12_transforming_spatial_3D.Rmd

# Transforming spatial data to 3D forms

## Questions

* What tools exist to convert data to meshes in R? 

## Overview

* Teaching: 10 min

## Transforming Spatial Data using modular 3D tools {.exercise}

1. Tell us if you use 3D visualization or meshes in analysis. 


This is hard! There's a few scattered tools. 

## On CRAN 

* `rgl::triangulate` and `decido::earcut` will triangulate polygons with holes but only suitable for plane-filling, because ugly triangles, no control over size and shape. 

* `RTriangle::triangulate` (and `sfdct::ct_triangulate`) do high-quality "near-Delaunay" triangulations

* `quadmesh::quadmesh` to create rgl-ready mesh3d from a raster

* `mapview::cubeView` does very compelling interactive raster-cube visualization

* `SymbolixAU/mapdeck`

* `tylermorganwall/rayshader`


Side note: there are many triangulation algorithms and many packages in R, but we need *constrained triangulation* to preserve all input edges - only a handful can do that, and RTriangle is the king (with a problematic license). 


## Quadmesh. 


```{r, quadmesh, eval=FALSE}
library(quadmesh)
quadmesh(anyRasterDEM, texture = anyRasterRGB)
rgl::shade3d()

## play with aspect3d, light3d, ...

```

Triangulations, `sfdct` is no good because it's very inefficient. `sf` is simply not suitable for mesh (a.k.a. *indexed*) forms of data. 


## Stuff not on CRAN!

* `hypertidy/silicate`, `hypertidy/anglr` - these are evolving together 

* this workshop!  https://github.com/MilesMcBain/gis_vs_web3D 

* `coolbutuseless/threed`


## Triangles or quads in hypertidy (WIP)

This is my work-in-progress approach to meshing any data structure. 


```{r, eval=FALSE}
## devtools::install_github("hypertidy/anglr")
## devtools::install_github("hypertidy/silicate")
library(anglr)  
triangles <- copy_down(TRI(anySFpolygon), anyRasterDEM)

mesh <- plot3d(triangles)

```