Comparison between SLOCs and number of files as size metrics for software evolution analysis

Abstract

There are some concerns in the research community about the convenience of using low-level metrics (such as SLOC, source lines of code) for characterizing the evolution of software, instead of the more traditional higher lever metrics (such as the number of modules or files). This issue has been raised in particular after some studies that suggest that libre (free, open source) software evolves differently than 'traditional' software, and therefore it does not conform to Lehman's laws of software evolution. Since those studies on libre software evolution use SLOCs as the base metric, while Lehman's and other traditional studies use modules or files, it is difficult to compare both cases. To overcome this difficulty, and to explore the differences between SLOC and files/modules counts in libre software projects, we have selected a large sample of programs and have calculated both size metrics over time. Our study shows that in those cases the evolution patterns in both cases (counting SLOCs or files) is the same, and that some patterns not conforming to Lehman's laws are indeed apparent