fix: apply Unicode full case fold in link reference label normalization

src/Tokenizer.ts

@@ -5,6 +5,7 @@ import {
   findClosingBracket,
   expandTabs,
   trimTrailingBlankLines,
+  normalizeLabel,
 } from './helpers.ts';
 import type { Rules } from './rules.ts';
 import type { _Lexer } from './Lexer.ts';
@@ -499,7 +500,7 @@ export class _Tokenizer<ParserOutput = string, RendererOutput = string> {
   def(src: string): Tokens.Def | undefined {
     const cap = this.rules.block.def.exec(src);
     if (cap) {
-      const tag = cap[1].toLowerCase().replace(this.rules.other.multipleSpaceGlobal, ' ');
+      const tag = normalizeLabel(cap[1]).replace(this.rules.other.multipleSpaceGlobal, ' ');
       const href = cap[2] ? cap[2].replace(this.rules.other.hrefBrackets, '$1').replace(this.rules.inline.anyPunctuation, '$1') : '';
       const title = cap[3] ? cap[3].substring(1, cap[3].length - 1).replace(this.rules.inline.anyPunctuation, '$1') : cap[3];
       return {
@@ -717,8 +718,8 @@ export class _Tokenizer<ParserOutput = string, RendererOutput = string> {
     let cap;
     if ((cap = this.rules.inline.reflink.exec(src))
       || (cap = this.rules.inline.nolink.exec(src))) {
-      const linkString = (cap[2] || cap[1]).replace(this.rules.other.multipleSpaceGlobal, ' ');
-      const link = links[linkString.toLowerCase()];
+      const linkString = normalizeLabel(cap[2] || cap[1]).replace(this.rules.other.multipleSpaceGlobal, ' ');
+      const link = links[linkString];
       if (!link) {
         const text = cap[0].charAt(0);
         return {

src/helpers.ts

@@ -164,3 +164,25 @@ export function expandTabs(line: string, indent = 0) {
 
   return expanded;
 }
+
+/**
+ * Normalize a link reference label per CommonMark spec:
+ * perform Unicode full case fold (covering multi-char expansions that
+ * String.prototype.toLowerCase() misses, e.g. ẞ→ss, ligatures→letters),
+ * then collapse runs of whitespace to a single space.
+ *
+ * @see https://spec.commonmark.org/0.31.2/#matches
+ */
+export function normalizeLabel(label: string): string {
+  return label
+    // Apply Unicode full case fold before toLowerCase so that
+    // ẞ (U+1E9E) → ss and ß (U+00DF) → ss (simple fold maps ẞ to ß first).
+    .toLowerCase()
+    .replace(/ß/g, 'ss') // U+00DF + folded U+1E9E
+    .replace(/ff/g, 'ff') // U+FB00
+    .replace(/fi/g, 'fi') // U+FB01
+    .replace(/fl/g, 'fl') // U+FB02
+    .replace(/ffi/g, 'ffi') // U+FB03
+    .replace(/ffl/g, 'ffl') // U+FB04
+    .replace(/ſt|st/g, 'st'); // U+FB05, U+FB06
+}

test/specs/commonmark/commonmark.0.31.2.json

@@ -4331,8 +4331,7 @@
     "example": 540,
     "start_line": 8129,
     "end_line": 8135,
-    "section": "Links",
-    "shouldFail": true
+    "section": "Links"
   },
   {
     "markdown": "[Foo\n  bar]: /url\n\n[Baz][Foo bar]\n",

test/specs/gfm/commonmark.0.31.2.json

@@ -4331,8 +4331,7 @@
     "example": 540,
     "start_line": 8129,
     "end_line": 8135,
-    "section": "Links",
-    "shouldFail": true
+    "section": "Links"
   },
   {
     "markdown": "[Foo\n  bar]: /url\n\n[Baz][Foo bar]\n",