Towards Modelling Acceptance Tests as a Support for Software Measurement

The DevOps paradigm emphasizes the need for a measurable feedback loop, starting from requirements and going as far as deployment in an automated way. In this context, a modelling challenge is to leverage the existing requirement engineering approaches to support measurements. Unfortunately, measurement methods are slow and costly by definition, preventing precisely measured requirements from being used in the DevOps loop. As a result, developers have to deal with grossly estimated elements, e.g., using story points promoted by agile methods. Thus, it is not possible to provide better support for the development team. We envision taking advantage of the artifacts that already exist in a DevOps context to provide better support for requirements measurement, making it available in an automated context such as the DevOps one. This paper focuses on the automated analysis of acceptance tests (e.g., expressed using the Gherkin language) to support functional measurement automation in a DevOps context. This proposition is illustrated by a scenario coming from an industrial partner, supporting the identification of four research challenges to be tackled.