xref: /freebsd/contrib/llvm-project/libcxx/include/__format/unicode.h (revision 1db9f3b21e39176dd5b67cf8ac378633b172463e)
1 // -*- C++ -*-
2 //===----------------------------------------------------------------------===//
3 //
4 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
5 // See https://llvm.org/LICENSE.txt for license information.
6 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
7 //
8 //===----------------------------------------------------------------------===//
9 
10 #ifndef _LIBCPP___FORMAT_UNICODE_H
11 #define _LIBCPP___FORMAT_UNICODE_H
12 
13 #include <__assert>
14 #include <__bit/countl.h>
15 #include <__concepts/same_as.h>
16 #include <__config>
17 #include <__format/extended_grapheme_cluster_table.h>
18 #include <__iterator/concepts.h>
19 #include <__iterator/readable_traits.h> // iter_value_t
20 #include <string_view>
21 
22 #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
23 #  pragma GCC system_header
24 #endif
25 
26 _LIBCPP_BEGIN_NAMESPACE_STD
27 
28 #if _LIBCPP_STD_VER >= 20
29 
30 namespace __unicode {
31 
32 // Helper struct for the result of a consume operation.
33 //
34 // The status value for a correct code point is 0. This allows a valid value to
35 // be used without masking.
36 // When the decoding fails it know the number of code units affected. For the
37 // current use-cases that value is not needed, therefore it is not stored.
38 // The escape routine needs the number of code units for both a valid and
39 // invalid character and keeps track of it itself. Doing it in this result
40 // unconditionally would give some overhead when the value is unneeded.
41 struct __consume_result {
42   // When __status == __ok it contains the decoded code point.
43   // Else it contains the replacement character U+FFFD
44   char32_t __code_point : 31;
45 
46   enum : char32_t {
47     // Consumed a well-formed code point.
48     __ok = 0,
49     // Encountered invalid UTF-8
50     __error = 1
51   } __status : 1 {__ok};
52 };
53 static_assert(sizeof(__consume_result) == sizeof(char32_t));
54 
55 #  ifndef _LIBCPP_HAS_NO_UNICODE
56 
57 /// Implements the grapheme cluster boundary rules
58 ///
59 /// These rules are used to implement format's width estimation as stated in
60 /// [format.string.std]/11
61 ///
62 /// The Standard refers to UAX \#29 for Unicode 12.0.0
63 /// https://www.unicode.org/reports/tr29/#Grapheme_Cluster_Boundary_Rules
64 ///
65 /// The data tables used are
66 /// https://www.unicode.org/Public/UCD/latest/ucd/auxiliary/GraphemeBreakProperty.txt
67 /// https://www.unicode.org/Public/UCD/latest/ucd/emoji/emoji-data.txt
68 /// https://www.unicode.org/Public/UCD/latest/ucd/auxiliary/GraphemeBreakTest.txt (for testing only)
69 
70 inline constexpr char32_t __replacement_character = U'\ufffd';
71 
72 // The error of a consume operation.
73 //
74 // This sets the code point to the replacement character. This code point does
75 // not participate in the grapheme clustering, so grapheme clustering code can
76 // ignore the error status and always use the code point.
77 inline constexpr __consume_result __consume_result_error{__replacement_character, __consume_result::__error};
78 
79 [[nodiscard]] _LIBCPP_HIDE_FROM_ABI constexpr bool __is_high_surrogate(char32_t __value) {
80   return __value >= 0xd800 && __value <= 0xdbff;
81 }
82 
83 [[nodiscard]] _LIBCPP_HIDE_FROM_ABI constexpr bool __is_low_surrogate(char32_t __value) {
84   return __value >= 0xdc00 && __value <= 0xdfff;
85 }
86 
87 // https://www.unicode.org/glossary/#surrogate_code_point
88 [[nodiscard]] _LIBCPP_HIDE_FROM_ABI inline constexpr bool __is_surrogate(char32_t __value) {
89   return __value >= 0xd800 && __value <= 0xdfff;
90 }
91 
92 // https://www.unicode.org/glossary/#code_point
93 [[nodiscard]] _LIBCPP_HIDE_FROM_ABI inline constexpr bool __is_code_point(char32_t __value) {
94   return __value <= 0x10ffff;
95 }
96 
97 // https://www.unicode.org/glossary/#unicode_scalar_value
98 [[nodiscard]] _LIBCPP_HIDE_FROM_ABI inline constexpr bool __is_scalar_value(char32_t __value) {
99   return __unicode::__is_code_point(__value) && !__unicode::__is_surrogate(__value);
100 }
101 
102 template <contiguous_iterator _Iterator>
103   requires same_as<iter_value_t<_Iterator>, char>
104 _LIBCPP_HIDE_FROM_ABI constexpr bool __is_continuation(_Iterator __char, int __count) {
105   do {
106     if ((*__char & 0b1100'0000) != 0b1000'0000)
107       return false;
108     --__count;
109     ++__char;
110   } while (__count);
111   return true;
112 }
113 
114 /// Helper class to extract a code unit from a Unicode character range.
