I am working on a large terrain and did a test to check if storing the vertices in a pointcloud could be more memory efficient than in a mesh with all faces stored too, but I see that in a test the mesh is 7MB and the pointcloud of the vertices is 11.4MB, that is more than 50% larger.
Why is that? Are meshes stored with a compression and pointclouds not, or are pointclouds of higher accuracy? Neither have colors stored to either vertices nor points.
Excellent questions.
Point clouds are stored in double precision, while meshes are stored in single precision. While it is possible to change the storage precision in meshes to double precision, this is not the default. This means that vertex storage takes 50%less space for meshes. Meshes also store face connectivity, so the difference is less than 50% overall.
So, an empty Mesh object, add only the vertices and nothing else (no faces, colors, normals, etc) would be the most “efficient” way to store a big list of points?
… if you don’t care about double precision…
(it would be an invalid mesh, of course…)
Keep in mind that compressing and uncompressing can be time consuming.
– Dale
Thanks, good to know. And that is a good point Riccardo, and mesh with no faces is a good idea. I’ll test it out. It should be just as accurate as the mesh we use for the terrain today.
@menno or @dale if I want to store a mesh object in Rhino with out adding it as a “physical object”, what is the best approach to do that?
From a plug-in? User data…
– Dale