Not All That Glitters Is Gold

Abstract

Type-2 clones are duplicated passages of source code that have identical sequences of tokens (whitespace and comments are ignored) except for the textual representation of identifiers and literals—parameters. Commonly, type-2 clones are detected by simply abstracting from the parameters’ textual representation. This normalization, however, frequently results in similar but otherwise unrelated code clones. We propose an alternative way to detect type-2 clones and compare the clones detected by our new approach to those detected by a traditional technique. Our results indicate that normalization of parameters is often inappropriate and should be applied with care.