如何实现模糊搜索之类的 sublime text？答案

【问题标题】：How to implement sublime text like fuzzy search?如何实现模糊搜索之类的 sublime text？
【发布时间】：2013-05-30 06:11:55
【问题描述】：

如何在 select2 上实现类似 sublime 的模糊搜索？

例如，输入“sta jav sub”将匹配“Stackoverflow javascript sublime like”

【问题讨论】：

标签： javascript jquery-select2

【解决方案1】：

这是一个备用匹配函数。 http://jsfiddle.net/trevordixon/pXzj3/4/

function match(search, text) {
    search = search.toUpperCase();
    text = text.toUpperCase();

    var j = -1; // remembers position of last found character

    // consider each search character one at a time
    for (var i = 0; i < search.length; i++) {
        var l = search[i];
        if (l == ' ') continue;     // ignore spaces

        j = text.indexOf(l, j+1);     // search for character & update position
        if (j == -1) return false;  // if it's not found, exclude this item
    }
    return true;
}

这个速度更快（根据 Chrome 中的 this test），如果您要过滤很多项目，这可能会开始重要。

【讨论】：

我喜欢你的实现，但你不会尝试将搜索字符串中的字符匹配在一起，ej ABS 会匹配 Any Bachelor Sunrise，这可能是也可能不是你所期望的 :)跨度>
这不正是模糊搜索所期望的吗？
@Kloar 他的意思是A B S应该匹配Any Bachelor Sunrise，但是ABS应该只匹配，比如说ABS Building Co.
@rvighne 我认为 ABS 应该两者都匹配，但是第二场比赛的排名应该比第一场高
我发现这永远不会搜索第一个字符，因为在执行indexOf 方法时 j 永远不会等于 0。最好按原样使用 j，然后在 if 条件后添加 j++ 来更新位置，而不是在 indexOf 方法中。

【解决方案2】：

select2 允许您实现自己的“匹配器”函数 (as seen on their docs)，使用它和一些正则表达式，您可以执行以下操作：

$("#element").select2({
    matcher: function(term, text, opt) {
        //We call to uppercase to do a case insensitive match
        //We replace every group of whitespace characters with a .+
        //matching any number of characters
        return text.toUpperCase().match(term.toUpperCase().replace(/\s+/g, '.+'));
    }
});

在过滤/搜索列表时，会针对每个 select2 列表元素调用匹配器函数，您可以使用它实现任何类型的自定义搜索。

【讨论】：

【解决方案3】：

我写了一些很长的 Sublime Text 的模糊匹配。实现这一点需要做一些事情。

首先，按顺序匹配模式中的所有字符。其次，对匹配进行评分，以使某些匹配的字符比其他字符更有价值。

我想出了几个要检查的因素。 “CamelCase”字母或分隔符（空格或下划线）后面的字母值得很多分。连续比赛的价值更高。在开始时找到的结果更有价值。

一个至关重要的技巧是找到最匹配的字符。这不一定是第一个。考虑模糊匹配（“tk”，“黑骑士”）。有两个可以匹配的K。第二个值得更多的分数，因为它跟随一个空格。

JavaScript 代码如下。有一些细微差别在博客文章中有更详细的描述。还有一个交互式演示。以及 GitHub 上的完整源代码（包括演示和 C++ 实现）。

Blog Post
Interactive Demo

GitHub

// Returns [bool, score, formattedStr]
// bool: true if each character in pattern is found sequentially within str
// score: integer; higher is better match. Value has no intrinsic meaning. Range varies with pattern. 
//        Can only compare scores with same search pattern.
// formattedStr: input str with matched characters marked in <b> tags. Delete if unwanted.

function fuzzy_match(pattern, str) {
    // Score consts
    var adjacency_bonus = 5;                // bonus for adjacent matches
    var separator_bonus = 10;               // bonus if match occurs after a separator
    var camel_bonus = 10;                   // bonus if match is uppercase and prev is lower
    var leading_letter_penalty = -3;        // penalty applied for every letter in str before the first match
    var max_leading_letter_penalty = -9;    // maximum penalty for leading letters
    var unmatched_letter_penalty = -1;      // penalty for every letter that doesn't matter

