1 // Copyright (c) 2011 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.
5 #include "base/i18n/rtl.h"
7 #include "base/files/file_path.h"
8 #include "base/logging.h"
9 #include "base/strings/string_util.h"
10 #include "base/strings/sys_string_conversions.h"
11 #include "base/strings/utf_string_conversions.h"
12 #include "third_party/icu/source/common/unicode/locid.h"
13 #include "third_party/icu/source/common/unicode/uchar.h"
14 #include "third_party/icu/source/common/unicode/uscript.h"
15 #include "third_party/icu/source/i18n/unicode/coll.h"
17 #if defined(TOOLKIT_GTK)
23 // Extract language, country and variant, but ignore keywords. For example,
24 // en-US, ca@valencia, ca-ES@valencia.
25 std::string
GetLocaleString(const icu::Locale
& locale
) {
26 const char* language
= locale
.getLanguage();
27 const char* country
= locale
.getCountry();
28 const char* variant
= locale
.getVariant();
31 (language
!= NULL
&& *language
!= '\0') ? language
: "und";
33 if (country
!= NULL
&& *country
!= '\0') {
38 if (variant
!= NULL
&& *variant
!= '\0') {
39 std::string
variant_str(variant
);
40 StringToLowerASCII(&variant_str
);
41 result
+= '@' + variant_str
;
47 // Returns LEFT_TO_RIGHT or RIGHT_TO_LEFT if |character| has strong
48 // directionality, returns UNKNOWN_DIRECTION if it doesn't. Please refer to
49 // http://unicode.org/reports/tr9/ for more information.
50 base::i18n::TextDirection
GetCharacterDirection(UChar32 character
) {
51 // Now that we have the character, we use ICU in order to query for the
52 // appropriate Unicode BiDi character type.
53 int32_t property
= u_getIntPropertyValue(character
, UCHAR_BIDI_CLASS
);
54 if ((property
== U_RIGHT_TO_LEFT
) ||
55 (property
== U_RIGHT_TO_LEFT_ARABIC
) ||
56 (property
== U_RIGHT_TO_LEFT_EMBEDDING
) ||
57 (property
== U_RIGHT_TO_LEFT_OVERRIDE
)) {
58 return base::i18n::RIGHT_TO_LEFT
;
59 } else if ((property
== U_LEFT_TO_RIGHT
) ||
60 (property
== U_LEFT_TO_RIGHT_EMBEDDING
) ||
61 (property
== U_LEFT_TO_RIGHT_OVERRIDE
)) {
62 return base::i18n::LEFT_TO_RIGHT
;
64 return base::i18n::UNKNOWN_DIRECTION
;
72 // Represents the locale-specific ICU text direction.
73 static TextDirection g_icu_text_direction
= UNKNOWN_DIRECTION
;
75 // Convert the ICU default locale to a string.
76 std::string
GetConfiguredLocale() {
77 return GetLocaleString(icu::Locale::getDefault());
80 // Convert the ICU canonicalized locale to a string.
81 std::string
GetCanonicalLocale(const char* locale
) {
82 return GetLocaleString(icu::Locale::createCanonical(locale
));
85 // Convert Chrome locale name to ICU locale name
86 std::string
ICULocaleName(const std::string
& locale_string
) {
87 // If not Spanish, just return it.
88 if (locale_string
.substr(0, 2) != "es")
90 // Expand es to es-ES.
91 if (LowerCaseEqualsASCII(locale_string
, "es"))
93 // Map es-419 (Latin American Spanish) to es-FOO depending on the system
94 // locale. If it's es-RR other than es-ES, map to es-RR. Otherwise, map
95 // to es-MX (the most populous in Spanish-speaking Latin America).
96 if (LowerCaseEqualsASCII(locale_string
, "es-419")) {
97 const icu::Locale
& locale
= icu::Locale::getDefault();
98 std::string language
= locale
.getLanguage();
99 const char* country
= locale
.getCountry();
100 if (LowerCaseEqualsASCII(language
, "es") &&
101 !LowerCaseEqualsASCII(country
, "es")) {
108 // Currently, Chrome has only "es" and "es-419", but later we may have
109 // more specific "es-RR".
110 return locale_string
;
113 void SetICUDefaultLocale(const std::string
& locale_string
) {
114 icu::Locale
locale(ICULocaleName(locale_string
).c_str());
115 UErrorCode error_code
= U_ZERO_ERROR
;
116 icu::Locale::setDefault(locale
, error_code
);
117 // This return value is actually bogus because Locale object is
118 // an ID and setDefault seems to always succeed (regardless of the
119 // presence of actual locale data). However,
120 // it does not hurt to have it as a sanity check.
