Does cloned code increase maintenance effort?

Abstract

In-spite of a number of in-depth investigations regarding the impact of clones in the maintenance phase there is no concrete answer to the long lived research question, “Does the presence of code clones increase maintenance effort?”. Existing studies have measured different change related metrics for cloned and non-cloned regions, however, no study calculates the maintenance effort spent for these code regions. In this paper, we perform an in-depth empirical study in order to compare the maintenance efforts required for cloned and non-cloned code. For the purpose of our study we implement a prototype tool which is capable of estimating the effort spent by a developer for changing a particular method. It can also predict effort that might need to be spent for making some changes to a particular method. Our estimation and prediction involve automatic extraction and analysis of the entire evolution history of a candidate software system. We applied our tool on hundreds of revisions of six open source subject systems written in three different programming languages for calculating the efforts spent for cloned and non-cloned code. According to our experimental results: (i) cloned code requires more effort in the maintenance phase than non-cloned code, and (ii) Type 2 and Type 3 clones require more effort compared to the efforts required by Type 1 clones. According to our findings, we should prioritize Type 2 and Type 3 clones when making clone management decisions.