Method for incremental control of consistency between structural and behavioral views of software architecture

Abstract

The development of software engineering poses a challenge for researchers to maintain the integrity of models stored simultaneously in a lightweight text format and in a formally rich metadata view. The existence of two views ensures broad compatibility with developer tools and accurate reproduction of semantics, but creates the risk of discrepancies between structural and behavioral descriptions. The relevance of the research is determined by the need for methods that prevent the accumulation of contradictions without significantly affecting the speed of design iterations. The aim of this work is to provide a theoretical justification for an incremental approach that can guarantee the consistency of a metamodel with two views during any successive changes. To achieve this goal, a generalized metamodel has been formed that distinguishes between a structural view for static entities and a behavioral view for dynamic aspects. A correspondence relationship has been introduced between views, which describes pairs of equivalent elements and sets rules for their mutual consistency. The set of rules is formalized in the language of object invariants. Incrementalism is ensured by localizing changes: after editing, only those fragments that are directly involved in the modification are checked, so that the time spent remains proportional to the volume of the updated part. The result of applying the method is to prove the correctness of the proposed restrictions, which excludes the possibility of inconsistent model states. An analytical assessment of the complexity of the procedure confirms a linear dependence on the number of changed elements, which indicates the suitability of the approach for industrial-scale models. A demonstration control example, built on a representative domain, showed that the method detects inconsistency immediately after a single edit and proposes a sequence of actions sufficient to eliminate it without involving outside expertise. As a result, the work proposes a new formal methodology for maintaining consistency between views of a single model, which comprehensively combines localized verification with a declarative description of dependencies. The practical significance is manifested in the reduction of error correction costs, increased reliability of documentation, and the ability to integrate the method into modern modeling and continuous development environments, making it a promising tool for the development and maintenance of large corporate systems.