About Unicode Normalizer

Normalize Unicode text between NFC, NFD, NFKC, and NFKD forms. Part of the DevTools Surf developer suite. Browse more tools in the Text / String collection.

Use Cases

• Backend developers fixing string comparison bugs across locales
• Data engineers deduplicating records with mixed Unicode forms
• macOS/Windows cross-platform teams resolving filename mismatches
• Search engineers normalizing user queries for consistent indexing

Tips

• Use NFC for web content to ensure consistent rendering
• Compare NFD and NFC to debug string matching failures
• Apply NFKC to normalize fullwidth characters to ASCII

Fun Facts

• Unicode normalization was formalized in Unicode Technical Report #15, first published in 1999 — the four forms (NFC, NFD, NFKC, NFKD) handle the 'e with acute' problem differently.
• The character 'fi' can be represented as one codepoint (U+FB01, the fi ligature) or two (f + i) — NFKD decomposes ligatures while NFC preserves them, causing silent search failures.
• Apple's macOS filesystem (APFS) stores filenames in NFD form, while Windows uses NFC — this mismatch causes cross-platform file sync bugs that have plagued Dropbox and Git for years.

FAQ

What are the four normalization forms?

NFC (composed, canonical — café as one character), NFD (decomposed — café as c + a + f + combining-acute + e), NFKC (composed + compatibility — ff ligature → ff), NFKD (decomposed + compatibility). NFC is the web default.

Which form should I use?

NFC for everything user-facing (storage, display, search indexing). NFD for typography processing. NFKC/NFKD only for search normalization — they're lossy.

Why does normalization matter?

Two visually-identical strings can compare as different if they use different normalization. Login comparisons, cache keys, and file name matching need consistent normalization.

Will it change visible characters?

NFC/NFD roundtrip losslessly. NFKC/NFKD may change ligatures, superscripts, and compatibility characters to their 'simple' forms — visible but usually equivalent.