OLD | NEW |
---|---|
1 // Copyright 2014 the V8 project authors. All rights reserved. | 1 // Copyright 2014 the V8 project authors. All rights reserved. |
2 // Use of this source code is governed by a BSD-style license that can be | 2 // Use of this source code is governed by a BSD-style license that can be |
3 // found in the LICENSE file. | 3 // found in the LICENSE file. |
4 | 4 |
5 | 5 |
6 #ifdef V8_I18N_SUPPORT | 6 #ifdef V8_I18N_SUPPORT |
7 #include "src/runtime/runtime-utils.h" | 7 #include "src/runtime/runtime-utils.h" |
8 | 8 |
9 #include "src/api.h" | 9 #include "src/api.h" |
10 #include "src/api-natives.h" | 10 #include "src/api-natives.h" |
(...skipping 11 matching lines...) Expand all Loading... | |
22 #include "unicode/dcfmtsym.h" | 22 #include "unicode/dcfmtsym.h" |
23 #include "unicode/decimfmt.h" | 23 #include "unicode/decimfmt.h" |
24 #include "unicode/dtfmtsym.h" | 24 #include "unicode/dtfmtsym.h" |
25 #include "unicode/dtptngen.h" | 25 #include "unicode/dtptngen.h" |
26 #include "unicode/locid.h" | 26 #include "unicode/locid.h" |
27 #include "unicode/numfmt.h" | 27 #include "unicode/numfmt.h" |
28 #include "unicode/numsys.h" | 28 #include "unicode/numsys.h" |
29 #include "unicode/rbbi.h" | 29 #include "unicode/rbbi.h" |
30 #include "unicode/smpdtfmt.h" | 30 #include "unicode/smpdtfmt.h" |
31 #include "unicode/timezone.h" | 31 #include "unicode/timezone.h" |
32 #include "unicode/translit.h" | |
32 #include "unicode/uchar.h" | 33 #include "unicode/uchar.h" |
33 #include "unicode/ucol.h" | 34 #include "unicode/ucol.h" |
34 #include "unicode/ucurr.h" | 35 #include "unicode/ucurr.h" |
35 #include "unicode/uloc.h" | 36 #include "unicode/uloc.h" |
37 #include "unicode/unistr.h" | |
36 #include "unicode/unum.h" | 38 #include "unicode/unum.h" |
37 #include "unicode/uversion.h" | 39 #include "unicode/uversion.h" |
38 | 40 |
39 | 41 |
40 namespace v8 { | 42 namespace v8 { |
41 namespace internal { | 43 namespace internal { |
42 | 44 |
43 RUNTIME_FUNCTION(Runtime_CanonicalizeLanguageTag) { | 45 RUNTIME_FUNCTION(Runtime_CanonicalizeLanguageTag) { |
44 HandleScope scope(isolate); | 46 HandleScope scope(isolate); |
45 Factory* factory = isolate->factory(); | 47 Factory* factory = isolate->factory(); |
(...skipping 696 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... | |
742 } else if (status >= UBRK_WORD_LETTER && status < UBRK_WORD_LETTER_LIMIT) { | 744 } else if (status >= UBRK_WORD_LETTER && status < UBRK_WORD_LETTER_LIMIT) { |
743 return *isolate->factory()->NewStringFromStaticChars("letter"); | 745 return *isolate->factory()->NewStringFromStaticChars("letter"); |
744 } else if (status >= UBRK_WORD_KANA && status < UBRK_WORD_KANA_LIMIT) { | 746 } else if (status >= UBRK_WORD_KANA && status < UBRK_WORD_KANA_LIMIT) { |
745 return *isolate->factory()->NewStringFromStaticChars("kana"); | 747 return *isolate->factory()->NewStringFromStaticChars("kana"); |
746 } else if (status >= UBRK_WORD_IDEO && status < UBRK_WORD_IDEO_LIMIT) { | 748 } else if (status >= UBRK_WORD_IDEO && status < UBRK_WORD_IDEO_LIMIT) { |
