1 | #ifndef DATA_LANG_UTF8_H
|
2 | #define DATA_LANG_UTF8_H
|
3 |
|
4 | #include <stdio.h>
|
5 |
|
6 | #include <stdint.h> // uint32_t
|
7 | #include <stddef.h> // size_t
|
8 |
|
9 | /**
|
10 | * ---- Quick reference about the encoding ----
|
11 | *
|
12 | * First, all valid UTF-8 sequences follow of bit "patterns" (Table 3-6.) The
|
13 | * first byte determines the length of the sequence and then the next 0-3 bytes
|
14 | * are "continuation bytes."
|
15 | *
|
16 | * +----------------------------+----------+----------+----------+----------+
|
17 | * | Scalar Value | 1st Byte | 2nd Byte | 3rd Byte | 4th Byte |
|
18 | * +----------------------------+----------+----------+----------+----------+
|
19 | * | 00000000 0xxxxxxx | 0xxxxxxx | | | |
|
20 | * | 00000yyy yyxxxxxx | 110yyyyy | 10xxxxxx | | |
|
21 | * | zzzzyyyy yyxxxxxx | 1110zzzz | 10yyyyyy | 10xxxxxx | |
|
22 | * | 000uuuuu zzzzyyyy yyxxxxxx | 11110uuu | 10uuzzzz | 10yyyyyy | 10xxxxxx |
|
23 | * +----------------------------+----------+----------+----------+----------+
|
24 | *
|
25 | * Table 3-6 from Unicode Standard 15.0.0 Ch3. UTF-8 bit patterns
|
26 | *
|
27 | * There are 3 further restrictions which make some valid bit patterns
|
28 | * *invalid*:
|
29 | * 1. Overlongs: eg, <0x41> and <0xC1 0x81> both store U+41, but the second
|
30 | * sequence is longer and thus an error.
|
31 | * 2. Surrogates: Any codepoint between U+D800 and U+DFFF (inclusive) is a
|
32 | * surrogate. It is an error to encode surrogates in UTF-8.
|
33 | * 3. Too Large: Any decoded value over 0x10FFFF is not a Unicode codepoint,
|
34 | * and must be rejected as an error.
|
35 | *
|
36 | * See https://aolsen.ca/writings/everything-about-utf8 for more details about
|
37 | * the encoding.
|
38 | */
|
39 |
|
40 | typedef enum Utf8Error {
|
41 | UTF8_OK = 0,
|
42 |
|
43 | // Encodes a codepoint in more bytes than necessary
|
44 | UTF8_ERR_OVERLONG,
|
45 |
|
46 | // Encodes a codepoint in the surrogate range (0xD800 to 0xDFFF, inclusive)
|
47 | UTF8_ERR_SURROGATE,
|
48 |
|
49 | // Encodes a value greater than the max codepoint U+10FFFF
|
50 | UTF8_ERR_TOO_LARGE,
|
51 |
|
52 | // Encoding doesn't conform to the UTF-8 bit patterns
|
53 | UTF8_ERR_BAD_ENCODING,
|
54 |
|
55 | // It looks like there is another codepoint, but it has been truncated.
|
56 | UTF8_ERR_TRUNCATED_BYTES,
|
57 |
|
58 | // We are at the end of the string. (input_len = 0)
|
59 | UTF8_ERR_END_OF_STREAM,
|
60 | } Utf8Error_t;
|
61 |
|
62 | typedef struct Utf8Result {
|
63 | Utf8Error_t error;
|
64 | uint32_t codepoint;
|
65 | size_t bytes_read;
|
66 | } Utf8Result_t;
|
67 |
|
68 | static inline void _cont(const unsigned char *input, Utf8Result_t *result) {
|
69 | if (result->error) return;
|
70 |
|
71 | int byte = input[result->bytes_read];
|
72 | if (byte == '\0') {
|
73 | result->error = UTF8_ERR_TRUNCATED_BYTES;
|
74 | return;
|
75 | }
|
76 | result->bytes_read += 1;
|
77 |
|
78 | // Continuation bytes follow the bit pattern 10xx_xxxx. We need to a)
|
79 | // validate the pattern and b) remove the leading '10'.
|
80 | if ((byte & 0xC0) == 0x80) {
|
81 | result->codepoint <<= 6;
|
82 | result->codepoint |= byte & 0x3F;
|
83 | } else {
|
84 | result->error = UTF8_ERR_BAD_ENCODING;
|
85 | }
|
86 | }
|
87 |
|
88 | /**
|
89 | * Given a nul-terminated string `input`, try to decode the next codepoint from
|
90 | * that string.
|
91 | *
|
92 | * If there was a surrogate, overlong or codepoint to large error then
|
93 | * `result.codepoint` will contain the recovered value.
|
94 | */
|
95 | static inline void utf8_decode(const unsigned char *input, Utf8Result_t *result) {
|
96 | result->error = UTF8_OK;
|
97 | result->codepoint = 0;
|
98 | result->bytes_read = 0;
|
99 |
|
100 | int first = *input;
|
101 | if (first == '\0') {
|
102 | result->error = UTF8_ERR_END_OF_STREAM;
|
103 | return;
|
104 | }
|
105 | result->bytes_read = 1;
|
106 |
|
107 | if ((first & 0x80) == 0) {
|
108 | // 1-byte long (ASCII subset)
|
109 | result->codepoint = first;
|
110 | return;
|
111 | }
|
112 |
|
113 | if ((first & 0xE0) == 0xC0) {
|
114 | // 2-bytes long
|
115 | result->codepoint = first & 0x1F;
|
116 |
|
117 | _cont(input, result);
|
118 | if (result->error) return;
|
119 |
|
120 | if (result->codepoint < 0x80) {
|
121 | result->error = UTF8_ERR_OVERLONG;
|
122 | }
|
123 |
|
124 | return;
|
125 | }
|
126 |
|
127 | if ((first & 0xF0) == 0xE0) {
|
128 | // 3-bytes long
|
129 | result->codepoint = first & 0x0F;
|
130 |
|
131 | _cont(input, result);
|
132 | _cont(input, result);
|
133 | if (result->error) return;
|
134 |
|
135 | if (result->codepoint < 0x800) {
|
136 | result->error = UTF8_ERR_OVERLONG;
|
137 | }
|
138 |
|
139 | if (0xD800 <= result->codepoint && result->codepoint <= 0xDFFF) {
|
140 | result->error = UTF8_ERR_SURROGATE;
|
141 | }
|
142 |
|
143 | return;
|
144 | }
|
145 |
|
146 | if ((first & 0xF8) == 0xF0) {
|
147 | // 4-bytes long
|
148 | result->codepoint = first & 0x07;
|
149 |
|
150 | _cont(input, result);
|
151 | _cont(input, result);
|
152 | _cont(input, result);
|
153 | if (result->error) return;
|
154 |
|
155 | if (result->codepoint < 0x10000) {
|
156 | result->error = UTF8_ERR_OVERLONG;
|
157 | }
|
158 |
|
159 | if (result->codepoint > 0x10FFFF) {
|
160 | result->error = UTF8_ERR_TOO_LARGE;
|
161 | }
|
162 |
|
163 | return;
|
164 | }
|
165 |
|
166 | result->error = UTF8_ERR_BAD_ENCODING;
|
167 | return;
|
168 | }
|
169 |
|
170 | #endif // DATA_LANG_UTF8_H
|