Line data Source code
1 : // Copyright 2013 the V8 project authors. All rights reserved.
2 : // Use of this source code is governed by a BSD-style license that can be
3 : // found in the LICENSE file.
4 :
5 : #ifndef V8_INTL_SUPPORT
6 : #error Internationalization is expected to be enabled.
7 : #endif // V8_INTL_SUPPORT
8 :
9 : #include "src/intl.h"
10 :
11 : #include <memory>
12 :
13 : #include "src/factory.h"
14 : #include "src/isolate.h"
15 : #include "src/objects-inl.h"
16 : #include "src/string-case.h"
17 : #include "unicode/calendar.h"
18 : #include "unicode/gregocal.h"
19 : #include "unicode/timezone.h"
20 : #include "unicode/uchar.h"
21 : #include "unicode/uvernum.h"
22 : #include "unicode/uversion.h"
23 :
24 : #if U_ICU_VERSION_MAJOR_NUM >= 59
25 : #include "unicode/char16ptr.h"
26 : #endif
27 :
28 : namespace v8 {
29 : namespace internal {
30 :
31 : namespace {
32 200090 : inline bool IsASCIIUpper(uint16_t ch) { return ch >= 'A' && ch <= 'Z'; }
33 :
34 : const uint8_t kToLower[256] = {
35 : 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, 0x0A, 0x0B,
36 : 0x0C, 0x0D, 0x0E, 0x0F, 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
37 : 0x18, 0x19, 0x1A, 0x1B, 0x1C, 0x1D, 0x1E, 0x1F, 0x20, 0x21, 0x22, 0x23,
38 : 0x24, 0x25, 0x26, 0x27, 0x28, 0x29, 0x2A, 0x2B, 0x2C, 0x2D, 0x2E, 0x2F,
39 : 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3A, 0x3B,
40 : 0x3C, 0x3D, 0x3E, 0x3F, 0x40, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67,
41 : 0x68, 0x69, 0x6A, 0x6B, 0x6C, 0x6D, 0x6E, 0x6F, 0x70, 0x71, 0x72, 0x73,
42 : 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7A, 0x5B, 0x5C, 0x5D, 0x5E, 0x5F,
43 : 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, 0x6A, 0x6B,
44 : 0x6C, 0x6D, 0x6E, 0x6F, 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77,
45 : 0x78, 0x79, 0x7A, 0x7B, 0x7C, 0x7D, 0x7E, 0x7F, 0x80, 0x81, 0x82, 0x83,
46 : 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0x8A, 0x8B, 0x8C, 0x8D, 0x8E, 0x8F,
47 : 0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99, 0x9A, 0x9B,
48 : 0x9C, 0x9D, 0x9E, 0x9F, 0xA0, 0xA1, 0xA2, 0xA3, 0xA4, 0xA5, 0xA6, 0xA7,
49 : 0xA8, 0xA9, 0xAA, 0xAB, 0xAC, 0xAD, 0xAE, 0xAF, 0xB0, 0xB1, 0xB2, 0xB3,
50 : 0xB4, 0xB5, 0xB6, 0xB7, 0xB8, 0xB9, 0xBA, 0xBB, 0xBC, 0xBD, 0xBE, 0xBF,
51 : 0xE0, 0xE1, 0xE2, 0xE3, 0xE4, 0xE5, 0xE6, 0xE7, 0xE8, 0xE9, 0xEA, 0xEB,
52 : 0xEC, 0xED, 0xEE, 0xEF, 0xF0, 0xF1, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xD7,
53 : 0xF8, 0xF9, 0xFA, 0xFB, 0xFC, 0xFD, 0xFE, 0xDF, 0xE0, 0xE1, 0xE2, 0xE3,
54 : 0xE4, 0xE5, 0xE6, 0xE7, 0xE8, 0xE9, 0xEA, 0xEB, 0xEC, 0xED, 0xEE, 0xEF,
55 : 0xF0, 0xF1, 0xF2, 0xF3, 0xF4, 0xF5, 0xF6, 0xF7, 0xF8, 0xF9, 0xFA, 0xFB,
56 : 0xFC, 0xFD, 0xFE, 0xFF,
57 : };
58 :
59 : inline uint16_t ToLatin1Lower(uint16_t ch) {
60 4943 : return static_cast<uint16_t>(kToLower[ch]);
61 : }
62 :
63 : inline uint16_t ToASCIIUpper(uint16_t ch) {
64 0 : return ch & ~((ch >= 'a' && ch <= 'z') << 5);
65 : }
66 :
67 : // Does not work for U+00DF (sharp-s), U+00B5 (micron), U+00FF.
