Casanova: A Language for Game Development

Giuseppe Maggiore's 2012 PhD thesis, produced at the Università Ca' Foscari di Venezia, argues that game development has become prohibitively costly and technically demanding for small teams, and that a purpose-built programming language can lower those barriers. The work centers on Casanova, a functional, F#-based language designed around a rules-script-draw pattern, with an explicit focus on indie, research, and serious games rather than AAA titles. Its design goals are simplicity, generality, composability, expressiveness, conciseness, efficiency, and readability, and the author frames the case for a dedicated language against mainstream systems such as Unity, Game Maker, and XNA, which the thesis finds limited in generality or expressive power.

The technical core covers the game loop, state machines modeled as discriminated unions, and monad-based coroutines for interruptible scripts. Implementation work addresses performance through query indexing, double-buffered rule storage, pre-allocated collections, memoized traversal, parallel rule evaluation, and per-rule update frequencies. The thesis reports benchmark gains, including framerate improvements of up to roughly two orders of magnitude in a city simulation from precomputed entity traversals, and compares Casanova scripts with Python, Lua, and C#, finding them about as concise as Python and Lua while running faster. Learning-complexity counts suggest Casanova has the smallest feature footprint among the systems compared, and classroom experience with high school and master's students was largely positive, though some students struggled with the shift from imperative to declarative thinking.

The thesis is candid about its limits. Its evaluation relies heavily on qualitative argument, sample games, and documentation-derived counts rather than measured development-cost figures. Casanova offers limited low-level optimization, unusual ML-style syntax, and weak support for advanced rendering such as shaders, so it is positioned for high-level abstraction rather than AAA-scale work. Appendices explore menu systems, prototype networking, goal-oriented and layered action planning for AI, and database-inspired techniques, several of which the author describes as incomplete at the time of writing.