Abstract Model‐Driven Development (MDD) is one of the important approaches to develop complex software systems. This approach tries to model a system in high‐abstraction level. Then through applying multiple transformations step by step, the model abstraction level is reduced and finally yields to executable code. As a result, Model transformation (MT) plays a pivotal role on the realization of MDD goals. Due to the increasing complexity of software systems, MTs naturally become more complex. Hence, qualitative technical issues may be overlooked or forgotten in these model transformations. To alleviate these issues in terms of technical debts/code smells in MTs, we can apply MT patterns. A main drawback on applying patterns is that most of them are cataloged in informal language. Additionally, construction of a conceptual framework to help MT designers through applying patterns requires a precise specification of the underlying MT patterns. With a formal basis, this paper is trying to realize the proposed framework. Hence, some of the existing well‐known MT patterns are formalized. Then based on the identified common technical debts/code smells in MTs, a designer can be directed to apply the appropriate patterns and resolve the detected problems iteratively. For the applicability and functionality of the proposed framework, several examples of problematic model transformations in terms of quality flaws were considered and resolved using the appropriate patterns. We consider the Epsilon Transformation Language (ETL) cases of model transformations in this paper, and other similar MT languages could be treated using the same measures and patterns as well.

Abstract Model transformation has become an established field, and it is important to improve the quality of specifications written in transformation languages. Different transformation patterns have been introduced in the model‐driven engineering (MDE) community to improve the quality of transformation specifications. However, due to the different definitions of pattern concepts by different authors, it is difficult for practitioners to understand how to apply patterns in practice. Therefore, there is a need to unify transformation pattern concepts by presenting a generic metamodel and formalizing patterns in terms of this metamodel, to define the meaning of pattern application. In this research a general metamodel for definition of different design patterns in model transformation is provided. The metamodel presents clear description of common aspects of transformation patterns, which facilitates the application of patterns on model transformations by validating the application against the underlying formalism. Additionally, a unified and precise terminology for the application and verification of model transformation patterns by using a formal model of model transformation patterns in the Z notation is presented. To show the applicability of the proposed formalism, four well‐known model transformation patterns are specified.

Summary Design patterns solve recurring design problems. One of the best practices, to solve a whole problem in a specific context, is to apply patterns in the form of a collection of related patterns, ie, a pattern language (PL). However, with the rapid growth of the number of patterns and their interrelationships in PLs and because manual verification of the applied PLs is a tedious and error‐prone task, when designing in PL‐based fashion, there is a need for tool support to verify the applied PLs. In our previous work, aimed at automating the development process of PL verification tools, a PL formalism was presented. This paper, based on the presented formalism, describes a framework called Pattern Language Verifier Generator (PLVGen) that automatically generates a pattern language verifier (PLV) for any given PL assuming its constituent patterns are represented in the Unified Modeling Language (UML). Using a modelware approach, two metamodels named pattern language specification language (PLSL) and pattern specification language (PSL) are developed to model the given PL and its patterns, respectively. By getting the models of a PL and its patterns, PLVGen automatically generates an epsilon validation language (EVL) module representing the expected PLV. By launching PLV, the applied PL on a UML model will be verified. To evaluate the applicability of PLVGen in practice, we automatically generated three PLVs. The statistics regarding the generated PLVs, as case studies, illustrate the scalability of PLVGen. Compared with similar tools, the generated PLVs are more portable, transparent, and free of ambiguities.