godot/core/translation.cpp

1176 lines
30 KiB
C++

/*************************************************************************/
/* translation.cpp */
/*************************************************************************/
/* This file is part of: */
/* GODOT ENGINE */
/* https://godotengine.org */
/*************************************************************************/
/* Copyright (c) 2007-2017 Juan Linietsky, Ariel Manzur. */
/* Copyright (c) 2014-2017 Godot Engine contributors (cf. AUTHORS.md) */
/* */
/* Permission is hereby granted, free of charge, to any person obtaining */
/* a copy of this software and associated documentation files (the */
/* "Software"), to deal in the Software without restriction, including */
/* without limitation the rights to use, copy, modify, merge, publish, */
/* distribute, sublicense, and/or sell copies of the Software, and to */
/* permit persons to whom the Software is furnished to do so, subject to */
/* the following conditions: */
/* */
/* The above copyright notice and this permission notice shall be */
/* included in all copies or substantial portions of the Software. */
/* */
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
/*************************************************************************/
#include "translation.h"
#include "io/resource_loader.h"
#include "os/os.h"
#include "project_settings.h"
static const char *locale_list[] = {
"aa", // Afar
"aa_DJ", // Afar (Djibouti)
"aa_ER", // Afar (Eritrea)
"aa_ET", // Afar (Ethiopia)
"af", // Afrikaans
"af_ZA", // Afrikaans (South Africa)
"agr_PE", // Aguaruna (Peru)
"ak_GH", // Akan (Ghana)
"am_ET", // Amharic (Ethiopia)
"an_ES", // Aragonese (Spain)
"anp_IN", // Angika (India)
"ar", // Arabic
"ar_AE", // Arabic (United Arab Emirates)
"ar_BH", // Arabic (Bahrain)
"ar_DZ", // Arabic (Algeria)
"ar_EG", // Arabic (Egypt)
"ar_IN", // Arabic (India)
"ar_IQ", // Arabic (Iraq)
"ar_JO", // Arabic (Jordan)
"ar_KW", // Arabic (Kuwait)
"ar_LB", // Arabic (Lebanon)
"ar_LY", // Arabic (Libya)
"ar_MA", // Arabic (Morocco)
"ar_OM", // Arabic (Oman)
"ar_QA", // Arabic (Qatar)
"ar_SA", // Arabic (Saudi Arabia)
"ar_SD", // Arabic (Sudan)
"ar_SS", // Arabic (South Soudan)
"ar_SY", // Arabic (Syria)
"ar_TN", // Arabic (Tunisia)
"ar_YE", // Arabic (Yemen)
"as_IN", // Assamese (India)
"ast_ES", // Asturian (Spain)
"ayc_PE", // Southern Aymara (Peru)
"ay_PE", // Aymara (Peru)
"az_AZ", // Azerbaijani (Azerbaijan)
"be", // Belarusian
"be_BY", // Belarusian (Belarus)
"bem_ZM", // Bemba (Zambia)
"ber_DZ", // Berber languages (Algeria)
"ber_MA", // Berber languages (Morocco)
"bg", // Bulgarian
"bg_BG", // Bulgarian (Bulgaria)
"bhb_IN", // Bhili (India)
"bho_IN", // Bhojpuri (India)
"bi_TV", // Bislama (Tuvalu)
"bn", // Bengali
"bn_BD", // Bengali (Bangladesh)
"bn_IN", // Bengali (India)
"bo", // Tibetan
"bo_CN", // Tibetan (China)
"bo_IN", // Tibetan (India)
"br_FR", // Breton (France)
"brx_IN", // Bodo (India)
"bs_BA", // Bosnian (Bosnia and Herzegovina)
"byn_ER", // Bilin (Eritrea)
"ca", // Catalan
"ca_AD", // Catalan (Andorra)
"ca_ES", // Catalan (Spain)
"ca_FR", // Catalan (France)
"ca_IT", // Catalan (Italy)
"ce_RU", // Chechen (Russia)
"chr_US", // Cherokee (United States)
"cmn_TW", // Mandarin Chinese (Taiwan)
"crh_UA", // Crimean Tatar (Ukraine)
"csb_PL", // Kashubian (Poland)
"cs", // Czech
"cs_CZ", // Czech (Czech Republic)
"cv_RU", // Chuvash (Russia)
"cy_GB", // Welsh (United Kingdom)
"da", // Danish
"da_DK", // Danish (Denmark)
"de", // German
"de_AT", // German (Austria)
"de_BE", // German (Belgium)