121 DCHECK(U_SUCCESS(error_code
));
122 g_icu_text_direction
= UNKNOWN_DIRECTION
;
126 #if defined(TOOLKIT_GTK)
127 GtkTextDirection gtk_dir
= gtk_widget_get_default_direction();
128 return gtk_dir
== GTK_TEXT_DIR_RTL
;
135 if (g_icu_text_direction
== UNKNOWN_DIRECTION
) {
136 const icu::Locale
& locale
= icu::Locale::getDefault();
137 g_icu_text_direction
= GetTextDirectionForLocale(locale
.getName());
139 return g_icu_text_direction
== RIGHT_TO_LEFT
;
142 TextDirection
GetTextDirectionForLocale(const char* locale_name
) {
143 UErrorCode status
= U_ZERO_ERROR
;
144 ULayoutType layout_dir
= uloc_getCharacterOrientation(locale_name
, &status
);
145 DCHECK(U_SUCCESS(status
));
146 // Treat anything other than RTL as LTR.
147 return (layout_dir
!= ULOC_LAYOUT_RTL
) ? LEFT_TO_RIGHT
: RIGHT_TO_LEFT
;
150 TextDirection
GetFirstStrongCharacterDirection(const string16
& text
) {
151 const UChar
* string
= text
.c_str();
152 size_t length
= text
.length();
154 while (position
< length
) {
156 size_t next_position
= position
;
157 U16_NEXT(string
, next_position
, length
, character
);
158 TextDirection direction
= GetCharacterDirection(character
);
159 if (direction
!= UNKNOWN_DIRECTION
)
161 position
= next_position
;
163 return LEFT_TO_RIGHT
;
166 TextDirection
GetLastStrongCharacterDirection(const string16
& text
) {
167 const UChar
* string
= text
.c_str();
168 size_t position
= text
.length();
169 while (position
> 0) {
171 size_t prev_position
= position
;
172 U16_PREV(string
, 0, prev_position
, character
);
173 TextDirection direction
= GetCharacterDirection(character
);
174 if (direction
!= UNKNOWN_DIRECTION
)
176 position
= prev_position
;
178 return LEFT_TO_RIGHT
;
181 TextDirection
GetStringDirection(const string16
& text
) {
182 const UChar
* string
= text
.c_str();
183 size_t length
= text
.length();
186 TextDirection
result(UNKNOWN_DIRECTION
);
187 while (position
< length
) {
189 size_t next_position
= position
;
190 U16_NEXT(string
, next_position
, length
, character
);
191 TextDirection direction
= GetCharacterDirection(character
);
192 if (direction
!= UNKNOWN_DIRECTION
) {
193 if (result
!= UNKNOWN_DIRECTION
&& result
!= direction
)
194 return UNKNOWN_DIRECTION
;
197 position
= next_position
;
200 // Handle the case of a string not containing any strong directionality
201 // characters defaulting to LEFT_TO_RIGHT.
202 if (result
== UNKNOWN_DIRECTION
)
203 return LEFT_TO_RIGHT
;
209 bool AdjustStringForLocaleDirection(string16
* text
) {
210 if (!IsRTL() || text
->empty())
213 // Marking the string as LTR if the locale is RTL and the string does not
214 // contain strong RTL characters. Otherwise, mark the string as RTL.
215 bool has_rtl_chars
= StringContainsStrongRTLChars(*text
);
217 WrapStringWithLTRFormatting(text
);
219 WrapStringWithRTLFormatting(text
);
224 bool UnadjustStringForLocaleDirection(string16
* text
) {
225 if (!IsRTL() || text
->empty())
228 *text
= StripWrappingBidiControlCharacters(*text
);
232 bool AdjustStringForLocaleDirection(string16
* text
) {
233 // On OS X & GTK the directionality of a label is determined by the first
234 // strongly directional character.
235 // However, we want to make sure that in an LTR-language-UI all strings are
236 // left aligned and vice versa.
237 // A problem can arise if we display a string which starts with user input.
238 // User input may be of the opposite directionality to the UI. So the whole
239 // string will be displayed in the opposite directionality, e.g. if we want to
240 // display in an LTR UI [such as US English]:
242 // EMAN_NOISNETXE is now installed.
244 // Since EXTENSION_NAME begins with a strong RTL char, the label's
245 // directionality will be set to RTL and the string will be displayed visually
248 // .is now installed EMAN_NOISNETXE
250 // In order to solve this issue, we prepend an LRM to the string. An LRM is a
251 // strongly directional LTR char.
252 // We also append an LRM at the end, which ensures that we're in an LTR
255 // Unlike Windows, Linux and OS X can correctly display RTL glyphs out of the
256 // box so there is no issue with displaying zero-width bidi control characters
257 // on any system. Thus no need for the !IsRTL() check here.
