1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
|
#ifndef SIMDUTF_UTF8_TO_UTF16_H
#define SIMDUTF_UTF8_TO_UTF16_H
namespace simdutf {
namespace scalar {
namespace {
namespace utf8_to_utf16 {
template <endianness big_endian>
inline size_t convert(const char *buf, size_t len, char16_t *utf16_output) {
const uint8_t *data = reinterpret_cast<const uint8_t *>(buf);
size_t pos = 0;
char16_t *start{utf16_output};
while (pos < len) {
// try to convert the next block of 16 ASCII bytes
if (pos + 16 <=
len) { // if it is safe to read 16 more bytes, check that they are ascii
uint64_t v1;
::memcpy(&v1, data + pos, sizeof(uint64_t));
uint64_t v2;
::memcpy(&v2, data + pos + sizeof(uint64_t), sizeof(uint64_t));
uint64_t v{v1 | v2};
if ((v & 0x8080808080808080) == 0) {
size_t final_pos = pos + 16;
while (pos < final_pos) {
*utf16_output++ = !match_system(big_endian)
? char16_t(utf16::swap_bytes(buf[pos]))
: char16_t(buf[pos]);
pos++;
}
continue;
}
}
uint8_t leading_byte = data[pos]; // leading byte
if (leading_byte < 0b10000000) {
// converting one ASCII byte !!!
*utf16_output++ = !match_system(big_endian)
? char16_t(utf16::swap_bytes(leading_byte))
: char16_t(leading_byte);
pos++;
} else if ((leading_byte & 0b11100000) == 0b11000000) {
// We have a two-byte UTF-8, it should become
// a single UTF-16 word.
if (pos + 1 >= len) {
return 0;
} // minimal bound checking
if ((data[pos + 1] & 0b11000000) != 0b10000000) {
return 0;
}
// range check
uint32_t code_point =
(leading_byte & 0b00011111) << 6 | (data[pos + 1] & 0b00111111);
if (code_point < 0x80 || 0x7ff < code_point) {
return 0;
}
if (!match_system(big_endian)) {
code_point = uint32_t(utf16::swap_bytes(uint16_t(code_point)));
}
*utf16_output++ = char16_t(code_point);
pos += 2;
} else if ((leading_byte & 0b11110000) == 0b11100000) {
// We have a three-byte UTF-8, it should become
// a single UTF-16 word.
if (pos + 2 >= len) {
return 0;
} // minimal bound checking
if ((data[pos + 1] & 0b11000000) != 0b10000000) {
return 0;
}
if ((data[pos + 2] & 0b11000000) != 0b10000000) {
return 0;
}
// range check
uint32_t code_point = (leading_byte & 0b00001111) << 12 |
(data[pos + 1] & 0b00111111) << 6 |
(data[pos + 2] & 0b00111111);
if (code_point < 0x800 || 0xffff < code_point ||
(0xd7ff < code_point && code_point < 0xe000)) {
return 0;
}
if (!match_system(big_endian)) {
code_point = uint32_t(utf16::swap_bytes(uint16_t(code_point)));
}
*utf16_output++ = char16_t(code_point);
pos += 3;
} else if ((leading_byte & 0b11111000) == 0b11110000) { // 0b11110000
// we have a 4-byte UTF-8 word.