"de_CH", // German (Switzerland)
"de_DE", // German (Germany)
"de_IT", // German (Italy)
"de_LU", // German (Luxembourg)
"doi_IN", // Dogri (India)
"dv_MV", // Dhivehi (Maldives)
"dz_BT", // Dzongkha (Bhutan)
"el", // Greek
"el_CY", // Greek (Cyprus)
"el_GR", // Greek (Greece)
"en", // English
"en_AG", // English (Antigua and Barbuda)
"en_AU", // English (Australia)
"en_BW", // English (Botswana)
"en_CA", // English (Canada)
"en_DK", // English (Denmark)
"en_GB", // English (United Kingdom)
"en_HK", // English (Hong Kong)
"en_IE", // English (Ireland)
"en_IL", // English (Israel)
"en_IN", // English (India)
"en_NG", // English (Nigeria)
"en_NZ", // English (New Zealand)
"en_PH", // English (Philippines)
"en_SG", // English (Singapore)
"en_US", // English (United States)
"en_ZA", // English (South Africa)
"en_ZM", // English (Zambia)
"en_ZW", // English (Zimbabwe)
"eo", // Esperanto
"es", // Spanish
"es_AR", // Spanish (Argentina)
"es_BO", // Spanish (Bolivia)
"es_CL", // Spanish (Chile)
"es_CO", // Spanish (Colombia)
"es_CR", // Spanish (Costa Rica)
"es_CU", // Spanish (Cuba)
"es_DO", // Spanish (Dominican Republic)
"es_EC", // Spanish (Ecuador)
"es_ES", // Spanish (Spain)
"es_GT", // Spanish (Guatemala)
"es_HN", // Spanish (Honduras)
"es_MX", // Spanish (Mexico)
"es_NI", // Spanish (Nicaragua)
"es_PA", // Spanish (Panama)
"es_PE", // Spanish (Peru)
"es_PR", // Spanish (Puerto Rico)
"es_PY", // Spanish (Paraguay)
"es_SV", // Spanish (El Salvador)
"es_US", // Spanish (United States)
"es_UY", // Spanish (Uruguay)
"es_VE", // Spanish (Venezuela)
"et", // Estonian
"et_EE", // Estonian (Estonia)
"eu", // Basque
"eu_ES", // Basque (Spain)
"fa", // Persian
"fa_IR", // Persian (Iran)
"ff_SN", // Fulah (Senegal)
"fi", // Finnish
"fi_FI", // Finnish (Finland)
"fil_PH", // Filipino (Philippines)
"fo_FO", // Faroese (Faroe Islands)
"fr", // French
"fr_BE", // French (Belgium)
"fr_CA", // French (Canada)
"fr_CH", // French (Switzerland)
"fr_FR", // French (France)
"fr_LU", // French (Luxembourg)
"fur_IT", // Friulian (Italy)
"fy_DE", // Western Frisian (Germany)
"fy_NL", // Western Frisian (Netherlands)
"ga", // Irish
"ga_IE", // Irish (Ireland)
"gd_GB", // Scottish Gaelic (United Kingdom)
"gez_ER", // Geez (Eritrea)
"gez_ET", // Geez (Ethiopia)
"gl_ES", // Galician (Spain)
"gu_IN", // Gujarati (India)
"gv_GB", // Manx (United Kingdom)
"hak_TW", // Hakka Chinese (Taiwan)
"ha_NG", // Hausa (Nigeria)
"he", // Hebrew
"he_IL", // Hebrew (Israel)
"hi", // Hindi
"hi_IN", // Hindi (India)
"hne_IN", // Chhattisgarhi (India)
"hr", // Croatian
"hr_HR", // Croatian (Croatia)
"hsb_DE", // Upper Sorbian (Germany)
"ht_HT", // Haitian (Haiti)
"hu", // Hungarian
"hu_HU", // Hungarian (Hungary)
"hus_MX", // Huastec (Mexico)
"hy_AM", // Armenian (Armenia)
"ia_FR", // Interlingua (France)
"id", // Indonesian
"id_ID", // Indonesian (Indonesia)
"ig_NG", // Igbo (Nigeria)
"ik_CA", // Inupiaq (Canada)
"is", // Icelandic
"is_IS", // Icelandic (Iceland)
"it", // Italian
"it_CH", // Italian (Switzerland)
"it_IT", // Italian (Italy)
"iu_CA", // Inuktitut (Canada)
"ja", // Japanese
"ja_JP", // Japanese (Japan)
"kab_DZ", // Kabyle (Algeria)
"ka_GE", // Georgian (Georgia)
"kk_KZ", // Kazakh (Kazakhstan)
"kl_GL", // Kalaallisut (Greenland)
"km_KH", // Central Khmer (Cambodia)
"kn_IN", // Kannada (India)
"kok_IN", // Konkani (India)
"ko", // Korean
"ko_KR", // Korean (South Korea)
"ks_IN", // Kashmiri (India)
"ku", // Kurdish
"ku_TR", // Kurdish (Turkey)
"kw_GB", // Cornish (United Kingdom)
"ky_KG", // Kirghiz (Kyrgyzstan)
