Use weighted algorithm for string matching when finding files in repo (#21370)

This PR is for:
* https://github.com/go-gitea/gitea/issues/20231

Now, when a user searches `word`, they always see `/{word}.txt` before
`/{w}e-g{o}t-{r}esult.{d}at`

Demo:

When searching "a", "a.ext" comes first. 

Then when searching "at", the longer matched "template" comes first.

<details>


![image](https://user-images.githubusercontent.com/2114189/194588738-3644d891-956f-40e4-b79b-b97d34265456.png)


![image](https://user-images.githubusercontent.com/2114189/194588797-9b124670-4e1e-4510-a170-780295ed89b8.png)

</details>

This PR also makes the frontend tests could import feature JS files by
introducing `jestSetup.js`

Co-authored-by: delvh <dev.lh@web.de>
Co-authored-by: silverwind <me@silverwind.io>

jest.config.js

@@ -1,3 +1,4 @@
+// to run tests with ES6 module, node must run with "--experimental-vm-modules", or see Makefile's "test-frontend" for reference
 export default {
   rootDir: 'web_src',
   setupFilesAfterEnv: ['jest-extended/all'],
@@ -7,6 +8,8 @@ export default {
   transform: {
     '\\.svg$': '<rootDir>/js/testUtils/jestRawLoader.js',
   },
+  setupFiles: [
+    './js/testUtils/jestSetup.js', // prepare global variables used by our code (eg: window.config)
+  ],
   verbose: false,
 };
-

web_src/js/features/repo-findfile.js

@@ -1,45 +1,96 @@
 import $ from 'jquery';
 
 import {svg} from '../svg.js';
-import {strSubMatch} from '../utils.js';
 const {csrf} = window.config;
 
 const threshold = 50;
 let files = [];
 let $repoFindFileInput, $repoFindFileTableBody, $repoFindFileNoResult;
 
-function filterRepoFiles(filter) {
-  const treeLink = $repoFindFileInput.attr('data-url-tree-link');
-  $repoFindFileTableBody.empty();
 
-  const fileRes = [];
+// return the case-insensitive sub-match result as an array:  [unmatched, matched, unmatched, matched, ...]
+// res[even] is unmatched, res[odd] is matched, see unit tests for examples
+// argument subLower must be a lower-cased string.
+export function strSubMatch(full, subLower) {
+  const res = [''];
+  let i = 0, j = 0;
+  const fullLower = full.toLowerCase();
+  while (i < subLower.length && j < fullLower.length) {
+    if (subLower[i] === fullLower[j]) {
+      if (res.length % 2 !== 0) res.push('');
+      res[res.length - 1] += full[j];
+      j++;
+      i++;
+    } else {
+      if (res.length % 2 === 0) res.push('');
+      res[res.length - 1] += full[j];
+      j++;
+    }
+  }
+  if (i !== subLower.length) {
+    // if the sub string doesn't match the full, only return the full as unmatched.
+    return [full];
+  }
+  if (j < full.length) {
+    // append remaining chars from full to result as unmatched
+    if (res.length % 2 === 0) res.push('');
+    res[res.length - 1] += full.substring(j);
+  }
+  return res;
+}
+
+export function calcMatchedWeight(matchResult) {
+  let weight = 0;
+  for (let i = 0; i < matchResult.length; i++) {
+    if (i % 2 === 1) { // matches are on odd indices, see strSubMatch
+      // use a function f(x+x) > f(x) + f(x) to make the longer matched string has higher weight.
+      weight += matchResult[i].length * matchResult[i].length;
+    }
+  }
+  return weight;
+}
+
+export function filterRepoFilesWeighted(files, filter) {
+  let filterResult = [];
   if (filter) {
-    for (let i = 0; i < files.length && fileRes.length < threshold; i++) {
+    const filterLower = filter.toLowerCase();
-      const subMatch = strSubMatch(files[i], filter);
+    // TODO: for large repo, this loop could be slow, maybe there could be one more limit:
-      if (subMatch.length > 1) {
+    // ... && filterResult.length < threshold * 20,  wait for more feedbacks
-        fileRes.push(subMatch);
+    for (let i = 0; i < files.length; i++) {
+      const res = strSubMatch(files[i], filterLower);
+      if (res.length > 1) { // length==1 means unmatched, >1 means having matched sub strings
+        filterResult.push({matchResult: res, matchWeight: calcMatchedWeight(res)});
       }
     }
+    filterResult.sort((a, b) => b.matchWeight - a.matchWeight);
+    filterResult = filterResult.slice(0, threshold);
   } else {
     for (let i = 0; i < files.length && i < threshold; i++) {
-      fileRes.push([files[i]]);
+      filterResult.push({matchResult: [files[i]], matchWeight: 0});
     }
   }
+  return filterResult;
+}
+
+function filterRepoFiles(filter) {
+  const treeLink = $repoFindFileInput.attr('data-url-tree-link');
+  $repoFindFileTableBody.empty();
 
