The Lie Your Runtime Tells You
'👨👩👧👦'.length // 11
That is not 11 characters. It is one family emoji. The .length property in JavaScript counts UTF-16UTF-16
每字符使用2或4个字节的可变长度Unicode编码方式，被JavaScript、Java和Windows内部使用。 code units, not visible characters. For most ASCII text this distinction is invisible, but emoji expose the gap brutally.
Understanding why this happens — and how to fix it — requires knowing what a grapheme cluster is.
Three Levels of "Character"
Unicode defines multiple levels of text unit, each useful for different purposes:
| Level | Unit | JavaScript | Python |
|---|---|---|---|
| Byte | Raw storage unit | — | len(s.encode('utf-8UTF-8 |
| Code unit | Encoding unit | s.length |
— |
| Code point | Unicode scalar value | [...s].length |
len(s) |
| Grapheme cluster | User-perceived character | Intl.Segmenter |
grapheme library |
A grapheme cluster is what a human thinks of as "one character" — one visible glyph on screen. For emoji, this is always at least one code point, but often several combined.
What Makes an Emoji "Long"?
Basic Emoji: 1 Code Point, 2 Code Units
😀 = U+1F600
UTF-16: 0xD83D 0xDE00 (surrogate pair)
.length in JS: 2
Python len(): 1
Emoji + Variation Selector: 2 Code Points
❤️ = U+2764 + U+FE0F
.length in JS: 2
Python len(): 2
Grapheme clusters: 1
Emoji + Skin Tone: 2 Code Points
👍🏽 = U+1F44D + U+1F3FD
.length in JS: 4 (two surrogate pairs)
Python len(): 2
Grapheme clusters: 1
ZWJ零宽连接符（ZWJ） (ZWJ)
一个不可见的Unicode字符（U+200D），用于将多个表情符号合并为单一复合表情符号，例如将人物与物体组合成职业表情符号。 Sequence: 3+ Code Points
👩💻 = U+1F469 + U+200D + U+1F4BB
.length in JS: 5
Python len(): 3
Grapheme clusters: 1
Family Emoji: 7 Code Points
👨👩👧👦 = 👨 + ZWJ + 👩 + ZWJ + 👧 + ZWJ + 👦
.length in JS: 11
Python len(): 7
Grapheme clusters: 1
Correct Counting in JavaScript
Using Intl.Segmenter (Modern, Built-In)
Intl.Segmenter is available in all modern browsers and Node.js 16+. It segments text by grapheme clusters:
function countGraphemes(str) {
const segmenter = new Intl.Segmenter();
return [...segmenter.segment(str)].length;
}
countGraphemes('Hello') // 5 — correct
countGraphemes('Hello 😀') // 7 — correct
countGraphemes('👨👩👧👦') // 1 — correct!
countGraphemes('👩🏽💻') // 1 — correct!
Slicing Text Correctly
Slicing with .slice() on emoji strings corrupts surrogate pairs:
const s = 'Hello 👋';
// WRONG: may cut through a surrogate pair
s.slice(0, 7) // 'Hello \uD83D' — broken!
// CORRECT: spread to code points first
[...s].slice(0, 7).join('') // 'Hello 👋' — correct
// EVEN BETTER: slice by grapheme clusters
function sliceByGrapheme(str, start, end) {
const segmenter = new Intl.Segmenter();
const segments = [...segmenter.segment(str)];
return segments.slice(start, end).map(s => s.segment).join('');
}
sliceByGrapheme('Hello 👋 World', 0, 7) // 'Hello 👋'
String Reversal
Reversing an emoji string with .split('').reverse().join('') produces garbage:
// WRONG: reverses code units, breaks surrogates
'Hello 😀'.split('').reverse().join('') // '😀 olleH' (actually garbled)
// CORRECT: reverse by code points
[...'Hello 😀'].reverse().join('') // '😀 olleH'
// BEST: reverse by grapheme clusters
function reverseGraphemes(str) {
const segmenter = new Intl.Segmenter();
return [...segmenter.segment(str)]
.map(s => s.segment)
.reverse()
.join('');
}
reverseGraphemes('Hello 👨👩👧👦') // '👨👩👧👦 olleH'
Correct Counting in Python
Python 3's len() counts code points, not grapheme clusters. For most ASCII text this is identical to grapheme count, but for emoji it is not.
text = '👨👩👧👦'
len(text) # 7 — code points (wrong for "visible characters")
Using the grapheme Library
import grapheme
grapheme.length('👨👩👧👦') # 1 — correct
grapheme.length('Hello 😀') # 7 — correct
# Slice by grapheme clusters
grapheme.slice('Hello 👨👩👧👦', 0, 7) # 'Hello 👨👩👧👦'
# Iterate graphemes
list(grapheme.graphemes('Hi 👋'))
# ['H', 'i', ' ', '👋']
Using the regex Module with \X
The \X pattern in the regex module matches grapheme clusters:
import regex
def count_graphemes(text):
return len(regex.findall(r'\X', text))
count_graphemes('👨👩👧👦') # 1
count_graphemes('Hello 👨👩👧👦') # 7
Other Languages
Java
// Java uses UTF-16 internally like JavaScript
"👨👩👧👦".length() // 11 (code units)
// Count code points
"👨👩👧👦".codePointCount(0, "👨👩👧👦".length()) // 7
// Count grapheme clusters (requires ICUICU (ICU)
国际Unicode组件（International Components for Unicode），广泛使用的开源库，提供Unicode和国际化支持，包括表情符号处理。)
BreakIterator bi = BreakIterator.getCharacterInstance();
bi.setText("👨👩👧👦");
int count = 0;
while (bi.next() != BreakIterator.DONE) count++;
// count = 1
Swift
Swift's String is designed around grapheme clusters — count returns grapheme count by default:
"👨👩👧👦".count // 1 — correct out of the box!
This is one area where Swift's string model is superior to most other languages.
Rust
// Rust len() returns bytes
"👨👩👧👦".len() // 25 (bytes)
// chars() iterates code points
"👨👩👧👦".chars().count() // 7
// For grapheme clusters, use the unicode-segmentation crate
use unicode_segmentation::UnicodeSegmentation;
"👨👩👧👦".graphemes(true).count() // 1
Practical Implications
Input Validation / Character Limits
If you show users "max 280 characters" and count by .length in JavaScript, users can enter far fewer emoji than expected (or you silently truncate their input). Count by grapheme clusters for user-facing limits.
Database Column Sizing
Ensure column size limits account for multi-code-point emoji. A VARCHAR(10) that accepts emoji-heavy input needs to be measured in code points (PostgreSQL) or bytes (MySQL with utf8mb4), not graphemes.
Text Rendering
Layout engines that use pixel widths handle emoji correctly by nature — they measure bounding boxes, not code units. But manual text layout (canvas, PDF generation, terminal output) needs grapheme-aware iteration.
Use our Sequence Analyzer to see exactly how many code points, code units, and grapheme clusters any emoji string contains — making these abstract concepts concrete.