summaryrefslogtreecommitdiffhomepage
path: root/include/unicode.h
blob: eb872ecf9e853f5fb6f772a77afb6bb547564f2d (plain)
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
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
// libunicode
//
// Author: Roland Reichwein <mail@reichwein.it>
//
// Available under the conditions of CC0 1.0 Universal
// https://creativecommons.org/publicdomain/zero/1.0/

#pragma once

#include "unicode/endian.h"
#include "unicode/iso.h"
#include "unicode/predicate.h"
#include "unicode/types.h"
#include "unicode/type_traits.h"
#include "unicode/utf.h"

#include <algorithm>
#include <array>
#include <cstdint>
#include <iterator>
#include <memory>
#include <stdexcept>
#include <string>
#include <type_traits>
#include <utility>

namespace unicode {

 // std::distance doesn't work here: it is based on "output" distance of iterators
 template<class Iterator>
 inline size_t input_distance(const Iterator& it1, const Iterator& it2)
 {
  return it2 - it1;
 }
 
 template<class Iterator>
 inline size_t input_distance_bytes(const Iterator& it1, const Iterator& it2)
 {
  return input_distance(it1, it2) * sizeof(typename Iterator::value_type);
 }

 // Optimizations following:
 static const size_t accu_size {sizeof(size_t)};

 template<int value_size>
 struct ConvertInputOptimizer {};

 template<> struct ConvertInputOptimizer<1>
 {
  static const uint32_t ascii_mask { 0x80808080 };
 };
 
 template<> struct ConvertInputOptimizer<2>
 {
  static const uint32_t ascii_mask { 0xFF80FF80 };
 };
 
 template<> struct ConvertInputOptimizer<4>
 {
  static const uint32_t ascii_mask { 0xFFFFFF80 };
 };

 template<int AccuSize, class ConvertInputOptimizer>
 struct ArchitectureOptimizer {};

 template<class ConvertInputOptimizer>
 struct ArchitectureOptimizer<4, ConvertInputOptimizer>
 {
  typedef ConvertInputOptimizer input_optimizer;
  typedef uint32_t accu_type;
  static const accu_type addr_mask {accu_size - 1};
  static const accu_type ascii_mask { (accu_type)input_optimizer::ascii_mask };
  static const accu_type ascii_value { 0ULL };
  
  template<typename input_value_type, class output_string_type>
  inline static void append(const input_value_type* addr, output_string_type& s)
  {
   if constexpr(sizeof(input_value_type) == sizeof(typename output_string_type::value_type)) {
    s.append(reinterpret_cast<const typename output_string_type::value_type*>(addr), accu_size / sizeof(input_value_type));
   } else if constexpr(is_utf_8_v<input_value_type>) {
    s.append({static_cast<typename output_string_type::value_type>(addr[0]),
              static_cast<typename output_string_type::value_type>(addr[1]),
              static_cast<typename output_string_type::value_type>(addr[2]),
              static_cast<typename output_string_type::value_type>(addr[3])});
   } else if constexpr(is_utf_16_v<input_value_type>) {
    s.append({static_cast<typename output_string_type::value_type>(addr[0]),
              static_cast<typename output_string_type::value_type>(addr[1])});
   } else if constexpr(is_utf_32_v<input_value_type>) {
    s.append({static_cast<typename output_string_type::value_type>(addr[0])});
   }
  }
 };

 template<class ConvertInputOptimizer>
 struct ArchitectureOptimizer<8, ConvertInputOptimizer>
 {
  typedef ConvertInputOptimizer input_optimizer;
  typedef uint64_t accu_type;
  static const accu_type addr_mask {accu_size - 1};
  static const accu_type ascii_mask { ((accu_type)input_optimizer::ascii_mask) << 32 | (accu_type)input_optimizer::ascii_mask };
  static const accu_type ascii_value { 0ULL };
  
