|
| 1 | +--- |
| 2 | +title: The Burrows-Wheeler Transform 块排序压缩算法 |
| 3 | +date: 2026-01-01 22:11:23 |
| 4 | +updated: 2026-01-01 22:11:23 |
| 5 | +categories: |
| 6 | + - 杂项 |
| 7 | +tag: |
| 8 | + - 算法 |
| 9 | +math: true |
| 10 | +description: 一种非常有趣的块排序压缩算法 |
| 11 | +hide: false |
| 12 | +sticky: false |
| 13 | +--- |
| 14 | + |
| 15 | +# 前言 |
| 16 | + |
| 17 | +最近发现了一个非常有意思的算法:The Burrows-Wheeler Transform(块排序压缩算法),所以稍微花点时间简单记录一下 |
| 18 | + |
| 19 | +说是一个压缩算法,实际上这个算法主要做的事情是把数据重排序,实际上并没有减少数据的长度(通常反而因为增加了行尾标识而增长了) |
| 20 | + |
| 21 | +但是它完成了一个非常有意思的效果:在几乎不增加字符串长度的情况下,把一个字符串的重复的部分尽可能聚合到一块了 |
| 22 | + |
| 23 | +而 Gzip 压缩算法基于Deflate算法,通过结合LZ77算法(查找并替换重复字符串)和霍夫曼编码,通过将接近的字符串聚合到一块,能够有效增加压缩率 |
| 24 | + |
| 25 | +# 算法操作方式 |
| 26 | + |
| 27 | +我做了一个简单的演示工具,我们就用经典的 `banana` 作为示例 |
| 28 | + |
| 29 | +<div id="v"> |
| 30 | + <div class="demo-block"> |
| 31 | + <div>在原串后添加结束结束符号<span class="demo-mono red">$</span>,且此符号认为是最小的字符</div> |
| 32 | + <div class="demo-window"> |
| 33 | + banana<span class="red">\$</span> |
| 34 | + </div> |
| 35 | + </div> |
| 36 | + <div class="demo-block"> |
| 37 | + <div>生成字符串的全部循环序列</div> |
| 38 | + <div class="demo-window"> |
| 39 | + banana<span class="red">\$</span><br> |
| 40 | + anana<span class="red">\$</span>b<br> |
| 41 | + nana<span class="red">\$</span>ba<br> |
| 42 | + ana<span class="red">\$</span>ban<br> |
| 43 | + na<span class="red">\$</span>bana<br> |
| 44 | + a<span class="red">\$</span>banan<br> |
| 45 | + <span class="red">\$</span>banana |
| 46 | + </div> |
| 47 | + </div> |
| 48 | + <div class="demo-block"> |
| 49 | + <div>将这几个字符串排序</div> |
| 50 | + <div class="demo-window"> |
| 51 | + <span class="red">\$</span>banana<br> |
| 52 | + a<span class="red">\$</span>banan<br> |
| 53 | + ana<span class="red">\$</span>ban<br> |
| 54 | + anana<span class="red">\$</span>b<br> |
| 55 | + banana<span class="red">\$</span><br> |
| 56 | + na<span class="red">\$</span>bana<br> |
| 57 | + nana<span class="red">\$</span>ba |
| 58 | + </div> |
| 59 | + </div> |
| 60 | + <div class="demo-block"> |
| 61 | + <div>取出最后一列字符串</div> |
| 62 | + <div class="demo-window"> |
| 63 | + <span class="red">\$</span>banan<span class="green">a</span><br> |
| 64 | + a<span class="red">\$</span>bana<span class="green">n</span><br> |
| 65 | + ana<span class="red">\$</span>ba<span class="green">n</span><br> |
| 66 | + anana<span class="red">\$</span><span class="green">b</span><br> |
| 67 | + banana<span class="green">\$</span><br> |
| 68 | + na<span class="red">\$</span>ban<span class="green">a</span><br> |
| 69 | + nana<span class="red">\$</span>b<span class="green">a</span> |
| 70 | + </div> |
| 71 | + </div> |
| 72 | + <div class="demo-block"> |
| 73 | + <div>得到结果</div> |
| 74 | + <div class="demo-window"> |
| 75 | + annb<span class="red">\$</span>aa |
| 76 | + </div> |
| 77 | + </div> |
| 78 | + <div class="demo-block"> |
| 79 | + <div>下面将进行还原操作</div> |
| 80 | + <div class="demo-window"> |
| 81 | + annb<span class="red">\$</span>aa |
| 82 | + </div> |
| 83 | + </div> |
| 84 | + <div class="demo-block"> |
| 85 | + <div>将结果排成一列</div> |
| 86 | + <div class="demo-window"> |
| 87 | + <span class="green">a</span><br> |
| 88 | + <span class="green">n</span><br> |
| 89 | + <span class="green">n</span><br> |
| 90 | + <span class="green">b</span><br> |
| 91 | + <span class="green">\$</span><br> |
| 92 | + <span class="green">a</span><br> |
| 93 | + <span class="green">a</span> |
| 94 | + </div> |
| 95 | + </div> |
| 96 | + <div class="demo-block"> |
| 97 | + <div>排序</div> |
| 98 | + <div class="demo-window"> |
| 99 | + <span class="green">\$</span><br> |
| 100 | + <span class="green">a</span><br> |
