An empirical study on the frequency of disciplined and undisciplined annotations in preprocessor –based systems in C and C++

Abstract

The C and C++ programming languages allow a preprocessor tool to write conditional annotations. Relevant software communities such as Linux and Apache have used conditional annotations in their software projects. However, developers write these conditional annotations in an undisciplined way. The application of undisciplined conditional annotations or undisciplined annotations can have a negative effect on code readability and increasing the error proneness. Over the last ten years, empirical studies have investigated how the use of undisciplined annotations can affect the software quality and how to discipline these annotations. A proposed strategy to solve undisciplined annotation is to apply refactorings. Refactoring is a code transformation that aims to improve the code quality, preserving the program behavior. In the context of undisciplined annotations, one refactoring aims to discipline it. However, there is no empirical evidence on to what extent the number of disciplined annotations increased and/or decreased over the last ten years. Mainly, if developers applied refactorings on undisciplined annotations and these refactorings disciplined these annotations. Based on those literature limitations, our study aims to verify if the frequency of disciplined and undisciplined annotations during the last ten years, if developers apply refactorings on undisciplined annotations in the practice. We investigated 23 software projects that were investigated ten years ago on undisciplined annotations. Our results presented that only ten software projects had an increase in the number of disciplined annotations over the last ten years, and we presented 19 refactorings that were applied in practice. We observed that developers refactor not only undisciplined annotations but also disciplined ones. These results can motivate future studies to investigate if existing findings and recommendations of refactorings have helped developers to discipline annotations or if developers have used the existing empirical knowledge to solve undisciplined annotations.