  template<typename input_value_type, class output_string_type>
  inline static void append(const input_value_type* addr, output_string_type& s)
  {
   if constexpr(sizeof(input_value_type) == sizeof(typename output_string_type::value_type)) {
    s.append(reinterpret_cast<const typename output_string_type::value_type*>(addr), accu_size / sizeof(input_value_type));
   } else if constexpr(is_utf_8_v<input_value_type>) {
    s.append({static_cast<typename output_string_type::value_type>(addr[0]),
              static_cast<typename output_string_type::value_type>(addr[1]),
              static_cast<typename output_string_type::value_type>(addr[2]),
              static_cast<typename output_string_type::value_type>(addr[3]),
              static_cast<typename output_string_type::value_type>(addr[4]),
              static_cast<typename output_string_type::value_type>(addr[5]),
              static_cast<typename output_string_type::value_type>(addr[6]),
              static_cast<typename output_string_type::value_type>(addr[7])});
   } else if constexpr(is_utf_16_v<input_value_type>) {
    s.append({static_cast<typename output_string_type::value_type>(addr[0]),
              static_cast<typename output_string_type::value_type>(addr[1]),
              static_cast<typename output_string_type::value_type>(addr[2]),
              static_cast<typename output_string_type::value_type>(addr[3])});
   } else if constexpr(is_utf_32_v<input_value_type>) {
    s.append({static_cast<typename output_string_type::value_type>(addr[0]),
              static_cast<typename output_string_type::value_type>(addr[1])});
   }
  }

 }; // class ArchitectureOptimizer

 // Optimize for the case of all ASCII (7-bit) data in a accu size row
 // From and To are Encodings
 template<typename From, typename To, std::enable_if_t<is_encoding_v<From> && is_encoding_v<To>, bool> = true>
 typename To::string_type convert_optimized(const typename From::string_type& s)
 {
  typename To::string_type result;
  typedef ConvertInputOptimizer<sizeof(typename From::value_type)> input_optimizer;
  typedef ArchitectureOptimizer<accu_size, input_optimizer> arch_optimizer;

  auto begin{From::begin(s)};
  auto end{From::end(s)};
  auto back_inserter{To::back_inserter(result)};
  auto addr{reinterpret_cast<const typename arch_optimizer::accu_type*>(&s.data()[s.size() - input_distance(begin, end)])};
  while (input_distance_bytes(begin, end) >= accu_size) {
   if (((uintptr_t)(void*)addr & arch_optimizer::addr_mask) == 0) {
    while (input_distance_bytes(begin, end) >= accu_size) {
     typename arch_optimizer::accu_type data{*addr};
     if ((data & arch_optimizer::ascii_mask) == arch_optimizer::ascii_value) {
      arch_optimizer::template append(reinterpret_cast<const typename From::value_type*>(addr), result);
      begin += accu_size / sizeof(typename From::value_type);
      ++addr;
     } else {
      // just advance one code unit for now and break to trigger unoptimized
      // version until next accu boundary
      back_inserter = *begin;
      ++begin;
      break;
     }
    }
   }

   // keep up after unaligned Non-ASCII code points
   while (begin != end && (uintptr_t)(void*)(addr = reinterpret_cast<const typename arch_optimizer::accu_type*>(&s.data()[s.size() - input_distance(begin, end)])) & arch_optimizer::addr_mask) {
    back_inserter = *begin;
    ++begin;
   }
  }

  // remainder < 8 bytes   
  while (begin != end) {
   back_inserter = *begin;
   ++begin;
  }

  return result;
 }

 template<size_t bits_to_compare = 32, typename To, typename std::enable_if_t<is_utf_8_v<To>, bool> = true>
 inline void append_utf(std::basic_string<To>& result, const char32_t& value)
 {
  using From = char32_t;
  if (bits_to_compare <= 7 || value < 0x80) { // 1 byte
   result.push_back(static_cast<To>(value));
  } else if (bits_to_compare <= 11 || value < 0x800) { // 2 bytes
   result.append({utf8_byte_n_of_m<0,2,From,To>(value), utf8_byte_n_of_m<1,2,From,To>(value)});
  } else if (bits_to_compare <= 16 || value < 0x10000) { // 3 bytes
   result.append({utf8_byte_n_of_m<0,3,From,To>(value), utf8_byte_n_of_m<1,3,From,To>(value), utf8_byte_n_of_m<2,3,From,To>(value)});
  } else { // 4 bytes
   // expect value to be already valid Unicode values
   result.append({utf8_byte_n_of_m<0,4,From,To>(value), utf8_byte_n_of_m<1,4,From,To>(value), utf8_byte_n_of_m<2,4,From,To>(value), utf8_byte_n_of_m<3,4,From,To>(value)});
  }
 }

