Biblio

Background: Researchers and practitioners have been using code complexity metrics for decades to predict how developers comprehend a program. While it is plausible and tempting to use code metrics for this purpose, their validity is debated, since they rely on simple code properties and rarely consider particularities of human cognition. Aims: We investigate whether and how code complexity metrics reflect difficulty of program comprehension. Method: We have conducted a functional magnetic resonance imaging (fMRI) study with 19 participants observing program comprehension of short code snippets at varying complexity levels. We dissected four classes of code complexity metrics and their relationship to neuronal, behavioral, and subjective correlates of program comprehension, overall analyzing more than 41 metrics. Results: While our data corroborate that complexity metrics can-to a limited degree-explain programmers' cognition in program comprehension, fMRI allowed us to gain insights into why some code properties are difficult to process. In particular, a code's textual size drives programmers' attention, and vocabulary size burdens programmers' working memory. Conclusion: Our results provide neuro-scientific evidence supporting warnings of prior research questioning the validity of code complexity metrics and pin down factors relevant to program comprehension. Future Work: We outline several follow-up experiments investigating fine-grained effects of code complexity and describe possible refinements to code complexity metrics.

Inferring behavior model of a running software system is quite useful for several automated software engineering tasks, such as program comprehension, anomaly detection, and testing. Most existing dynamic model inference techniques are white-box, i.e., they require source code to be instrumented to get run-time traces. However, in many systems, instrumenting the entire source code is not possible (e.g., when using black-box third-party libraries) or might be very costly. Unfortunately, most black-box techniques that detect states over time are either univariate, or make assumptions on the data distribution, or have limited power for learning over a long period of past behavior. To overcome the above issues, in this paper, we propose a hybrid deep neural network that accepts as input a set of time series, one per input/output signal of the system, and applies a set of convolutional and recurrent layers to learn the non-linear correlations between signals and the patterns, over time. We have applied our approach on a real UAV auto-pilot solution from our industry partner with half a million lines of C code. We ran 888 random recent system-level test cases and inferred states, over time. Our comparison with several traditional time series change point detection techniques showed that our approach improves their performance by up to 102%, in terms of finding state change points, measured by F1 score. We also showed that our state classification algorithm provides on average 90.45% F1 score, which improves traditional classification algorithms by up to 17%.