+  const filterResult = filterRepoFilesWeighted(files, filter);
   const tmplRow = `<tr><td><a></a></td></tr>`;
 
-  $repoFindFileNoResult.toggle(fileRes.length === 0);
+  $repoFindFileNoResult.toggle(filterResult.length === 0);
-  for (const matchRes of fileRes) {
+  for (const r of filterResult) {
     const $row = $(tmplRow);
     const $a = $row.find('a');
-    $a.attr('href', `${treeLink}/${matchRes.join('')}`);
+    $a.attr('href', `${treeLink}/${r.matchResult.join('')}`);
     const $octiconFile = $(svg('octicon-file')).addClass('mr-3');
     $a.append($octiconFile);
-    // if the target file path is "abc/xyz", to search "bx", then the matchRes is ['a', 'b', 'c/', 'x', 'yz']
+    // if the target file path is "abc/xyz", to search "bx", then the matchResult is ['a', 'b', 'c/', 'x', 'yz']
-    // the matchRes[odd] is matched and highlighted to red.
+    // the matchResult[odd] is matched and highlighted to red.
-    for (let j = 0; j < matchRes.length; j++) {
+    for (let j = 0; j < r.matchResult.length; j++) {
-      if (!matchRes[j]) continue;
+      if (!r.matchResult[j]) continue;
-      const $span = $('<span>').text(matchRes[j]);
+      const $span = $('<span>').text(r.matchResult[j]);
       if (j % 2 === 1) $span.addClass('ui text red');
       $a.append($span);
     }

web_src/js/features/repo-findfile.test.js

@@ -0,0 +1,34 @@
+import {strSubMatch, calcMatchedWeight, filterRepoFilesWeighted} from './repo-findfile.js';
+
+describe('Repo Find Files', () => {
+  test('strSubMatch', () => {
+    expect(strSubMatch('abc', '')).toEqual(['abc']);
+    expect(strSubMatch('abc', 'a')).toEqual(['', 'a', 'bc']);
+    expect(strSubMatch('abc', 'b')).toEqual(['a', 'b', 'c']);
+    expect(strSubMatch('abc', 'c')).toEqual(['ab', 'c']);
+    expect(strSubMatch('abc', 'ac')).toEqual(['', 'a', 'b', 'c']);
+    expect(strSubMatch('abc', 'z')).toEqual(['abc']);
+    expect(strSubMatch('abc', 'az')).toEqual(['abc']);
+
+    expect(strSubMatch('ABc', 'ac')).toEqual(['', 'A', 'B', 'c']);
+    expect(strSubMatch('abC', 'ac')).toEqual(['', 'a', 'b', 'C']);
+
+    expect(strSubMatch('aabbcc', 'abc')).toEqual(['', 'a', 'a', 'b', 'b', 'c', 'c']);
+    expect(strSubMatch('the/directory', 'hedir')).toEqual(['t', 'he', '/', 'dir', 'ectory']);
+  });
+
+  test('calcMatchedWeight', () => {
+    expect(calcMatchedWeight(['a', 'b', 'c', 'd']) < calcMatchedWeight(['a', 'bc', 'c'])).toBeTruthy();
+  });
+
+  test('filterRepoFilesWeighted', () => {
+    // the first matched result should always be the "word.txt"
+    let res = filterRepoFilesWeighted(['word.txt', 'we-got-result.dat'], 'word');
+    expect(res).toHaveLength(2);
+    expect(res[0].matchResult).toEqual(['', 'word', '.txt']);
+
+    res = filterRepoFilesWeighted(['we-got-result.dat', 'word.txt'], 'word');
+    expect(res).toHaveLength(2);
+    expect(res[0].matchResult).toEqual(['', 'word', '.txt']);
+  });
+});

