Awesome to see people excited about nontrivial proceduralism and willing to explain/open-source their code, but it's unfortunate they don't give any credit for their technique. It's a nearly direct implementation of the 2001 Siggraph paper "Procedural Modeling of Cities" by Parish and Müller. (http://dl.acm.org/citation.cfm?id=383292 pdf easily available via google)