WebWorker计算慢regexp匹配明显慢(3倍)-仅firefox

Question

首先，我刚刚创建了一个正则表达式，它将匹配项目中所有头文件列表中所有唯一的外部库路径。一周前我问了一个关于制作regexp的问题。

我开始摆弄它，看看它在异步和变成web worker时的行为。为了方便和可靠，我创建了这个在所有三种模式下运行的通用文件:

/** Will call result() callback with every match it founds. Asynchronous unless called 
 *  with interval = -1.
 *  Javadoc style comment for Arnold Rimmer and other Java programmers:
 *  
 * @param regex regular expression to match in string
 * @param string guess what
 * @param result callback function that accepts one parameter, string match
 * @param done callback on finish, has no parameters
 * @param interval delay (not actual interval) between finding matches. If -1, 
 *        function  will be blocking
 * @property working false if loop isn't running, otherwise contains timeout ID
 *           for use with clearTimeout
 * @property done copy of done parameter
 * @throws heavy boulders
**/
function processRegex(regex, string, result, done, interval) {
  var m;
  //Please tell me interpreter optimizes this
  interval = typeof interval!='number'?1:interval;
  //And this
  processRegex.done = done;
  while ((m = regex.exec(string))) {
    Array.prototype.splice.call(m,0,1);
    var path = m.join("");
    //It's good to keep in mind that result() slows down the process
    result(path);
    if (interval>=0) {
      processRegex.working = setTimeout(processRegex, 
                              interval, regex, string, 
                              result, done, interval);
      // Comment these out for maximum speed
      processRegex.progress = regex.lastIndex/string.length;
      console.log("Progress: "+Math.round(processRegex.progress*100)+"%");
      return;
    }
  }
  processRegex.working = false;
  processRegex.done = null;
  if (typeof done=="function")
    done();
}
processRegex.working = false; 

我创建了一个测试文件，而不是粘贴在这里，我把它上传到非常可靠的网络托管:Demo -测试数据。

我发现非常令人惊讶的是，在web worker和浏览器执行RegExp之间存在如此显著的差异。我得到的结果:

• Mozilla Firefox
  • [WORKER]: Time elapsed:16.860s
  • [WORKER-SYNC]: Time elapsed:16.739s
  • [TIMEOUT]: Time elapsed:5.186s
  • [LOOP]: Time elapsed:5.028s

您还可以看到，对于我的特定正则表达式，同步循环和异步循环之间的区别是微不足道的。我试图使用一个匹配列表，而不是一个向前看的表达式，结果改变了很多。下面是对旧函数的修改:

function processRegexUnique(regex, string, result, done, interval) {
  var matchList = arguments[5]||[];
  ... same as before ...
  while ((m = regex.exec(string))) {
    ... same as before ...
    if (matchList.indexOf(path)==-1) {
      result(path);
      matchList.push(path);
    }
    if (interval>=0) {
      processRegex.working = setTimeout(processRegex, interval, 
                               regex, string, result, 
                               done, interval, matchList);
      ... same as before ...
    }
  }
  ... same as before ...
}

和结果:

• Mozilla Firefox
  • [WORKER]: Time elapsed:0.062s
  • [WORKER-SYNC]: Time elapsed:0.023s
  • [TIMEOUT]: Time elapsed:12.250s(自我提醒:每分钟都变得越来越奇怪)
  • [LOOP]: Time elapsed:0.006s

谁能解释这种速度上的差异?

Answer 1

经过一系列测试，我确认这是一个Mozilla Firefox问题(它影响了我尝试过的所有windows桌面版本)。对于Google Chrome, Opera，甚至Firefox移动版，regexp匹配都是相同的，无论是否工作。

如果您需要修复此问题，请务必在bugzilla上的bug报告中投票。