1 // Copyright (c) 2009 The Chromium 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 // File utilities that use the ICU library go in this file. 6 7 #include "base/i18n/file_util_icu.h" 8 9 #include "base/file_path.h" 10 #include "base/scoped_ptr.h" 11 #include "base/singleton.h" 12 #include "base/string_util.h" 13 #include "base/sys_string_conversions.h" 14 #include "build/build_config.h" 15 #include "unicode/coll.h" 16 #include "unicode/uniset.h" 17 18 namespace { 19 20 class IllegalCharacters { 21 public: 22 bool contains(UChar32 ucs4) { 23 return !!set->contains(ucs4); 24 } 25 26 bool containsNone(const string16 &s) { 27 return !!set->containsNone(icu::UnicodeString(s.c_str(), s.size())); 28 } 29 30 private: 31 friend class Singleton<IllegalCharacters>; 32 friend struct DefaultSingletonTraits<IllegalCharacters>; 33 34 IllegalCharacters(); 35 ~IllegalCharacters() { } 36 37 scoped_ptr<icu::UnicodeSet> set; 38 39 DISALLOW_COPY_AND_ASSIGN(IllegalCharacters); 40 }; 41 42 IllegalCharacters::IllegalCharacters() { 43 UErrorCode status = U_ZERO_ERROR; 44 // Control characters, formatting characters, non-characters, and 45 // some printable ASCII characters regarded as dangerous ('"*/:<>?\\'). 46 // See http://blogs.msdn.com/michkap/archive/2006/11/03/941420.aspx 47 // and http://msdn2.microsoft.com/en-us/library/Aa365247.aspx 48 // TODO(jungshik): Revisit the set. ZWJ and ZWNJ are excluded because they 49 // are legitimate in Arabic and some S/SE Asian scripts. However, when used 50 // elsewhere, they can be confusing/problematic. 51 // Also, consider wrapping the set with our Singleton class to create and 52 // freeze it only once. Note that there's a trade-off between memory and 53 // speed. 54 #if defined(WCHAR_T_IS_UTF16) 55 set.reset(new icu::UnicodeSet(icu::UnicodeString( 56 L"[[\"*/:<>?\\\\|][:Cc:][:Cf:] - [\u200c\u200d]]"), status)); 57 #else 58 set.reset(new icu::UnicodeSet(UNICODE_STRING_SIMPLE( 59 "[[\"*/:<>?\\\\|][:Cc:][:Cf:] - [\\u200c\\u200d]]").unescape(), 60 status)); 61 #endif 62 DCHECK(U_SUCCESS(status)); 63 // Add non-characters. If this becomes a performance bottleneck by 64 // any chance, do not add these to |set| and change IsFilenameLegal() 65 // to check |ucs4 & 0xFFFEu == 0xFFFEu|, in addiition to calling 66 // containsNone(). 67 set->add(0xFDD0, 0xFDEF); 68 for (int i = 0; i <= 0x10; ++i) { 69 int plane_base = 0x10000 * i; 70 set->add(plane_base + 0xFFFE, plane_base + 0xFFFF); 71 } 72 set->freeze(); 73 } 74 75 class LocaleAwareComparator { 76 public: 77 LocaleAwareComparator() { 78 UErrorCode error_code = U_ZERO_ERROR; 79 // Use the default collator. The default locale should have been properly 80 // set by the time this constructor is called. 81 collator_.reset(icu::Collator::createInstance(error_code)); 82 DCHECK(U_SUCCESS(error_code)); 83 // Make it case-sensitive. 84 collator_->setStrength(icu::Collator::TERTIARY); 85 // Note: We do not set UCOL_NORMALIZATION_MODE attribute. In other words, we 86 // do not pay performance penalty to guarantee sort order correctness for 87 // non-FCD (http://unicode.org/notes/tn5/#FCD) file names. This should be a 88 // reasonable tradeoff because such file names should be rare and the sort 89 // order doesn't change much anyway. 