| 101 | + <span class="green">a</span><br> |
| 102 | + <span class="green">a</span><br> |
| 103 | + <span class="green">b</span><br> |
| 104 | + <span class="green">n</span><br> |
| 105 | + <span class="green">n</span> |
| 106 | + </div> |
| 107 | + </div> |
| 108 | + <div class="demo-block"> |
| 109 | + <div>在当前的列之前添加 BWT 结果</div> |
| 110 | + <div class="demo-window"> |
| 111 | + <span class="green">a</span><span class="gray">\$</span><br> |
| 112 | + <span class="green">n</span><span class="gray">a</span><br> |
| 113 | + <span class="green">n</span><span class="gray">a</span><br> |
| 114 | + <span class="green">b</span><span class="gray">a</span><br> |
| 115 | + <span class="green">\$</span><span class="gray">b</span><br> |
| 116 | + <span class="green">a</span><span class="gray">n</span><br> |
| 117 | + <span class="green">a</span><span class="gray">n</span> |
| 118 | + </div> |
| 119 | + </div> |
| 120 | + <div class="demo-block"> |
| 121 | + <div>再次排序</div> |
| 122 | + <div class="demo-window"> |
| 123 | + <span class="green">\$</span><span class="gray">b</span><br> |
| 124 | + <span class="green">a</span><span class="gray">\$</span><br> |
| 125 | + <span class="green">a</span><span class="gray">n</span><br> |
| 126 | + <span class="green">a</span><span class="gray">n</span><br> |
| 127 | + <span class="green">b</span><span class="gray">a</span><br> |
| 128 | + <span class="green">n</span><span class="gray">a</span><br> |
| 129 | + <span class="green">n</span><span class="gray">a</span> |
| 130 | + </div> |
| 131 | + </div> |
| 132 | + <div class="demo-block"> |
| 133 | + <div>重复上述步骤</div> |
| 134 | + <div class="demo-window"> |
| 135 | + <span class="green">a</span><span class="gray">\$b</span><br> |
| 136 | + <span class="green">n</span><span class="gray">a\$</span><br> |
| 137 | + <span class="green">n</span><span class="gray">an</span><br> |
| 138 | + <span class="green">b</span><span class="gray">an</span><br> |
| 139 | + <span class="green">\$</span><span class="gray">ba</span><br> |
| 140 | + <span class="green">a</span><span class="gray">na</span><br> |
| 141 | + <span class="green">a</span><span class="gray">na</span> |
| 142 | + </div> |
| 143 | + </div> |
| 144 | + <div class="demo-block"> |
| 145 | + <div>再次排序</div> |
| 146 | + <div class="demo-window"> |
| 147 | + <span class="green">\$</span><span class="gray">ba</span><br> |
| 148 | + <span class="green">a</span><span class="gray">\$b</span><br> |
| 149 | + <span class="green">a</span><span class="gray">na</span><br> |
| 150 | + <span class="green">a</span><span class="gray">na</span><br> |
| 151 | + <span class="green">b</span><span class="gray">an</span><br> |
| 152 | + <span class="green">n</span><span class="gray">a\$</span><br> |
| 153 | + <span class="green">n</span><span class="gray">an</span> |
| 154 | + </div> |
| 155 | + </div> |
| 156 | + <div class="demo-block"> |
| 157 | + <div>重复上述步骤</div> |
| 158 | + <div class="demo-window"> |
| 159 | + <span class="green">a</span><span class="gray">\$ba</span><br> |
| 160 | + <span class="green">n</span><span class="gray">a\$b</span><br> |
| 161 | + <span class="green">n</span><span class="gray">ana</span><br> |
| 162 | + <span class="green">b</span><span class="gray">ana</span><br> |
| 163 | + <span class="green">\$</span><span class="gray">ban</span><br> |
| 164 | + <span class="green">a</span><span class="gray">na\$</span><br> |
| 165 | + <span class="green">a</span><span class="gray">nan</span> |
| 166 | + </div> |
| 167 | + </div> |
| 168 | + <div class="demo-block"> |
| 169 | + <div>再次排序</div> |
| 170 | + <div class="demo-window"> |
| 171 | + <span class="green">\$</span><span class="gray">ban</span><br> |
