fix: prevent quadratic complexity in emStrongLDelim regex

src/Tokenizer.ts

@@ -729,11 +729,12 @@ export class _Tokenizer<ParserOutput = string, RendererOutput = string> {
   emStrong(src: string, maskedSrc: string, prevChar = ''): Tokens.Em | Tokens.Strong | undefined {
     let match = this.rules.inline.emStrongLDelim.exec(src);
     if (!match) return;
+    if (!match[1] && !match[2] && !match[3] && !match[4]) return;
 
     // _ can't be between two alphanumerics. \p{L}\p{N} includes non-english alphabet/numbers as well
-    if (match[3] && prevChar.match(this.rules.other.unicodeAlphaNumeric)) return;
+    if (match[4] && prevChar.match(this.rules.other.unicodeAlphaNumeric)) return;
 
-    const nextChar = match[1] || match[2] || '';
+    const nextChar = match[1] || match[3] || '';
 
     if (!nextChar || !prevChar || this.rules.inline.punctuation.exec(prevChar)) {
       // unicode Regex counts emoji as 1 char; spread into array for proper count (used multiple times below)

src/rules.ts

@@ -290,7 +290,7 @@ const blockSkip = edit(/link|precode-code|html/, 'g')
   .replace('html', /<(?! )[^<>]*?>/)
   .getRegex();
 
-const emStrongLDelimCore = /^(?:\*+(?:((?!\*)punct)|[^\s*]))|^_+(?:((?!_)punct)|([^\s_]))/;
+const emStrongLDelimCore = /^(?:\*+(?:((?!\*)punct)|([^\s*]))?)|^_+(?:((?!_)punct)|([^\s_]))?/;
 
 const emStrongLDelim = edit(emStrongLDelimCore, 'u')
   .replace(/punct/g, _punctuation)

test/specs/redos/quadratic_emstrong_delim.cjs

@@ -0,0 +1,4 @@
+module.exports = {
+  markdown: '_'.repeat(10000) + ' a',
+  html: `<p>${'_'.repeat(10000)} a</p>\n`,
+};