747 return *isolate->factory()->NewStringFromStaticChars("ideo"); | 749 return *isolate->factory()->NewStringFromStaticChars("ideo"); |
748 } else { | 750 } else { |
749 return *isolate->factory()->NewStringFromStaticChars("unknown"); | 751 return *isolate->factory()->NewStringFromStaticChars("unknown"); |
750 } | 752 } |
751 } | 753 } |
754 | |
755 namespace { | |
756 inline void LocaleConvertCaseHelper(icu::UnicodeString* s, bool is_to_upper, | |
757 const icu::Locale& locale) { | |
758 if (is_to_upper) | |
759 s->toUpper(locale); | |
760 else | |
761 s->toLower(locale); | |
762 } | |
763 void ConvertCaseWithTransliterator(icu::UnicodeString* input, | |
764 const char* transliterator_id) { | |
765 UErrorCode status = U_ZERO_ERROR; | |
766 base::SmartPointer<icu::Transliterator> translit( | |
767 icu::Transliterator::createInstance( | |
768 icu::UnicodeString(transliterator_id, -1, US_INV), UTRANS_FORWARD, | |
769 status)); | |
770 if (U_FAILURE(status)) return; | |
771 translit->transliterate(*input); | |
772 } | |
773 | |
774 MUST_USE_RESULT Object* LocaleConvertCase(Handle<String> s, Isolate* isolate, | |
775 bool is_to_upper, int locale_id) { | |
776 static const char* conversion_locales[] = { | |
777 "az", "el", "lt", "tr", | |
778 }; | |
779 RUNTIME_ASSERT(locale_id >= -1 && | |
780 locale_id < static_cast<int>(arraysize(conversion_locales))); | |
781 int32_t length = s->length(); | |
782 icu::UnicodeString converted; | |
783 | |
784 DisallowHeapAllocation no_gc; | |
jungshik at Google
2016/04/13 17:34:34
Getting this out of an isolated block to avoid an
Dan Ehrenberg
2016/04/13 17:43:43
I believe DisallowHeapAllocation's function is to
| |
785 DCHECK(s->IsFlat()); | |
786 String::FlatContent flat = s->GetFlatContent(); | |
787 | |
788 const UChar* src; | |
789 base::SmartArrayPointer<uc16> sap; | |
790 if (flat.IsOneByte()) { | |
791 sap = s->ToWideCString(); | |
792 src = reinterpret_cast<const UChar*>(sap.get()); | |
793 } else { | |
794 src = reinterpret_cast<const UChar*>(flat.ToUC16Vector().start()); | |
795 } | |
796 converted = icu::UnicodeString(false, src, length); | |
797 | |
798 if (locale_id == -1) { | |
799 LocaleConvertCaseHelper(&converted, is_to_upper, icu::Locale::getRoot()); | |
800 } else if (V8_UNLIKELY(locale_id == 1 && is_to_upper)) { | |
801 // TODO(jshin): Once http://bugs.icu-project.org/trac/ticket/10582 is | |
802 // fixed, remove this special-casing for uppercasing in Greek(el) locale. | |
803 // This is ~500 times slower than using the case conversion API. | |
804 ConvertCaseWithTransliterator(&converted, "el-Upper"); | |
805 } else { | |
806 LocaleConvertCaseHelper(&converted, is_to_upper, | |
807 icu::Locale(conversion_locales[locale_id])); | |
808 } | |
809 | |
810 Handle<String> result; | |
811 ASSIGN_RETURN_FAILURE_ON_EXCEPTION( | |
812 isolate, result, | |
813 isolate->factory()->NewStringFromTwoByte(Vector<const uint16_t>( | |
814 reinterpret_cast<const uint16_t*>(converted.getBuffer()), | |
815 converted.length()))); | |
816 return *result; | |
817 } | |
818 | |
819 inline bool IsASCIIUpper(uint16_t ch) { return ch >= 'A' && ch <= 'Z'; } | |
820 | |
821 inline uint16_t ToASCIILower(uint16_t ch) { | |
