Implement faster trimming of control characters

eslint.config.mjs

@@ -4,6 +4,7 @@ export default [
   {
     ignores: [
       "coverage/",
+      "_site/",
       "test/web-platform-tests/",
       "live-viewer/whatwg-url.mjs",
       "lib/VoidFunction.js",

lib/url-state-machine.js

@@ -441,8 +441,22 @@ function domainToASCII(domain, beStrict = false) {
   return result;
 }
 
-function trimControlChars(url) {
-  return url.replace(/^[\u0000-\u001F\u0020]+|[\u0000-\u001F\u0020]+$/ug, "");
+function trimControlChars(string) {
+  // Avoid using regexp because of this V8 bug: https://issues.chromium.org/issues/42204424
+
+  let start = 0;
+  let end = string.length;
+  for (; start < end; ++start) {
+    if (string.charCodeAt(start) > 0x20) {
+      break;
+    }
+  }
+  for (; end > start; --end) {
+    if (string.charCodeAt(end - 1) > 0x20) {
+      break;
+    }
+  }
+  return string.substring(start, end);
 }
 
 function trimTabAndNewline(url) {

package.json

@@ -33,7 +33,8 @@
     "prepare": "node scripts/transform.js",
     "pretest": "node scripts/get-latest-platform-tests.js && node scripts/transform.js",
     "build-live-viewer": "esbuild --bundle --format=esm --sourcemap --outfile=live-viewer/whatwg-url.mjs index.js",
-    "test": "node --test test/*.js"
+    "test": "node --test test/*.js",
+    "bench": "node scripts/benchmark.js"
   },
   "c8": {
     "reporter": [

scripts/benchmark.js

@@ -1,11 +1,11 @@
 "use strict";
 const { URL } = require("../");
 const Benchmark = require("benchmark");
-const testData = require("../test/web-platform-tests/resources/urltestdata.json");
 
+const testData = require("../test/web-platform-tests/resources/urltestdata.json");
 const testInputs = testData.filter(c => typeof c === "object").map(c => c.input);
 
-const benchmark = new Benchmark(() => {
+runBenchmark("URL constructor with WPT data", () => {
   for (const input of testInputs) {
     try {
       // eslint-disable-next-line no-new
@@ -16,5 +16,23 @@ const benchmark = new Benchmark(() => {
   }
 });
 
-benchmark.on("cycle", e => console.log(e.target.toString()));
-benchmark.run();
+runBenchmark("long input not starting or ending with control characters (GH-286)", () => {
+  try {
+    // eslint-disable-next-line no-new
+    new URL(`!!${"\u0000".repeat(100000)}A\rA`);
+  } catch {
+    // intentionally empty
+  }
+});
+
+function runBenchmark(name, fn) {
+  new Benchmark(name, fn, {
+    onComplete(event) {
+      console.log(`${name}:`);
+      console.log(`  ${event.target.hz.toFixed(0)} ops/second`);
+      console.log(`  ±${event.target.stats.rme.toFixed(2)}% relative margin of error`);
+      console.log(`  ${event.target.stats.sample.length} samples`);
+      console.log("");
+    }
+  }).run();
+}