fix: replace regex HTML pre-scan with linear scanner

packages/text-vide/src/__tests__/index.test.ts

@@ -321,6 +321,30 @@ describe('with html tags', () => {
     expect(textVide(text)).toBe(expected);
   });
 
+  it('preserves tag ranges after an unterminated HTML comment opener', () => {
+    const text = 'aa <!-- bb <div>cd</div>';
+    const expected = '<b>a</b>a <!-- bb <div><b>c</b>d</div>';
+    expect(textVide(text)).toBe(expected);
+  });
+
+  it('continues highlighting after an unterminated comment with multiple tags', () => {
+    const text = 'x <!-- unterminated <div>ab</div> yy <i>cd</i>';
+    const expected =
+      'x <!-- unterminated <div><b>a</b>b</div> <b>y</b>y <i><b>c</b>d</i>';
+    expect(textVide(text)).toBe(expected);
+  });
+
+  it('treats an unterminated comment opener like a regular tag opener', () => {
+    const text = 'a <!-- foo';
+    const expected = 'a <!-- <b>fo</b>o';
+    expect(textVide(text)).toBe(expected);
+  });
+
+  it('ignores empty angle brackets', () => {
+    const text = '<>';
+    expect(textVide(text)).toBe('<>');
+  });
+
   it('complex html tags', () => {
     const text = `<div class="bionic-reader-container">
 

packages/text-vide/src/useCheckIsHtmlTag.ts

@@ -1,10 +1,44 @@
-import { extractMatchRangeList } from './utils';
+import { MatchRange } from './utils';
 
-const HTML_TAG_REGEX = /<!--[^]*?-->|<[^>]+>/g;
+const getHtmlTagRangeList = (text: string): MatchRange[] => {
+  const htmlTagRangeList: MatchRange[] = [];
+  let cursor = 0;
+
+  while (cursor < text.length) {
+    const openIndex = text.indexOf('<', cursor);
+    if (openIndex === -1) {
+      break;
+    }
+
+    if (text.startsWith('<!--', openIndex)) {
+      const commentCloseIndex = text.indexOf('-->', openIndex + 4);
+      if (commentCloseIndex !== -1) {
+        htmlTagRangeList.push([openIndex, commentCloseIndex + 2]);
+        cursor = commentCloseIndex + 3;
+        continue;
+      }
+      // Unterminated `<!--`: fall through and treat the `<` as a normal
+      // tag opener so subsequent tags are still recognized (matches the
+      // behavior of the previous `/<!--[^]*?-->|<[^>]+>/g` regex).
+    }
+
+    const closeIndex = text.indexOf('>', openIndex + 1);
+    if (closeIndex === -1) {
+      break;
+    }
+
+    if (closeIndex > openIndex + 1) {
+      htmlTagRangeList.push([openIndex, closeIndex]);
+    }
+
+    cursor = closeIndex + 1;
+  }
+
+  return htmlTagRangeList;
+};
 
 export const useCheckIsHtmlTag = (text: string) => {
-  const htmlTagMatchList = text.matchAll(HTML_TAG_REGEX);
-  const htmlTagRangeList = extractMatchRangeList(htmlTagMatchList);
+  const htmlTagRangeList = getHtmlTagRangeList(text);
   const reversedHtmlTagRangeList = htmlTagRangeList.reverse();
 
   return (match: RegExpMatchArray) => {

packages/text-vide/src/utils.ts

@@ -1,6 +1,8 @@
+export type MatchRange = [number, number];
+
 export const extractMatchRangeList = (
   matchList: IterableIterator<RegExpMatchArray>,
-) =>
+): MatchRange[] =>
   Array.from(matchList).map(match => {
     const startIndex = match.index!;
     const [matchedWord] = match;