A systematic review on the code smell effect

Context: Code smell is a term commonly used to describe potential problems in the design of software. The concept is well accepted by the software engineering community. However, some studies have presented divergent findings about the usefulness of the smell concept as a tool to support software development tasks. The reasons of these divergences have not been considered because the studies are presented independently. Objective: To synthesize current knowledge related to the usefulness of the smell concept. We focused on empirical studies investigating how smells impact the software development, the code smell effect. Method: A systematic review about the smell effect is carried out. We grouped the primary studies findings in a thematic map. Result: The smell concept does not support the evaluation of quality design in practice activities of software development. There is no strong evidence correlating smells and some important software development attributes, such as effort in maintenance. Moreover, the studies point out that human agreement on smell detection is low. Conclusion: In order to improve analysis on the subject, the area needs to better outline: (i) factors affecting human evaluation of smells; and (ii) a classification of types of smells, grouping them according to relevant characteristics.