115 ///
116 /// The stored range is a view. There are multiple specialization for different
117 /// character types.
118 template <class _CharT>
119 class __code_point_view;
120 
121 /// UTF-8 specialization.
122 template <>
123 class __code_point_view<char> {
124   using _Iterator = basic_string_view<char>::const_iterator;
125 
126 public:
127   _LIBCPP_HIDE_FROM_ABI constexpr explicit __code_point_view(_Iterator __first, _Iterator __last)
128       : __first_(__first), __last_(__last) {}
129 
130   _LIBCPP_HIDE_FROM_ABI constexpr bool __at_end() const noexcept { return __first_ == __last_; }
131   _LIBCPP_HIDE_FROM_ABI constexpr _Iterator __position() const noexcept { return __first_; }
132 
133   // https://www.unicode.org/versions/latest/ch03.pdf#G7404
134   // Based on Table 3-7, Well-Formed UTF-8 Byte Sequences
135   //
136   // Code Points        First Byte Second Byte Third Byte Fourth Byte  Remarks
137   // U+0000..U+007F     00..7F                                         U+0000..U+007F 1 code unit range
138   //                    C0..C1     80..BF                              invalid overlong encoding
139   // U+0080..U+07FF     C2..DF     80..BF                              U+0080..U+07FF 2 code unit range
140   //                    E0         80..9F      80..BF                  invalid overlong encoding
141   // U+0800..U+0FFF     E0         A0..BF      80..BF                  U+0800..U+FFFF 3 code unit range
142   // U+1000..U+CFFF     E1..EC     80..BF      80..BF
143   // U+D000..U+D7FF     ED         80..9F      80..BF
144   // U+D800..U+DFFF     ED         A0..BF      80..BF                  invalid encoding of surrogate code point
145   // U+E000..U+FFFF     EE..EF     80..BF      80..BF
146   //                    F0         80..8F      80..BF     80..BF       invalid overlong encoding
147   // U+10000..U+3FFFF   F0         90..BF      80..BF     80..BF       U+10000..U+10FFFF 4 code unit range
148   // U+40000..U+FFFFF   F1..F3     80..BF      80..BF     80..BF
149   // U+100000..U+10FFFF F4         80..8F      80..BF     80..BF
150   //                    F4         90..BF      80..BF     80..BF       U+110000.. invalid code point range
151   //
152   // Unlike other parsers, these invalid entries are tested after decoding.
153   // - The parser always needs to consume these code units
154   // - The code is optimized for well-formed UTF-8
155   [[nodiscard]] _LIBCPP_HIDE_FROM_ABI constexpr __consume_result __consume() noexcept {
156     _LIBCPP_ASSERT_INTERNAL(__first_ != __last_, "can't move beyond the end of input");
157 
158     // Based on the number of leading 1 bits the number of code units in the
159     // code point can be determined. See
160     // https://en.wikipedia.org/wiki/UTF-8#Encoding
161     switch (std::countl_one(static_cast<unsigned char>(*__first_))) {
162     case 0:
163       return {static_cast<unsigned char>(*__first_++)};
164 
165     case 2: {
166       if (__last_ - __first_ < 2 || !__unicode::__is_continuation(__first_ + 1, 1)) [[unlikely]]
167         break;
168 
169       char32_t __value = static_cast<unsigned char>(*__first_++) & 0x1f;
170       __value <<= 6;
171       __value |= static_cast<unsigned char>(*__first_++) & 0x3f;
172 
173       // These values should be encoded in 1 UTF-8 code unit.