| 172 | + <span class="green">a</span><span class="gray">\$ba</span><br> |
| 173 | + <span class="green">a</span><span class="gray">na\$</span><br> |
| 174 | + <span class="green">a</span><span class="gray">nan</span><br> |
| 175 | + <span class="green">b</span><span class="gray">ana</span><br> |
| 176 | + <span class="green">n</span><span class="gray">a\$b</span><br> |
| 177 | + <span class="green">n</span><span class="gray">ana</span> |
| 178 | + </div> |
| 179 | + </div> |
| 180 | + <div class="demo-block"> |
| 181 | + <div>重复上述步骤</div> |
| 182 | + <div class="demo-window"> |
| 183 | + <span class="green">a</span><span class="gray">\$ban</span><br> |
| 184 | + <span class="green">n</span><span class="gray">a\$ba</span><br> |
| 185 | + <span class="green">n</span><span class="gray">ana\$</span><br> |
| 186 | + <span class="green">b</span><span class="gray">anan</span><br> |
| 187 | + <span class="green">\$</span><span class="gray">bana</span><br> |
| 188 | + <span class="green">a</span><span class="gray">na\$b</span><br> |
| 189 | + <span class="green">a</span><span class="gray">nana</span> |
| 190 | + </div> |
| 191 | + </div> |
| 192 | + <div class="demo-block"> |
| 193 | + <div>再次排序</div> |
| 194 | + <div class="demo-window"> |
| 195 | + <span class="green">\$</span><span class="gray">bana</span><br> |
| 196 | + <span class="green">a</span><span class="gray">\$ban</span><br> |
| 197 | + <span class="green">a</span><span class="gray">na\$b</span><br> |
| 198 | + <span class="green">a</span><span class="gray">nana</span><br> |
| 199 | + <span class="green">b</span><span class="gray">anan</span><br> |
| 200 | + <span class="green">n</span><span class="gray">a\$ba</span><br> |
| 201 | + <span class="green">n</span><span class="gray">ana\$</span> |
| 202 | + </div> |
| 203 | + </div> |
| 204 | + <div class="demo-block"> |
| 205 | + <div>重复上述步骤</div> |
| 206 | + <div class="demo-window"> |
| 207 | + <span class="green">a</span><span class="gray">\$bana</span><br> |
| 208 | + <span class="green">n</span><span class="gray">a\$ban</span><br> |
| 209 | + <span class="green">n</span><span class="gray">ana\$b</span><br> |
| 210 | + <span class="green">b</span><span class="gray">anana</span><br> |
| 211 | + <span class="green">\$</span><span class="gray">banan</span><br> |
| 212 | + <span class="green">a</span><span class="gray">na\$ba</span><br> |
| 213 | + <span class="green">a</span><span class="gray">nana\$</span> |
| 214 | + </div> |
| 215 | + </div> |
| 216 | + <div class="demo-block"> |
| 217 | + <div>再次排序</div> |
| 218 | + <div class="demo-window"> |
| 219 | + <span class="green">\$</span><span class="gray">banan</span><br> |
| 220 | + <span class="green">a</span><span class="gray">\$bana</span><br> |
| 221 | + <span class="green">a</span><span class="gray">na\$ba</span><br> |
| 222 | + <span class="green">a</span><span class="gray">nana\$</span><br> |
| 223 | + <span class="green">b</span><span class="gray">anana</span><br> |
| 224 | + <span class="green">n</span><span class="gray">a\$ban</span><br> |
| 225 | + <span class="green">n</span><span class="gray">ana\$b</span> |
| 226 | + </div> |
| 227 | + </div> |
| 228 | + <div class="demo-block"> |
| 229 | + <div>重复上述步骤</div> |
| 230 | + <div class="demo-window"> |
| 231 | + <span class="green">a</span><span class="gray">\$banan</span><br> |
| 232 | + <span class="green">n</span><span class="gray">a\$bana</span><br> |
| 233 | + <span class="green">n</span><span class="gray">ana\$ba</span><br> |
| 234 | + <span class="green">b</span><span class="gray">anana\$</span><br> |
| 235 | + <span class="green">\$</span><span class="gray">banana</span><br> |
