Towards a Corpus of Use-Cases for Model-Driven Engineering Courses

Dimitrios S. Kolovos, Jordi CabotEduSymp/OSS4MDE@MoDELS 2016 · OpenReview

Abstract

Model-Driven Engineering (MDE) is a powerful paradigm in industry, yet it remains a difficult subject to teach effectively in academic settings. Students often perceive MDE as an unnecessary layer of complexity when applied to "toy" problems. This paper addresses these pedagogical challenges by identifying common pitfalls in current curricula and proposing a shift toward a community-driven corpus of pragmatic use-cases. By focusing on scenarios where MDE provides a clear advantage over manual coding—such as a Wedding Organisation Domain-Specific Language (DSL)—educators can better demonstrate the reasoning power and long-term return on investment of modeling.

Introduction

Model-Driven Engineering (MDE) involves the use of high-level abstractions, such as diagrams or domain-specific languages (DSLs), to automate code generation and system analysis. Despite its industrial utility, MDE often fails to inspire students. The primary hurdle is that classroom examples frequently feel less efficient than traditional imperative programming. This paper reflects on years of teaching experience to propose a new direction for MDE pedagogy.

The "Deadly Sins" of MDE Pedagogy

The authors identify three core reasons why students often reject MDE:

  1. Unrealistic Domains: Using models for tasks where standard tools (like relational databases) are clearly superior. Asking students to build a library system with models when they know a relational database is the "right" tool for the job sets MDE up for failure.
  2. Artificial Processes: Teaching complex transformations (like UML to SQL) that produce low-quality results compared to modern frameworks like Hibernate.
  3. One-and-Done Scenarios: MDE pays off over multiple iterations. When students only build a system once, manual coding always seems faster, leading them to question the cost-effectiveness of the entire discipline.

Proposed Use-Cases for a New Corpus

To overcome these hurdles, the paper suggests use-cases that are pragmatic, motivating, and demonstrate a clear advantage over manual imperative programming:

  • Auto-Synchronised References in Java: Solving a genuine language limitation (the lack of built-in bidirectional pointers) using model-to-text transformation.
  • State Machines for Logic: Using models to reason about complex behavior that is difficult to "trace" in raw Java code.
  • Wedding Organisation DSL: Creating a custom language for a niche domain where no standard software exists. This avoids the "Database Trap" because students lack preconceived notions of how "wedding software" should look, making them more open to the custom MDE approach.
Metamodel for a Wedding Organisation DSL, illustrating the relationships between weddings, guests, tables, and potential conflicts.

As shown in Figure 1, the Wedding DSL allows students to model complex constraints, such as guest conflicts and seating arrangements. This model can then be used to generate invitations, seating charts, and validate "guest conflicts" in a relatable, non-industrial context.

Key Insights and Ideas

  • Models as Reasoning Tools: MDE isn't just about "generating code." Use-cases should show how models help us understand system reachability or unreachable states—things that are hard to see in 1,000 lines of if-else statements.
  • The Iteration Argument: To prove MDE's value, assignments must include "change-adapt" cycles. Students need to see how a small model change can instantly update a massive codebase, which is where the "aha!" moment happens.
  • Niche is Better: A "Wedding DSL" works better than a "Bank DSL" because it targets a domain where the MDE solution is the most practical tool available.

Conclusion and Future Work

The authors propose building a community-wide "Corpus of MDE Use-Cases" on platforms like GitHub to prevent educators from "reinventing the wheel." The primary results of this work are the qualitative framework for better MDE education and the initial set of proposed use-cases. Future steps involve validating these specific scenarios in university courses and developing the technical infrastructure to make these examples easy for other professors to adopt.

Summary updated

BibTeX
@inproceedings{kolovos2016towards,
  author = {Dimitrios S. Kolovos and Jordi Cabot},
  title = {{Towards a Corpus of Use-Cases for Model-Driven Engineering Courses}},
  booktitle = {EduSymp/OSS4MDE@MoDELS},
  year = {2016},
  url = {https://openreview.net/forum?id=gOiNTHK0kq},
}
Figure