Code Confabulator: Harnessing LLMs to Compile Code for Visualization

Code visualization tools play a core role for visualization of the code and helping novices in understanding, analyzing, and optimization of their programs. Tools such as Algorithm Visualizer and PythonTutor are developed to make algorithm visualization possible across different programming languages like Python, Java, C, and $\mathrm{C++}$. Despite their advantages, these tools suffer from unnecessary data collection, increased time complexity, and limitation in visualizing a few data structures. In our research, we want to introduce a more humanlike view of understanding and solve the shortcomings mentioned above by proposing a framework, which will make use of LLMs for extracting data from the code in Python, Java, C, or C++ to be visualized. In our framework, we have used LLMs instead of the traditional compiler for feasibility and adaptability. Our objective is the appropriateness evaluation of LLMs for this task, starting from freeform prompts. The LLM is asked to generate outputs in a given format; hence, these outputs are taken and converted into visualizations with a rule-based approach. The task will be done to enhance quality visualization and reduce time complexity while allowing the user to view the code at a human level view. We have also publicly released the framework so that people are able to provide further contributions and improvements.