<https://github.com/StefanSalewski/RBR>
That router project was the most important reason in 2014 for leaving Ruby and searching for another compiled and statically typed language. The router itself needs at least a delaunay triangulation and a reliable convexHullOfDisk implementation, and used initially also a fibonacci priority queue with decrese_key() function. And for an EDA tool to using that router a GUI is necessary of course. Well now after nearly 7 years I finally managed to convert the core module with the rubberband routing from Ruby to Nim. Took me less than 200 hours for the 2k lines of Ruby code, which is faster than expected. The Nim code size is not much larger, and looks not worse than the Ruby code. Static typing makes all much cleaner. Speed with Nim is much larger of course. Porting was really not that hard: I did all the conversions manually. First I converted very carefully to make it compile and partly work. Then I fixed a few bugs. Then I compared one more the Nim and Ruby code very carefully. And finally I had to find an ugly and difficult bug, which resulted from the fact that the Ruby code used the BOOST fibonacci queue for the dijstra routing, while we now use the heapqueue without decrese_key() operation and so have to record and check distances for each node carefully. Well 200 hours for 2k Ruby code is fast for me, but it may sound slow to you: But note that we now use different libs with different API for delaunay, hull and priority queue. And after 7 years the own code looks partly strange as foreign code :-)