| 236 | + <span class="green">a</span><span class="gray">na\$ban</span><br> |
| 237 | + <span class="green">a</span><span class="gray">nana\$b</span> |
| 238 | + </div> |
| 239 | + </div> |
| 240 | + <div class="demo-block"> |
| 241 | + <div>再次排序</div> |
| 242 | + <div class="demo-window"> |
| 243 | + <span class="green">\$</span><span class="gray">banana</span><br> |
| 244 | + <span class="green">a</span><span class="gray">\$banan</span><br> |
| 245 | + <span class="green">a</span><span class="gray">na\$ban</span><br> |
| 246 | + <span class="green">a</span><span class="gray">nana\$b</span><br> |
| 247 | + <span class="green">b</span><span class="gray">anana\$</span><br> |
| 248 | + <span class="green">n</span><span class="gray">a\$bana</span><br> |
| 249 | + <span class="green">n</span><span class="gray">ana\$ba</span> |
| 250 | + </div> |
| 251 | + </div> |
| 252 | + <div class="demo-block"> |
| 253 | + <div>回到最初的矩阵</div> |
| 254 | + <div class="demo-window"> |
| 255 | + <span class="red">\$</span>banana<br> |
| 256 | + a<span class="red">\$</span>banan<br> |
| 257 | + ana<span class="red">\$</span>ban<br> |
| 258 | + anana<span class="red">\$</span>b<br> |
| 259 | + banana<span class="red">\$</span><br> |
| 260 | + na<span class="red">\$</span>bana<br> |
| 261 | + nana<span class="red">\$</span>ba |
| 262 | + </div> |
| 263 | + </div> |
| 264 | +</div> |
| 265 | + |
| 266 | +<div class="demo-button-block"> |
| 267 | + <div></div> |
| 268 | + <button class="demo-button" id="prev" onclick="p()">上一步</button> |
| 269 | + <div></div> |
| 270 | + <button class="demo-button" id="next" onclick="n()">下一步</button> |
| 271 | + <div></div> |
| 272 | +</div> |
| 273 | + |
| 274 | +<style> |
| 275 | + .demo-button { |
| 276 | + padding: 6px 14px; |
| 277 | + border: 1px solid #ccc; |
| 278 | + background: #fff; |
| 279 | + cursor: pointer; |
| 280 | + border-radius: 4px; |
| 281 | + } |
| 282 | + .demo-button:disabled { |
| 283 | + opacity: .4; |
| 284 | + cursor: not-allowed; |
| 285 | + } |
| 286 | + .demo-block { |
| 287 | + display: grid; |
| 288 | + place-items: center; |
| 289 | + } |
| 290 | + .demo-button-block { |
| 291 | + display: grid; |
| 292 | + grid-template-columns: 1fr auto 30px auto 1fr; |
| 293 | + } |
| 294 | + .demo-window { |
| 295 | + height: 200px; |
| 296 | + font-family: monospace; |
| 297 | + border: 1px solid gray; |
| 298 | + padding: 5px 30px 5px 30px; |
| 299 | + margin: 5px 0 5px 0; |
| 300 | + } |
| 301 | + .red { |
| 302 | + color: red; |
| 303 | + } |
| 304 | + .green { |
| 305 | + color: green; |
| 306 | + } |
| 307 | + .gray { |
| 308 | + color: gray; |
| 309 | + } |
| 310 | +</style> |
| 311 | + |
| 312 | +<script> |
| 313 | + let i = 0, |
| 314 | + v = document.getElementById('v'), |
| 315 | + c = v.children, |
| 316 | + prev = document.getElementById('prev'), |
| 317 | + next = document.getElementById('next'); |
| 318 | + |
| 319 | + let u = () => { |
| 320 | + [...c].forEach((d, n) => d.hidden = n !== i); |
| 321 | + prev.disabled = i === 0; |
| 322 | + next.disabled = i === c.length - 1; |
| 323 | + }; |
| 324 | + |
| 325 | + let p = () => { if (i > 0) i--; u() }; |
| 326 | + let n = () => { if (i < c.length - 1) i++; u() }; |
| 327 | + |
| 328 | + u() |
| 329 | +</script> |
| 330 | + |
| 331 | + |
| 332 | +# 算法原理 |
| 333 | + |
| 334 | +觉得这个算法有意思的地方,可能并不是它的实用价值。这里有一个非常有意思:为什么这样排序了几次之后,就会回到最初的矩阵 |
| 335 | + |
| 336 | +这里蕴藏了一个非常有意思的字符串排序逻辑。 |
| 337 | + |
| 338 | +通常情况下,我们会使用字符串从第一个字符开始比较,如果相同则比下一个字符 |
| 339 | + |
| 340 | +而在这个问题下,假定所有字符串长度相同,那其实完全可以从最后一位比起,然后逐次比较新增加的字符,可以实现类似桶排序的方式,达成最终排序结果 |
| 341 | + |
| 342 | +由于后排序的结果会覆盖先排序的结果,使得实际上达成了“从第一个字符开始比较,如果相同则比下一个字符”的效果 |
0 commit comments