174       if (__value < 0x0080) [[unlikely]]
175         return __consume_result_error;
176 
177       return {__value};
178     }
179 
180     case 3: {
181       if (__last_ - __first_ < 3 || !__unicode::__is_continuation(__first_ + 1, 2)) [[unlikely]]
182         break;
183 
184       char32_t __value = static_cast<unsigned char>(*__first_++) & 0x0f;
185       __value <<= 6;
186       __value |= static_cast<unsigned char>(*__first_++) & 0x3f;
187       __value <<= 6;
188       __value |= static_cast<unsigned char>(*__first_++) & 0x3f;
189 
190       // These values should be encoded in 1 or 2 UTF-8 code units.
191       if (__value < 0x0800) [[unlikely]]
192         return __consume_result_error;
193 
194       // A surrogate value is always encoded in 3 UTF-8 code units.
195       if (__unicode::__is_surrogate(__value)) [[unlikely]]
196         return __consume_result_error;
197 
198       return {__value};
199     }
200 
201     case 4: {
202       if (__last_ - __first_ < 4 || !__unicode::__is_continuation(__first_ + 1, 3)) [[unlikely]]
203         break;
204 
205       char32_t __value = static_cast<unsigned char>(*__first_++) & 0x07;
206       __value <<= 6;
207       __value |= static_cast<unsigned char>(*__first_++) & 0x3f;
208       __value <<= 6;
209       __value |= static_cast<unsigned char>(*__first_++) & 0x3f;
210       __value <<= 6;
211       __value |= static_cast<unsigned char>(*__first_++) & 0x3f;
212 
213       // These values should be encoded in 1, 2, or 3 UTF-8 code units.
214       if (__value < 0x10000) [[unlikely]]
215         return __consume_result_error;
216 
217       // A value too large is always encoded in 4 UTF-8 code units.
218       if (!__unicode::__is_code_point(__value)) [[unlikely]]
219         return __consume_result_error;
220 
221       return {__value};
222     }
223     }
224     // An invalid number of leading ones can be garbage or a code unit in the
225     // middle of a code point. By consuming one code unit the parser may get
226     // "in sync" after a few code units.
227     ++__first_;
228     return __consume_result_error;
229   }
230 
231 private:
232   _Iterator __first_;
233   _Iterator __last_;
234 };
235 
236 #    ifndef _LIBCPP_HAS_NO_WIDE_CHARACTERS
237 _LIBCPP_HIDE_FROM_ABI constexpr bool __is_surrogate_pair_high(wchar_t __value) {
238   return __value >= 0xd800 && __value <= 0xdbff;
239 }
240 
241 _LIBCPP_HIDE_FROM_ABI constexpr bool __is_surrogate_pair_low(wchar_t __value) {
242   return __value >= 0xdc00 && __value <= 0xdfff;
243 }
244 
245 /// This specialization depends on the size of wchar_t
246 /// - 2 UTF-16 (for example Windows and AIX)
247 /// - 4 UTF-32 (for example Linux)
248 template <>
249 class __code_point_view<wchar_t> {
250   using _Iterator = typename basic_string_view<wchar_t>::const_iterator;
251 
252 public:
253   static_assert(sizeof(wchar_t) == 2 || sizeof(wchar_t) == 4, "sizeof(wchar_t) has a not implemented value");
254 
255   _LIBCPP_HIDE_FROM_ABI constexpr explicit __code_point_view(_Iterator __first, _Iterator __last)
256       : __first_(__first), __last_(__last) {}
257 
258   _LIBCPP_HIDE_FROM_ABI constexpr _Iterator __position() const noexcept { return __first_; }
259   _LIBCPP_HIDE_FROM_ABI constexpr bool __at_end() const noexcept { return __first_ == __last_; }
260 
261   [[nodiscard]] _LIBCPP_HIDE_FROM_ABI constexpr __consume_result __consume() noexcept {
262     _LIBCPP_ASSERT_INTERNAL(__first_ != __last_, "can't move beyond the end of input");
263 
264     char32_t __value = static_cast<char32_t>(*__first_++);
265     if constexpr (sizeof(wchar_t) == 2) {
266       if (__unicode::__is_low_surrogate(__value)) [[unlikely]]
267         return __consume_result_error;
268 
269       if (__unicode::__is_high_surrogate(__value)) {
270         if (__first_ == __last_ || !__unicode::__is_low_surrogate(static_cast<char32_t>(*__first_))) [[unlikely]]
271           return __consume_result_error;
272 
273         __value -= 0xd800;
274         __value <<= 10;
275         __value += static_cast<char32_t>(*__first_++) - 0xdc00;
276         __value += 0x10000;
277 
278         if (!__unicode::__is_code_point(__value)) [[unlikely]]
279           return __consume_result_error;