    // Loop variables
    var score = 0;
    var patternIdx = 0;
    var patternLength = pattern.length;
    var strIdx = 0;
    var strLength = str.length;
    var prevMatched = false;
    var prevLower = false;
    var prevSeparator = true;       // true so if first letter match gets separator bonus

    // Use "best" matched letter if multiple string letters match the pattern
    var bestLetter = null;
    var bestLower = null;
    var bestLetterIdx = null;
    var bestLetterScore = 0;

    var matchedIndices = [];

    // Loop over strings
    while (strIdx != strLength) {
        var patternChar = patternIdx != patternLength ? pattern.charAt(patternIdx) : null;
        var strChar = str.charAt(strIdx);

        var patternLower = patternChar != null ? patternChar.toLowerCase() : null;
        var strLower = strChar.toLowerCase();
        var strUpper = strChar.toUpperCase();

        var nextMatch = patternChar && patternLower == strLower;
        var rematch = bestLetter && bestLower == strLower;

        var advanced = nextMatch && bestLetter;
        var patternRepeat = bestLetter && patternChar && bestLower == patternLower;
        if (advanced || patternRepeat) {
            score += bestLetterScore;
            matchedIndices.push(bestLetterIdx);
            bestLetter = null;
            bestLower = null;
            bestLetterIdx = null;
            bestLetterScore = 0;
        }

        if (nextMatch || rematch) {
            var newScore = 0;

            // Apply penalty for each letter before the first pattern match
            // Note: std::max because penalties are negative values. So max is smallest penalty.
            if (patternIdx == 0) {
                var penalty = Math.max(strIdx * leading_letter_penalty, max_leading_letter_penalty);
                score += penalty;
            }

            // Apply bonus for consecutive bonuses
            if (prevMatched)
                newScore += adjacency_bonus;

            // Apply bonus for matches after a separator
            if (prevSeparator)
                newScore += separator_bonus;

            // Apply bonus across camel case boundaries. Includes "clever" isLetter check.
            if (prevLower && strChar == strUpper && strLower != strUpper)
                newScore += camel_bonus;

            // Update patter index IFF the next pattern letter was matched
            if (nextMatch)
                ++patternIdx;

            // Update best letter in str which may be for a "next" letter or a "rematch"
            if (newScore >= bestLetterScore) {

                // Apply penalty for now skipped letter
                if (bestLetter != null)
                    score += unmatched_letter_penalty;

                bestLetter = strChar;
                bestLower = bestLetter.toLowerCase();
                bestLetterIdx = strIdx;
                bestLetterScore = newScore;
            }

            prevMatched = true;
        }
        else {
            // Append unmatch characters
            formattedStr += strChar;

            score += unmatched_letter_penalty;
            prevMatched = false;
        }

        // Includes "clever" isLetter check.
        prevLower = strChar == strLower && strLower != strUpper;
        prevSeparator = strChar == '_' || strChar == ' ';

        ++strIdx;
    }

    // Apply score for last match
    if (bestLetter) {
        score += bestLetterScore;
        matchedIndices.push(bestLetterIdx);
    }

    // Finish out formatted string after last pattern matched
    // Build formated string based on matched letters
    var formattedStr = "";
    var lastIdx = 0;
    for (var i = 0; i < matchedIndices.length; ++i) {
        var idx = matchedIndices[i];
        formattedStr += str.substr(lastIdx, idx - lastIdx) + "<b>" + str.charAt(idx) + "</b>";
        lastIdx = idx + 1;
    }
    formattedStr += str.substr(lastIdx, str.length - lastIdx);

    var matched = patternIdx == patternLength;
    return [matched, score, formattedStr];
}