 template<size_t bits_to_compare = 32, typename To, typename std::enable_if_t<is_utf_16_v<To>, bool> = true>
 inline void append_utf(std::basic_string<To>& result, const char32_t& value)
 {
  if (bits_to_compare <= 16 || value <= 0xFFFF) { // expect value to be already valid Unicode values
   result.push_back(static_cast<To>(value));
  } else {
   char32_t value_reduced{value - 0x10000};
   result.append({static_cast<To>((value_reduced >> 10) + 0xD800), static_cast<To>((value_reduced & 0x3FF) + 0xDC00)});
  }
 }

 template<size_t bits_to_compare = 32, typename To, typename std::enable_if_t<is_utf_32_v<To>, bool> = true>
 inline void append_utf(std::basic_string<To>& result, const char32_t& value)
 {
  // expect value to be already valid Unicode values (checked in input iterator)
  result.push_back(static_cast<To>(value));
 }

 // Little Endian optimized version for UTF-8
 // In block_mode, at least 4 bytes are in accu. On first call, even 8.
 // otherwise, at least one code unit is in accu
 template<typename From, typename To, bool block_mode = true, typename std::enable_if_t<is_utf_8_v<From>, bool> = true>
 inline static void append_accu(std::basic_string<To>& result, uint64_t& accu, int& bytes_in_accu)
 {
#if 1
  if (block_mode && bytes_in_accu == 8 && (accu & 0x8080808080808080) == 0) {
   result.append({
                 static_cast<To>(accu & 0x7F),
                 static_cast<To>((accu >> 8) & 0x7F),
                 static_cast<To>((accu >> 16) & 0x7F),
                 static_cast<To>((accu >> 24) & 0x7F),
                 static_cast<To>((accu >> 32) & 0x7F),
                 static_cast<To>((accu >> 40) & 0x7F),
                 static_cast<To>((accu >> 48) & 0x7F),
                 static_cast<To>((accu >> 56) & 0x7F),
                 });
   accu = 0;
   bytes_in_accu = 0;
  } else
#endif
  if ((accu & 0x80) == 0) { // 1 byte sequence
   append_utf<7>(result, static_cast<char32_t>(accu & 0x7F));
   accu >>= 8;
   bytes_in_accu -= 1;
  } else if ((block_mode || bytes_in_accu >= 2) && (accu & 0xC0E0) == 0x80C0) { // 2 byte sequence
   char32_t value {static_cast<char32_t>(((accu & 0x1F) << 6) | ((accu >> 8) & 0x3f))};
   accu >>= 16;
   bytes_in_accu -= 2;
   if (is_valid_unicode<11>(value))
    append_utf<11>(result, value);
   else
    throw std::invalid_argument("Invalid Unicode character in 2 byte UTF-8 sequence");
  } else if ((block_mode || bytes_in_accu >= 3) && (accu & 0xC0C0F0) == 0x8080E0) { // 3 byte sequence
   char32_t value {static_cast<char32_t>(((accu & 0x0F) << 12) | ((accu >> 2) & 0x0FC0) | ((accu >> 16) & 0x3f))};
   accu >>= 24;
   bytes_in_accu -= 3;
   if (is_valid_unicode<16>(value))
    append_utf<16>(result, value);
   else
    throw std::invalid_argument("Invalid Unicode character in 3 byte UTF-8 sequence");
  } else if ((block_mode || bytes_in_accu >= 4) && (accu & 0xC0C0C0F8) == 0x808080F0) { // 4 byte sequence
   char32_t value {static_cast<char32_t>(((accu & 0x07) << 18) | ((accu << 4) & 0x3f000) | ((accu >> 10) & 0xFC0) | ((accu >> 24) & 0x3f))};
   accu >>= 32;
   bytes_in_accu -= 4;
   if (is_valid_unicode<21>(value))
    append_utf(result, value);
   else
    throw std::invalid_argument("Invalid Unicode character in 4 byte UTF-8 sequence");
  } else
   throw std::invalid_argument("Invalid UTF-8 byte sequence");
 }

