Hacker News Reader

Computational geometry is a real challenge for python because it requires object oriented design and is speed critical. It is a marvel of technical engineering that this was possible with CGAL, but I don't see this working out unless you have a robust python compiler that is able to remove the various `if` statements that are handled by templates in C++.

If I was to do computational geometry in python, I would wrap CGAL.

View on HN

Out of curiosity, what specific 'object-oriented' features are needed to robustly implement any given computational geometry algorithm?

The goal is to avoid `if` statements. It is true that a good system will let you deal with many dimensions and data types, but you might be thinking, "I only want this for (x,y,z) floating point data", but there are a few algorithms that require slicing the data into different views and this precisely where things like iterators that are eventually vectorized make the scheme computational tenable.

There are already SWIG generated python bindings for CGAL: https://code.google.com/p/cgal-bindings/

Reply on news.ycombinator.com