Uma investigação sobre Test Smells em códigos de testes JavaScript.

Abstract

Software testing is an essential activity in system development, as it minimizes the possibility of manifesting problems and maximizes the chances that the production code meets its specifications. However, just as a poor design choice (bad smell) affects the production code, tests are also susceptible to the same type of problem. Thus, a test indicating possible design defects (test smell) can lead to challenges ranging from code readability, making maintenance difficult, to being prone to giving false positives or simply ceasing to function. Indeed, test smells are anti-patterns adopted in the test code implementation. These smells can vary or trigger the emergence of new types, depending on the language or technology used for test writing. In parallel, the JavaScript language has developed over decades since its incep tion and a notable evolution since 2015 with the ECMAScript 6 (ES6) specification. JavaScript is a high-level language, which includes the following peculiarities: being multi-paradigm, interpreted, using dynamic typing, having prototype-based object ori entation, and even first-class functions. Given this, we focused our study on smells in JavaScript, because it is a constantly evolving language and widely used to develop software on various platforms, such as: servers, desktop, command line, mobile systems, and the internet of things. In summary, the aim of this study is to investigate the occurrence and severity of classic and specific test smells in the JavaScript language, and to check if these anti patterns might be correlated with each other and with the quality metrics of the test code. To achieve this goal, we initially selected 15 classic smells from the literature and one inherent to the language, to assess their occurrence in the test code. To facilitate the detection of smells and validate our study, we developed an automated detection tool for these 16 smells using the static code analysis technique and conducted an exploratory study with 65 projects, 61 open-source and 4 proprietary. As results, we concluded that the incidence of some of these test smells is moderately correlated with each other and with quality metrics in a positive sense. Additionally, we conducted a complementary study involving developers and testers with the aim of investigating the impact of experience on the recognition of these smells 1 2 in JavaScript. This study aimed to understand not only the prevalence of specific test smells in the perception of these professionals, but also the strategies and tools they used to mitigate such anti-patterns. This aspect of the study allowed us to assess the awareness and practices adopted in the real context of software development and testing, enriching the understanding of the challenges faced and the possible applicable solutions. With this study, we hope to have contributed to the initial formalization of test smell studies in the JavaScript language, considering that although it has been increas ingly used in various development contexts, it had not yet been academically explored regarding its specifics and ramifications in the realm of test smells.