[Pcb2gcode-devel] Douglas-Peucker simplification algorithm

[fritz <fri...@gm...>]

I have a new branch "douglas" to push to the repository.

In it I have an implementation of the Douglas-Peucker simplification 
algorithm.  My work (in C++) is based off of a bit of python code used 
in the "image-to-gcode" feature of linuxcnc (http://linuxcnc.org).

Current notes:
     - a new boolean option "--smooth" added.
     - I've subclassed NGC_Exporter.  I realize I could've simply added 
in logic to NGC_exporter to act on an option, and that will probably be 
what the finished version will have.  For now, the switch between the 
base and derived class occurs with what ends up in the 
shared_ptr<NGC_Exporter> in main.cpp.
     - The python version (which is in lib/python/rs274/author.py of 
linuxcnc's repository), can produce arcs - G02/G03 moves.  My code 
currently does not - I am still learning the precise semantics of using 
float numbers in C++ vs Python.
My code produces lines that closely approximate arcs, but somehow things 
do pass the logic test to make an arc.  This is something I'm most 
focused on.
     - The DPA does not like closed paths.  When given a path whose 
endpoints match, the DPA will simplify it right out of existence.  To 
combat this, I've made my buffer object not only break a path into two, 
but also do it when it receives moves that are larger than the tolerance.
     - At 1000dpi, DPA output is a little rough.  At 5000dpi, it looks 
really good, and the file size is much smaller than the regular output 
of pcb2gcode.