90 } 91 92 // Note: A similar function is available in l10n_util. 93 // We cannot use it because base should not depend on l10n_util. 94 // TODO(yuzo): Move some of l10n_util to base. 95 int Compare(const string16& a, const string16& b) { 96 // We are not sure if Collator::compare is thread-safe. 97 // Use an AutoLock just in case. 98 AutoLock auto_lock(lock_); 99 100 UErrorCode error_code = U_ZERO_ERROR; 101 UCollationResult result = collator_->compare( 102 static_cast<const UChar*>(a.c_str()), 103 static_cast<int>(a.length()), 104 static_cast<const UChar*>(b.c_str()), 105 static_cast<int>(b.length()), 106 error_code); 107 DCHECK(U_SUCCESS(error_code)); 108 return result; 109 } 110 111 private: 112 scoped_ptr<icu::Collator> collator_; 113 Lock lock_; 114 friend struct DefaultSingletonTraits<LocaleAwareComparator>; 115 116 DISALLOW_COPY_AND_ASSIGN(LocaleAwareComparator); 117 }; 118 119 } // namespace 120 121 namespace file_util { 122 123 bool IsFilenameLegal(const string16& file_name) { 124 return Singleton<IllegalCharacters>()->containsNone(file_name); 125 } 126 127 void ReplaceIllegalCharactersInPath(FilePath::StringType* file_name, 128 char replace_char) { 129 DCHECK(file_name); 130 131 DCHECK(!(Singleton<IllegalCharacters>()->contains(replace_char))); 132 133 // Remove leading and trailing whitespace. 134 TrimWhitespace(*file_name, TRIM_ALL, file_name); 135 136 IllegalCharacters* illegal = Singleton<IllegalCharacters>::get(); 137 int cursor = 0; // The ICU macros expect an int. 138 while (cursor < static_cast<int>(file_name->size())) { 139 int char_begin = cursor; 140 uint32 code_point; 141 #if defined(OS_MACOSX) 142 // Mac uses UTF-8 encoding for filenames. 143 U8_NEXT(file_name->data(), cursor, static_cast<int>(file_name->length()), 144 code_point); 145 #elif defined(OS_WIN) 146 // Windows uses UTF-16 encoding for filenames. 147 U16_NEXT(file_name->data(), cursor, static_cast<int>(file_name->length()), 148 code_point); 149 #elif defined(OS_POSIX) 150 // Linux doesn't actually define an encoding. It basically allows anything 151 // except for a few special ASCII characters. 152 unsigned char cur_char = static_cast<unsigned char>((*file_name)[cursor++]); 153 if (cur_char >= 0x80) 154 continue; 155 code_point = cur_char; 156 #else 157 NOTREACHED(); 158 #endif 159 160 if (illegal->contains(code_point)) { 161 file_name->replace(char_begin, cursor - char_begin, 1, replace_char); 162 // We just made the potentially multi-byte/word char into one that only 163 // takes one byte/word, so need to adjust the cursor to point to the next 164 // character again. 165 cursor = char_begin + 1; 166 } 167 } 168 } 169 170 bool LocaleAwareCompareFilenames(const FilePath& a, const FilePath& b) { 171 #if defined(OS_WIN) 172 return Singleton<LocaleAwareComparator>()->Compare(a.value().c_str(), 173 b.value().c_str()) < 0; 174 175 #elif defined(OS_POSIX) 176 // On linux, the file system encoding is not defined. We assume 177 // SysNativeMBToWide takes care of it. 178 // 179 // ICU's collator can take strings in OS native encoding. But we convert the 180 // strings to UTF-16 ourselves to ensure conversion consistency. 181 // TODO(yuzo): Perhaps we should define SysNativeMBToUTF16? 182 return Singleton<LocaleAwareComparator>()->Compare( 183 WideToUTF16(base::SysNativeMBToWide(a.value().c_str())), 184 WideToUTF16(base::SysNativeMBToWide(b.value().c_str()))) < 0; 185 #else 186 #error Not implemented on your system 187 #endif 188 } 189 190 } // namespace 191