68 : inline uint16_t ToLatin1Upper(uint16_t ch) {
69 : DCHECK(ch != 0xDF && ch != 0xB5 && ch != 0xFF);
70 : return ch &
71 2113 : ~(((ch >= 'a' && ch <= 'z') || (((ch & 0xE0) == 0xE0) && ch != 0xF7))
72 2113 : << 5);
73 : }
74 :
75 : template <typename Char>
76 0 : bool ToUpperFastASCII(const Vector<const Char>& src,
77 : Handle<SeqOneByteString> result) {
78 : // Do a faster loop for the case where all the characters are ASCII.
79 : uint16_t ored = 0;
80 : int32_t index = 0;
81 0 : for (auto it = src.begin(); it != src.end(); ++it) {
82 0 : uint16_t ch = static_cast<uint16_t>(*it);
83 0 : ored |= ch;
84 0 : result->SeqOneByteStringSet(index++, ToASCIIUpper(ch));
85 : }
86 0 : return !(ored & ~0x7F);
87 : }
88 :
89 : const uint16_t sharp_s = 0xDF;
90 :
91 : template <typename Char>
92 2467 : bool ToUpperOneByte(const Vector<const Char>& src, uint8_t* dest,
93 : int* sharp_s_count) {
94 : // Still pretty-fast path for the input with non-ASCII Latin-1 characters.
95 :
96 : // There are two special cases.
97 : // 1. U+00B5 and U+00FF are mapped to a character beyond U+00FF.
98 : // 2. Lower case sharp-S converts to "SS" (two characters)
99 329 : *sharp_s_count = 0;
100 3618 : for (auto it = src.begin(); it != src.end(); ++it) {
101 1581 : uint16_t ch = static_cast<uint16_t>(*it);
102 1581 : if (V8_UNLIKELY(ch == sharp_s)) {
103 160 : ++(*sharp_s_count);
104 160 : continue;
105 : }
106 1421 : if (V8_UNLIKELY(ch == 0xB5 || ch == 0xFF)) {
107 : // Since this upper-cased character does not fit in an 8-bit string, we
108 : // need to take the 16-bit path.
109 : return false;
110 : }
111 2640 : *dest++ = ToLatin1Upper(ch);
112 : }
113 :
114 : return true;
115 : }
116 :
117 : template <typename Char>
118 1213 : void ToUpperWithSharpS(const Vector<const Char>& src,
119 : Handle<SeqOneByteString> result) {
120 : int32_t dest_index = 0;
121 2166 : for (auto it = src.begin(); it != src.end(); ++it) {
122 953 : uint16_t ch = static_cast<uint16_t>(*it);
123 953 : if (ch == sharp_s) {
124 160 : result->SeqOneByteStringSet(dest_index++, 'S');
125 160 : result->SeqOneByteStringSet(dest_index++, 'S');
126 : } else {
127 793 : result->SeqOneByteStringSet(dest_index++, ToLatin1Upper(ch));
128 : }
129 : }
130 130 : }
131 :
132 194510 : inline int FindFirstUpperOrNonAscii(Handle<String> s, int length) {
133 203608 : for (int index = 0; index < length; ++index) {
134 : uint16_t ch = s->Get(index);
135 200090 : if (V8_UNLIKELY(IsASCIIUpper(ch) || ch & ~0x7F)) {
136 : return index;
137 : }
138 : }
139 : return length;
140 : }
141 :
142 : } // namespace
143 :
144 584292 : const UChar* GetUCharBufferFromFlat(const String::FlatContent& flat,
145 : std::unique_ptr<uc16[]>* dest,
146 : int32_t length) {
147 : DCHECK(flat.IsFlat());
148 292202 : if (flat.IsOneByte()) {
149 283982 : if (!*dest) {
150 283870 : dest->reset(NewArray<uc16>(length));
151 : CopyChars(dest->get(), flat.ToOneByteVector().start(), length);
152 : }
153 283982 : return reinterpret_cast<const UChar*>(dest->get());
154 : } else {
