组合字符

What is a Combining Character?

A combining character is a Unicode character that has no independent visual form of its own — instead, it attaches to and modifies the preceding character (called the base character). Combining characters implement diacritical marks, tone marks, vowel signs, and other modifier symbols that, in Unicode's model, are logically separate from the letter they modify.

The key insight is that Unicode separates the identity of a character from its rendering. A letter with a diacritic can be represented either as a single precomposed code point (é = U+00E9) or as a base letter followed by a combining mark (e + ◌́ = U+0065 U+0301). Both sequences represent the same abstract character and produce the same rendered output.

How Combining Characters Work

Combining characters have a General Category of Mn (Non-spacing Mark), Mc (Spacing Mark), or Me (Enclosing Mark). Non-spacing marks are the most common — they occupy zero advance width and position themselves relative to the base character's glyph using the font's anchor points.

When a text shaping engine encounters a base character followed by one or more combining marks, it: 1. Retrieves the base glyph from the font 2. Positions each combining glyph at the appropriate anchor (top, bottom, left, right) 3. Renders them as a single grapheme cluster

Multiple combining characters can stack on a single base:

a + ◌̂ + ◌̄ = â̄  (a with circumflex and macron)

Unicode defines a canonical ordering for combining marks based on their Combining Class value (0–255). Marks with lower combining class values (e.g., below-base nuktas) appear before marks with higher values (e.g., above-base accents) in normalized text.

Important Combining Character Ranges

Grapheme Clusters

A base character plus all its combining marks form a grapheme cluster — the unit that users perceive as a single character. Programming languages must account for this:

import unicodedata

# "e" + combining acute = é
s = "e\u0301"
print(len(s))           # 2 (two code points)
print(s)                # é (looks like 1 character)

# Precomposed é
s2 = "\u00e9"
print(len(s2))          # 1 (one code point)
print(s == s2)          # False! Different code points

# Normalize to compare
import unicodedata
print(unicodedata.normalize("NFC", s) == s2)  # True

JavaScript's Intl.Segmenter and Swift's String.count handle grapheme clusters correctly; many other APIs count code points instead.

Quick Facts