280       }
281     } else {
282       if (!__unicode::__is_scalar_value(__value)) [[unlikely]]
283         return __consume_result_error;
284     }
285 
286     return {__value};
287   }
288 
289 private:
290   _Iterator __first_;
291   _Iterator __last_;
292 };
293 #    endif // _LIBCPP_HAS_NO_WIDE_CHARACTERS
294 
295 _LIBCPP_HIDE_FROM_ABI constexpr bool __at_extended_grapheme_cluster_break(
296     bool& __ri_break_allowed,
297     bool __has_extened_pictographic,
298     __extended_grapheme_custer_property_boundary::__property __prev,
299     __extended_grapheme_custer_property_boundary::__property __next) {
300   using __extended_grapheme_custer_property_boundary::__property;
301 
302   __has_extened_pictographic |= __prev == __property::__Extended_Pictographic;
303 
304   // https://www.unicode.org/reports/tr29/tr29-39.html#Grapheme_Cluster_Boundary_Rules
305 
306   // *** Break at the start and end of text, unless the text is empty. ***
307 
308   _LIBCPP_ASSERT_INTERNAL(__prev != __property::__sot, "should be handled in the constructor"); // GB1
309   _LIBCPP_ASSERT_INTERNAL(__prev != __property::__eot, "should be handled by our caller");      // GB2
310 
311   // *** Do not break between a CR and LF. Otherwise, break before and after controls. ***
312   if (__prev == __property::__CR && __next == __property::__LF) // GB3
313     return false;
314 
315   if (__prev == __property::__Control || __prev == __property::__CR || __prev == __property::__LF) // GB4
316     return true;
317 
318   if (__next == __property::__Control || __next == __property::__CR || __next == __property::__LF) // GB5
319     return true;
320 
321   // *** Do not break Hangul syllable sequences. ***
322   if (__prev == __property::__L && (__next == __property::__L || __next == __property::__V ||
323                                     __next == __property::__LV || __next == __property::__LVT)) // GB6
324     return false;
325 
326   if ((__prev == __property::__LV || __prev == __property::__V) &&
327       (__next == __property::__V || __next == __property::__T)) // GB7
328     return false;
329 
330   if ((__prev == __property::__LVT || __prev == __property::__T) && __next == __property::__T) // GB8
331     return false;
332 
333   // *** Do not break before extending characters or ZWJ. ***
334   if (__next == __property::__Extend || __next == __property::__ZWJ)
335     return false; // GB9
336 
337   // *** Do not break before SpacingMarks, or after Prepend characters. ***
338   if (__next == __property::__SpacingMark) // GB9a
339     return false;
340 
341   if (__prev == __property::__Prepend) // GB9b
342     return false;
343 
344   // *** Do not break within emoji modifier sequences or emoji zwj sequences. ***
345 
346   // GB11 \p{Extended_Pictographic} Extend* ZWJ x \p{Extended_Pictographic}
347   //
348   // Note that several parts of this rule are matched by GB9: Any x (Extend | ZWJ)
349   // - \p{Extended_Pictographic} x Extend
350   // - Extend x Extend
351   // - \p{Extended_Pictographic} x ZWJ
352   // - Extend x ZWJ
353   //
354   // So the only case left to test is
355   // - \p{Extended_Pictographic}' x ZWJ x \p{Extended_Pictographic}
356   //   where  \p{Extended_Pictographic}' is stored in __has_extened_pictographic
357   if (__has_extened_pictographic && __prev == __property::__ZWJ && __next == __property::__Extended_Pictographic)
358     return false;
359 
360   // *** Do not break within emoji flag sequences ***
361 
362   // That is, do not break between regional indicator (RI) symbols if there
363   // is an odd number of RI characters before the break point.
364 
365   if (__prev == __property::__Regional_Indicator && __next == __property::__Regional_Indicator) { // GB12 + GB13
366     __ri_break_allowed = !__ri_break_allowed;
367     return __ri_break_allowed;
368   }
369 
370   // *** Otherwise, break everywhere. ***
371   return true; // GB999
372 }
373 
374 /// Helper class to extract an extended grapheme cluster from a Unicode character range.
375 ///
376 /// This function is used to determine the column width of an extended grapheme
377 /// cluster. In order to do that only the first code point is evaluated.
378 /// Therefore only this code point is extracted.
