Phi ký tự

What is a Noncharacter?

A noncharacter is one of 66 specific Unicode code points that are permanently reserved and will never be assigned to any character. Unlike unassigned code points (which are available for future use), noncharacters are deliberately and irrevocably excluded from representing text data.

The Unicode Standard explicitly states: "Noncharacters are code points that are permanently reserved for internal use. They are not characters and must not be used for data interchange."

The 66 Noncharacters

The 66 noncharacters fall into two groups:

Group 1 — Arabic Presentation Forms-A block (34 values):

U+FDD0–U+FDEF  (32 code points)

Group 2 — Last two positions of each plane (32 values):

U+FFFE,  U+FFFF   (BMP, Plane 0)
U+1FFFE, U+1FFFF  (Plane 1)
U+2FFFE, U+2FFFF  (Plane 2)
...
U+10FFFE, U+10FFFF (Plane 16)

Each of the 17 planes contributes 2 noncharacters at its very end: 0xFFFE and 0xFFFF in the low 16 bits. This gives 17 × 2 = 34 noncharacters. Plus the 32 in U+FDD0–U+FDEF = 66 total.

Why Noncharacters Exist

The most important noncharacters are U+FFFE and U+FFFF:

U+FFFE is the mirror image of U+FEFF (the Byte Order Mark). If a UTF-16 stream opens with bytes FF FE, the BOM is at U+FEFF (little-endian byte order). If it opens with FE FF, the BOM would decode as U+FFFE — which signals that the byte order is wrong. Readers can detect and correct the byte order. This clever design means U+FFFE was never available for character use.

U+FFFF was historically used as a "no character" sentinel in character arrays and file reading loops, similar to how -1 (EOF) is used in C's fgetc(). Reserving it prevents confusion between the sentinel and actual text.

The FDD0–FDEF range and the per-plane pairs are reserved to give implementations maximum flexibility for internal sentinels and markers on all planes.

Conformance Requirements

Unicode conformance has specific rules about noncharacters:

• Conformant applications may use noncharacters internally
• Conformant applications should not generate noncharacters in open data interchange
• Conformant applications may accept and pass through noncharacters without error (they are not required to reject them)

This is subtler than it sounds: noncharacters are not invalid UTF-8 or UTF-16 sequences — they are valid encoded values of code points that are simply not intended for data interchange.

Noncharacters vs Surrogates

Detecting Noncharacters

def is_noncharacter(cp: int) -> bool:
    # FDD0-FDEF range
    if 0xFDD0 <= cp <= 0xFDEF:
        return True
    # xFFFE and xFFFF for each plane
    last_two = cp & 0xFFFF
    return last_two in (0xFFFE, 0xFFFF) and cp <= 0x10FFFF

# Examples
print(is_noncharacter(0xFDD0))    # True
print(is_noncharacter(0xFFFF))    # True
print(is_noncharacter(0x1FFFF))   # True
print(is_noncharacter(0x10FFFF))  # True
print(is_noncharacter(0x0041))    # False (LATIN CAPITAL LETTER A)

Quick Facts