[raster] ST_Drape

[Bborie Park]

Taking a geometry and a raster, return a new XYM geometry resulting from overlaying the geometry on the raster and populating M with values from the raster.

Reference:

​http://vterrain.org/Misc/draping.html

geometry ST_Drape(
  rast raster, nband int,
  geom geometry,
  computed_vertices boolean DEFAULT TRUE
)

Some things to think about before implementing...

1. XYM or XYZ output geometry: Assuming the raster is DEM, units are typically meter or foot. If X/Y are degrees, populating Z in the DEM's units don't make sense. Maybe a function parameter flag indicating output should be XYZ or XYM?

2. 3D geometry with rasters: The geometry is treated as 2D. Should pixel values be put into Z or M?

3. computed_vertices: The returning geometry should be as detailed as the input raster. So, returning geometry will have the same vertices of the input geometry and possibly.

4. Polygons: How to handle the interior of the polygon? Boundary is easy. Multipolygon? TIN?

[Bborie Park]

For polygons, use Delaunay Triangles capability from GEOS. Should have something similar to how ST_DelaunayTriangles() allows the user to control output between a TIN, MULTILINESTRING or COLLECTION of polygons.

[smathermather]

Also relevant: ​http://blog.mathieu-leplatre.info/drape-lines-on-a-dem-with-postgis.html

Re: 2) I'd say keep 3D, add M, and if the user wants to make it Z, then they should convert to 2D before draping.

[dmcclean]

Amen. It would be nice to have a tolerance parameter like the one to ST_Simplify that adaptively downsamples from the full DEM-resolution/input-resolution result.

[robe]

Milestone renamed

[robe]

This was done in PostGIS 3.2 and is called ST_SetM ​http://postgis.net/docs/manual-3.2/RT_ST_SetM.html