Variance of source code quality change caused by version control operations

Abstract

Software maintenance consumes huge efforts. Its cost strongly depends on the quality of the source code: an easy-to-maintain code needs much less effort than the maintenance of a more problematic one. Based on experiences, the maintainability of the source code tends to decrease during its lifetime. However, in most of the cases, this decrease is not a smooth linear one, but there are larger and smaller ups and downs, and the net root of these changes generally results in a negative tendency. Detecting common development patterns which similarly influence the maintainability could help to stop or even turn back source code erosion. In this research the scale of the ups and downs are investigated, namely that which version control operations cause bigger and which smaller changes in the maintainability. We calculated the maintainability and collected the cardinality of each version control operation for every revision of four inspected software systems. With the help of these data we checked which version control operation causes higher absolute code quality changes and which lower. We found clear connection between version control operations and the variance of the maintainability changes. File Additions and file Deletions caused significantly higher variance in maintainability changes compared to file Updates. Commits containing higher number of operations - regardless of the type of the operation - caused higher variance in maintainability changes than those commits containing lower number of operations. As a practical conclusion, it is recommended to pay special attention to the quality of commits containing new file additions, e.g. with the help of a mandatory code review.