Towards systematic mutations for and with ATL model transformations

Abstract

Model transformation is a key technique to automate software engineering tasks, such as generating implementations of software systems from higher-level models. To enable this automation, transformation engines are used to synthesize various types of software artifacts from models, where the rules according to which these artifacts are generated are implemented by means of dedicated model transformation languages. Hence, the quality of the generated software artifacts depends on the quality of the transformation rules applied to generate them. Thus, there is the need for approaches to certify their behavior for a selected set of test models. As mutation analysis has proven useful as a practical testing approach, we propose a set of mutation operators for the ATLAS Transformation Language (ATL) derived by a comprehensive language-centric synthesis approach. We describe the rationale behind each of the mutation operators and propose an automated process to generate mutants for ATL transformations based on a combination of generic mutation operators and higher-order transformations. Finally, we describe a cost-effective solution for executing the obtained mutants.

Mutation Testing Advances: An Analysis and Survey

“…In this approach the authors introduce specific mutant operators for the Atlas Transformation Language. Later, Troya et al [298] focuses on the same subject and presented an extensive set of mutant operators that uses both first and higher-order mutation transformations [299]. Another study that tests model transformations using mutation is due to Aranega et al [300], who focuses on how to support the generation of mutation-adequate test cases for checking model transformations.…”

Systematic Engineering of Mutation Operators.

“…The authors establish a criteria based on semantic faults in the navigation, the filtering, the output model creation, and the input model modification to create such set. Mutation operators have also been defined using Henshin in (Burdusel et al 2019), and ATL in (Troya et al 2015). Instead, WODEL is a DSL targeted to define mutation operators, giving support for specific mutation actions (e.g., retyping, cloning), the automatic initialization of object features and containers, and the configuration of the number of mutants to generate.…”

“…Instead, WODEL is a DSL targeted to define mutation operators, giving support for specific mutation actions (e.g., retyping, cloning), the automatic initialization of object features and containers, and the configuration of the number of mutants to generate. Works like (Troya et al 2015) miss such policies and only produce one mutant per input model.…”

Spectrum-Based Fault Localization in Model Transformations

“…To test the usability and effectiveness of our approach, we apply mutation analysis [58], so that we have produced mutants for all model transformations, where artificial bugs have been seeded. We have used the operators presented in Reference [98] and have applied them [98] it is materialized as binding feature change.…”

“…For instance, the number of rules ranges from 8 to 39 and the lines of code from 53 to 1055. We have defined 117 OCL assertions for the four case studies, many of them taken from Reference [18], and have applied mutation testing by creating 158 mutants using the operators presented in Reference [98], where each mutant is a faulty variation of the original model transformation. Experimental results reveal that the best techniques place the faulty transformation rule among the three most suspicious rules in around 74% of the cases.…”