docspell/website/site/static/js/searchhelper.js

// Taken from mdbook
// The strategy is as follows:
// First, assign a value to each word in the document:
//  Words that correspond to search terms (stemmer aware): 40
//  Normal words: 2
//  First word in a sentence: 8
// Then use a sliding window with a constant number of words and count the
// sum of the values of the words within the window. Then use the window that got the
// maximum sum. If there are multiple maximas, then get the last one.
// Enclose the terms in *.
function makeTeaser(body, terms) {
  var TERM_WEIGHT = 40;
  var NORMAL_WORD_WEIGHT = 2;
  var FIRST_WORD_WEIGHT = 8;
  var TEASER_MAX_WORDS = 30;

  var stemmedTerms = terms.map(function (w) {
    return elasticlunr.stemmer(w.toLowerCase());
  });
  var termFound = false;
  var index = 0;
  var weighted = []; // contains elements of ["word", weight, index_in_document]

  // split in sentences, then words
  var sentences = body.toLowerCase().split(". ");

  for (var i in sentences) {
    var words = sentences[i].split(" ");
    var value = FIRST_WORD_WEIGHT;

    for (var j in words) {
      var word = words[j];

      if (word.length > 0) {
        for (var k in stemmedTerms) {
          if (elasticlunr.stemmer(word).startsWith(stemmedTerms[k])) {
            value = TERM_WEIGHT;
            termFound = true;
          }
        }
        weighted.push([word, value, index]);
        value = NORMAL_WORD_WEIGHT;
      }

      index += word.length;
      index += 1;  // ' ' or '.' if last word in sentence
    }

    index += 1;  // because we split at a two-char boundary '. '
  }

  if (weighted.length === 0) {
    return body;
  }

  var windowWeights = [];
  var windowSize = Math.min(weighted.length, TEASER_MAX_WORDS);
  // We add a window with all the weights first
  var curSum = 0;
  for (var i = 0; i < windowSize; i++) {
    curSum += weighted[i][1];
  }
  windowWeights.push(curSum);

  for (var i = 0; i < weighted.length - windowSize; i++) {
    curSum -= weighted[i][1];
    curSum += weighted[i + windowSize][1];
    windowWeights.push(curSum);
  }

  // If we didn't find the term, just pick the first window
  var maxSumIndex = 0;
  if (termFound) {
    var maxFound = 0;
    // backwards
    for (var i = windowWeights.length - 1; i >= 0; i--) {
      if (windowWeights[i] > maxFound) {
        maxFound = windowWeights[i];
        maxSumIndex = i;
      }
    }
  }

  var teaser = [];
  var startIndex = weighted[maxSumIndex][2];
  for (var i = maxSumIndex; i < maxSumIndex + windowSize; i++) {
    var word = weighted[i];
    if (startIndex < word[2]) {
      // missing text from index to start of `word`
      teaser.push(body.substring(startIndex, word[2]));
      startIndex = word[2];
    }

    // add <em/> around search terms
    if (word[1] === TERM_WEIGHT) {
      teaser.push("**");
    }
    startIndex = word[2] + word[0].length;
    teaser.push(body.substring(word[2], startIndex));

    if (word[1] === TERM_WEIGHT) {
      teaser.push("**");
    }
  }
  teaser.push("…");
  return teaser.join("");
}

var index = elasticlunr.Index.load(window.searchIndex);


var initElmSearch = function(elmSearch) {
    var options = {
        bool: "AND",
        fields: {
            title: {boost: 2},
            body: {boost: 1},
        }
    };

    elmSearch.ports.doSearch.subscribe(function(str) {
        var results = index.search(str, options);
        for (var i = 0; i < results.length; i ++) {
            var teaser = makeTeaser(results[i].doc.body, str.split(" "));
            results[i].doc.body = teaser;
        }
        elmSearch.ports.receiveSearch.send(results);
    });

};