web_src/js/testUtils/jestSetup.js

@@ -0,0 +1,5 @@
+window.config = {
+  csrfToken: 'jest-test-csrf-token-123456',
+  pageData: {},
+  i18n: {},
+};

web_src/js/utils.js

@@ -59,36 +59,6 @@ export function parseIssueHref(href) {
   return {owner, repo, type, index};
 }
 
-// return the sub-match result as an array:  [unmatched, matched, unmatched, matched, ...]
-// res[even] is unmatched, res[odd] is matched, see unit tests for examples
-export function strSubMatch(full, sub) {
-  const res = [''];
-  let i = 0, j = 0;
-  const subLower = sub.toLowerCase(), fullLower = full.toLowerCase();
-  while (i < subLower.length && j < fullLower.length) {
-    if (subLower[i] === fullLower[j]) {
-      if (res.length % 2 !== 0) res.push('');
-      res[res.length - 1] += full[j];
-      j++;
-      i++;
-    } else {
-      if (res.length % 2 === 0) res.push('');
-      res[res.length - 1] += full[j];
-      j++;
-    }
-  }
-  if (i !== sub.length) {
-    // if the sub string doesn't match the full, only return the full as unmatched.
-    return [full];
-  }
-  if (j < full.length) {
-    // append remaining chars from full to result as unmatched
-    if (res.length % 2 === 0) res.push('');
-    res[res.length - 1] += full.substring(j);
-  }
-  return res;
-}
-
 // pretty-print a number using locale-specific separators, e.g. 1200 -> 1,200
 export function prettyNumber(num, locale = 'en-US') {
   if (typeof num !== 'number') return '';

web_src/js/utils.test.js

@@ -1,5 +1,5 @@
 import {
-  basename, extname, isObject, uniq, stripTags, joinPaths, parseIssueHref, strSubMatch,
+  basename, extname, isObject, uniq, stripTags, joinPaths, parseIssueHref,
   prettyNumber, parseUrl,
 } from './utils.js';
 
@@ -86,22 +86,6 @@ test('parseIssueHref', () => {
   expect(parseIssueHref('')).toEqual({owner: undefined, repo: undefined, type: undefined, index: undefined});
 });
 
-test('strSubMatch', () => {
-  expect(strSubMatch('abc', '')).toEqual(['abc']);
-  expect(strSubMatch('abc', 'a')).toEqual(['', 'a', 'bc']);
-  expect(strSubMatch('abc', 'b')).toEqual(['a', 'b', 'c']);
-  expect(strSubMatch('abc', 'c')).toEqual(['ab', 'c']);
-  expect(strSubMatch('abc', 'ac')).toEqual(['', 'a', 'b', 'c']);
-  expect(strSubMatch('abc', 'z')).toEqual(['abc']);
-  expect(strSubMatch('abc', 'az')).toEqual(['abc']);
-
-  expect(strSubMatch('abc', 'aC')).toEqual(['', 'a', 'b', 'c']);
-  expect(strSubMatch('abC', 'ac')).toEqual(['', 'a', 'b', 'C']);
-
-  expect(strSubMatch('aabbcc', 'abc')).toEqual(['', 'a', 'a', 'b', 'b', 'c', 'c']);
-  expect(strSubMatch('the/directory', 'hedir')).toEqual(['t', 'he', '/', 'dir', 'ectory']);
-});
-
 test('prettyNumber', () => {
   expect(prettyNumber()).toEqual('');
   expect(prettyNumber(null)).toEqual('');