WebGL 文本 使用字符纹理
在上一篇文章中我们复习了在 WebGL 场景中如何使用纹理绘制文本。技术是很常见的,对一些事物也是极重要的,例如在多人游戏中你想在一个头像上放置一个名字。同时这个名字也不能影响它的完美性。
比方说你想呈现大量的文本,这需要经常改变 UI 之类的事物。前一篇文章给出的最后一个例子中,一个明显的解决方案是给每个字母加纹理。我们来尝试一下改变上一个例子。
var names = [
"anna", // 0
"colin", // 1
"james", // 2
"danny", // 3
"kalin", // 4
"hiro", // 5
"eddie", // 6
"shu",// 7
"brian", // 8
"tami", // 9
"rick", // 10
"gene", // 11
"natalie",// 12,
"evan", // 13,
"sakura", // 14,
"kai",// 15,
];
// create text textures, one for each letter
var textTextures = [
"a",// 0
"b",// 1
"c",// 2
"d",// 3
"e",// 4
"f",// 5
"g",// 6
"h",// 7
"i",// 8
"j",// 9
"k",// 10
"l",// 11
"m",// 12,
"n",// 13,
"o",// 14,
"p",// 14,
"q",// 14,
"r",// 14,
"s",// 14,
"t",// 14,
"u",// 14,
"v",// 14,
"w",// 14,
"x",// 14,
"y",// 14,
"z",// 14,
].map(function(name) {
var textCanvas = makeTextCanvas(name, 10, 26);
相对于为每个名字呈现一个四元组,我们将为每个名字的每个字母呈现一个四元组。
// setup to draw the text.
// Because every letter uses the same attributes and the same progarm
// we only need to do this once.
gl.useProgram(textProgramInfo.program);
setBuffersAndAttributes(gl, textProgramInfo.attribSetters, textBufferInfo);
textPositions.forEach(function(pos, ndx) {
var name = names[ndx];
// for each leter
for (var ii = 0; ii < name.length; ++ii) {
var letter = name.charCodeAt(ii);
var letterNdx = letter - "a".charCodeAt(0);
// select a letter texture
var tex = textTextures[letterNdx];
// use just the position of the 'F' for the text
// because pos is in view space that means it's a vector from the eye to
// some position. So translate along that vector back toward the eye some distance
var fromEye = normalize(pos);
var amountToMoveTowardEye = 150; // because the F is 150 units long
var viewX = pos[0] - fromEye[0] * amountToMoveTowardEye;
var viewY = pos[1] - fromEye[1] * amountToMoveTowardEye;
var viewZ = pos[2] - fromEye[2] * amountToMoveTowardEye;
var desiredTextScale = -1 / gl.canvas.height; // 1x1 pixels
var scale = viewZ * desiredTextScale;
var textMatrix = makeIdentity();
textMatrix = matrixMultiply(textMatrix, makeTranslation(ii, 0, 0));
textMatrix = matrixMultiply(textMatrix, makeScale(tex.width * scale, tex.height * scale, 1));
textMatrix = matrixMultiply(textMatrix, makeTranslation(viewX, viewY, viewZ));
textMatrix = matrixMultiply(textMatrix, projectionMatrix);
// set texture uniform
textUniforms.u_texture = tex.texture;
copyMatrix(textMatrix, textUniforms.u_matrix);
setUniforms(textProgramInfo.uniformSetters, textUniforms);
// Draw the text.
gl.drawElements(gl.TRIANGLES, textBufferInfo.numElements, gl.UNSIGNED_SHORT, 0);
}
});
你可以看到它是如何工作的:
不幸的是它很慢。下面的例子:单独绘制 73 个四元组,还看不出来差别。我们计算 73 个矩阵和 292 个矩阵倍数。一个典型的 UI 可能有 1000 个字母要显示。这是众多工作可以得到一个合理的帧速率的方式。
解决这个问题通常的方法是构造一个纹理图谱,其中包含所有的字母。我们讨论给立方体的 6 面加纹理时,复习了纹理图谱。
下面的代码构造了字符的纹理图谱。
function makeGlyphCanvas(ctx, maxWidthOfTexture, heightOfLetters, baseLine, padding, letters) {
var rows = 1; // number of rows of glyphs
var x = 0; // x position in texture to draw next glyph
var y = 0; // y position in texture to draw next glyph
var glyphInfos = { // info for each glyph
};
// Go through each letter, measure it, remember its width and position
for (var ii = 0; ii < letters.length; ++ii) {
var letter = letters[ii];
var t = ctx.measureText(letter);
// Will this letter fit on this row?
