Leveraging product line engineering for the development of domain-specific metamodeling languages

Highlights

• We propose a workflow to develop a domain-specific metamodeling language (DSM2L) by software product line engineering (SPLE).

• To put this idea into practice, the MoDEBiTE metamodeling language is engineered based on the SPLE paradigm.

• The applicability, usefulness, and validity of MoDEBiTE is presented with an experiment on six case studies, which shows the validity of the proposed workflow.

Abstract

A domain-specific metamodeling language (DSM2L) enables language engineers to define a family of similar metamodel-based languages. In recent years, several DSM2Ls have been developed for various domains, e.g., traceability, variability management, process modeling, and metamodeling feature models. However, there is no consensus on an engineering approach for constructing a DSM2L. To address this problem, we consider a DSM2L as a software product line (SPL), in which the software products are the family of languages. Based on this assumption, we propose a roadmap to develop a DSM2L using an SPL engineering framework. To investigate the pros and cons of the roadmap, the MoDEBiTE metamodeling language is engineered in the domain of bidirectional transformations. In order to validate the proposal of engineering a DSM2L, an experiment with six transformation cases is performed on MoDEBiTE. The results of the experiment show the applicability, usefulness, and validity of MoDEBiTE, demonstrating the validity of the proposal.

Introduction

Domain-specific languages (DSLs) provide more focused syntax and more precise semantics for a particular domain rather than general-purpose languages. A DSL can comprise a family of similar languages, which have a common core with different features [1]. While a DSL may be developed in an ad-hoc manner, it is possible to produce the DSL variants automatically to facilitate reusability. In order to generate the syntax, semantics, and tooling aspects of a family of DSL variants, Zschaler et al. [2] proposed the idea of domain-specific metamodeling language (DSM2L). A DSM2L enables language engineers defining a reusable core with a set of variation points to generate language-specific constructs. The development architecture of a wide range of the DSM2L languages [3], [4], [5], [6], [7] is based on the promotion transformation. Promotion, which is considered in this paper due to its wide use, is a transformation from a model of the DSM2L language to the metamodel of a DSL variant [5].

Developing a DSM2L language faces some issues. Most existing approaches for defining DSM2Ls [2], [3], [8], [9], [10] concentrate on technical aspects of implementing domain-specific metamodeling languages. Moreover, increasing the complexity of problem domains makes DSM2Ls to become complex as well. This fact hinders the analysis, design, and development of DSM2Ls in a systematic manner. To alleviate these issues, we propose to consider a DSM2L language as a software product line (SPL) due to the similar concepts in both paradigms. A product line is an infrastructure consisting of a set of common core assets with well-defined variation points to produce a family of similar applications [11].

This paper presents a roadmap for engineering DSM2L languages based on a mature framework of software product line engineering (SPLE) [11]. The main benefit of the SPLE framework is separation of concerns in constructing the product line (domain engineering) and building DSL variants (application engineering) [11]. The proposed roadmap is applicable to DSM2L development when the metamodeling language is defined based on the promotion architecture. To investigate the pros and cons of the proposal, we use the roadmap to develop MoDEBiTE¹ [7], which is a metamodeling language for producing languages specific to the domain of bidirectional transformations (Bx). A bidirectional transformation provides a mechanism to preserve inter-model consistency between two sources of data [12]. We apply the whole life cycle of the proposed roadmap on engineering the MoDEBiTE language, which leads to a feature model, a set of design patterns, a development process, and a testing strategy. To validate that the engineering process proposed by the roadmap works fine for MoDEBiTE, this DSM2L tool is applied on six Bx examples to develop six DSL variants. The results show the applicability, usefulness, and validity of the MoDEBiTE language.

The contribution of this paper targets two groups of engineers: DSM2L developers and Bx designers. For DSM2L developers, this article provides an instructive roadmap to present a systematic way for engineering a metamodeling language. In turn, the Bx designer can find the paper a practical guide to learn how to apply MoDEBiTE for the development of a case-specific bidirectional transformation language.

The rest of the paper is organized as follows. In Section 2, we propose a roadmap for developing DSM2Ls based on the product line engineering framework. Section 3 describes several steps of the roadmap for engineering the MoDEBiTE metamodeling language as an illustrative case. Section 4 investigates the results of applying a systematic methodology of case study research which is focused on evaluating the contributions of the MoDEBiTE product line. In Section 5, the related work is studied. After that, a discussion is provided in Section 6. Finally, the paper is concluded in Section 7.