if (pos + 3 >= len) {
return 0;
} // minimal bound checking
if ((data[pos + 1] & 0b11000000) != 0b10000000) {
return 0;
}
if ((data[pos + 2] & 0b11000000) != 0b10000000) {
return 0;
}
if ((data[pos + 3] & 0b11000000) != 0b10000000) {
return 0;
}
// range check
uint32_t code_point = (leading_byte & 0b00000111) << 18 |
(data[pos + 1] & 0b00111111) << 12 |
(data[pos + 2] & 0b00111111) << 6 |
(data[pos + 3] & 0b00111111);
if (code_point <= 0xffff || 0x10ffff < code_point) {
return 0;
}
code_point -= 0x10000;
uint16_t high_surrogate = uint16_t(0xD800 + (code_point >> 10));
uint16_t low_surrogate = uint16_t(0xDC00 + (code_point & 0x3FF));
if (!match_system(big_endian)) {
high_surrogate = utf16::swap_bytes(high_surrogate);
low_surrogate = utf16::swap_bytes(low_surrogate);
}
*utf16_output++ = char16_t(high_surrogate);
*utf16_output++ = char16_t(low_surrogate);
pos += 4;
} else {
return 0;
}
}
return utf16_output - start;
}
template <endianness big_endian>
inline result convert_with_errors(const char *buf, size_t len,
char16_t *utf16_output) {
const uint8_t *data = reinterpret_cast<const uint8_t *>(buf);
size_t pos = 0;
char16_t *start{utf16_output};
while (pos < len) {
// try to convert the next block of 16 ASCII bytes
if (pos + 16 <=
len) { // if it is safe to read 16 more bytes, check that they are ascii
uint64_t v1;
::memcpy(&v1, data + pos, sizeof(uint64_t));
uint64_t v2;
::memcpy(&v2, data + pos + sizeof(uint64_t), sizeof(uint64_t));
uint64_t v{v1 | v2};
if ((v & 0x8080808080808080) == 0) {
size_t final_pos = pos + 16;
while (pos < final_pos) {
*utf16_output++ = !match_system(big_endian)
? char16_t(utf16::swap_bytes(buf[pos]))
: char16_t(buf[pos]);
pos++;
}
continue;
}
}
uint8_t leading_byte = data[pos]; // leading byte
if (leading_byte < 0b10000000) {
// converting one ASCII byte !!!
*utf16_output++ = !match_system(big_endian)
? char16_t(utf16::swap_bytes(leading_byte))
: char16_t(leading_byte);
pos++;
} else if ((leading_byte & 0b11100000) == 0b11000000) {
// We have a two-byte UTF-8, it should become
// a single UTF-16 word.
if (pos + 1 >= len) {
return result(error_code::TOO_SHORT, pos);
} // minimal bound checking
if ((data[pos + 1] & 0b11000000) != 0b10000000) {
return result(error_code::TOO_SHORT, pos);
}
// range check
uint32_t code_point =
(leading_byte & 0b00011111) << 6 | (data[pos + 1] & 0b00111111);
if (code_point < 0x80 || 0x7ff < code_point) {
return result(error_code::OVERLONG, pos);
}
if (!match_system(big_endian)) {
code_point = uint32_t(utf16::swap_bytes(uint16_t(code_point)));
}
*utf16_output++ = char16_t(code_point);
pos += 2;
} else if ((leading_byte & 0b11110000) == 0b11100000) {
// We have a three-byte UTF-8, it should become
// a single UTF-16 word.
if (pos + 2 >= len) {
return result(error_code::TOO_SHORT, pos);
} // minimal bound checking
if ((data[pos + 1] & 0b11000000) != 0b10000000) {
return result(error_code::TOO_SHORT, pos);
}
if ((data[pos + 2] & 0b11000000) != 0b10000000) {
return result(error_code::TOO_SHORT, pos);
}
// range check
uint32_t code_point = (leading_byte & 0b00001111) << 12 |
(data[pos + 1] & 0b00111111) << 6 |
(data[pos + 2] & 0b00111111);
if ((code_point < 0x800) || (0xffff < code_point)) {
return result(error_code::OVERLONG, pos);
}
if (0xd7ff < code_point && code_point < 0xe000) {
return result(error_code::SURROGATE, pos);
}
if (!match_system(big_endian)) {
code_point = uint32_t(utf16::swap_bytes(uint16_t(code_point)));
}
*utf16_output++ = char16_t(code_point);
pos += 3;
} else if ((leading_byte & 0b11111000) == 0b11110000) { // 0b11110000
// we have a 4-byte UTF-8 word.