if (x + t.width + padding > maxWidthOfTexture) {
// so move to the start of the next row
x = 0;
y += heightOfLetters;
++rows;
}
// Remember the data for this letter
glyphInfos[letter] = {
x: x,
y: y,
width: t.width,
};
// advance to space for next letter.
x += t.width + padding;
}
// Now that we know the size we need set the size of the canvas
// We have to save the canvas settings because changing the size
// of a canvas resets all the settings
var settings = saveProperties(ctx);
ctx.canvas.width = (rows == 1) ? x : maxWidthOfTexture;
ctx.canvas.height = rows * heightOfLetters;
restoreProperties(settings, ctx);
// Draw the letters into the canvas
for (var ii = 0; ii < letters.length; ++ii) {
var letter = letters[ii];
var glyphInfo = glyphInfos[letter];
var t = ctx.fillText(letter, glyphInfo.x, glyphInfo.y + baseLine);
}
return glyphInfos;
}
现在我们试试看:
var ctx = document.createElement("canvas").getContext("2d");
ctx.font = "20px sans-serif";
ctx.fillStyle = "white";
var maxTextureWidth = 256;
var letterHeight = 22;
var baseline = 16;
var padding = 1;
var letters = "0123456789.abcdefghijklmnopqrstuvwxyz";
var glyphInfos = makeGlyphCanvas(
ctx,
maxTextureWidth,
letterHeight,
baseline,
padding,
letters);
结果如下
现在,我们已经创建了一个我们需要使用的字符纹理。看看效果怎样,我们为每个字符建四个顶点。这些顶点将使用纹理坐标来选择特殊的字符。
给定一个字符串,来建立顶点:
function makeVerticesForString(fontInfo, s) {
var len = s.length;
var numVertices = len * 6;
var positions = new Float32Array(numVertices * 2);
var texcoords = new Float32Array(numVertices * 2);
var offset = 0;
var x = 0;
for (var ii = 0; ii < len; ++ii) {
var letter = s[ii];
var glyphInfo = fontInfo.glyphInfos[letter];
if (glyphInfo) {
var x2 = x + glyphInfo.width;
var u1 = glyphInfo.x / fontInfo.textureWidth;
var v1 = (glyphInfo.y + fontInfo.letterHeight) / fontInfo.textureHeight;
var u2 = (glyphInfo.x + glyphInfo.width) / fontInfo.textureWidth;
var v2 = glyphInfo.y / fontInfo.textureHeight;
// 6 vertices per letter
positions[offset + 0] = x;
positions[offset + 1] = 0;
texcoords[offset + 0] = u1;
texcoords[offset + 1] = v1;
positions[offset + 2] = x2;
positions[offset + 3] = 0;
texcoords[offset + 2] = u2;
texcoords[offset + 3] = v1;
positions[offset + 4] = x;
positions[offset + 5] = fontInfo.letterHeight;
texcoords[offset + 4] = u1;
texcoords[offset + 5] = v2;
positions[offset + 6] = x;
positions[offset + 7] = fontInfo.letterHeight;
texcoords[offset + 6] = u1;
texcoords[offset + 7] = v2;
positions[offset + 8] = x2;
positions[offset + 9] = 0;
texcoords[offset + 8] = u2;
texcoords[offset + 9] = v1;
positions[offset + 10] = x2;
positions[offset + 11] = fontInfo.letterHeight;
texcoords[offset + 10] = u2;
texcoords[offset + 11] = v2;
x += glyphInfo.width;
offset += 12;
} else {
// we don't have this character so just advance
x += fontInfo.spaceWidth;
}
}
// return ArrayBufferViews for the portion of the TypedArrays
// that were actually used.