"lb_LU", // Luxembourgish (Luxembourg)
"lg_UG", // Ganda (Uganda)
"li_BE", // Limburgan (Belgium)
"li_NL", // Limburgan (Netherlands)
"lij_IT", // Ligurian (Italy)
"ln_CD", // Lingala (Congo)
"lo_LA", // Lao (Laos)
"lt", // Lithuanian
"lt_LT", // Lithuanian (Lithuania)
"lv", // Latvian
"lv_LV", // Latvian (Latvia)
"lzh_TW", // Literary Chinese (Taiwan)
"mag_IN", // Magahi (India)
"mai_IN", // Maithili (India)
"mg_MG", // Malagasy (Madagascar)
"mh_MH", // Marshallese (Marshall Islands)
"mhr_RU", // Eastern Mari (Russia)
"mi_NZ", // Maori (New Zealand)
"miq_NI", // Mískito (Nicaragua)
"mk", // Macedonian
"mk_MK", // Macedonian (Macedonia)
"ml_IN", // Malayalam (India)
"mni_IN", // Manipuri (India)
"mn_MN", // Mongolian (Mongolia)
"mr_IN", // Marathi (India)
"ms", // Malay
"ms_MY", // Malay (Malaysia)
"mt", // Maltese
"mt_MT", // Maltese (Malta)
"my_MM", // Burmese (Myanmar)
"myv_RU", // Erzya (Russia)
"nah_MX", // Nahuatl languages (Mexico)
"nan_TW", // Min Nan Chinese (Taiwan)
"nb", // Norwegian Bokmål
"nb_NO", // Norwegian Bokmål (Norway)
"nds_DE", // Low German (Germany)
"nds_NL", // Low German (Netherlands)
"ne_NP", // Nepali (Nepal)
"nhn_MX", // Central Nahuatl (Mexico)
"niu_NU", // Niuean (Niue)
"niu_NZ", // Niuean (New Zealand)
"nl", // Dutch
"nl_AW", // Dutch (Aruba)
"nl_BE", // Dutch (Belgium)
"nl_NL", // Dutch (Netherlands)
"nn", // Norwegian Nynorsk
"nn_NO", // Norwegian Nynorsk (Norway)
"nr_ZA", // South Ndebele (South Africa)
"nso_ZA", // Pedi (South Africa)
"oc_FR", // Occitan (France)
"om", // Oromo
"om_ET", // Oromo (Ethiopia)
"om_KE", // Oromo (Kenya)
"or_IN", // Oriya (India)
"os_RU", // Ossetian (Russia)
"pa_IN", // Panjabi (India)
"pap", // Papiamento
"pap_AN", // Papiamento (Netherlands Antilles)
"pap_AW", // Papiamento (Aruba)
"pap_CW", // Papiamento (Curaçao)
"pa_PK", // Panjabi (Pakistan)
"pl", // Polish
"pl_PL", // Polish (Poland)
"pr", // Pirate
"ps_AF", // Pushto (Afghanistan)
"pt", // Portuguese
"pt_BR", // Portuguese (Brazil)
"pt_PT", // Portuguese (Portugal)
"quy_PE", // Ayacucho Quechua (Peru)
"quz_PE", // Cusco Quechua (Peru)
"raj_IN", // Rajasthani (India)
"ro", // Romanian
"ro_RO", // Romanian (Romania)
"ru", // Russian
"ru_RU", // Russian (Russia)
"ru_UA", // Russian (Ukraine)
"rw_RW", // Kinyarwanda (Rwanda)
"sa_IN", // Sanskrit (India)
"sat_IN", // Santali (India)
"sc_IT", // Sardinian (Italy)
"sco", // Scots
"sd_IN", // Sindhi (India)
"se_NO", // Northern Sami (Norway)
"sgs_LT", // Samogitian (Lithuania)
"shs_CA", // Shuswap (Canada)
"sid_ET", // Sidamo (Ethiopia)
"si_LK", // Sinhala (Sri Lanka)
"sk", // Slovak
"sk_SK", // Slovak (Slovakia)
"sl", // Slovenian
"sl_SI", // Slovenian (Slovenia)
"so", // Somali
"so_DJ", // Somali (Djibouti)
"so_ET", // Somali (Ethiopia)
"so_KE", // Somali (Kenya)
"so_SO", // Somali (Somalia)
"son_ML", // Songhai languages (Mali)
"sq", // Albanian
"sq_AL", // Albanian (Albania)
"sq_KV", // Albanian (Kosovo)
"sq_MK", // Albanian (Macedonia)
"sr", // Serbian
"sr_Cyrl", // Serbian (Cyrillic)
"sr_ME", // Serbian (Montenegro)
"sr_RS", // Serbian (Serbia)
"ss_ZA", // Swati (South Africa)
"st_ZA", // Southern Sotho (South Africa)
"sv", // Swedish
"sv_FI", // Swedish (Finland)
"sv_SE", // Swedish (Sweden)
"sw_KE", // Swahili (Kenya)
"sw_TZ", // Swahili (Tanzania)
"szl_PL", // Silesian (Poland)
"ta", // Tamil
"ta_IN", // Tamil (India)
"ta_LK", // Tamil (Sri Lanka)
"tcy_IN", // Tulu (India)
"te_IN", // Telugu (India)
"tg_TJ", // Tajik (Tajikistan)
"the_NP", // Chitwania Tharu (Nepal)
"th", // Thai
