Vault import: decode ZIP names as UTF-8 and NFC-normalize them

Real vault ZIPs broke umlauts in page titles ("Fußball zum Götzen" →
mojibake, slug fua-ball…goi-tzen): fflate honors only the ZIP UTF-8
flag, which common archivers omit, and decodes unflagged names as
Latin-1. That decoding is byte-lossless, so parseVaultZip now re-reads
any name whose chars all fit one byte as UTF-8 (a strict decoder —
genuine Latin-1 and flag-decoded precomposed chars fall back
unchanged), then NFC-normalizes: macOS zips store umlauts decomposed,
which silently broke slugify's ä→ae digraphs, wikilink matching, and
duplicate-basename detection. slugify itself also precomposes first as
defense in depth for NFD input from other paths.

Unit tests pin both cases: a hand-patched ZIP whose UTF-8 name bytes
carry no UTF-8 flag, and an NFD-named note that must come out
precomposed with an ueber- slug.

Pages already imported with garbled titles stay as they are — delete
the imported subtree and re-import after this lands (or rename by
hand).

Fixes #127

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Fb2VzvcoBPHkjh8bZ6PzQn

apps/api/src/import-export/obsidian-vault.test.ts

@@ -54,6 +54,41 @@ describe('parseVaultZip (issue #116)', () => {
     expect(vault.assets[0]!.extension).toBe('png');
   });
 
+  it('decodes UTF-8 entry names even without the ZIP UTF-8 flag (#127)', () => {
+    // Real-world archivers often omit the UTF-8 flag; fflate then decodes
+    // the name bytes as Latin-1 ("Fußball" → mojibake in page titles).
+    // Build such a ZIP by zipping an ASCII placeholder of the same byte
+    // length (keeps the flag clear) and patching in the raw UTF-8 bytes.
+    const utf8Name = new TextEncoder().encode('Grüße.md');
+    const placeholder = 'GrAABBe.md';
+    expect(utf8Name.length).toBe(placeholder.length);
+    const zipped = zipSync({ [placeholder]: new TextEncoder().encode('# Hallo') });
+    const placeholderBytes = new TextEncoder().encode(placeholder);
+    for (let i = 0; i <= zipped.length - placeholderBytes.length; i += 1) {
+      if (placeholderBytes.every((byte, j) => zipped[i + j] === byte)) {
+        zipped.set(utf8Name, i); // local header + central directory
+      }
+    }
+
+    const vault = parseVaultZip(zipped);
+    expect(vault.notes.map((n) => n.name)).toEqual(['Grüße']);
+  });
+
+  it('NFC-normalizes decomposed names so umlaut digraphs survive (#127)', () => {
+    // macOS zips store umlauts decomposed (NFD): "\u00dc" as U + combining mark.
+    const nfdName = 'U\u0308ber NFD.md';
+    expect(nfdName.normalize('NFC')).not.toBe(nfdName); // really decomposed
+    const vault = parseVaultZip(fixtureZip({ [nfdName]: new TextEncoder().encode('# NFD') }));
+    const note = vault.notes.find((n) => n.path.includes('NFD'));
+    expect(note?.name).toBe('\u00dcber NFD'); // precomposed
+    const plan = planVaultImport(fixtureZip({ [nfdName]: new TextEncoder().encode('x') }), {
+      frontmatterMode: 'strip',
+      existingSlugs: new Set(),
+      mountDepth: 0,
+    });
+    expect(plan.notes.map((n) => n.slug)).toContain('ueber-nfd');
+  });
+
   it('rejects zip-slip paths, garbage, and oversized vaults', () => {
     expect(() => parseVaultZip(new Uint8Array([1, 2, 3]))).toThrowError(VaultError);
     expect(() =>

apps/api/src/import-export/obsidian-vault.ts

@@ -99,6 +99,32 @@ function safeRelativePath(rawPath: string): string {
   return path;
 }
 
+/**
+ * Repairs a ZIP entry name that fflate decoded as Latin-1 (#127). fflate
+ * honors only the ZIP UTF-8 flag (bit 11), which common archivers omit —
+ * the UTF-8 bytes of "Fußball" then arrive as one mojibake char per byte
+ * and end up in page titles. Latin-1 decoding is byte-lossless (charCode ==
+ * byte), so when every char fits a byte we re-read them as UTF-8; names the
+ * flag DID mark as UTF-8 contain either multi-byte chars (> 0xFF, left
+ * alone) or precomposed Latin-1 chars whose bytes are invalid UTF-8 (the
+ * decoder throws → keep the original). Finally NFC-normalize: macOS zips
+ * store umlauts decomposed (NFD), which breaks slugify's ä→ae digraphs,
+ * wikilink matching, and duplicate-basename detection.
+ */
+function decodeZipName(raw: string): string {
+  let name = raw;
+  // eslint-disable-next-line no-control-regex
+  if (/^[\u0000-\u00ff]*$/.test(raw)) {
+    const bytes = Uint8Array.from(raw, (char) => char.charCodeAt(0));
+    try {
+      name = new TextDecoder('utf-8', { fatal: true }).decode(bytes);
+    } catch {
+      // Genuinely Latin-1 (or already plain ASCII) — keep as decoded.
+    }
+  }
+  return name.normalize('NFC');
+}
+
 /**
  * Unpacks the vault ZIP into notes (`.md`) and assets (everything else),
  * enforcing the unpacked-size ceiling incrementally. Hidden housekeeping
@@ -119,7 +145,8 @@ export function parseVaultZip(
   const notes: VaultNote[] = [];
   const assets: VaultAsset[] = [];
   let unpacked = 0;
-  for (const [rawPath, bytes] of Object.entries(entries)) {
+  for (const [zipName, bytes] of Object.entries(entries)) {
+    const rawPath = decodeZipName(zipName);
     if (rawPath.endsWith('/')) continue; // directory entries
     const path = safeRelativePath(rawPath);
     const segments = path.split('/');

packages/shared/src/ponds.ts

@@ -112,16 +112,21 @@ const MAX_SLUG_LENGTH = 60;
  * Uniqueness (numeric suffixes) is the caller's job, not slugify's.
  */
 export function slugify(input: string): string {
-  return input
-    .toLowerCase()
-    .replace(/ä/g, 'ae')
-    .replace(/ö/g, 'oe')
-    .replace(/ü/g, 'ue')
-    .replace(/ß/g, 'ss')
-    .normalize('NFKD')
-    .replace(/[\u0300-\u036f]/g, '')
-    .replace(/[^a-z0-9]+/g, '-')
-    .replace(/^-+|-+$/g, '')
-    .slice(0, MAX_SLUG_LENGTH)
-    .replace(/-+$/, '');
+  return (
+    input
+      // Precompose first (#127): macOS file names arrive NFD-decomposed, and
+      // the umlaut digraphs below only match the precomposed forms.
+      .normalize('NFC')
+      .toLowerCase()
+      .replace(/ä/g, 'ae')
+      .replace(/ö/g, 'oe')
+      .replace(/ü/g, 'ue')
+      .replace(/ß/g, 'ss')
+      .normalize('NFKD')
+      .replace(/[\u0300-\u036f]/g, '')
+      .replace(/[^a-z0-9]+/g, '-')
+      .replace(/^-+|-+$/g, '')
+      .slice(0, MAX_SLUG_LENGTH)
+      .replace(/-+$/, '')
+  );
 }