return {
arrays: {
position: new Float32Array(positions.buffer, 0, offset),
texcoord: new Float32Array(texcoords.buffer, 0, offset),
},
numVertices: offset / 2,
};
}
为了使用它,我们手动创建一个 bufferInfo。(如果你已经不记得了,可以查看前面的文章:bufferInfo 是什么)。
// Maunally create a bufferInfo
var textBufferInfo = {
attribs: {
a_position: { buffer: gl.createBuffer(), numComponents: 2, },
a_texcoord: { buffer: gl.createBuffer(), numComponents: 2, },
},
numElements: 0,
};
使用 bufferInfo 中的字符创建画布的 fontInfo 和纹理:
var ctx = document.createElement("canvas").getContext("2d");
ctx.font = "20px sans-serif";
ctx.fillStyle = "white";
var maxTextureWidth = 256;
var letterHeight = 22;
var baseline = 16;
var padding = 1;
var letters = "0123456789.,abcdefghijklmnopqrstuvwxyz";
var glyphInfos = makeGlyphCanvas(
ctx,
maxTextureWidth,
letterHeight,
baseline,
padding,
letters);
var fontInfo = {
glyphInfos: glyphInfos,
letterHeight: letterHeight,
baseline: baseline,
spaceWidth: 5,
textureWidth: ctx.canvas.width,
textureHeight: ctx.canvas.height,
};
然后渲染我们将更新缓冲的文本。我们也可以构成动态的文本:
textPositions.forEach(function(pos, ndx) {
var name = names[ndx];
var s = name + ":" + pos[0].toFixed(0) + "," + pos[1].toFixed(0) + "," + pos[2].toFixed(0);
var vertices = makeVerticesForString(fontInfo, s);
// update the buffers
textBufferInfo.attribs.a_position.numComponents = 2;
gl.bindBuffer(gl.ARRAY_BUFFER, textBufferInfo.attribs.a_position.buffer);
gl.bufferData(gl.ARRAY_BUFFER, vertices.arrays.position, gl.DYNAMIC_DRAW);
gl.bindBuffer(gl.ARRAY_BUFFER, textBufferInfo.attribs.a_texcoord.buffer);
gl.bufferData(gl.ARRAY_BUFFER, vertices.arrays.texcoord, gl.DYNAMIC_DRAW);
setBuffersAndAttributes(gl, textProgramInfo.attribSetters, textBufferInfo);
// use just the position of the 'F' for the text
var textMatrix = makeIdentity();
// because pos is in view space that means it's a vector from the eye to
// some position. So translate along that vector back toward the eye some distance
var fromEye = normalize(pos);
var amountToMoveTowardEye = 150; // because the F is 150 units long
textMatrix = matrixMultiply(textMatrix, makeTranslation(
pos[0] - fromEye[0] * amountToMoveTowardEye,
pos[1] - fromEye[1] * amountToMoveTowardEye,
pos[2] - fromEye[2] * amountToMoveTowardEye));
textMatrix = matrixMultiply(textMatrix, projectionMatrix);
// set texture uniform
copyMatrix(textMatrix, textUniforms.u_matrix);
setUniforms(textProgramInfo.uniformSetters, textUniforms);
// Draw the text.
gl.drawArrays(gl.TRIANGLES, 0, vertices.numVertices);
});
即:
这是使用字符纹理集的基本技术。可以添加一些明显的东西或方式来改进它。
- 重用相同的数组。
目前,每次被调用时,makeVerticesForString 就会分配新的 32 位浮点型数组。这最终可能会导致垃圾收集出现问题。重用相同的数组可能会更好。如果不是足够大,你也放大数组,但是保留原来的大小。 - 添加支持回车
当生成顶点时,检查 \n 是否存在从而实现换行。这将使文本分隔段落更容易。 - 添加对各种格式的支持。
如果你想文本居中,或调整你添加的一切文本的格式。 - 添加对顶点颜色的支持。
你可以为文本的每个字母添加不同的颜色。当然你必须决定如何指定何时改变颜色。
这里不打算涉及的另一个大问题是:纹理大小有限,但字体实际上是无限的。如果你想支持所有的 unicode,你就必须处理汉语、日语和阿拉伯语等其他所有语言,2015 年在 unicode 有超过 110000 个符号!你不可能在纹理中适配所有这些,也没有足够的空间供你这样做。
操作系统和浏览器 GPU 加速处理这个问题的方式是:通过使用一个字符纹理缓存实现。上面的实现他们是把纹理处理成纹理集,但他们为每个 glpyh 布置一个固定大小的区域,保留纹理集中最近使用的符号。如果需要绘制一个字符,而这个字符不在纹理集中,他们就用他们需要的这个新的字符取代最近最少使用的一个。当然如果他们即将取代的字符仍被有待绘制的四元组引用,他们需要绘制他们之前所取代的字符。
虽然我不推荐它,但是还有另一件事你可以做,将这项技术和以前的技术结合在一起。你可以直接渲染另一种纹理的符号。当然 GPU 加速画布已经这样做了,你可能没有自己动手的理由。
另一种在 WebGL 中绘制文本的方法实际上是使用了 3D 文本。在上面所有的例子中 “F” 是一个 3D 的字母。你已经为每个字母都构成了一个相应的 3D 字符。3D 字母常见于标题和电影标志,此外的用处就少了。
我希望在 WebGL 这可以覆盖文本。