Investigating Test Smells in JavaScript Test Code

• Dalton Jorge UFCG
• Patricia Machado UFCG
• Wilkerson Andrade UFCG

Resumo

Writing automated test cases is a challenging and demanding activity. The test case itself is software that requires proper design to ensure it can be implemented and maintained as long as the production code evolves. Like code smells, test smells may indicate violations of principles that negatively affect the quality of test code design, making it difficult to comprehend and, consequently, impairing its proper use and evolution. This work aims to investigate the occurrence of test smells in JavaScript test code and whether their presence can be correlated with test code quality. We perform an empirical study using the STEEL tool where the test suites of 11 open-source JavaScript projects from the Github repository are analyzed to detect a set of previously cataloged test smells. We then investigate: i) which ones occur more frequently; ii) whether given test smells are likely to occur together, and iii) if the presence of certain test smells is related to classical bad design indicators on the test code. We found that the Duplicate Assert, Magic Number Test, Unknown Test and Conditional Test Logic smells are the most common in JavaScript test code, whereas the Mystery Guest, Ignored Test and Resource Optimism smells are the least common. Moreover, the Conditional Test Logic, Magic Number Test, Duplicate Assert and the Exception Handling smells may often appear together. Furthermore, there is a moderate to a strong positive correlation between some smells count and quality measures in the test code. We can conclude that test smells are frequently found in JavaScript test code, and their presence may be an indicator of low design quality.

Palavras-chave: Javascript, quality metrics, test smells