A Study on the Metrics-Based Duplicated Code Type Smell Detection Techniques Relating the Metrics to Its Quality

Abstract

AbstractThe presence of bad smells in code hampers software’s maintainability, comprehensibility, and extensibility. A type of code smell, which is common in software projects, is “duplicated code” bad smell, also known as code clones. These types of smells generally arise in a software system due to the copy-paste-modify actions of software developers. They can either be exact copies or copies with certain modifications. Different clone detection techniques exist, which can be broadly classified as text-based, token-based, abstract syntax tree-based (AST-based), metrics-based, or program dependence graph-based (PDG-based) approaches based on the amount of preprocessing required on the input source code. Researchers have also built clone detection techniques using a hybrid of two or more approaches described above. In this paper, we did a narrative review of the metrics-based techniques (solo or hybrid) reported in the previously published studies and analyzed them for their quality in terms of run-time efficiency, accuracy values, and the types of clones they detect. This study can be helpful for practitioners to select an appropriate set of metrics, measuring all the code characteristics required for clone detection in a particular scenario.KeywordsClone detectionMetrics-based techniquesHybrid clone detection techniquesCategorizationQualitative analysis