GSoC 2026 Draft: Modernizing Mesa-Geo (Flyweight Architecture & SolaraViz)

[AdMub]

Hi @wang-boyu and @jackiekazil!

Following up on my recently merged PR (#309) implementing get_random_xy, I have spent the last week deeply analyzing the architectural needs for the "Modernizing Mesa-Geo" project.

Following the community discussions around PropertyLayer volatility and the object-identity issues seen in recent community PR attempts, my GSoC proposal centers on a Flyweight architecture. The goal is to unlock the C-speed of vectorized PropertyLayer arrays while creating a strict abstraction layer that completely insulates Mesa-Geo's public API from any future internal changes in Mesa-Core. This ensures backward compatibility and preserves mesa.Agent lifecycle hooks.

I have finished my full draft covering the Raster Refactor, the Pandas-style Aggregation API, and the SolaraViz GeoSpace overhaul. To keep this thread clean (and to keep my draft private before applications officially open next week), I am sending you both a direct message on Matrix with the link for any early feedback.

Looking forward to the community dev meeting on March 17th!

[jackiekazil]

@AdMub, I struggle with providing early feedback, because of the number of requests we get. So, I defer to only providing feedback to things that are completely public out of fairness to provide discoverability.

I did look at your PR, and appreciated it.

I am going to close this discussion. @wang-boyu, feel free to reopen if you see fit.