Development and Evolution of Xtext-based DSLs on GitHub: An Empirical Investigation

TL;DR

This study systematically analyzes 1002 GitHub repositories of Xtext-based DSLs, revealing development practices, evolution patterns, and domain-specific usage trends to inform future MDE tool and methodology improvements.

Contribution

It provides the first large-scale empirical analysis of Xtext-based DSL development and evolution, including a dataset of repositories and insights into practices and trends.

Findings

DSLs are more used and evolve faster in domains like Data Management.

Most DSLs follow a grammar-driven approach, with frequent updates to grammar and instances.

Xtext is retrofitted for existing languages, showing its flexibility.

Abstract

Domain-specific languages (DSLs) play a crucial role in facilitating a wide range of software development activities in the context of model-driven engineering (MDE). However, a systematic understanding of their evolution is lacking, which hinders methodology and tool development. To address this gap, we performed a comprehensive investigation into the development and evolution of textual DSLs created with Xtext, a particularly widely used language workbench in the MDE. We systematically identified and analyzed 1002 GitHub repositories containing Xtext-related projects. A manual classification of the repositories brought forward 226 ones that contain a fully developed language. These were further categorized into 18 application domains, where we examined DSL artifacts and the availability of example instances. We explored DSL development practices, including development scenarios, evolution activities, and co-evolution of related artifacts. We observed that DSLs are used more, evolve faster, and are maintained longer in specific domains, such as Data Management and Databases. We identified DSL grammar definitions in 722 repositories, but only a third provided textual instances, with most utilizing over 60% of grammar rules. We found that most analyzed DSLs followed a grammar-driven approach, though some adopted a metamodel-driven approach. Additionally, we observed a trend of retrofitting existing languages in Xtext, demonstrating its flexibility beyond new DSL creation. We found that in most DSL development projects, updates to grammar definitions and example instances are very frequent, and most of the evolution activities can be classified as ``perfective'' changes. To support the research in the model-driven engineering community, we contribute a dataset of repositories with meta-information, helping to develop improved tools for DSL evolution.