How to use multiple textures in WebGL?

Question

I wanted to use 6 different textures on a cube, one per side, but can\'t find the mistake. Here\'s my current code:

   var texturen = new Array();
function          


        

Answer 1

First off, the short answer: Replace the last 10 lines of your code with the following and I think it should work.

gl.activeTexture(gl.TEXTURE0);
gl.bindTexture(gl.TEXTURE_2D, texturen[1]); 
gl.vertexAttribPointer(textureCoordAttribute, cubeVertexTextureCoordBuffer1.itemSize,  gl.FLOAT, false, 0, 0);
gl.bindBuffer(gl.ARRAY_BUFFER, cubeVertexPositionBuffer1);
gl.vertexAttribPointer(shaderProgram.vertexPositionAttribute,cubeVertexPositionBuffer1.itemSize, gl.FLOAT, false, 0, 0);

gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, cubeVertexIndexBuffer);
gl.drawElements(gl.TRIANGLES, cubeVertexIndexBuffer.numItems, gl.UNSIGNED_SHORT, 0);

The key changes are:

• activeTexture is still TEXTURE0 (you're only using one texture at a time here)
• Using cubeVertexIndexBuffer instead of cubeVertexIndexBuffer1, which contained out-of-range indicies

For a better explanation of what the indices are actually used for I'll refer you to this SO question as I'd rather not repeat all that here.

Now, on to the more general answer.

There are two basic ways to handle the issue of using different textures on different faces. The first and simpler method is to do exactly what you are doing here: Render the object in pieces, binding a different texture for each piece. Though not strictly the most efficient way to accomplish the effect it is nevertheless the most common way to handle it in high-performance applications such as games simply because it offers a lot of flexibility, especially when your materials are more complex than a simple diffuse texture.

There is a simple way that you can improve the performance of code like yours for a case like this, though. The vertices/indices don't have to be broken into separate buffers per-texture. You could instead combine them all into a single buffer and render them with different offsets like so:

// On Init
var vertBuffer = gl.createBuffer();
gl.bindBuffer(gl.ARRAY_BUFFER, vertBuffer);
var vertices = [
    // Vertex values for all 6 faces
];
gl.bufferData(gl.ARRAY_BUFFER, new Float32Array(vertices), gl.STATIC_DRAW);

var indexBuffer = gl.createBuffer();
gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, indexBuffer);
var indices = [
    // Index values for all 6 faces
];
gl.bufferData(gl.ELEMENT_ARRAY_BUFFER, new Uint16Array(indices), gl.STATIC_DRAW);

// On Draw
// Setup the shader and uniforms and all that jazz

gl.bindBuffer(gl.ARRAY_BUFFER, vertBuffer);
gl.bindBuffer(gl.ELEMENT_ARRAY_BUFFER, indexBuffer);
gl.vertexAttribPointer(// Blah blah blah...);

// Draw face 0
gl.bindTexture(gl.TEXTURE_2D, texture[0]);
gl.drawElements(gl.TRIANGLES, 6, gl.UNSIGNED_SHORT, 0);

// Draw face 1
gl.bindTexture(gl.TEXTURE_2D, texture[1]);
gl.drawElements(gl.TRIANGLES, 6, gl.UNSIGNED_SHORT, 12);

// Draw face 2
gl.bindTexture(gl.TEXTURE_2D, texture[2]);
gl.drawElements(gl.TRIANGLES, 6, gl.UNSIGNED_SHORT, 24);

// .. And so on to face 5
// Draw face 2
gl.bindTexture(gl.TEXTURE_2D, texture[5]);
gl.drawElements(gl.TRIANGLES, 6, gl.UNSIGNED_SHORT, 60);

What's going on here is that each call to drawElements only draws 2 triangles (6 vertices, the second parameter of the call), but each call is offset into the index buffer so that it starts on a different face. (fourth parameter of the call, which indicates a byte offset. Each index is a Uint16, so 2 bytes per index. 12 == "start on index[6]") This way all of the binding and setup only happens once, and each draw call only has to change the state that is actually necessary (the texture).

The other way to handle this, which is faster but harder to generalize, is to bind an array of textures to a shader uniform and use another vertex attribute to index into the array within the shader. I'm not going to detail the code for this approach but it should be relatively easy to set up once you're comfortable with WebGL shader use. If you have further questions about this specific approach it would be best to ask in a different SO question so as not to overload this one.