 // Little Endian optimized version for UTF-16
 // In block_mode, at least 4 bytes are in accu. On first call, even 8.
 // otherwise, at least one code unit is in accu
 template<typename From, typename To, bool block_mode = true, typename std::enable_if_t<is_utf_16_v<From>, bool> = true>
 inline static void append_accu(std::basic_string<To>& result, uint64_t& accu, int& bytes_in_accu)
 {
#if 1
  if ((accu & 0xFF80FF80FF80FF80) == 0) {
   auto number_of_values{bytes_in_accu / sizeof(From)};
   result.resize(result.size() + number_of_values);
   for (int i = 0; i < number_of_values; i++) {
    result[result.size() - number_of_values + i] = static_cast<To>(accu & 0x7F);
    accu >>= 16;
   }
   bytes_in_accu = 0;
  } else
#endif
  if ((accu & 0xFC00FC00FC00FC00) == 0xDC00D800DC00D800) {
   // found 4 code units forming 3 code points in UTF-16;
   // by definition of UTF-16, we have valid unicode values at this point
   if constexpr(is_utf_32_v<To>) {
    //result.resize(result.size() + 2);
    //*reinterpret_cast<uint64_t*>(&result[result.size() - 2]) = (((accu & 0x03FF000003FF) << 10) | ((accu >> 16) & 0x03FF000003FF)) + 0x0001000000010000;
    result.append({
                  static_cast<To>(((accu & 0x03FF) << 10 | ((accu >> 16) & 0x03FF)) + 0x10000),
                  static_cast<To>(((accu & 0x03FF00000000) >> 22 | ((accu >> 48) & 0x03FF)) + 0x10000)
                  });
   } else {
    append_utf(result, ((accu & 0x03FF) << 10 | ((accu >> 16) & 0x03FF)) + 0x10000);
    append_utf(result, ((accu & 0x03FF00000000) >> 22 | ((accu >> 48) & 0x03FF)) + 0x10000);
   }
   accu = 0;
   bytes_in_accu = 0;
  } else
  if (From unit0 {static_cast<From>(accu & 0xFFFF)}; is_valid_unicode<16>(unit0)) {
   append_utf<16>(result, unit0);
   accu >>= 16;
   bytes_in_accu -= 2;
  } else
  if ((accu & 0xFC00FC00) == 0xDC00D800) {
   // found 2 code units forming 1 code point in UTF-16;
   // by definition of UTF-16, we have a valid unicode value at this point
   append_utf(result, ((accu & 0x03FF) << 10 | ((accu >> 16) & 0x03FF)) + 0x10000);
   accu >>= 32;
   bytes_in_accu -= 4;
  } else
   throw std::invalid_argument("Bad input: Continuation of first UTF-16 unit missing");
 }

 // Little Endian optimized version
 template<typename From, typename To, std::enable_if_t<is_encoding_v<From> && is_encoding_v<To>, bool> = true>
 typename To::string_type convert_optimized_utf(const typename From::string_type& s)
 {
  typename To::string_type result;
  if constexpr(is_utf_32_v<typename From::value_type>) {
   for (const auto value: s) {
    if (is_valid_unicode(value))
     append_utf(result, value);
    else
     throw std::invalid_argument("Invalid Unicode character in UTF-32");
   }
#if 0
  } else if constexpr(is_utf_16_v<typename From::value_type>) {
   for (int i = 0; i < s.size(); i++) {
    typename From::value_type unit0{s[i]};
    if (is_valid_unicode(unit0)) {
     append_utf(result, unit0);
    } else {
     i++;
     if (i < s.size()) {
      typename From::value_type unit1 {s[i]};
      if ((unit0 & 0xFC00) != 0xD800 || (unit1 & 0xFC00) != 0xDC00)
       throw std::invalid_argument("Bad input: 2 malformed UTF-16 surrogates");

      append_utf(result, (static_cast<char32_t>(unit0 & 0x03FF) << 10 | (unit1 & 0x03FF)) + 0x10000);
     } else
      throw std::invalid_argument("Invalid code unit at end of UTF-16 string");
    }
   }
#endif
  } else {
   uint64_t accu{};
   int bytes_in_accu{};

   size_t s_index{};
   size_t s_size{s.size()};
   while (s_index + 8 / sizeof(typename From::value_type) <= s_size) {
    // read input
    // assume: bytes_in_accu < 8
    accu |= (*reinterpret_cast<const uint64_t*>(&(s.data()[s_index]))) << (bytes_in_accu * 8);
    s_index += (8 - bytes_in_accu) / sizeof(typename From::value_type);
    bytes_in_accu = 8;

    while (bytes_in_accu >= 4) {
     append_accu<typename From::value_type, typename To::value_type, true>(result, accu, bytes_in_accu);
    }
   }