"th_TH", // Thai (Thailand)
"ti", // Tigrinya
"ti_ER", // Tigrinya (Eritrea)
"ti_ET", // Tigrinya (Ethiopia)
"tig_ER", // Tigre (Eritrea)
"tk_TM", // Turkmen (Turkmenistan)
"tl_PH", // Tagalog (Philippines)
"tn_ZA", // Tswana (South Africa)
"tr", // Turkish
"tr_CY", // Turkish (Cyprus)
"tr_TR", // Turkish (Turkey)
"ts_ZA", // Tsonga (South Africa)
"tt_RU", // Tatar (Russia)
"ug_CN", // Uighur (China)
"uk", // Ukrainian
"uk_UA", // Ukrainian (Ukraine)
"unm_US", // Unami (United States)
"ur", // Urdu
"ur_IN", // Urdu (India)
"ur_PK", // Urdu (Pakistan)
"uz", // Uzbek
"uz_UZ", // Uzbek (Uzbekistan)
"ve_ZA", // Venda (South Africa)
"vi", // Vietnamese
"vi_VN", // Vietnamese (Vietnam)
"wa_BE", // Walloon (Belgium)
"wae_CH", // Walser (Switzerland)
"wal_ET", // Wolaytta (Ethiopia)
"wo_SN", // Wolof (Senegal)
"xh_ZA", // Xhosa (South Africa)
"yi_US", // Yiddish (United States)
"yo_NG", // Yoruba (Nigeria)
"yue_HK", // Yue Chinese (Hong Kong)
"zh", // Chinese
"zh_CN", // Chinese (China)
"zh_HK", // Chinese (Hong Kong)
"zh_SG", // Chinese (Singapore)
"zh_TW", // Chinese (Taiwan)
"zu_ZA", // Zulu (South Africa)
0
};
static const char *locale_names[] = {
"Afar",
"Afar (Djibouti)",
"Afar (Eritrea)",
"Afar (Ethiopia)",
"Afrikaans",
"Afrikaans (South Africa)",
"Aguaruna (Peru)",
"Akan (Ghana)",
"Amharic (Ethiopia)",
"Aragonese (Spain)",
"Angika (India)",
"Arabic",
"Arabic (United Arab Emirates)",
"Arabic (Bahrain)",
"Arabic (Algeria)",
"Arabic (Egypt)",
"Arabic (India)",
"Arabic (Iraq)",
"Arabic (Jordan)",
"Arabic (Kuwait)",
"Arabic (Lebanon)",
"Arabic (Libya)",
"Arabic (Morocco)",
"Arabic (Oman)",
"Arabic (Qatar)",
"Arabic (Saudi Arabia)",
"Arabic (Sudan)",
"Arabic (South Soudan)",
"Arabic (Syria)",
"Arabic (Tunisia)",
"Arabic (Yemen)",
"Assamese (India)",
"Asturian (Spain)",
"Southern Aymara (Peru)",
"Aymara (Peru)",
"Azerbaijani (Azerbaijan)",
"Belarusian",
"Belarusian (Belarus)",
"Bemba (Zambia)",
"Berber languages (Algeria)",
"Berber languages (Morocco)",
"Bulgarian",
"Bulgarian (Bulgaria)",
"Bhili (India)",
"Bhojpuri (India)",
"Bislama (Tuvalu)",
"Bengali",
"Bengali (Bangladesh)",
"Bengali (India)",
"Tibetan",
"Tibetan (China)",
"Tibetan (India)",
"Breton (France)",
"Bodo (India)",
"Bosnian (Bosnia and Herzegovina)",
"Bilin (Eritrea)",
"Catalan",
"Catalan (Andorra)",
"Catalan (Spain)",
"Catalan (France)",
"Catalan (Italy)",
"Chechen (Russia)",
"Cherokee (United States)",
"Mandarin Chinese (Taiwan)",
"Crimean Tatar (Ukraine)",
"Kashubian (Poland)",
"Czech",
"Czech (Czech Republic)",
"Chuvash (Russia)",
"Welsh (United Kingdom)",
"Danish",
"Danish (Denmark)",
"German",
"German (Austria)",
"German (Belgium)",
"German (Switzerland)",
"German (Germany)",
"German (Italy)",
"German (Luxembourg)",
"Dogri (India)",
"Dhivehi (Maldives)",
"Dzongkha (Bhutan)",
"Greek",
"Greek (Cyprus)",
"Greek (Greece)",
"English",
"English (Antigua and Barbuda)",
"English (Australia)",
"English (Botswana)",
"English (Canada)",
"English (Denmark)",
"English (United Kingdom)",
"English (Hong Kong)",
"English (Ireland)",
"English (Israel)",
"English (India)",
"English (Nigeria)",
"English (New Zealand)",
"English (Philippines)",
"English (Singapore)",
"English (United States)",
"English (South Africa)",
"English (Zambia)",
"English (Zimbabwe)",
"Esperanto",
"Spanish",
"Spanish (Argentina)",
"Spanish (Bolivia)",
"Spanish (Chile)",
"Spanish (Colombia)",
"Spanish (Costa Rica)",
"Spanish (Cuba)",
"Spanish (Dominican Republic)",
"Spanish (Ecuador)",
"Spanish (Spain)",
"Spanish (Guatemala)",
"Spanish (Honduras)",
"Spanish (Mexico)",