if (pos + 3 >= len) {
return result(error_code::TOO_SHORT, pos);
} // minimal bound checking
if ((data[pos + 1] & 0b11000000) != 0b10000000) {
return result(error_code::TOO_SHORT, pos);
}
if ((data[pos + 2] & 0b11000000) != 0b10000000) {
return result(error_code::TOO_SHORT, pos);
}
if ((data[pos + 3] & 0b11000000) != 0b10000000) {
return result(error_code::TOO_SHORT, pos);
}
// range check
uint32_t code_point = (leading_byte & 0b00000111) << 18 |
(data[pos + 1] & 0b00111111) << 12 |
(data[pos + 2] & 0b00111111) << 6 |
(data[pos + 3] & 0b00111111);
if (code_point <= 0xffff) {
return result(error_code::OVERLONG, pos);
}
if (0x10ffff < code_point) {
return result(error_code::TOO_LARGE, pos);
}
code_point -= 0x10000;
uint16_t high_surrogate = uint16_t(0xD800 + (code_point >> 10));
uint16_t low_surrogate = uint16_t(0xDC00 + (code_point & 0x3FF));
if (!match_system(big_endian)) {
high_surrogate = utf16::swap_bytes(high_surrogate);
low_surrogate = utf16::swap_bytes(low_surrogate);
}
*utf16_output++ = char16_t(high_surrogate);
*utf16_output++ = char16_t(low_surrogate);
pos += 4;
} else {
// we either have too many continuation bytes or an invalid leading byte
if ((leading_byte & 0b11000000) == 0b10000000) {
return result(error_code::TOO_LONG, pos);
} else {
return result(error_code::HEADER_BITS, pos);
}
}
}
return result(error_code::SUCCESS, utf16_output - start);
}
/**
* When rewind_and_convert_with_errors is called, we are pointing at 'buf' and
* we have up to len input bytes left, and we encountered some error. It is
* possible that the error is at 'buf' exactly, but it could also be in the
* previous bytes (up to 3 bytes back).
*
* prior_bytes indicates how many bytes, prior to 'buf' may belong to the
* current memory section and can be safely accessed. We prior_bytes to access
* safely up to three bytes before 'buf'.
*
* The caller is responsible to ensure that len > 0.
*
* If the error is believed to have occurred prior to 'buf', the count value
* contain in the result will be SIZE_T - 1, SIZE_T - 2, or SIZE_T - 3.
*/
template <endianness endian>
inline result rewind_and_convert_with_errors(size_t prior_bytes,
const char *buf, size_t len,
char16_t *utf16_output) {
size_t extra_len{0};
// We potentially need to go back in time and find a leading byte.
// In theory '3' would be sufficient, but sometimes the error can go back
// quite far.
size_t how_far_back = prior_bytes;
// size_t how_far_back = 3; // 3 bytes in the past + current position
// if(how_far_back >= prior_bytes) { how_far_back = prior_bytes; }
bool found_leading_bytes{false};
// important: it is i <= how_far_back and not 'i < how_far_back'.
for (size_t i = 0; i <= how_far_back; i++) {
unsigned char byte = buf[-static_cast<std::ptrdiff_t>(i)];
found_leading_bytes = ((byte & 0b11000000) != 0b10000000);
if (found_leading_bytes) {
if (i > 0 && byte < 128) {
// If we had to go back and the leading byte is ascii
// then we can stop right away.
return result(error_code::TOO_LONG, 0 - i + 1);
}
buf -= i;
extra_len = i;
break;
}
}
//
// It is possible for this function to return a negative count in its result.
// C++ Standard Section 18.1 defines size_t is in <cstddef> which is described
// in C Standard as <stddef.h>. C Standard Section 4.1.5 defines size_t as an
// unsigned integral type of the result of the sizeof operator
//
// An unsigned type will simply wrap round arithmetically (well defined).
//
if (!found_leading_bytes) {
// If how_far_back == 3, we may have four consecutive continuation bytes!!!
// [....] [continuation] [continuation] [continuation] | [buf is
// continuation] Or we possibly have a stream that does not start with a
// leading byte.
return result(error_code::TOO_LONG, 0 - how_far_back);
}
result res = convert_with_errors<endian>(buf, len + extra_len, utf16_output);
if (res.error) {
res.count -= extra_len;
}
return res;
}
} // namespace utf8_to_utf16
} // unnamed namespace
} // namespace scalar
} // namespace simdutf
#endif
|
