How color attributes work in VBO?

Question

I am coding to OpenGL ES 2.0 (Webgl). I am using VBOs to draw primitives. I have vertex array, color array and array of indices. I have looked at sample codes, books and tutorial, but one thing I don't get - if color is defined per vertex how does it affect the polygonal surfaces adjacent to those vertices? (I am a newbie to OpenGL(ES))

I will explain with an example. I have a cube to draw. From what I read in OpenGLES book, the color is defined as an vertex attribute. In that case, if I want to draw 6 faces of the cube with 6 different colors how should I define the colors. The source of my confusion is: each vertex is common to 3 faces, then how will it help defining a color per vertex? (Or should the color be defined per index?). The fact that we need to subdivide these faces into triangles, makes it harder for me to understand how this relationship works. The same confusion goes for edges. Instead of drawing triangles, let's say I want to draw edges using LINES primitives. Each edge of different color. How am I supposed to define color attributes in that case?

I have seen few working examples. Specifically this tutorial: http://learningwebgl.com/blog/?p=370

I see how color array is defined in the above example to draw a cube with 6 different colored faces, but I don't understand why is defined that way. (Why is each color copied 4 times into unpackedColors for instance?)

Can someone explain how color attributes work in VBO?

[The link above seems inaccessible, so I will post the relevant code here]

cubeVertexPositionBuffer = gl.createBuffer();
gl.bindBuffer(gl.ARRAY_BUFFER, cubeVertexPositionBuffer);
vertices = [
  // Front face
  -1.0, -1.0,  1.0,
   1.0, -1.0,  1.0,
   1.0,  1.0,  1.0,
  -1.0,  1.0,  1.0,

  // Back face
  -1.0, -1.0, -1.0,
  -1.0,  1.0, -1.0,
   1.0,  1.0, -1.0,
   1.0, -1.0, -1.0,

  // Top face
  -1.0,  1.0, -1.0,
  -1.0,  1.0,  1.0,
   1.0,  1.0,  1.0,
   1.0,  1.0, -1.0,

  // Bottom face
  -1.0, -1.0, -1.0,
   1.0, -1.0, -1.0,
   1.0, -1.0,  1.0,
  -1.0, -1.0,  1.0,

  // Right face
   1.0, -1.0, -1.0,
   1.0,  1.0, -1.0,
   1.0,  1.0,  1.0,
   1.0, -1.0,  1.0,

  // Left face
  -1.0, -1.0, -1.0,
  -1.0, -1.0,  1.0,
  -1.0,  1.0,  1.0,
  -1.0,  1.0, -1.0,
];
gl.bufferData(gl.ARRAY_BUFFER, new WebGLFloatArray(vertices), gl.STATIC_DRAW);
cubeVertexPositionBuffer.itemSize = 3;
cubeVertexPositionBuffer.numItems = 24;

cubeVertexColorBuffer = gl.createBuffer();
gl.bindBuffer(gl.ARRAY_BUFFER, cubeVertexColorBuffer);
var colors = [
  [1.0, 0.0, 0.0, 1.0],     // Front face
  [1.0, 1.0, 0.0, 1.0],     // Back face
  [0.0, 1.0, 0.0, 1.0],     // Top face
  [1.0, 0.5, 0.5, 1.0],     // Bottom face
  [1.0, 0.0, 1.0, 1.0],     // Right face
  [0.0, 0.0, 1.0, 1.0],     // Left face
];
var unpackedColors = []
for (var i in colors) {
  var color = colors[i];
  for (var j=0; j < 4; j++) {
    unpackedColors = unpackedColors.concat(color);
  }
}
gl.bufferData(gl.ARRAY_BUFFER, new WebGLFloatArray(unpackedColors), gl.STATIC_DRAW);
cubeVertexColorBuffer.itemSize = 4;
cubeVertexColorBuffer.numItems = 24;

cubeVertexIndexBuffer = gl.createBuffer();
gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, cubeVertexIndexBuffer);
var cubeVertexIndices = [
  0, 1, 2,      0, 2, 3,    // Front face
  4, 5, 6,      4, 6, 7,    // Back face
  8, 9, 10,     8, 10, 11,  // Top face
  12, 13, 14,   12, 14, 15, // Bottom face
  16, 17, 18,   16, 18, 19, // Right face
  20, 21, 22,   20, 22, 23  // Left face
]
gl.bufferData(gl.ELEMENT_ARRAY_BUFFER, new WebGLUnsignedShortArray(cubeVertexIndices), gl.STATIC_DRAW);
cubeVertexIndexBuffer.itemSize = 1;
cubeVertexIndexBuffer.numItems = 36;

Ok I figured the above example code defines same vertices multiple times. As you can see, a cube has 8 vertices, but this code has 24 of them. That's how it's specifying 3 different colors for the same vertex depending upon which face it is part of. Is this the only way to do it? Isn't it too wasteful? — Jayesh

Unknown Unknown · Accepted Answer · 2010-05-06T16:13:18

The way I like to look at it is that each vertex is not a point in space but rather a bundle of attributes. These generally (but not always) include its location and may include its colour, texture coordinates, etc., etc., etc. A triangle (or line, or other primitive) is defined by specifying a set of vertices, then generating values for each attribute at each pixel by linearly interpolating the per-vertex values.

As Liam says, and as you've realised in your comment, this means that if you want to have a point in space that is used by a vertex for multiple primitives -- for example, the corner of a cube -- with other non-location attributes varying on a per-primitive basis, you need a separate vertex for each combination of attributes.

This is wasteful of memory to some degree -- but the complexity involved in doing it any other way would make things worse, and would require the graphics hardware to do a lot more work unpacking and repacking data. To me, it feels like the waste is comparable to the waste we get by using 32-bit RGBA values for each pixel in our video memory instead of keeping a "palette" lookup table of every colour we want to use and then just storing an index into that per-pixel (which is, of course, what we used to do when RAM was more expensive).

How color attributes work in VBO?

2 Answers