"Spanish (Nicaragua)",
"Spanish (Panama)",
"Spanish (Peru)",
"Spanish (Puerto Rico)",
"Spanish (Paraguay)",
"Spanish (El Salvador)",
"Spanish (United States)",
"Spanish (Uruguay)",
"Spanish (Venezuela)",
"Estonian",
"Estonian (Estonia)",
"Basque",
"Basque (Spain)",
"Persian",
"Persian (Iran)",
"Fulah (Senegal)",
"Finnish",
"Finnish (Finland)",
"Filipino (Philippines)",
"Faroese (Faroe Islands)",
"French",
"French (Belgium)",
"French (Canada)",
"French (Switzerland)",
"French (France)",
"French (Luxembourg)",
"Friulian (Italy)",
"Western Frisian (Germany)",
"Western Frisian (Netherlands)",
"Irish",
"Irish (Ireland)",
"Scottish Gaelic (United Kingdom)",
"Geez (Eritrea)",
"Geez (Ethiopia)",
"Galician (Spain)",
"Gujarati (India)",
"Manx (United Kingdom)",
"Hakka Chinese (Taiwan)",
"Hausa (Nigeria)",
"Hebrew",
"Hebrew (Israel)",
"Hindi",
"Hindi (India)",
"Chhattisgarhi (India)",
"Croatian",
"Croatian (Croatia)",
"Upper Sorbian (Germany)",
"Haitian (Haiti)",
"Hungarian",
"Hungarian (Hungary)",
"Huastec (Mexico)",
"Armenian (Armenia)",
"Interlingua (France)",
"Indonesian",
"Indonesian (Indonesia)",
"Igbo (Nigeria)",
"Inupiaq (Canada)",
"Icelandic",
"Icelandic (Iceland)",
"Italian",
"Italian (Switzerland)",
"Italian (Italy)",
"Inuktitut (Canada)",
"Japanese",
"Japanese (Japan)",
"Kabyle (Algeria)",
"Georgian (Georgia)",
"Kazakh (Kazakhstan)",
"Kalaallisut (Greenland)",
"Central Khmer (Cambodia)",
"Kannada (India)",
"Konkani (India)",
"Korean",
"Korean (South Korea)",
"Kashmiri (India)",
"Kurdish",
"Kurdish (Turkey)",
"Cornish (United Kingdom)",
"Kirghiz (Kyrgyzstan)",
"Luxembourgish (Luxembourg)",
"Ganda (Uganda)",
"Limburgan (Belgium)",
"Limburgan (Netherlands)",
"Ligurian (Italy)",
"Lingala (Congo)",
"Lao (Laos)",
"Lithuanian",
"Lithuanian (Lithuania)",
"Latvian",
"Latvian (Latvia)",
"Literary Chinese (Taiwan)",
"Magahi (India)",
"Maithili (India)",
"Malagasy (Madagascar)",
"Marshallese (Marshall Islands)",
"Eastern Mari (Russia)",
"Maori (New Zealand)",
"Mískito (Nicaragua)",
"Macedonian",
"Macedonian (Macedonia)",
"Malayalam (India)",
"Manipuri (India)",
"Mongolian (Mongolia)",
"Marathi (India)",
"Malay",
"Malay (Malaysia)",
"Maltese",
"Maltese (Malta)",
"Burmese (Myanmar)",
"Erzya (Russia)",
"Nahuatl languages (Mexico)",
"Min Nan Chinese (Taiwan)",
"Norwegian Bokmål",
"Norwegian Bokmål (Norway)",
"Low German (Germany)",
"Low German (Netherlands)",
"Nepali (Nepal)",
"Central Nahuatl (Mexico)",
"Niuean (Niue)",
"Niuean (New Zealand)",
"Dutch",
"Dutch (Aruba)",
"Dutch (Belgium)",
"Dutch (Netherlands)",
"Norwegian Nynorsk",
"Norwegian Nynorsk (Norway)",
"South Ndebele (South Africa)",
"Pedi (South Africa)",
"Occitan (France)",
"Oromo",
"Oromo (Ethiopia)",
"Oromo (Kenya)",
"Oriya (India)",
"Ossetian (Russia)",
"Panjabi (India)",
"Papiamento",
"Papiamento (Netherlands Antilles)",
"Papiamento (Aruba)",
"Papiamento (Curaçao)",
"Panjabi (Pakistan)",
"Polish",
"Polish (Poland)",
"Pirate",
"Pushto (Afghanistan)",
"Portuguese",
"Portuguese (Brazil)",
"Portuguese (Portugal)",
"Ayacucho Quechua (Peru)",
"Cusco Quechua (Peru)",
"Rajasthani (India)",
"Romanian",
"Romanian (Romania)",
"Russian",
"Russian (Russia)",
"Russian (Ukraine)",
"Kinyarwanda (Rwanda)",
"Sanskrit (India)",
"Santali (India)",
"Sardinian (Italy)",
"Scots (Scotland)",
"Sindhi (India)",
"Northern Sami (Norway)",
"Samogitian (Lithuania)",
"Shuswap (Canada)",
"Sidamo (Ethiopia)",
"Sinhala (Sri Lanka)",
"Slovak",
"Slovak (Slovakia)",
"Slovenian",
"Slovenian (Slovenia)",
"Somali",
"Somali (Djibouti)",
"Somali (Ethiopia)",
"Somali (Kenya)",