261 bool ui_direction_is_rtl
= IsRTL();
263 bool has_rtl_chars
= StringContainsStrongRTLChars(*text
);
264 if (!ui_direction_is_rtl
&& has_rtl_chars
) {
265 WrapStringWithRTLFormatting(text
);
266 text
->insert(static_cast<size_t>(0), static_cast<size_t>(1),
268 text
->push_back(kLeftToRightMark
);
269 } else if (ui_direction_is_rtl
&& has_rtl_chars
) {
270 WrapStringWithRTLFormatting(text
);
271 text
->insert(static_cast<size_t>(0), static_cast<size_t>(1),
273 text
->push_back(kRightToLeftMark
);
274 } else if (ui_direction_is_rtl
) {
275 WrapStringWithLTRFormatting(text
);
276 text
->insert(static_cast<size_t>(0), static_cast<size_t>(1),
278 text
->push_back(kRightToLeftMark
);
286 bool UnadjustStringForLocaleDirection(string16
* text
) {
290 size_t begin_index
= 0;
291 char16 begin
= text
->at(begin_index
);
292 if (begin
== kLeftToRightMark
||
293 begin
== kRightToLeftMark
) {
297 size_t end_index
= text
->length() - 1;
298 char16 end
= text
->at(end_index
);
299 if (end
== kLeftToRightMark
||
300 end
== kRightToLeftMark
) {
304 string16 unmarked_text
=
305 text
->substr(begin_index
, end_index
- begin_index
+ 1);
306 *text
= StripWrappingBidiControlCharacters(unmarked_text
);
312 bool StringContainsStrongRTLChars(const string16
& text
) {
313 const UChar
* string
= text
.c_str();
314 size_t length
= text
.length();
316 while (position
< length
) {
318 size_t next_position
= position
;
319 U16_NEXT(string
, next_position
, length
, character
);
321 // Now that we have the character, we use ICU in order to query for the
322 // appropriate Unicode BiDi character type.
323 int32_t property
= u_getIntPropertyValue(character
, UCHAR_BIDI_CLASS
);
324 if ((property
== U_RIGHT_TO_LEFT
) || (property
== U_RIGHT_TO_LEFT_ARABIC
))
327 position
= next_position
;
333 void WrapStringWithLTRFormatting(string16
* text
) {
337 // Inserting an LRE (Left-To-Right Embedding) mark as the first character.
338 text
->insert(static_cast<size_t>(0), static_cast<size_t>(1),
339 kLeftToRightEmbeddingMark
);
341 // Inserting a PDF (Pop Directional Formatting) mark as the last character.
342 text
->push_back(kPopDirectionalFormatting
);
345 void WrapStringWithRTLFormatting(string16
* text
) {
349 // Inserting an RLE (Right-To-Left Embedding) mark as the first character.
350 text
->insert(static_cast<size_t>(0), static_cast<size_t>(1),
351 kRightToLeftEmbeddingMark
);
353 // Inserting a PDF (Pop Directional Formatting) mark as the last character.
354 text
->push_back(kPopDirectionalFormatting
);
357 void WrapPathWithLTRFormatting(const FilePath
& path
,
358 string16
* rtl_safe_path
) {
359 // Wrap the overall path with LRE-PDF pair which essentialy marks the
360 // string as a Left-To-Right string.
361 // Inserting an LRE (Left-To-Right Embedding) mark as the first character.
362 rtl_safe_path
->push_back(kLeftToRightEmbeddingMark
);
363 #if defined(OS_MACOSX)
364 rtl_safe_path
->append(UTF8ToUTF16(path
.value()));
365 #elif defined(OS_WIN)
366 rtl_safe_path
->append(path
.value());
367 #else // defined(OS_POSIX) && !defined(OS_MACOSX)
368 std::wstring wide_path
= base::SysNativeMBToWide(path
.value());
369 rtl_safe_path
->append(WideToUTF16(wide_path
));
371 // Inserting a PDF (Pop Directional Formatting) mark as the last character.
372 rtl_safe_path
->push_back(kPopDirectionalFormatting
);
375 string16
GetDisplayStringInLTRDirectionality(const string16
& text
) {
376 // Always wrap the string in RTL UI (it may be appended to RTL string).
377 // Also wrap strings with an RTL first strong character direction in LTR UI.
378 if (IsRTL() || GetFirstStrongCharacterDirection(text
) == RIGHT_TO_LEFT
) {
379 string16
text_mutable(text
);
380 WrapStringWithLTRFormatting(&text_mutable
);
386 string16
StripWrappingBidiControlCharacters(const string16
& text
) {
389 size_t begin_index
= 0;
390 char16 begin
= text
[begin_index
];
391 if (begin
== kLeftToRightEmbeddingMark
||
392 begin
== kRightToLeftEmbeddingMark
||
393 begin
== kLeftToRightOverride
||
394 begin
== kRightToLeftOverride
)
396 size_t end_index
= text
.length() - 1;
397 if (text
[end_index
] == kPopDirectionalFormatting
)
399 return text
.substr(begin_index
, end_index
- begin_index
+ 1);