Clone Wars

Code clones, the result of the ”copy&paste” programming technique, have negative impact on software quality and on the efficiency of the software maintenance process. Although copying may be the fastest way of creating a new feature, after a while it is really hard to detect and maintain the multiple instances of the same code snippets. Based on static source code analysis, clone detectors try to identify code clones automatically. Several clone detectors exist [Roy et al. 2009] applying different techniques to select the clones. These techniques include string, token, syntax and also semantics based approaches. In the context of the Erlang programming language [Armstrong 2007], there are three clone detectors [Li and Thompson 2009; Fördős and Tóth 2014b; 2013] implementing different techniques to select duplicated code. Although the clones identified by these techniques can be considered duplicates, some of them are irrelevant in certain points of view. The filtering system of Clone IdentifiErl allows users to tailor the result in different ways using domain specific knowledge about the language. This filtering technique can be easily applied on duplicate code detectors that yield clone pairs [Fördős and Tóth 2014b; 2013]: it simply leaves out the pairs which do not fulfil the requirements. When a clone detector groups the identified clones [Baker 1996; Koschke 2012; Fördős and Tóth 2014a], the result is more comprehensible, but makes the filtering less straightforward. Filtering out some part of a group of clones results in smaller groups of clones. Sometimes smaller means that we have less group members, in other cases we have smaller clones – or both.