Revealing Patterns for Programs in Graphical Visual Languages Using Clustering

As the number of developers around the world continues to grow, there is a noticeable trend towards the reuse of code components. One interesting development in this area is the rise of graphical or block languages, which allow users to piece together programs by simply dragging and dropping pre-made components, much like a puzzle. While many of these languages may be seen as niche tools by the larger programming community, there are already examples of their practical applications. One standout example is App Inventor, a visual tool developed at MIT that is specifically designed to be a user-friendly introduction to programming for children. This platform enables users to easily create Android applications in a web-based environment. What sets App Inventor apart is its robust set of teaching materials, created by the developers themselves, as well as the contributions made by students and external persons who have developed problem-oriented case studies using the platform. This collaborative approach to learning and programming highlights the potential of graphical languages in bridging the gap between developers and non-developers in the creation of useful applications. In this paper, we introduce a comprehensive dataset containing 95 open-source graphical programs designed for App Inventor. We then suggest an approach for automatically analyzing this dataset with augmentation, aimed at detecting prevalent patterns found within these programs. This analysis serves as a pioneering effort in the evaluation of graphic programs and their corresponding metrics. Our study primarily relies on clustering methods and Markov chains to achieve this analysis.