【讨论】：

【解决方案4】：

albertein 的答案与 Trevor 的版本不匹配，因为原始函数基于字符而不是单词执行匹配。这是一个更简单的基于字符的匹配：

$("#element").select2({
  matcher: function(term, text, opts) {
    var pattern = term.replace(/\s+/g, '').split('').join('.*');
    text.match(new RegExp(pattern, 'i'))
  }
})

【讨论】：

【解决方案5】：

var fuzzysearch = function (querystrings, values) {
    return !querystrings.some(function (q) {
        return !values.some(function (v) {
            return v.toLocaleLowerCase().indexOf(q) !== -1;
        });
    });
}

在图书收藏中搜索标题和作者的示例 http://jsfiddle.net/runjep/r887etnh/2/

对于对搜索结果进行排名的 9kb 替代方案：http://kiro.me/projects/fuse.html

你可能需要一个 polyfill 来实现 'some' 函数 https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Array/some

var books = [{
    id: 1,
    title: 'The Great Gatsby',
    author: 'F. Scott Fitzgerald'
}, {
    id: 2,
    title: 'The DaVinci Code',
    author: 'Dan Brown'
}, {
    id: 3,
    title: 'Angels & Demons',
    author: 'Dan Brown'
}];
search = function () {
    var queryarray = document.getElementById('inp').value.trim().toLowerCase().split(' ');
    var res = books.filter(function (b) {
        return fs(queryarray, [b.title, b.author]);
    });
    document.getElementById('res').innerHTML = res.map(function (b) {
        return b.title + ' <i> ' + b.author + '</i>';
    }).join('<br/> ');
}
fs = function (qs, vals) {
    return !qs.some(function (q) {
        return !vals.some(function (v) {
            return v.toLocaleLowerCase().indexOf(q) !== -1;
        });
    });
}

<input id="inp" />
<button id="but" onclick="search()">Search</button>
<div id="res"></div>

【讨论】：

这是原版 javascript，没有选择 2 - 抱歉

【解决方案6】：

function fuzzyMe(term, query) {
  var score = 0;
  var termLength = term.length;
  var queryLength = query.length;
  var highlighting = '';
  var ti = 0;
  // -1 would not work as this would break the calculations of bonus
  // points for subsequent character matches. Something like
  // Number.MIN_VALUE would be more appropriate, but unfortunately
  // Number.MIN_VALUE + 1 equals 1...
  var previousMatchingCharacter = -2;
  for (var qi = 0; qi < queryLength && ti < termLength; qi++) {
    var qc = query.charAt(qi);
    var lowerQc = qc.toLowerCase();

    for (; ti < termLength; ti++) {
      var tc = term.charAt(ti);

      if (lowerQc === tc.toLowerCase()) {
        score++;

        if ((previousMatchingCharacter + 1) === ti) {
          score += 2;
        }

        highlighting += "<em>" + tc + "</em>";
        previousMatchingCharacter = ti;
        ti++;
        break;
      } else {
        highlighting += tc;
      }
    }
  }

  highlighting += term.substring(ti, term.length);

  return {
    score: score,
    term: term,
    query: query,
    highlightedTerm: highlighting
  };
}

以上处理了模糊性。然后您可以遍历所有选择的 2 个元素

$("#element").select2({
  matcher: function(term, text, opt) {
    return fuzzyMe(term, text).highlightedTerm;
  }
});

模糊代码的功劳-：https://github.com/bripkens/fuzzy.js

【讨论】：

【解决方案7】：

在新的 select2 上遇到了困难，这里有什么效果

 $("#foo").select2({
   matcher: matcher
 });

function matcher(params, data) {
  // return all opts if seachbox is empty
  if(!params.term) {
    return data;
  } else if(data) {
    var term = params.term.toUpperCase();
    var option = data.text.toUpperCase();
    var j = -1; // remembers position of last found character

    // consider each search character one at a time
    for (var i = 0; i < term.length; i++) {
      var l = term[i];
      if (l == ' ') continue;     // ignore spaces

      j = option.indexOf(l, j+1);     // search for character & update position
      if (j == -1) return false;  // if it's not found, exclude this item
    }
    return data; // return option
  }
}

【讨论】：