fix: dedupe extracted comments in linear time (#682)

Author: alexander-akait

.changeset/dedupe-extracted-comments-linear.md

@@ -0,0 +1,5 @@
+---
+"minimizer-webpack-plugin": patch
+---
+
+deduplicate extracted comments in linear time, so builds stay fast when an asset contains many distinct preserved comments

.github/workflows/nodejs.yml

@@ -118,6 +118,22 @@ jobs:
         if: matrix.node-version == '14.x' || matrix.node-version == '16.x'
         run: npm install -D --no-save cssnano@^6 --force
 
+      - name: Pin terser to the locked version (Node 10/12/14)
+        # npm 6 (bundled with Node <= 14) can't read the lockfileVersion 3
+        # package-lock.json, so the install step re-resolves `terser` to the
+        # latest match of its range instead of the locked version (and
+        # rewrites package-lock.json to v1 in the process). A newer terser
+        # changes the minified output, which breaks the snapshots. Read the
+        # locked version from the pristine committed lockfile and reinstall
+        # that exact version so these rows match the snapshots.
+        if: matrix.node-version == '10.x' || matrix.node-version == '12.x' || matrix.node-version == '14.x'
+        shell: bash
+        run: |
+          git show HEAD:package-lock.json > pristine-package-lock.json
+          TERSER_VERSION="$(node -e "process.stdout.write(require('./pristine-package-lock.json').packages['node_modules/terser'].version)")"
+          rm -f pristine-package-lock.json
+          npm install -D --no-save --ignore-scripts --force "terser@${TERSER_VERSION}"
+
       - name: Install webpack ${{ matrix.webpack-version }}
         if: matrix.webpack-version != 'latest'
         run: npm i webpack@${{ matrix.webpack-version }}

jest.config.js

@@ -1,10 +1,23 @@
 // The bundled CSS minimizers (`cssnano@7`, `@swc/css`, `lightningcss`,
 // `esbuild@0.27`) require modern Node and don't reliably install on the
 // Windows agents. Skip the dedicated CSS test file outright on rows that
-// can't run them so we don't end up with stale or missing snapshots.
+// can't run them so we don't end up with stale or missing snapshots. The
+// `@swc/html` minimizers require Node >= 14, so the dedicated swc-html file
+// is skipped on older Node for the same reason.
 const NODE_MAJOR = Number(process.versions.node.split(".")[0]);
 const IS_WINDOWS = process.platform === "win32";
 const RUN_CSS_TESTS = NODE_MAJOR >= 18 && !IS_WINDOWS;
+const RUN_SWC_HTML_TESTS = NODE_MAJOR >= 14;
+
+const testPathIgnorePatterns = [];
+
+if (!RUN_CSS_TESTS) {
+  testPathIgnorePatterns.push("/test/css-minify-option\\.test\\.js$");
+}
+
+if (!RUN_SWC_HTML_TESTS) {
+  testPathIgnorePatterns.push("/test/swc-html-minify-option\\.test\\.js$");
+}
 
 module.exports = {
   testEnvironment: "node",
@@ -13,7 +26,6 @@ module.exports = {
   // routinely take longer than that.
   testTimeout: 60000,
   coveragePathIgnorePatterns: ["src/serialize-javascript.js"],
-  testPathIgnorePatterns: RUN_CSS_TESTS
-    ? []
-    : ["/test/css-minify-option\\.test\\.js$"],
+  snapshotSerializers: ["<rootDir>/test/helpers/snapshotHashSerializer.js"],
+  testPathIgnorePatterns,
 };

package-lock.json

@@ -116,7 +116,7 @@
     "serialize-javascript": "^7.0.5",
     "typescript": "^6.0.3",
     "uglify-js": "^3.19.3",
-    "webpack": "^5.101.0",
+    "webpack": "^5.107.2",
     "webpack-cli": "^4.10.0",
     "worker-loader": "^3.0.8"
   },

package.json

@@ -223,6 +223,8 @@ async function terserMinify(
 
     // Redefine the comments function to extract and preserve
     // comments according to the two conditions
+    const seenComments = new Set(extractedComments);
+
     return (astNode, comment) => {
       if (
         /** @type {{ extract: ExtractCommentsFunction }} */
@@ -234,7 +236,8 @@ async function terserMinify(
             : `//${comment.value}`;
 
         // Don't include duplicate comments
-        if (!extractedComments.includes(commentText)) {
+        if (!seenComments.has(commentText)) {
+          seenComments.add(commentText);
           extractedComments.push(commentText);
         }
       }
@@ -477,6 +480,8 @@ async function uglifyJsMinify(
 
     // Redefine the comments function to extract and preserve
     // comments according to the two conditions
+    const seenComments = new Set(extractedComments);
+
     return (astNode, comment) => {
       if (
         /** @type {{ extract: ExtractCommentsFunction }} */
@@ -488,7 +493,8 @@ async function uglifyJsMinify(
             : `//${comment.value}`;
 
         // Don't include duplicate comments
-        if (!extractedComments.includes(commentText)) {
+        if (!seenComments.has(commentText)) {
+          seenComments.add(commentText);
           extractedComments.push(commentText);
         }
       }