Evaluating Interactive Detection of Code Smells on Software Development Activities
Danyllo Wagner Albuquerque, Everton Tavares Guimarães, Mirko Perkusich, Hyggo Oliveira de Almeida, Ângelo Perkusich · 2023
Traditional code smell detection techniques rely on Non-Interactive Detection (NID), which only allows developers to identify smells in the entire source code upon request. However, studies suggest that NID may lead to fewer correctly identified smells, increasing remaining code smell instances. To address NID’s limitations, Interactive Detection (ID) has emerged, allowing developers to reveal smell instances without explicit requests and promoting early detection and correction. In this study, we aimed to evaluate the impact of ID on detecting code smells during software development by conducting a controlled experiment using Eclipse ConCAD with software developers and students. Our results indicated that ID could decrease up to 40% of remaining smell instances compared to NID during software development activities. Our findings suggest that ID is an effective technique to help developers quickly detect and fix code smell instances, improving overall code quality.