155 : return reinterpret_cast<const UChar*>(flat.ToUC16Vector().start());
156 : }
157 : }
158 :
159 3233 : MUST_USE_RESULT Object* LocaleConvertCase(Handle<String> s, Isolate* isolate,
160 : bool is_to_upper, const char* lang) {
161 3233 : auto case_converter = is_to_upper ? u_strToUpper : u_strToLower;
162 : int32_t src_length = s->length();
163 : int32_t dest_length = src_length;
164 : UErrorCode status;
165 : Handle<SeqTwoByteString> result;
166 3233 : std::unique_ptr<uc16[]> sap;
167 :
168 3233 : if (dest_length == 0) return isolate->heap()->empty_string();
169 :
170 : // This is not a real loop. It'll be executed only once (no overflow) or
171 : // twice (overflow).
172 286 : for (int i = 0; i < 2; ++i) {
173 : // Case conversion can increase the string length (e.g. sharp-S => SS) so
174 : // that we have to handle RangeError exceptions here.
175 7038 : ASSIGN_RETURN_FAILURE_ON_EXCEPTION(
176 : isolate, result, isolate->factory()->NewRawTwoByteString(dest_length));
177 : DisallowHeapAllocation no_gc;
178 : DCHECK(s->IsFlat());
179 3519 : String::FlatContent flat = s->GetFlatContent();
180 3519 : const UChar* src = GetUCharBufferFromFlat(flat, &sap, src_length);
181 3519 : status = U_ZERO_ERROR;
182 3519 : dest_length = case_converter(reinterpret_cast<UChar*>(result->GetChars()),
183 3519 : dest_length, src, src_length, lang, &status);
184 3519 : if (status != U_BUFFER_OVERFLOW_ERROR) break;
185 : }
186 :
187 : // In most cases, the output will fill the destination buffer completely
188 : // leading to an unterminated string (U_STRING_NOT_TERMINATED_WARNING).
189 : // Only in rare cases, it'll be shorter than the destination buffer and
190 : // |result| has to be truncated.
191 : DCHECK(U_SUCCESS(status));
192 3233 : if (V8_LIKELY(status == U_STRING_NOT_TERMINATED_WARNING)) {
193 : DCHECK(dest_length == result->length());
194 3023 : return *result;
195 : }
196 210 : if (U_SUCCESS(status)) {
197 : DCHECK(dest_length < result->length());
198 : return *Handle<SeqTwoByteString>::cast(
199 420 : SeqString::Truncate(result, dest_length));
200 : }
201 0 : return *s;
202 : }
203 :
204 201128 : MUST_USE_RESULT Object* ConvertToLower(Handle<String> s, Isolate* isolate) {
205 201128 : if (!s->HasOnlyOneByteChars()) {
206 : // Use a slower implementation for strings with characters beyond U+00FF.
207 2105 : return LocaleConvertCase(s, isolate, false, "");
208 : }
209 :
210 : int length = s->length();
211 :
212 : // We depend here on the invariant that the length of a Latin1
213 : // string is invariant under ToLowerCase, and the result always
214 : // fits in the Latin1 range in the *root locale*. It does not hold
215 : // for ToUpperCase even in the root locale.
216 :
217 : // Scan the string for uppercase and non-ASCII characters for strings
218 : // shorter than a machine-word without any memory allocation overhead.
219 : // TODO(jshin): Apply this to a longer input by breaking FastAsciiConvert()
220 : // to two parts, one for scanning the prefix with no change and the other for
221 : // handling ASCII-only characters.
