码元

What is a Code Unit?

A code unit is the minimal bit combination used in a Unicode encoding. Different encodings use different code unit sizes:

• UTF-8: 8-bit code units (bytes)
• UTF-16: 16-bit code units (2-byte words)
• UTF-32: 32-bit code units (4-byte words)

A code unit is not the same as a code point. Code points are abstract Unicode values (U+0000–U+10FFFF); code units are the concrete building blocks that encodings use to represent those values. One code point may require one or more code units depending on the encoding and the code point's value.

Code Units by Encoding

UTF-8 (8-bit code units)

UTF-8 uses 1 to 4 bytes per code point, following a variable-length scheme:

'A'  (U+0041) → 1 code unit:  0x41
'é'  (U+00E9) → 2 code units: 0xC3 0xA9
'中' (U+4E2D) → 3 code units: 0xE4 0xB8 0xAD
'😀' (U+1F600) → 4 code units: 0xF0 0x9F 0x98 0x80

UTF-16 (16-bit code units)

BMP characters (U+0000–U+FFFF) use 1 code unit. Supplementary characters use 2 code units (a surrogate pair):

'A'  (U+0041)  → 1 code unit:  0x0041
'中' (U+4E2D)  → 1 code unit:  0x4E2D
'😀' (U+1F600) → 2 code units: 0xD83D 0xDE00  (surrogate pair)

UTF-32 (32-bit code units)

Every code point uses exactly 1 code unit — UTF-32 is the only fixed-width Unicode encoding:

'A'  (U+0041)  → 0x00000041
'中' (U+4E2D)  → 0x00004E2D
'😀' (U+1F600) → 0x0001F600

Why Code Units Matter in Programming

Many programming languages expose string length in terms of code units, not code points or grapheme clusters:

s = "😀"
len(s)               # Python 3: 1 — counts code points (Unicode scalars)
len(s.encode("utf-8"))  # 4 — UTF-8 code units (bytes)

"😀".length          // 2 — JavaScript counts UTF-16 code units
[..."😀"].length     // 1 — spread iterator counts code points

String s = "😀";
s.length()           // 2 — Java String.length() counts UTF-16 code units
s.codePointCount(0, s.length()) // 1 — code point count

This is a common source of bugs: naive string slicing by index in JavaScript or Java can split a surrogate pair, producing invalid text.

Common Pitfalls

Confusing code units with bytes: A UTF-16 code unit is 2 bytes, not 1. A string of length n in a Java String occupies at least 2n bytes.

Assuming 1 code unit = 1 character: A single user-visible character (grapheme cluster) may require multiple code points, each potentially requiring multiple code units.

Slicing strings at byte offsets: UTF-8 continuation bytes begin with 10xxxxxx. Slicing between them produces invalid UTF-8 sequences.

Quick Facts