BEHAVIORAL VIEW IN XMI FORMAT WITHIN THE UML METAMODEL WITH TWO VIEWS

Abstract

The article proposes an approach to representing behavioral models in the Unified Modeling Language (UML) in a format based on cross-platform XML exchange (XMI). This format is part of the UML metamodel, which consists of structural and behavioral views, where structural information is stored in a format based on JavaScript Object Notation. This division of model views provides greater modularity, better support for tool-based verification, and the ability to incrementally control consistency between structural and behavioral components. A specialized profile has been developed for behavioral models, containing mechanisms for tracing between behavioral and structural elements, in particular using JsonRef, TriggerSource, and other extensions that indicate logical relationships with structural objects. This implementation allows the semantic integrity of the model to be preserved even when the views are physically divided into two parts.

The paper defines formalized sets of elements for describing the structure and behavior of the model, as well as the mapping between them, which are specified in the form of correspondence relations. For the main types of behavioral diagrams (state, activity, and sequence diagrams), a system of logical invariants is formulated that express formal requirements for the correctness of models, in particular regarding the completeness of conditions, the reachability of states, the correspondence of message calls, and the typing of parameters. An algorithm for localized verification of changes in the behavioral view of a model is proposed, which uses the construction of a graph of dependencies between model elements to limit the scope of re-verification. As a result, this significantly reduces the computational cost of editing large models during development, particularly in continuous integration or sequential refactoring environments.

The theoretical complexity of the algorithm is evaluated, and the possibility of generalizing it to other types of behavioral diagrams is analyzed. The presented results can be applied in modeling tools focused on supporting formal model verification, modular storage, and automated integrity control. The proposed approach forms the basis for further research in the direction of automated behavior generation from structural specifications, as well as bidirectional transformation between the traditional XMI form and the proposed UML metamodel in order to ensure compatibility with external UML tools.