Add comments explaining compute_bowyer_envelope!
#38
Conversation
| for i in 2:num_triangles(all_triangles) | ||
| v = get_adjacent(temp_adjacent, I(n + 1), v) | ||
| #= | ||
| Get the vertex associated with the edge `(point, points[v])` such that the | ||
| returned point (the new `v`) forms a positively oriented triangle | ||
| with that edge. Then, add the newly found `v` to the envelope. | ||
| To be clear, `n+1` is the index of the newly inserted vertex in the triangulation. | ||
| =# | ||
| v = get_adjacent(temp_adjacent, I(n + 1), v) | ||
| push!(envelope, v) | ||
| end |
There was a problem hiding this comment.
I'm not 100% sure what this loop is doing - is there some guarantee that all triangles added have n+1 as a vertex, and we're just getting each individual natural neighbour from this loop?
There was a problem hiding this comment.
That loop is building up the "envelope". All triangles will have n+1 as a vertex. I'll draw a picture
There was a problem hiding this comment.
Gotcha - that's what I thought was happening but just wanted to confirm. Thanks!
There was a problem hiding this comment.
By pivoting around n+1 (which I've called i+1 in the image for some reason), we can traverse the boundary of the envelope counter-clockwise by looping over the triangles. The orientation is important since we compute areas later
|
I also have some notes on this here https://juliageometry.github.io/DelaunayTriangulation.jl/dev/applications/interpolation/ and https://danielvandh.github.io/NaturalNeighbours.jl/dev/interpolation_math/. I don't know how much I talk about the internals in either of those links though. The documentation here isn't my best unfortunately |
|
Thanks! |
I was reading the code here and struggled a bit when interpreting it, so added some comments. Feel free to accept/reject/edit as necessary - some of these may be unnecessary.