222 : int index_to_first_unprocessed = length;
223 : const bool is_short = length < static_cast<int>(sizeof(uintptr_t));
224 199023 : if (is_short) {
225 194510 : index_to_first_unprocessed = FindFirstUpperOrNonAscii(s, length);
226 : // Nothing to do if the string is all ASCII with no uppercase.
227 198028 : if (index_to_first_unprocessed == length) return *s;
228 : }
229 :
230 : Handle<SeqOneByteString> result =
231 391010 : isolate->factory()->NewRawOneByteString(length).ToHandleChecked();
232 :
233 : DisallowHeapAllocation no_gc;
234 : DCHECK(s->IsFlat());
235 195505 : String::FlatContent flat = s->GetFlatContent();
236 195505 : uint8_t* dest = result->GetChars();
237 195505 : if (flat.IsOneByte()) {
238 : const uint8_t* src = flat.ToOneByteVector().start();
239 195505 : bool has_changed_character = false;
240 : index_to_first_unprocessed = FastAsciiConvert<true>(
241 : reinterpret_cast<char*>(dest), reinterpret_cast<const char*>(src),
242 195505 : length, &has_changed_character);
243 : // If not ASCII, we keep the result up to index_to_first_unprocessed and
244 : // process the rest.
245 195505 : if (index_to_first_unprocessed == length)
246 386892 : return has_changed_character ? *result : *s;
247 :
248 4943 : for (int index = index_to_first_unprocessed; index < length; ++index) {
249 9886 : dest[index] = ToLatin1Lower(static_cast<uint16_t>(src[index]));
250 : }
251 : } else {
252 0 : if (index_to_first_unprocessed == length) {
253 : DCHECK(!is_short);
254 0 : index_to_first_unprocessed = FindFirstUpperOrNonAscii(s, length);
255 : }
256 : // Nothing to do if the string is all ASCII with no uppercase.
257 0 : if (index_to_first_unprocessed == length) return *s;
258 : const uint16_t* src = flat.ToUC16Vector().start();
259 0 : CopyChars(dest, src, index_to_first_unprocessed);
260 0 : for (int index = index_to_first_unprocessed; index < length; ++index) {
261 0 : dest[index] = ToLatin1Lower(static_cast<uint16_t>(src[index]));
262 : }
263 : }
264 :
265 2059 : return *result;
266 : }
267 :
268 8932 : MUST_USE_RESULT Object* ConvertToUpper(Handle<String> s, Isolate* isolate) {
269 : int32_t length = s->length();
270 8932 : if (s->HasOnlyOneByteChars() && length > 0) {
271 : Handle<SeqOneByteString> result =
272 17014 : isolate->factory()->NewRawOneByteString(length).ToHandleChecked();
273 :
274 : DCHECK(s->IsFlat());
275 : int sharp_s_count;
276 : bool is_result_single_byte;
277 : {
278 : DisallowHeapAllocation no_gc;
279 8507 : String::FlatContent flat = s->GetFlatContent();
280 8507 : uint8_t* dest = result->GetChars();
281 8507 : if (flat.IsOneByte()) {
282 : Vector<const uint8_t> src = flat.ToOneByteVector();
283 8507 : bool has_changed_character = false;
284 : int index_to_first_unprocessed =
285 : FastAsciiConvert<false>(reinterpret_cast<char*>(result->GetChars()),
286 : reinterpret_cast<const char*>(src.start()),
287 8507 : length, &has_changed_character);
288 8507 : if (index_to_first_unprocessed == length)
289 16356 : return has_changed_character ? *result : *s;
290 : // If not ASCII, we keep the result up to index_to_first_unprocessed and
291 : // process the rest.
292 : is_result_single_byte =
293 : ToUpperOneByte(src.SubVector(index_to_first_unprocessed, length),
294 658 : dest + index_to_first_unprocessed, &sharp_s_count);
295 : } else {
296 : DCHECK(flat.IsTwoByte());
297 0 : Vector<const uint16_t> src = flat.ToUC16Vector();
298 0 : if (ToUpperFastASCII(src, result)) return *result;
299 0 : is_result_single_byte = ToUpperOneByte(src, dest, &sharp_s_count);
300 : }
301 : }
302 :
303 : // Go to the full Unicode path if there are characters whose uppercase
304 : // is beyond the Latin-1 range (cannot be represented in OneByteString).