"Somali (Somalia)",
"Songhai languages (Mali)",
"Albanian",
"Albanian (Albania)",
"Albanian (Kosovo)",
"Albanian (Macedonia)",
"Serbian",
"Serbian (Cyrillic)",
"Serbian (Montenegro)",
"Serbian (Serbia)",
"Swati (South Africa)",
"Southern Sotho (South Africa)",
"Swedish",
"Swedish (Finland)",
"Swedish (Sweden)",
"Swahili (Kenya)",
"Swahili (Tanzania)",
"Silesian (Poland)",
"Tamil",
"Tamil (India)",
"Tamil (Sri Lanka)",
"Tulu (India)",
"Telugu (India)",
"Tajik (Tajikistan)",
"Chitwania Tharu (Nepal)",
"Thai",
"Thai (Thailand)",
"Tigrinya",
"Tigrinya (Eritrea)",
"Tigrinya (Ethiopia)",
"Tigre (Eritrea)",
"Turkmen (Turkmenistan)",
"Tagalog (Philippines)",
"Tswana (South Africa)",
"Turkish",
"Turkish (Cyprus)",
"Turkish (Turkey)",
"Tsonga (South Africa)",
"Tatar (Russia)",
"Uighur (China)",
"Ukrainian",
"Ukrainian (Ukraine)",
"Unami (United States)",
"Urdu",
"Urdu (India)",
"Urdu (Pakistan)",
"Uzbek",
"Uzbek (Uzbekistan)",
"Venda (South Africa)",
"Vietnamese",
"Vietnamese (Vietnam)",
"Walloon (Belgium)",
"Walser (Switzerland)",
"Wolaytta (Ethiopia)",
"Wolof (Senegal)",
"Xhosa (South Africa)",
"Yiddish (United States)",
"Yoruba (Nigeria)",
"Yue Chinese (Hong Kong)",
"Chinese",
"Chinese (China)",
"Chinese (Hong Kong)",
"Chinese (Singapore)",
"Chinese (Taiwan)",
"Zulu (South Africa)",
0
};
static const char *locale_renames[][2] = {
{ "no", "nb" },
{ NULL, NULL }
};
static String get_trimmed_locale(const String &p_locale) {
return p_locale.substr(0, 2);
}
///////////////////////////////////////////////
PoolVector<String> Translation::_get_messages() const {
PoolVector<String> msgs;
msgs.resize(translation_map.size() * 2);
int idx = 0;
for (const Map<StringName, StringName>::Element *E = translation_map.front(); E; E = E->next()) {
msgs.set(idx + 0, E->key());
msgs.set(idx + 1, E->get());
idx += 2;
}
return msgs;
}
PoolVector<String> Translation::_get_message_list() const {
PoolVector<String> msgs;
msgs.resize(translation_map.size());
int idx = 0;
for (const Map<StringName, StringName>::Element *E = translation_map.front(); E; E = E->next()) {
msgs.set(idx, E->key());
idx += 1;
}
return msgs;
}
void Translation::_set_messages(const PoolVector<String> &p_messages) {
int msg_count = p_messages.size();
ERR_FAIL_COND(msg_count % 2);
PoolVector<String>::Read r = p_messages.read();
for (int i = 0; i < msg_count; i += 2) {
add_message(r[i + 0], r[i + 1]);
}
}
void Translation::set_locale(const String &p_locale) {
String univ_locale = TranslationServer::standardize_locale(p_locale);
if (!TranslationServer::is_locale_valid(univ_locale)) {
String trimmed_locale = get_trimmed_locale(univ_locale);
ERR_EXPLAIN("Invalid locale: " + trimmed_locale);
ERR_FAIL_COND(!TranslationServer::is_locale_valid(trimmed_locale));
locale = trimmed_locale;
} else {
locale = univ_locale;
}
if (OS::get_singleton()->get_main_loop()) {
OS::get_singleton()->get_main_loop()->notification(MainLoop::NOTIFICATION_TRANSLATION_CHANGED);
}
}
void Translation::add_message(const StringName &p_src_text, const StringName &p_xlated_text) {
translation_map[p_src_text] = p_xlated_text;
}
StringName Translation::get_message(const StringName &p_src_text) const {
const Map<StringName, StringName>::Element *E = translation_map.find(p_src_text);
if (!E)
return StringName();
return E->get();
}
void Translation::erase_message(const StringName &p_src_text) {
translation_map.erase(p_src_text);
}
void Translation::get_message_list(List<StringName> *r_messages) const {
for (const Map<StringName, StringName>::Element *E = translation_map.front(); E; E = E->next()) {
