Experience report: growing and shrinking polygons for random testing of computational geometry algorithms

This paper documents our experience of adapting and using the QuickCheck-style approach for extensive randomised property-based testing of computational geometry algorithms. The need in rigorous evaluation of computational geometry procedures has naturally arisen in our quest of organising a medium-size programming contest for second year university students—an experiment we conducted as an attempt to introduce them to computational geometry. The main effort in organising the event was implementation of a solid infrastructure for testing and ranking solutions. For this, we employed functional programming techniques. The choice of the language and the paradigm made it possible for us to engineer, from scratch and in a very short period of time, a series of robust geometric primitives and algorithms, as well as implement a scalable framework for their randomised testing. We describe the main insights, enabling efficient random testing of geometric procedures, and report on our experience of using the testing framework, which helped us to detect and fix a number of issues not just in our programming artefacts, but also in the published algorithms we had implemented.