feat: detect duplicate/similar commit subjects

.github/workflows/test.yml

@@ -7,9 +7,22 @@ on:
     branches: [ '*' ]
 
 jobs:
+  unit-tests:
+    name: Unit tests
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v2
+      - uses: actions/setup-node@v3
+        with:
+          node-version: '18'
+          cache: 'npm'
+      - run: npm ci
+      - run: npm test
+
   test-action:
     name: Enforce convention commit
     runs-on: ubuntu-latest
+    needs: unit-tests
     steps:
       - uses: actions/checkout@v2
       - name: Test the action locally

action.yml

@@ -26,6 +26,14 @@ inputs:
     required: false
     description: Whether to scrub old lint reports from comments
     default: 'true'
+  check_double_colon:
+    required: false
+    description: 'Fail when a commit header contains a redundant colon after the type(scope): separator, e.g. "fix: : description"'
+    default: 'true'
+  duplicate_threshold:
+    required: false
+    description: 'Similarity threshold (0.0–1.0) above which two commit subjects are considered duplicates and the action fails. Leave empty to disable.'
+    default: ''
 
 runs:
   using: node12

index.js

@@ -13,19 +13,23 @@ const moment = require('moment');
 const lint = lintLib.default;
 const defaultConfigRules = defaultConfig.rules;
 
+const { normalizeSubject, diceSimilarity, hasDoubleColon } = require('./src/similarity');
+
 const validEvent = ['pull_request'];
 
 async function run() {
   try {
     const token = core.getInput('token', { required: true });
     const createComment = core.getBooleanInput('comment');
     const deleteComment = core.getBooleanInput('delete_comment');
+    const checkDoubleColon = core.getBooleanInput('check_double_colon');
+    const duplicateThresholdRaw = core.getInput('duplicate_threshold');
+    const duplicateThreshold = duplicateThresholdRaw !== '' ? parseFloat(duplicateThresholdRaw) : null;
 
     // load rules from file, if there
     const configPath = core.getInput('config_path', { required: true });
 
-    // relative to dist/index.js
-    const filename = path.join(__dirname, '/../', configPath);
+    const filename = path.join(process.env.GITHUB_WORKSPACE || process.cwd(), configPath);
     core.debug(`Loading config from path: ${filename}`);
     const config = fs.existsSync(filename) ? JSON.parse(fs.readFileSync(filename, 'utf8')) : {};
     core.debug(`Loaded config is: ${JSON.stringify(config)}`);
@@ -109,8 +113,9 @@ async function run() {
         `By **[${commit.author.name} (${meta.committer.login})](https://github.com/${meta.committer.login})** _${relativeTime}_` :
         `By **${commit.author.name} (Unknown Login) _${relativeTime}_`;
 
-      const headerIcon = report.valid ? '✅' :
-        report.errors.length ? '❌' : '⚠️';
+      const doubleColon = checkDoubleColon && hasDoubleColon(commit.message);
+      const headerIcon = (report.valid && !doubleColon) ? '✅' :
+        (report.errors.length || doubleColon) ? '❌' : '⚠️';
       let commitReportText = `
 ### ${headerIcon} [Commit ${shaShort}](https://github.com/${owner}/${repo.name}/commit/${sha})
 
@@ -146,14 +151,40 @@ ${commit.message}
         commitReportText += `
 ${errorReportText}
 ${warningReportText}
+`;
+      }
+
+      if (doubleColon) {
+        core.error(`header-colon-count: commit header contains a redundant colon ("${commit.message}")`);
+        countErrors++;
+        commitReportText += `
+❌ **ERROR**: commit header must contain exactly one colon — remove the extra \`:\` before the description
 `;
       }
       commitReports.push(commitReportText);
       core.endGroup();
     });
 
-    if (countErrors) {
-      core.setFailed(`Action failed with ${countErrors} errors (and ${countWarnings} warnings)`);
+    // Duplicate / similar commit detection
+    const duplicatePairs = [];
+    if (duplicateThreshold !== null) {
+      core.debug(`Checking for duplicate commit subjects (threshold: ${duplicateThreshold})...`);
+      for (let i = 0; i < commits.data.length; i++) {
+        for (let j = i + 1; j < commits.data.length; j++) {
+          const subjectA = normalizeSubject(commits.data[i].commit.message);
+          const subjectB = normalizeSubject(commits.data[j].commit.message);
+          const similarity = diceSimilarity(subjectA, subjectB);
+          core.debug(`  Similarity [${i}↔${j}]: ${similarity.toFixed(3)} — "${subjectA}" vs "${subjectB}"`);
+          if (similarity >= duplicateThreshold) {
+            duplicatePairs.push({ a: commits.data[i], b: commits.data[j], similarity });
+          }
+        }
+      }
+      core.debug(`Found ${duplicatePairs.length} duplicate pair(s)`);
+    }
+
+    if (countErrors || duplicatePairs.length) {
+      core.setFailed(`Action failed with ${countErrors} error(s), ${countWarnings} warning(s), and ${duplicatePairs.length} duplicate commit subject(s)`);
     }
 
     if (deleteComment) {
@@ -187,7 +218,26 @@ ${warningReportText}
     }
 
     if (createComment) {
-      if (countErrors || countWarnings) {
+      const duplicateReport = duplicatePairs.length ? `
+## Duplicate Commit Subjects
+
+The following commit pairs have subjects that are too similar (threshold: ${duplicateThreshold}):
+
+${duplicatePairs.map(({ a, b, similarity }) => {
+    const shaA = a.sha.substring(0, 7);
+    const shaB = b.sha.substring(0, 7);
+    const pct = Math.round(similarity * 100);
+    return `### 🔁 ${pct}% similar
+
+| | Commit | Subject |
+|---|---|---|
+| A | [\`${shaA}\`](https://github.com/${owner}/${repo.name}/commit/${a.sha}) | \`${a.commit.message.split('\n')[0]}\` |
+| B | [\`${shaB}\`](https://github.com/${owner}/${repo.name}/commit/${b.sha}) | \`${b.commit.message.split('\n')[0]}\` |
+`;
+  }).join('\n')}
+` : '';
+
+      if (countErrors || countWarnings || duplicatePairs.length) {
         const finalReport = `
 # 🚨🚔 Unconventional Commit 👮‍♀️🙅‍♂️
 
@@ -199,9 +249,10 @@ ${warningReportText}
 - 👤  **${authors.length} author(s)**
 - ❌  **${countErrors} lint error(s)**
 - ⚠️  **${countWarnings} lint warning(s)**
+- 🔁  **${duplicatePairs.length} duplicate subject pair(s)**
 
 ${commitReports.join('\n')}
-
+${duplicateReport}
 ## Tips
 
 Be sure to follow the [Conventional Commit](https://www.conventionalcommits.org/en/v1.0.0/) guideline when authoring your commits.