379 template <class _CharT>
380 class __extended_grapheme_cluster_view {
381   using _Iterator = typename basic_string_view<_CharT>::const_iterator;
382 
383 public:
384   _LIBCPP_HIDE_FROM_ABI constexpr explicit __extended_grapheme_cluster_view(_Iterator __first, _Iterator __last)
385       : __code_point_view_(__first, __last),
386         __next_code_point_(__code_point_view_.__consume().__code_point),
387         __next_prop_(__extended_grapheme_custer_property_boundary::__get_property(__next_code_point_)) {}
388 
389   struct __cluster {
390     /// The first code point of the extended grapheme cluster.
391     ///
392     /// The first code point is used to estimate the width of the extended
393     /// grapheme cluster.
394     char32_t __code_point_;
395 
396     /// Points one beyond the last code unit in the extended grapheme cluster.
397     ///
398     /// It's expected the caller has the start position and thus can determine
399     /// the code unit range of the extended grapheme cluster.
400     _Iterator __last_;
401   };
402 
403   _LIBCPP_HIDE_FROM_ABI constexpr __cluster __consume() {
404     _LIBCPP_ASSERT_INTERNAL(__next_prop_ != __extended_grapheme_custer_property_boundary::__property::__eot,
405                             "can't move beyond the end of input");
406 
407     char32_t __code_point = __next_code_point_;
408     if (!__code_point_view_.__at_end())
409       return {__code_point, __get_break()};
410 
411     __next_prop_ = __extended_grapheme_custer_property_boundary::__property::__eot;
412     return {__code_point, __code_point_view_.__position()};
413   }
414 
415 private:
416   __code_point_view<_CharT> __code_point_view_;
417 
418   char32_t __next_code_point_;
419   __extended_grapheme_custer_property_boundary::__property __next_prop_;
420 
421   _LIBCPP_HIDE_FROM_ABI constexpr _Iterator __get_break() {
422     bool __ri_break_allowed         = true;
423     bool __has_extened_pictographic = false;
424     while (true) {
425       _Iterator __result                                              = __code_point_view_.__position();
426       __extended_grapheme_custer_property_boundary::__property __prev = __next_prop_;
427       if (__code_point_view_.__at_end()) {
428         __next_prop_ = __extended_grapheme_custer_property_boundary::__property::__eot;
429         return __result;
430       }
431       __next_code_point_ = __code_point_view_.__consume().__code_point;
432       __next_prop_       = __extended_grapheme_custer_property_boundary::__get_property(__next_code_point_);
433 
434       __has_extened_pictographic |=
435           __prev == __extended_grapheme_custer_property_boundary::__property::__Extended_Pictographic;
436 
437       if (__at_extended_grapheme_cluster_break(__ri_break_allowed, __has_extened_pictographic, __prev, __next_prop_))
438         return __result;
439     }
440   }
441 };
442 
443 template <contiguous_iterator _Iterator>
444 __extended_grapheme_cluster_view(_Iterator, _Iterator) -> __extended_grapheme_cluster_view<iter_value_t<_Iterator>>;
445 
446 #  else //  _LIBCPP_HAS_NO_UNICODE
447 
448 // For ASCII every character is a "code point".
449 // This makes it easier to write code agnostic of the _LIBCPP_HAS_NO_UNICODE define.
450 template <class _CharT>
451 class __code_point_view {
452   using _Iterator = typename basic_string_view<_CharT>::const_iterator;
453 
454 public:
455   _LIBCPP_HIDE_FROM_ABI constexpr explicit __code_point_view(_Iterator __first, _Iterator __last)
456       : __first_(__first), __last_(__last) {}
457 
458   _LIBCPP_HIDE_FROM_ABI constexpr bool __at_end() const noexcept { return __first_ == __last_; }
459   _LIBCPP_HIDE_FROM_ABI constexpr _Iterator __position() const noexcept { return __first_; }
460 
461   [[nodiscard]] _LIBCPP_HIDE_FROM_ABI constexpr __consume_result __consume() noexcept {
462     _LIBCPP_ASSERT_INTERNAL(__first_ != __last_, "can't move beyond the end of input");
463     return {static_cast<char32_t>(*__first_++)};
464   }
465 
466 private:
467   _Iterator __first_;
468   _Iterator __last_;
469 };
470 
471 #  endif //  _LIBCPP_HAS_NO_UNICODE
472 
473 } // namespace __unicode
474 
475 #endif //_LIBCPP_STD_VER >= 20
476 
477 _LIBCPP_END_NAMESPACE_STD
478 
479 #endif // _LIBCPP___FORMAT_UNICODE_H
480