822 return ch | ((ch >= 'A' && ch <= 'Z') << 5); | |
823 } | |
824 | |
825 inline uint16_t ToASCIIUpper(uint16_t ch) { | |
826 return ch & ~((ch >= 'a' && ch <= 'z') << 5); | |
827 } | |
828 | |
829 } // namespace | |
830 | |
831 RUNTIME_FUNCTION(Runtime_StringToLowerCaseI18N) { | |
832 HandleScope scope(isolate); | |
833 DCHECK_EQ(args.length(), 1); | |
834 CONVERT_ARG_HANDLE_CHECKED(String, s, 0); | |
835 | |
836 int length = s->length(); | |
837 s = String::Flatten(s); | |
838 // First scan the string for uppercase and non-ASCII characters: | |
839 if (s->HasOnlyOneByteChars()) { | |
840 unsigned first_index_to_lower = length; | |
841 for (int index = 0; index < length; ++index) { | |
842 // Blink specializes this path for one-byte strings, so it | |
843 // does not need to do a generic get, but can do the equivalent | |
844 // of SeqOneByteStringGet. | |
845 uint16_t ch = s->Get(index); | |
846 if (V8_UNLIKELY(IsASCIIUpper(ch) || ch & ~0x7F)) { | |
847 first_index_to_lower = index; | |
848 break; | |
849 } | |
850 } | |
851 | |
852 // Nothing to do if the string is all ASCII with no uppercase. | |
853 if (first_index_to_lower == length) return *s; | |
854 | |
855 // We depend here on the invariant that the length of a Latin1 | |
856 // string is invariant under ToLowerCase, and the result always | |
857 // fits in the Latin1 range (untrue for ToUpperCase, and might | |
858 // be untrue in some locales, but this is the root locale) | |
859 Handle<SeqOneByteString> result; | |
860 ASSIGN_RETURN_FAILURE_ON_EXCEPTION( | |
861 isolate, result, isolate->factory()->NewRawOneByteString(length)); | |
862 if (s->IsSeqOneByteString()) { | |
863 SeqOneByteString* source = SeqOneByteString::cast(*s); | |
864 CopyChars(result->GetChars(), source->GetChars(), first_index_to_lower); | |
865 } else { | |
866 // Do we have to worry about External{One,Two}ByteString? | |
867 DCHECK(s->IsSeqTwoByteString()); | |
868 SeqTwoByteString* source = SeqTwoByteString::cast(*s); | |
869 CopyChars(result->GetChars(), source->GetChars(), first_index_to_lower); | |
870 } | |
871 | |
872 for (int index = first_index_to_lower; index < length; ++index) { | |
873 uint16_t ch = s->Get(index); | |
874 result->SeqOneByteStringSet( | |
875 index, V8_UNLIKELY(ch & ~0x7F) ? static_cast<uint16_t>(u_tolower(ch)) | |
876 : ToASCIILower(ch)); | |
877 } | |
878 | |
879 return *result; | |
880 } | |
881 | |
882 // Blink had an additional case here for ASCII 2-byte strings, but | |
883 // that is subsumed by the above code (assuming there isn't a false | |
884 // negative for HasOnlyOneByteChars). | |
885 | |
886 // Do a slower implementation for cases that include non-ASCII characters. | |
887 return LocaleConvertCase(s, isolate, false, -1); | |
888 } | |
889 | |
890 RUNTIME_FUNCTION(Runtime_StringToUpperCaseI18N) { | |
891 HandleScope scope(isolate); | |
892 DCHECK_EQ(args.length(), 1); | |
893 CONVERT_ARG_HANDLE_CHECKED(String, s, 0); | |
894 | |
895 // This function could be optimized for no-op cases the way lowercase | |
896 // counterpart is, but in empirical testing, few actual calls to upper() | |
