An empirical study on code comprehension: DCI compared to OO

Comprehension of source code affects software development, especially its maintenance where reading code is the most time consuming performed activity. A programming paradigm imposes a style of arranging the source code that is aligned with a way of thinking toward a computable solution. Then, a programming paradigm with a programming language represents an important factor for source code comprehension. Object-Oriented (OO) is the dominant paradigm today. Although, it was criticized from its beginning and recently an alternative has been proposed. In an OO source code, system functions cannot escape outside the definition of classes and their descriptions live inside multiple class declarations. This results in an obfuscated code, a lost sense the run-time, and in a lack of global knowledge that weaken the understandability of the source code at system level. A new paradigm is emerging to address these and other OO issues, this is the Data Context Interaction (DCI) paradigm. We conducted the first human subject related controlled experiment to evaluate the effects of DCI on code comprehension compared to OO. We looked for correctness, time consumption, and focus of attention during comprehension tasks. We also present a novel approach using metrics from Social Network Analysis to analyze what we call the Cognitive Network of Language Elements (CNLE) that is built by programmers while comprehending a system. We consider this approach useful to understand source code properties uncovered from code reading cognitive tasks. The results obtained are preliminary in nature but indicate that DCI-trygve approach produces more comprehensible source code and promotes a stronger focus the attention in important files when programmers are reading code during program comprehension. Regarding reading time spent on files, we were not able to indicate with statistical significance which approach allows programmers to consume less time