Reflections on Pyramid

[Punn]

http://www.babylonjs-playground.com/#TPE3U

 

Why is there a strange reflection on a plane pyramid side?

[Ahiru]

I think Jerome had a good explanation for it lately:

 

http://www.html5gamedevs.com/topic/17040-the-mystery-of-computenormals/?p=95871

[adam]

To fix it, make the subdivision 1.

 

http://www.babylonjs-playground.com/#TPE3U#1

[jerome]

subdivision = 1 http://www.babylonjs-playground.com/#TPE3U#3

[jerome]

However it should work with the mast version of the cylinder implementation, even with many subdivisions now.

I don't know if it's been pushed in the PG engine

[GameMonetize]

I think so

[jerome]

mmmh.. the distorded reflection is typically the problem encountered with computeNormals() on vertices belonging to exacly 3 coplanar faces : http://www.babylonjs-playground.com/#TPE3U#4

 

However the new cylinder re-implementation doesn't use computeNormal() any longer but dedicated computed normals from the cylinder geometry itself.

I tested it, there were no artifact, even with many subdivisions (I didn't test tessellation = 4 ...) .

[jerome]

Ok, got it : http://www.babylonjs-playground.com/#TPE3U#5

 

Well, it's rightly the same problem than encountered with computeNormals() about 2 coplanar facets having vertices with different normals due to their but ... without computeNormals() !

Because there is only a quad per face and many subdivisions, we create here this specific case : 2 coplanar triangles, but different normalson each, so the quad plane has no 4 similar normals => visual artifact.

 

I'm afraid, in this very case, the solutions are to set the subdivisions to 1 OR to set your normals afterwards with a computation dedicated to the specific geometry.

 

 

NOTE : this means computeNormals() is not magic, but dedicated computed computed are neither

[jerome]

If you really need to keep many cylinder subdivisions (instead of 1 to avoid the artifact), you could convert your mesh to a flat shaded one : http://www.babylonjs-playground.com/#TPE3U#6

 

Each mesh face will then have its own normals not depending from the other side vertices. This is a really simple way to deal with artifacts on meshes which have edge angles lower than 120°

[Punn]

To fix it, make the subdivision 1.

 

http://www.babylonjs-playground.com/#TPE3U#1

 

Still some artefact? And why is the tip of the pyramid cut now?

[jerome]

oops

I need to check this

It seems the 0 top diameter isn't taken in account now, I don't understand why ...

I'll check tomorrow (I have a lead in vertexdata.createcylinder default parameter handling)

 

In this case, the artifact is due to this not zero top diameter : http://www.babylonjs-playground.com/#TPE3U#7

[jerome]

gotcha : https://github.com/BabylonJS/Babylon.js/blob/master/src/Mesh/babylon.mesh.vertexData.ts#L790

 

JS lazy evaluation :

diameterTop = diameterTop || options.diameterTop || 1;

if diameterTop is falsy then options.diameterTop else 1 ...

But 0 is falsy in JS !

I'm fixing this right now.

 

 

@DK : I think we've got the same problem on each sideOrientation evaluation :

sideOrientation = sideOrientation || options.sideOrientation || Mesh.DEFAULTSIDE;

This will work only as long as the constant DEFAULTSIDE is set to zero.

If we change it some day, the lazy evaluation might come to strange results.

ex : if DEFAULT is set to 1 and we call the method by requesting sideOrientation: FRONTSIDE, what is zero => we'll get DEFAULT (== 1) instead

 

Maybe something like this would be better :

            if (sideOrientation === 0 || options.sideOrientation === 0) {                sideOrientation = 0;            } else {                sideOrientation = sideOrientation || options.sideOrientation || Mesh.DEFAULTSIDE;            }

[jerome]

fixed and PRed.

 

and the subdivisions parameter of CreateCylinder is optional again