897 // are no-ops. So, it wouldn't be worth the extra time for pre-scanning. | |
898 | |
899 int32_t length = s->length(); | |
900 s = String::Flatten(s); | |
901 static const uint16_t sharp_s = 0x00DFu; | |
902 | |
903 if (s->HasOnlyOneByteChars()) { | |
904 Handle<SeqOneByteString> result; | |
905 ASSIGN_RETURN_FAILURE_ON_EXCEPTION( | |
906 isolate, result, isolate->factory()->NewRawOneByteString(length)); | |
907 | |
908 // Do a faster loop for the case where all the characters are ASCII. | |
909 uint16_t ored = 0; | |
910 for (int index = 0; index < length; ++index) { | |
911 uint16_t ch = s->Get(index); | |
912 ored |= ch; | |
913 result->SeqOneByteStringSet(index, ToASCIIUpper(ch)); | |
914 } | |
915 if (!(ored & ~0x7F)) return *result; | |
916 | |
917 // Do a slower implementation for cases that include non-ASCII Latin-1 | |
918 // characters. | |
919 int sharp_s_count = 0; | |
920 | |
921 // There are two special cases. | |
922 // 1. latin-1 characters when converted to upper case are 16 bit | |
923 // characters. | |
924 // 2. Lower case sharp-S converts to "SS" (two characters) | |
925 for (int32_t index = 0; index < length; ++index) { | |
926 uint16_t ch = s->Get(index); | |
927 if (V8_UNLIKELY(ch == sharp_s)) { | |
928 ++sharp_s_count; | |
929 continue; | |
930 } | |
931 uint16_t upper = static_cast<uint16_t>(u_toupper(static_cast<UChar>(ch))); | |
932 if (V8_UNLIKELY(upper > 0xff)) { | |
933 // Since this upper-cased character does not fit in an 8-bit string, we | |
934 // need to take the 16-bit path. | |
935 return LocaleConvertCase(s, isolate, true, -1); | |
936 } | |
937 result->SeqOneByteStringSet(index, upper); | |
938 } | |
939 | |
940 if (sharp_s_count == 0) return *result; | |
941 | |
942 // We have sharp_s_count sharp-s characters, but none of the other special | |
943 // characters. | |
944 ASSIGN_RETURN_FAILURE_ON_EXCEPTION( | |
945 isolate, result, | |
946 isolate->factory()->NewRawOneByteString(length + sharp_s_count)); | |
947 for (int32_t index = 0, dest_index = 0; index < length; ++index) { | |
948 uint16_t ch = s->Get(index); | |
949 if (ch == sharp_s) { | |
950 result->SeqOneByteStringSet(dest_index++, 'S'); | |
951 result->SeqOneByteStringSet(dest_index++, 'S'); | |
952 } else { | |
953 uint16_t upper = | |
954 static_cast<uint16_t>(u_toupper(static_cast<UChar>(ch))); | |
955 result->SeqOneByteStringSet(dest_index++, upper); | |
956 } | |
957 } | |
958 | |
959 return *result; | |
960 } | |
961 | |
962 return LocaleConvertCase(s, isolate, true, -1); | |
963 } | |
964 | |
965 RUNTIME_FUNCTION(Runtime_StringLocaleConvertCase) { | |
966 HandleScope scope(isolate); | |
967 DCHECK_EQ(args.length(), 3); | |
968 CONVERT_ARG_HANDLE_CHECKED(String, s, 0); | |
969 CONVERT_BOOLEAN_ARG_CHECKED(is_upper, 1); | |
970 CONVERT_NUMBER_CHECKED(int, lang_id, Int32, args[2]); | |
971 | |
972 return LocaleConvertCase(s, isolate, is_upper, lang_id); | |
973 } | |
974 | |
752 } // namespace internal | 975 } // namespace internal |
753 } // namespace v8 | 976 } // namespace v8 |
754 | 977 |
755 #endif // V8_I18N_SUPPORT | 978 #endif // V8_I18N_SUPPORT |
OLD | NEW |