Exisiting Triangulation

[rmchurch]

Nice work, this was quite the undertaking. I don't have a firm grasp on the code yet, but could it be sped up if there is an exisiting triangulation? (i.e. can I get rid of the Delaunay calls if I have by own triObj = Triangulation(x,y) and I normally call plt.tricontourf(triObj,data)). My hope was in cases such as animations of a stack of images of data on a triangulated mesh, this could perhaps significantly speed up the animation.

[asher-pembroke]

Thanks! Yes, I think it's feasible to use a pre-computed triangulation and call tricontourf instead (I didn't even know that function existed!). I'll look at the output from tricontourf: hopefully the paths will work the same. Thanks for the suggestion!

[asher-pembroke]

This enhancement seems a bit subtler than I thought. The problem is that the contour vertices are not in the graph of your triangulation, but they probably lie somewhere along on the triangulation edges. Maybe there's a way to identify which edges each contour vertex belongs to, then find the paths that join different contours. I will have to think about it more.