| 1 | const res = await fetch(`${window.basePath}/index.json`);
|
| 2 | const index = await res.json();
|
| 3 |
|
| 4 | /**
|
| 5 | * True Damerau–Levenshtein distance (allows multiple transpositions).
|
| 6 | * Adapted from [0] with help from ChatGPT 5.2.
|
| 7 | *
|
| 8 | * Computes the DL distance between textA and textB. This is the number of edits between A and B. The possible edits are:
|
| 9 | * - Insertion (adding a single char)
|
| 10 | * - Deletion (removing a single char)
|
| 11 | * - Substitution (changing a single char)
|
| 12 | * - Transposition (swapping the locations of 2 chars)
|
| 13 | *
|
| 14 | * [0] https://en.wikipedia.org/wiki/Damerau%E2%80%93Levenshtein_distance#Distance_with_adjacent_transpositions
|
| 15 | */
|
| 16 | function damerauLevenshteinDistance(textA, textB) {
|
| 17 | const lenA = textA.length;
|
| 18 | const lenB = textB.length;
|
| 19 |
|
| 20 | const INF = lenA + lenB;
|
| 21 |
|
| 22 | // lastRowByChar == da in the pseudocode:
|
| 23 | // maps a character -> last row index (1..lenA) where it appeared in textA
|
| 24 | const lastRowByChar = new Map();
|
| 25 |
|
| 26 | // dist is the shifted version of d. Size: (lenA+2) x (lenB+2)
|
| 27 | const dist = Array.from({ length: lenA + 2 }, () => new Array(lenB + 2).fill(0));
|
| 28 |
|
| 29 | // Initialize sentinel borders
|
| 30 | dist[0][0] = INF;
|
| 31 | for (let i = 0; i <= lenA; i++) {
|
| 32 | dist[i + 1][0] = INF; // d[i, -1]
|
| 33 | dist[i + 1][1] = i; // d[i, 0]
|
| 34 | }
|
| 35 | for (let j = 0; j <= lenB; j++) {
|
| 36 | dist[0][j + 1] = INF; // d[-1, j]
|
| 37 | dist[1][j + 1] = j; // d[ 0, j]
|
| 38 | }
|
| 39 |
|
| 40 | // Main dynamic programming algorithm
|
| 41 | for (let i = 1; i <= lenA; i++) {
|
| 42 | let lastMatchingColInB = 0; // db in the pseudocode (last column j where a[i] matched)
|
| 43 | const charA = textA[i - 1];
|
| 44 |
|
| 45 | for (let j = 1; j <= lenB; j++) {
|
| 46 | const charB = textB[j - 1];
|
| 47 |
|
| 48 | const lastRowWithCharBInA = lastRowByChar.get(charB) ?? 0; // k := da[b[j]]
|
| 49 | const lastMatchingColForThisRow = lastMatchingColInB; // ℓ := db
|
| 50 |
|
| 51 | const cost = (charA === charB) ? 0 : 1;
|
| 52 | if (cost === 0) lastMatchingColInB = j;
|
| 53 |
|
| 54 | // Shifted accesses:
|
| 55 | // d[i-1, j-1] -> dist[i][j]
|
| 56 | // d[i, j-1] -> dist[i+1][j]
|
| 57 | // d[i-1, j ] -> dist[i][j+1]
|
| 58 | const substitution = dist[i][j] + cost;
|
| 59 | const insertion = dist[i + 1][j] + 1;
|
| 60 | const deletion = dist[i][j + 1] + 1;
|
| 61 |
|
| 62 | // Transposition term (shifted):
|
| 63 | // d[k-1, ℓ-1] in pseudocode -> dist[k][ℓ]
|
| 64 | const transposition =
|
| 65 | dist[lastRowWithCharBInA][lastMatchingColForThisRow] +
|
| 66 | (i - lastRowWithCharBInA - 1) +
|
| 67 | cost +
|
| 68 | (j - lastMatchingColForThisRow - 1);
|
| 69 |
|
| 70 | dist[i + 1][j + 1] = Math.min(substitution, insertion, deletion, transposition);
|
| 71 | }
|
| 72 |
|
| 73 | // da[a[i]] := i
|
| 74 | lastRowByChar.set(charA, i);
|
| 75 | }
|
| 76 |
|
| 77 | // return d[lenA, lenB] -> dist[lenA+1][lenB+1]
|
| 78 | return dist[lenA + 1][lenB + 1];
|
| 79 | }
|
| 80 |
|
| 81 | /**
|
| 82 | * Search the global `index` against `query`.
|
| 83 | *
|
| 84 | * The matching algorithm is as follows, include all items with:
|
| 85 | * - An exact match
|
| 86 | * - Damerau-Levenshtein distance <= 5
|
| 87 | *
|
| 88 | * Note that all matching is case-insensitive.
|
| 89 | *
|
| 90 | * Only the top 25 results are given. We first show the exact matches, and then
|
| 91 | * the inexact matches ranked by decreasing DL distance.
|
| 92 | */
|
| 93 | function search(query) {
|
| 94 | if (!query) {
|
| 95 | return [];
|
| 96 | }
|
| 97 |
|
| 98 | // We do case insensitive matching.
|
| 99 | query = query.toLowerCase();
|
| 100 |
|
| 101 | // Results is a list of pairs [entry, rank] where a lower rank is better.
|
| 102 | const results = [];
|
| 103 | for (const entry of index) {
|
| 104 | const symbol = entry.symbol.toLowerCase();
|
| 105 | if (symbol.includes(query)) {
|
| 106 | results.push([entry, 0]);
|
| 107 | continue;
|
| 108 | }
|
| 109 |
|
| 110 | const distance = damerauLevenshteinDistance(query, entry.symbol);
|
| 111 | if (distance > 5) {
|
| 112 | continue;
|
| 113 | }
|
| 114 |
|
| 115 | results.push([entry, distance]);
|
| 116 | }
|
| 117 |
|
| 118 | results.sort(([_a, aRank], [_b, bRank]) => aRank - bRank);
|
| 119 | return results.map(([entry, _rank]) => entry).slice(0, 25);
|
| 120 | }
|
| 121 |
|
| 122 |
|
| 123 | const searchDiv = document.getElementById("search");
|
| 124 | const searchbar = document.getElementById("searchbar");
|
| 125 |
|
| 126 | let resultsList = null;
|
| 127 | searchbar.addEventListener('input', async (event) => {
|
| 128 | const query = event.target.value;
|
| 129 | const results = search(query);
|
| 130 |
|
| 131 | if (resultsList) {
|
| 132 | resultsList.remove();
|
| 133 | }
|
| 134 |
|
| 135 | resultsList = document.createElement('ul');
|
| 136 | searchDiv.appendChild(resultsList);
|
| 137 |
|
| 138 | for (const result of results) {
|
| 139 | const item = document.createElement('li');
|
| 140 | const link = document.createElement('a');
|
| 141 | link.innerHTML = result.symbol;
|
| 142 | link.href = window.basePath + '/' + result.anchor;
|
| 143 | item.appendChild(link);
|
| 144 | resultsList.appendChild(item);
|
| 145 | }
|
| 146 | });
|