   // 0..3 bytes left in accu
   // 0..7 bytes left in s

   while (s_index < s_size || bytes_in_accu > 0) {
    while (s_index < s_size && bytes_in_accu < 8) {
     accu |= static_cast<uint64_t>(*reinterpret_cast<const typename From::value_type*>(&(s.data()[s_index]))) << (bytes_in_accu * 8);
     ++s_index;
     bytes_in_accu += sizeof(typename From::value_type);
    }

    append_accu<typename From::value_type, typename To::value_type, false>(result, accu, bytes_in_accu);
   }
  }
  return result;
 }

 // From and To are Encodings
 template<typename From, typename To, std::enable_if_t<is_encoding_v<From> && is_encoding_v<To>, bool> = true>
 typename To::string_type convert(const typename From::string_type& s)
 {
  // if input type == output type, only validate and return input, if appropriate
  if constexpr(sizeof(typename From::value_type) == sizeof(typename To::value_type) &&
               is_utf_encoding_v<From> && is_utf_encoding_v<To>) {
   if (validate_utf<typename From::value_type>(s)) {
    return s;
   } else {
    throw std::invalid_argument("Invalid UTF input");
   }
  } else if constexpr(accu_size == 8 && is_little_endian() && is_utf_8_v<typename From::value_type> &&
                      is_utf_encoding_v<From> && is_utf_encoding_v<To>) { // endian specific optimization
   return convert_optimized_utf<From, To>(s);
  } else if constexpr(accu_size == 4 || accu_size == 8) { // accu size specific optimization with speedup for 7bit input
   return convert_optimized<From, To>(s);
  } else {
   typename To::string_type result;
   std::copy(From::begin(s), From::end(s), To::back_inserter(result));
   return result;
  }
 }

 // From and To are from: utf8_t (i.e. char or char8_t (C++20)), char16_t and char32_t, char, wchar_t, uint8_t, uint16_t, uint32_t
 template<typename From, typename To,
  typename FromContainer=std::basic_string<From>,
  typename ToContainer=std::basic_string<To>,
  std::enable_if_t<is_char_v<From> && is_char_v<To>, bool> = true>
 ToContainer convert(const FromContainer& s)
 {
  typedef UTF<utf_iterator<From>, utf_back_insert_iterator<To>> UTF_Trait;
  
  ToContainer result;

  std::copy(UTF_Trait::begin(s), UTF_Trait::end(s), UTF_Trait::back_inserter(result));

  return result;
 }

 // From and To are containers
 template<typename FromContainer, typename ToContainer,
  std::enable_if_t<is_container_v<FromContainer> && is_container_v<ToContainer>, bool> = true
 >
 ToContainer convert(const FromContainer& s)
 {
  typedef UTF<utf_iterator<typename FromContainer::value_type, FromContainer>, utf_back_insert_iterator<typename ToContainer::value_type, ToContainer>> UTF_Trait;
  
  ToContainer result;

  std::copy(UTF_Trait::begin(s), UTF_Trait::end(s), UTF_Trait::back_inserter(result));

  return result;
 }

 // Container version
 template<typename Container, std::enable_if_t<is_container_v<Container>, bool> = true>
 bool is_valid_utf(const Container& s)
 {
  typedef UTF<utf_iterator<typename Container::value_type, Container>, utf_back_insert_iterator<typename Container::value_type, Container>> UTF_Trait;
  
  try {
   std::for_each(UTF_Trait::begin(s), UTF_Trait::end(s), [](const char32_t& c){});
  } catch (const std::invalid_argument&) {
   return false;
  }
  return true;
 }

 // basic type version
 template<typename T,
  typename Container=std::basic_string<T>,
  std::enable_if_t<is_char_v<T>, bool> = true>
 bool is_valid_utf(const Container& s)
 {
  typedef UTF<utf_iterator<T>, utf_back_insert_iterator<T>> UTF_Trait;
  
  try {
   std::for_each(UTF_Trait::begin(s), UTF_Trait::end(s), [](const char32_t& c){});
  } catch (const std::invalid_argument&) {
   return false;
  }
  return true;
 }

 // Encoding version
 template<typename Encoding, std::enable_if_t<is_encoding_v<Encoding>, bool> = true>
 bool is_valid_utf(const typename Encoding::string_type& s)
 {
  return validate_utf<typename Encoding::value_type>(s);
 }

} // namespace unicode