NFKC (Compatibility Composition)

NFKC: Normalization for Identifiers and Search

NFKC (Normalization Form KC — Compatibility Decomposition followed by Canonical Composition) is the most useful normalization form for search engines, programming language identifiers, and any system that needs to treat visually or semantically similar characters as equivalent.

The "K" in NFKC stands for "Kompatibility" — beyond the canonical decompositions that NFC handles, NFKC also folds compatibility characters: Unicode code points that exist for historical or typographic reasons but represent the same abstract character as something already encoded.

What Compatibility Folding Does

Compatibility decompositions cover hundreds of character categories:

import unicodedata

# Ligature folding
assert unicodedata.normalize("NFKC", "file") == "file"

# Fullwidth ASCII folding (common in CJK input)
fw = "Ａ１Ｂ２"   # fullwidth
ascii_eq = unicodedata.normalize("NFKC", fw)
print(ascii_eq)  # "A1B2"

# Superscript folding
assert unicodedata.normalize("NFKC", "x²") == "x2"

# Combined with casefold for case-insensitive search
def normalize_for_search(s: str) -> str:
    return unicodedata.normalize("NFKC", s).casefold()

print(normalize_for_search("ＨＥＬＬＯＷＯＲＬＤﾞ"))  # "helloworld゛" → further processing needed

NFKC in Standards

Python identifiers: Python 3 uses NFKC normalization for identifiers. That means ＭｙＣｌａｓｓ (fullwidth) is valid and equivalent to MyClass.

PRECIS (RFC 8264): The successor to stringprep for usernames and passwords uses NFKC as a normalization step.

IDNA (Internationalized Domain Names): Domain name processing uses NFKC.

Password hashing: Many systems apply NFKC before hashing to ensure finance and finance hash the same way.

NFKC vs NFC: The Trade-off

NFKC loses information — ² and 2 are distinct characters with different semantic roles, but NFKC makes them identical. This is intentional for search and identifiers but wrong for document storage. Never use NFKC as your storage format if you need to preserve superscripts, ligatures, or enclosed numbers.

Quick Facts