraints and reaction rules) • Agent-based aspects 4

Agent-Object-Relationship modeling (AOR) • AOR treats an agent type as a distinctive kind of object type. • This entails treating the following as basic modeling elements: – action types – event types because … 1. Agents continuously perform three functions: 1. perception of events (including incoming messages) 2. action to communicate, and affect conditions in the environment 3. reasoning to interpret perceptions, draw inferences, and determine appropriate actions • A further option is to introduce commitments and claims, to better model social interaction processes. 5

• An interaction process type in AOR is specified by defining, for each participating agent (type), a set of inter-related behavior rules, called Reaction Rules. • A Reaction Rule has four parts: – Event – Condition – Action – Postcondition • The modeling principles of AOR apply to UML and to ORM 6

Extending ORM with Agents and Actions 7

Lexical action (message communication) Non-lexical action (physical) Passive Entity Involvement in Action Order of actions 8

Interaction Sequence Diagram for library loans 9

ORM and Reaction Rules Sample reaction rule, expressed informally: When the library receives a book request from a member, it checks whether a copy of that book is available, and if so, the request is confirmed, a new loan object is created, and a copy of the book is delivered to the faculty member. Note: Each loan is for exactly one book copy. The rule does not specify any time limits for the actions to take place. In this sense, it is not an operational rule. 10

Graphical part of ORM schema: 11

Simple derivation rule 12

Time-point viewed as transition between two states 13

Textual constraint 14

Temporal order in an interaction model may be captured in an ORM schema using • Explicit subset/equality constraints • Implicit subset constraints (implied by mandatory role constraints) • Subtyping (with formal subtype definitions – not illustrated here) However the interaction pattern and temporal order are more readily grasped from the interaction diagram. Interaction diagrams should be used in conjunction with data use cases to seed ORM models. 15

Some current research topics What is the best underlying logic to unify static and dynamic aspects? Temporal and Dynamic logics can express “leads to” semantics (e. g. confirmation of a book request leads to delivery of a book copy in some possible future state) but pragmatically, the constraint is not operational unless some deadline is specified for the delivery. To what extent can static and dynamic models be transformed into each other, or used for cross-referencing? How are actions involving three or more agents best modeled? How are viewpoints from the perspective of a single agent type best integrated to a global viewpoint with multiple autonomous agents? 16

Further Reading v www. orm. net -- Terry’s ORM website v www. AOR. rezearch. info -- Gerd’s AOR website v Halpin, T. A. 2001, Information Modeling and Relational Databases, Morgan Kaufmann, San Francisco (ISBN 1 -55860 -672 -6). v Wagner, G. 2003, ‘The Agent-Object-Relationship Metamodel: Towards a Unified View of State and Behavior’. Information Systems, 28(5), pp. 475 -504. 17