ThingML

Nicolas Harrand, Franck Fleurey, Brice Morin, Knut Eilif Husa

2016 · 119 citations · 7 references

Concepts

TL;DR

Model‑driven software engineering promises productivity gains through automatic code generation, yet real‑world adoption remains limited because models are often ill‑suited and generators are perceived as opaque black boxes. This paper shares the authors’ experiences applying the ThingML approach across multiple domains. ThingML comprises a dedicated modeling language and an open‑source, highly customizable multi‑platform code‑generation framework that supports heterogeneous target platforms. Case‑study evaluations demonstrate that ThingML has been successfully employed in the development of a commercial ambient‑assisted‑living system.

Abstract

One of the selling points of Model-Driven Software Engineering (MDSE) is the increase in productivity offered by automatically generating code from models. However, the practical adoption of code generation remains relatively slow and limited to niche applications. Tooling issues are often pointed out but more fundamentally, experience shows that: (i) models and modeling languages used for other purposes are not necessarily well suited for code generation and (ii) code generators are often seen as black-boxes which are not easy to trust and produce sub-optimal code. This paper presents and discusses our experiences applying the ThingML approach to different domains. ThingML includes a modeling language and tool designed for supporting code generation and a highly customizable multi-platform code generation framework. The approach is implemented in an open-source tool providing a family of code generators targeting heterogeneous platforms. It has been evaluated through several case studies and is being used for in the development of a commercial ambient assisted living system.

References

7