Comparing the influence of using feature-oriented programming and conditional compilation on comprehending feature-oriented software

Several variability representations have been proposed over the years. Software maintenance in the presence of variability is known to be hard. One of the reasons is that maintenance tasks require a large amount of cognitive effort for program comprehension. In fact, the different ways of representing variability in source code might influence the comprehension process in different ways. Despite the differences, there is little evidence about how these variability representations – such as conditional-compilation directives or feature-oriented programming – influence program comprehension. Existing research has focused primarily on either understanding how code using modern paradigms evolves compared to the traditional way of realizing variability, namely conditional compilation, or on the aspects influencing the comprehension of conditional compilation only. We used two different programs implemented in Java and each of these variability representations. As Java does not support conditional compilation natively, we relied on the mimicking (i.e., preprocessing annotations in comments) that has been used in the literature. Our results show no significant statistical differences regarding the evaluated measures (correctness, understanding, or response time) in the tasks. Our heterogeneous sample allowed us to produce evidence about the influence of using CC and FOP variability representations on the aspects involved in the comprehension of feature-oriented software, while addressing bug-finding tasks.