Supporting Domain Experts by using Model-Based Equivalence Class Partitioning for Efficient Test Data Generation

Context. Production Systems Engineering (PSE) faces a growing complexity of software, i.a., due to increasing capabilities of the hardware, requiring efficient approaches for designing test data and test cases. Apart from this, for long-running legacy systems where often no tests are available tests need to be added belatedly during system maintenance. Equivalence Class Partitioning (ECP) can help to systematically cluster test input and result data as a foundation for test case definition. However, different tools and technologies in PSE often hinder the precise definition and creation of Equivalence Classes (EC) based on the existing source code. Objective. In this paper, we present a model-based approach for deriving ECs based on abstract representations of existing source code and evaluate the approach in a real-world industry use case. Method. We build on best-practices from software testing for developing a model-based approach for deriving ECs as a foundation for test data and test case generation and on the Abstract Syntax Tree (AST) to describe the structure of the underlying source code. Results. First results were promising in the evaluation use case to improve test activities by systematically deriving test data and test cases based on ECs. Conclusions. We observed additional benefits, such as increased test coverage and capabilities for testing error cases.