A Token-based Code Clone Detection Technique and Its Evaluation

A code clone is a code port ion in source f i les that is identical or similar to another. Since code clones generally reduce maintainabili ty of soft ware, several code clone detection techniques and tools have been proposed. This paper proposes a new clone detection technique, which consists of transformation of input source text and t o ken-bytoken comparison. Based on the proposed code clone detection technique, we developed a tool named CC Finder, which extracts code clones in C/C++ or Java source files. As well metrics for code clones were developed. In order to evaluate the usefulness of the tool and metrics, we con ducted several exper iments . As the resul ts , the tool found several subsystems in two operat ing sys tems, namely FreeBSD and Linux, that could be traced t o the same original. As well, the pro posed met r ics found interesting clones in a Java library, JDK.