r_messages->push_back(E->key());
}
}
int Translation::get_message_count() const {
return translation_map.size();
};
void Translation::_bind_methods() {
ClassDB::bind_method(D_METHOD("set_locale", "locale"), &Translation::set_locale);
ClassDB::bind_method(D_METHOD("get_locale"), &Translation::get_locale);
ClassDB::bind_method(D_METHOD("add_message", "src_message", "xlated_message"), &Translation::add_message);
ClassDB::bind_method(D_METHOD("get_message", "src_message"), &Translation::get_message);
ClassDB::bind_method(D_METHOD("erase_message", "src_message"), &Translation::erase_message);
ClassDB::bind_method(D_METHOD("get_message_list"), &Translation::_get_message_list);
ClassDB::bind_method(D_METHOD("get_message_count"), &Translation::get_message_count);
ClassDB::bind_method(D_METHOD("_set_messages"), &Translation::_set_messages);
ClassDB::bind_method(D_METHOD("_get_messages"), &Translation::_get_messages);
ADD_PROPERTY(PropertyInfo(Variant::POOL_STRING_ARRAY, "messages", PROPERTY_HINT_NONE, "", PROPERTY_USAGE_NOEDITOR), "_set_messages", "_get_messages");
ADD_PROPERTY(PropertyInfo(Variant::STRING, "locale"), "set_locale", "get_locale");
}
Translation::Translation() :
locale("en") {
}
///////////////////////////////////////////////
bool TranslationServer::is_locale_valid(const String &p_locale) {
const char **ptr = locale_list;
while (*ptr) {
if (*ptr == p_locale)
return true;
ptr++;
}
return false;
}
String TranslationServer::standardize_locale(const String &p_locale) {
// Replaces '-' with '_' for macOS Sierra-style locales
String univ_locale = p_locale.replace("-", "_");
// Handles known non-ISO locale names used e.g. on Windows
int idx = 0;
while (locale_renames[idx][0] != NULL) {
if (locale_renames[idx][0] == univ_locale) {
univ_locale = locale_renames[idx][1];
break;
}
idx++;
}
return univ_locale;
}
void TranslationServer::set_locale(const String &p_locale) {
String univ_locale = standardize_locale(p_locale);
if (!is_locale_valid(univ_locale)) {
String trimmed_locale = get_trimmed_locale(univ_locale);
ERR_EXPLAIN("Invalid locale: " + trimmed_locale);
ERR_FAIL_COND(!is_locale_valid(trimmed_locale));
locale = trimmed_locale;
} else {
locale = univ_locale;
}
if (OS::get_singleton()->get_main_loop()) {
OS::get_singleton()->get_main_loop()->notification(MainLoop::NOTIFICATION_TRANSLATION_CHANGED);
}
ResourceLoader::reload_translation_remaps();
}
String TranslationServer::get_locale() const {
return locale;
}
String TranslationServer::get_locale_name(const String &p_locale) const {
if (!locale_name_map.has(p_locale)) return String();
return locale_name_map[p_locale];
}
Vector<String> TranslationServer::get_all_locales() {
Vector<String> locales;
const char **ptr = locale_list;
while (*ptr) {
locales.push_back(*ptr);
ptr++;
}
return locales;
}
Vector<String> TranslationServer::get_all_locale_names() {
Vector<String> locales;
const char **ptr = locale_names;
while (*ptr) {
locales.push_back(String::utf8(*ptr));
ptr++;
}
return locales;
}
void TranslationServer::add_translation(const Ref<Translation> &p_translation) {
translations.insert(p_translation);
}
void TranslationServer::remove_translation(const Ref<Translation> &p_translation) {
translations.erase(p_translation);
}
void TranslationServer::clear() {
translations.clear();
};
StringName TranslationServer::translate(const StringName &p_message) const {
//translate using locale
if (!enabled)
return p_message;
StringName res;
bool near_match = false;
const CharType *lptr = &locale[0];
