Is Mesh Coordinate Precision Single?

I am using OnMesh Class for geometric construction, and I have realized that the vertex coordinates are represented as "float" in VC#.

Could anyone tell me whether this means that a mesh has only single precision or there is no difference between "float" and "double"?
Also, does this also generally apply to mesh in Rhino?

Thank you.

Replies to This Discussion

Permalink Reply by Damien Alomar on September 23, 2009 at 8:30am

Since meshes will tend to have a large number of vertices, the programmers at Rhino decided to make the x,y and z values made of of single precision floating point numbers as opposed to double precision floating point numbers that are used every where else. Its more or less, just using a slightly less accurate number to save space. If you dig deeper, you'll find that the vertices are actually a different class than normal points... On3fPoint as opposed to On3dPoint. Does that cover it?

Permalink Reply by Tomohiro Tachi on September 23, 2009 at 9:20am

Thank you for the answer. That makes sense.

What should be the standard way to construct a polysurface from vertex coordinates and connectivity?
(I have been constructing simple polyhedral models by constructing meshes and converting them into polysurfaces, but it doesn't seem to be a good idea in the sense of precision.)

For example is using
RhUtil.RhinoCreateSurfaceFromCorners
to construct facets and then connecting them by
RhUtil.RhinoJoinBreps
an efficient way?

Permalink Reply by Damien Alomar on September 23, 2009 at 10:31am

Well, from you're description I don't see any benefit of having NURBS over meshes. The precision issue is not really a very big one, unless you're dealing with extremely small objects and/or a lot of very precise calculations.

If you absolutely must have NURBS, then either the BrepForm method will give you access to it, or you can use OnUtil.ON_BrepFromMesh function.

Permalink Reply by Tomohiro Tachi on September 23, 2009 at 12:33pm

I was comparing these two processes:
coordinates -> mesh -> polysurface
coordinates -> polysurface
The former is what I have been doing, and I was curious about the latter possibility.

I want to make it precise as possible, since I am programming calculations that can easily accumulate errors, although thus far, I have not yet experienced any precision problem.

Thanks for the advice, anyway. ON_BrepFromMesh sounds nice for the mesh -> polysurface part.

RSS