@aeva it can either only be this or max(abs(dx),abs(dy)). and if you think it's neither, then i'd say it still is, just hidden within whatever it is you did.
The rendering is orthographic, and I'm drawing a cone per seed point. If the base of each cone were to be round, it would look like a regular voronoi diagram, but you can change the distance metric by dropping the vertex count of the cones. These have just 3 verts in the base :3
so idk what this is worth, but I realized that you can also just give every voronoi cell a different metric if you want and it'll just work lol. I'll have to try that out some other time.
@aeva@ruojake a nice trick you can use for drawing voronoi diagrams is to use the depth buffer. You draw a shape around the cell, and give each pixel a depth value based on distance to the center of the cell. Then let the depth buffer figure out which cell is closest. And you can have each cell compute distance however it wants 🙃
@lritter lots of overdraw in this case because I'm erring towards very large cone bases but I think it still comes out ok because of early z elimination
@aeva wow, I never thought of doing voronoi that way. Is that an established optimization? something you came up with just now? also it makes for some really cool looking isometric cityscapes apparently (at least that's what my brain sees)
@peacememories I have no idea if this is an established method. I just sat down to implement JFA with point drawing and then realized that was way overkill.
@aeva this is like nerd sniping catered entirely for me... I want to guess it's a distance calculation being wrong somewhere given the diagonal angles?
@aeva I WAS supposed to doing some relatively boring routine api-database coding today, but instead I seem to be reading up on voroni diagrams and taxicab geometry and ... well... I obviously have no idea of why yours look wonky, but this is still way more fun than yet an other rest api.
Add comment