305 329 : if (V8_UNLIKELY(!is_result_single_byte)) {
306 101 : return LocaleConvertCase(s, isolate, true, "");
307 : }
308 :
309 326 : if (sharp_s_count == 0) return *result;
310 :
311 : // We have sharp_s_count sharp-s characters, but the result is still
312 : // in the Latin-1 range.
313 260 : ASSIGN_RETURN_FAILURE_ON_EXCEPTION(
314 : isolate, result,
315 : isolate->factory()->NewRawOneByteString(length + sharp_s_count));
316 : DisallowHeapAllocation no_gc;
317 130 : String::FlatContent flat = s->GetFlatContent();
318 130 : if (flat.IsOneByte()) {
319 130 : ToUpperWithSharpS(flat.ToOneByteVector(), result);
320 : } else {
321 0 : ToUpperWithSharpS(flat.ToUC16Vector(), result);
322 : }
323 :
324 130 : return *result;
325 : }
326 :
327 425 : return LocaleConvertCase(s, isolate, true, "");
328 : }
329 :
330 206312 : MUST_USE_RESULT Object* ConvertCase(Handle<String> s, bool is_upper,
331 : Isolate* isolate) {
332 206312 : return is_upper ? ConvertToUpper(s, isolate) : ConvertToLower(s, isolate);
333 : }
334 :
335 42 : ICUTimezoneCache::ICUTimezoneCache() : timezone_(nullptr) { Clear(); }
336 :
337 28 : ICUTimezoneCache::~ICUTimezoneCache() { Clear(); }
338 :
339 28 : const char* ICUTimezoneCache::LocalTimezone(double time_ms) {
340 28 : bool is_dst = DaylightSavingsOffset(time_ms) != 0;
341 28 : char* name = is_dst ? dst_timezone_name_ : timezone_name_;
342 28 : if (name[0] == '\0') {
343 : icu::UnicodeString result;
344 56 : GetTimeZone()->getDisplayName(is_dst, icu::TimeZone::LONG, result);
345 : result += '\0';
346 :
347 56 : icu::CheckedArrayByteSink byte_sink(name, kMaxTimezoneChars);
348 28 : result.toUTF8(byte_sink);
349 56 : CHECK(!byte_sink.Overflowed());
350 : }
351 28 : return const_cast<const char*>(name);
352 : }
353 :
354 0 : icu::TimeZone* ICUTimezoneCache::GetTimeZone() {
355 105 : if (timezone_ == nullptr) {
356 14 : timezone_ = icu::TimeZone::createDefault();
357 : }
358 105 : return timezone_;
359 : }
360 :
361 77 : bool ICUTimezoneCache::GetOffsets(double time_ms, int32_t* raw_offset,
362 : int32_t* dst_offset) {
363 77 : UErrorCode status = U_ZERO_ERROR;
364 77 : GetTimeZone()->getOffset(time_ms, false, *raw_offset, *dst_offset, status);
365 154 : return U_SUCCESS(status);
366 : }
367 :
368 63 : double ICUTimezoneCache::DaylightSavingsOffset(double time_ms) {
369 : int32_t raw_offset, dst_offset;
370 63 : if (!GetOffsets(time_ms, &raw_offset, &dst_offset)) return 0;
371 63 : return dst_offset;
372 : }
373 :
374 14 : double ICUTimezoneCache::LocalTimeOffset() {
375 : int32_t raw_offset, dst_offset;
376 14 : if (!GetOffsets(icu::Calendar::getNow(), &raw_offset, &dst_offset)) return 0;
377 14 : return raw_offset;
378 : }
379 :
380 14 : void ICUTimezoneCache::Clear() {
381 28 : delete timezone_;
382 42 : timezone_ = nullptr;
383 42 : timezone_name_[0] = '\0';
384 42 : dst_timezone_name_[0] = '\0';
385 14 : }
386 :
387 : } // namespace internal
388 : } // namespace v8
|