Quality Management for Deterministic Code Generators
Abstract
Generative programming has become an increasingly relevant research and application area, especially in model-driven engineering. While generative approaches offer promising advantages in terms of engineering efficiency and quality of the end-product, the quality of the generator itself is often neglected. Current testing solutions mostly relate to the front end, i.e., processing the input, or are either tailor-made for specific generation use cases or largely focus on end-to-end validation, in which exemplary target systems are tested but not the generator itself. In this paper, we present a detailed analysis of intricate code-generation components and their potential for quality assurance by focusing on testability and their distinct characteristics and properties. Based on a comprehensive reference architecture, we analyse various testing strategies for different transformation stages and components. Our analysis supports ensuring the quality of code generators, which is essential in making generative programming approaches applicable at an industrial scale.