[Master’s thesis] Evaluation of visualization approaches to comprehend the outline of program behavior

Kaixie Lyu, a master’s student of our laboratory, presented his master’s thesis on Aug. 1st.

Title: Evaluation of visualization approaches to comprehend the outline of program behavior

Reverse-engineered sequence diagram is valuable to understand the running behavior of programs by visualizing software execution traces. However, the execution traces of realworld software can be very massive, so the generated sequence diagram would be too large to read.

Recent years, researchers have proposed several techniques for reducing the size of sequence diagrams. Even though these techniques usually reach high compression ratios, the resulting diagram may be either still large or losing valuable information. Whether the compression techniques are helpful to understand sequence diagrams is not well evaluated yet. Moreover, by investigating related studies, we found that the effectiveness of comprehension by reverse-engineered sequence diagrams is not well validated yet.

The goal of this research is to explore the effectiveness of reverse-engineered sequence diagram in the aspect of program comprehension.

Meanwhile, we want to evaluate the influence of compression techniques.

To achieve this goal, we present a toolkit SDExplorer to effectively and interactively explore massive-scale sequence diagrams. Using SDExplorer, we did a user experiment by inviting students from graduate school to solve software comprehension questions with sequence diagrams. Along with their answers to the questions, we took questionnaires after the experiment and recorded users’ operations during the experiment.

Each comprehension question contains one or two comprehending tasks. From the user experiment, we find that comprehending tasks have different suitabilities for sequence diagrams. We then indicate the suitability of each task and conclude the obstacles.

Meanwhile, our experiment shows that classic compression techniques only have a slight influence on the comprehension results, even if they reach a very high compression ratio (97.43%). However, the compression techniques are necessary for users according to their feedback.

Finally, we give some suggestions to improve the understandability of sequence diagrams
based on the result of our experiment.