5

I'm using Blender 2.9. I created a wedge, I sliced a box diagonally. I turned on the statistic, I see:

  • 1 object
  • 6 vertices
  • 8 edges
  • 4 faces 8 triangles

But why there are 8 triangles, can someone explain?

enter image description here

RonPringadi
  • 151
  • 1
  • 4
  • 4
    Related https://blender.stackexchange.com/questions/102597/finding-vertices-edges-faces-and-tris-using-python as noted in answer below a mesh is tessellated into triangles. A face made up of n verts will be split into (n - 2) triangles. Eg for a a triangular prism (eg a 3 vert cylinder) has (6verts 5 faces 9 edges) with 3 quad faces (3 x (4 - 2)) and two triangle faces (1 x (3 - 2)) = 8 triangles. – batFINGER Feb 21 '21 at 04:04
  • @susu wouldn't have bothered deleting answer over my query (think it answers the gist of the question) Simply intrigued by the geom count. is it 4 quads all sharing 2 edges? or is it..? – batFINGER Feb 21 '21 at 16:57
  • 1
    The quad faces are naturally split by blender’s internal system. As I understand it, no rendering engine really draws four sided faces. They always draw triangles, then maybe outline “defined” edges. This is why if you bend a four sided face in blender, it comes out as two triangles with a ghost edge across the bent surface. – TheLabCat Feb 21 '21 at 17:28
  • Any 3D software, under the hood actually only understands faces with 3 sides (tris). This is because for any 3 distinct points in space there will always be a single plane that contains all 3 points, so a tri is always planar. The same cannot be said about 4 points in space, as the fourth point might be offset away from the plane. So every quad is actually made up of two tris "under the hood", even if the edge that splits the quad into two tris is not displayed to the user. The same is true for n-gons, they are always broken up into tris even if you can't see them. – Alexandre Marcati Aug 17 '21 at 16:47
  • Something is wrong, but it's not the tri count. A prism should have 5 faces and 9 edges, yours has 4 faces and 8 edges. Can you show us a wireframe so we can see all edges of the shape? – Alexandre Marcati Aug 17 '21 at 17:08

3 Answers3

4

All render engines use triangles.

In blender, each face with 4 vertices is in reality two triangles.

Each of the rectangular faces in the shape you show are comprised of two triangles (see the green lines in the example)

enter image description here

enter image description here

susu
  • 14,002
  • 3
  • 25
  • 48
2

Faces are tessellated into triangles.

Due to sense of deja vu re comment on question and since removed answer thought I'd add an answer

All faces are tessellated into triangles on render. The statistics gives us this conversion count using formula outlined here

Finding Vertices, Edges, Faces, and Tris using Python

A face made up of n verts will be split into (n - 2) triangles. Eg for a a triangular prism (eg a 3 vert cylinder) has (6verts 5 faces 9 edges) with 3 quad faces, and two triangles $$(3 \times(4 - 2)) + (2 \times (3 - 2)) = 8$$ triangles

Re question image.

One way to achieve the counts in question image

enter image description here

is to dissolve the edge between quads effectively making an ugly duck 6 vert ngon

>>> for f in C.object.data.polygons:
...     len(f.vertices)
...     
6
3
3
4

which once again

$$4 + 1 + 1 + 2 = 8$$

batFINGER
  • 84,216
  • 10
  • 108
  • 233
  • if you dissolve the edge, blender removes everything but one edge and two vertices. How did you manage to accomplish the equivalent? – Marty Fouts Aug 17 '21 at 21:27
  • This can be done by dissolving edge but not dissolving vertices in the adjust last operator. – Tiles Aug 22 '21 at 07:53
  • @MartyFouts missed comment, prob should have elaborated eg dissolve edge between two quads creates a right angle bent 6 vert ngon, 4, 4 -> 6 (or between quad and tri. 4, 3 -> 5) Resultt is an "ugly duck". non-manifold geometry, which tessellates weirdly when transformed in view. For those that cannot see the deleted answer of susu, which FWIW I upvoted, goes over the theory and explains the tri count using a tri prism example. Made a comment re the q geometry, susu chose to remove answer, which IMO was unnecessary, since it's (IMO) the gist of the question. Have attempted to cover both – batFINGER Aug 22 '21 at 08:15
  • So then you agree with me that it's non-manifold geometry. Do you agree that the original description "sliced a box diagonally" doesn't produce the result? – Marty Fouts Aug 22 '21 at 15:33
  • Does it matter? My guess is OP was wondering why 8 tris not 2. Would need to ask OP & ditto for any info re method for object creation. Certainly not worth downvoting any of the answers is it? – batFINGER Aug 22 '21 at 16:03
  • At some level, none of this really matters; but if stack exchange is about identifying the best answer, then, as any good reference librarian will tell you, it matters that we don't know what the OP really did and are speculating about both that and what they really wanted to know. This question doesn't really show it, but I see a lot of misleading answers, including some of my own, based on bad guesses about the OP's intent. It also seems to matter to the persons who have voted down my answer. – Marty Fouts Aug 23 '21 at 16:09
1

Any 3D software under the hood actually only 'understands' faces with 3 sides (tris). This is because for any 3 distinct points in space there will always be a single plane that contains all 3 points, so a tri is always planar. The same cannot be said about 4 points in space, as the fourth point might be offset away from the plane. So every quad is actually made up of two tris "under the hood", even if the edge that splits the quad into two tris is not displayed to the user. The same is true for n-gons, they are always broken up into tris even if you can't see them.

So in the stats the tri count includes the tris that make up the quads, each quad is 2 tris.

That's why n-gons are considered bad topology, especially for meshes that will be deformed by an armature: you don't know how the software will split the n-gon into tris and so you don't have as much control over how exactly it will deform, and different softwares might triangulate differently from each other, so if you have to import the mesh to a different software things might not look the same.

Alexandre Marcati
  • 1,805
  • 8
  • 31
  • but that doesn't work in this case, because there are five faces if you do what the OP claimed to have done, and the stats in the question say 4. – Marty Fouts Aug 17 '21 at 21:29
  • Well, the question was about the tri count and that's the answer. But you are right, something is wrong in his face and vert count (not the tri count). I asked him in the comments for a wireframe image of his mesh. Something is wrong, but it's not the tri count. He probably has a rogue vertex (when he sliced he might have missed the vertex and created a new one very close to the other), or maybe his geometry is non-manifold. – Alexandre Marcati Aug 18 '21 at 13:20
  • The problem is that given that the other counts are wrong, we can only speculate on what's hidden behind the visible faces, so we can't really say that the tri count is right or wrong. – Marty Fouts Aug 18 '21 at 13:52
  • We can say it's right or wrong based on the description. He said he made a "wedge" by slicing a cube diagonally in half. The tri count is what you'd expect from that shape. The vert and edge count on the other hand are not. And there is no 'correct' (manifold, no rogue verts, etc) that can have those counts, it's geometrically impossible. – Alexandre Marcati Aug 18 '21 at 14:05