Uma abordagem visual para a evolução de test smells em software Java

Abstract

Software testing is part of the software development cycle, being the process of performing system tasks whose main objective is to identify whether the software behaves as specified. Testing also aims to identify defects in the code produced. During this process, developers and testers may include poor design or implementation choices in the test code, leading to the insertion of so-called test smells. Test smells are bad design or implementation choices for test code and their presence can make it ineffective at finding bugs, generating rework and additional costs. Test smells can also make code difficult to understand and maintain. Also, like any software artifact, test code requires quality and maintainability assessment. Test code maintenance can occur due to the identification of code problems and the evolution of the production code. As the test code evolves, some test smells may be removed and others may be inserted. To visualize the evolution of software systems, software visualization techniques have been used, whose objective is to transform the production code into a visual representation to facilitate its understanding. Although these techniques help to visualize the evolution of software systems, this approach does not cover test code to visualize test smells. Thus, the objective of this work was to propose an approach to visualize the evolution of test smells in the test code and its behavior, for example, if there was an increase or decrease of test smells, and visualize the possible authors for their inclusion. An approach was proposed that defined three visual strategies (TSInstant, TSEvolution and TSAuthor) and a tool that implements this approach through three software visualization techniques (Graph View, Treemap View and Timeline View) for 21 types of test smells. This approach was implemented in a tool called TSVizzEvolution and was evaluated through a controlled experiment. These views can help testers and others involved in software testing activities improve the quality of test code.