Comprehending a codebase requires developers to construct mental models spanning multiple levels of abstraction, from architectural overviews to individual components. These components differ across paradigms (e.g., classes in object-oriented systems, functions in functional programming). Software visualization supports this process by translating abstract software structures into perceptually meaningful forms that align with human cognitive processes. When grounded in familiar metaphors, such as cities or landscapes, visualization can further narrow the gap between abstract program constructs and developers’ mental models. How can language-agnostic expressive metaphors, direct manipulation, and extended-reality be integrated to create immersive, explorable virtual-city environments that help developers explore, understand, and refactor complex, large software systems? Our research advances this idea by extending the city metaphor to represent classes and packages as buildings and districts, where architectural details encode structural and behavioral properties. We classify classes into Class Archetypes using heuristics based on metrics, intra-class, and inter-class relationships. We then develop a specific building for each archetype: Utility classes become factories, data classes become depots, constant definers become archives, and adapter classes appear as surrounding structures. Building on this foundation, we aim to develop a unified, interactive, and immersive visual language for program comprehension.

@inproceedings{paperbot11851,
  title = {System Comprehension and Source Code Manipulation in eXtended Reality},
  author = {Mattia Giannaccari},
  booktitle = {ICSE},
  year = {2026},
  doi = {10.1145/3774748.3787665}
}