for (const Set<Ref<Translation> >::Element *E = translations.front(); E; E = E->next()) {
const Ref<Translation> &t = E->get();
String l = t->get_locale();
if (lptr[0] != l[0] || lptr[1] != l[1])
continue; // locale not match
//near match
bool match = (l != locale);
if (near_match && !match)
continue; //only near-match once
StringName r = t->get_message(p_message);
if (!r)
continue;
res = r;
if (match)
break;
else
near_match = true;
}
if (!res) {
//try again with fallback
if (fallback.length() >= 2) {
const CharType *fptr = &fallback[0];
bool near_match = false;
for (const Set<Ref<Translation> >::Element *E = translations.front(); E; E = E->next()) {
const Ref<Translation> &t = E->get();
String l = t->get_locale();
if (fptr[0] != l[0] || fptr[1] != l[1])
continue; // locale not match
//near match
bool match = (l != fallback);
if (near_match && !match)
continue; //only near-match once
StringName r = t->get_message(p_message);
if (!r)
continue;
res = r;
if (match)
break;
else
near_match = true;
}
}
}
if (!res)
return p_message;
return res;
}
TranslationServer *TranslationServer::singleton = NULL;
bool TranslationServer::_load_translations(const String &p_from) {
if (ProjectSettings::get_singleton()->has_setting(p_from)) {
PoolVector<String> translations = ProjectSettings::get_singleton()->get(p_from);
int tcount = translations.size();
if (tcount) {
PoolVector<String>::Read r = translations.read();
for (int i = 0; i < tcount; i++) {
//print_line( "Loading translation from " + r[i] );
Ref<Translation> tr = ResourceLoader::load(r[i]);
if (tr.is_valid())
add_translation(tr);
}
}
return true;
}
return false;
}
void TranslationServer::setup() {
String test = GLOBAL_DEF("locale/test", "");
test = test.strip_edges();
if (test != "")
set_locale(test);
else
set_locale(OS::get_singleton()->get_locale());
fallback = GLOBAL_DEF("locale/fallback", "en");
#ifdef TOOLS_ENABLED
{
String options = "";
int idx = 0;
while (locale_list[idx]) {
if (idx > 0)
options += ",";
options += locale_list[idx];
idx++;
}
ProjectSettings::get_singleton()->set_custom_property_info("locale/fallback", PropertyInfo(Variant::STRING, "locale/fallback", PROPERTY_HINT_ENUM, options));
}
#endif
//load translations
}
void TranslationServer::set_tool_translation(const Ref<Translation> &p_translation) {
tool_translation = p_translation;
}
StringName TranslationServer::tool_translate(const StringName &p_message) const {
if (tool_translation.is_valid()) {
StringName r = tool_translation->get_message(p_message);
if (r) {
return r;
}
}
return p_message;
}
void TranslationServer::_bind_methods() {
ClassDB::bind_method(D_METHOD("set_locale", "locale"), &TranslationServer::set_locale);
ClassDB::bind_method(D_METHOD("get_locale"), &TranslationServer::get_locale);
ClassDB::bind_method(D_METHOD("get_locale_name", "locale"), &TranslationServer::get_locale_name);
ClassDB::bind_method(D_METHOD("translate", "message"), &TranslationServer::translate);
ClassDB::bind_method(D_METHOD("add_translation", "translation"), &TranslationServer::add_translation);
ClassDB::bind_method(D_METHOD("remove_translation", "translation"), &TranslationServer::remove_translation);
ClassDB::bind_method(D_METHOD("clear"), &TranslationServer::clear);
}
void TranslationServer::load_translations() {
String locale = get_locale();
bool found = _load_translations("locale/translations"); //all
if (_load_translations("locale/translations_" + locale.substr(0, 2)))
found = true;
if (locale.substr(0, 2) != locale) {
if (_load_translations("locale/translations_" + locale))
found = true;
}
}
TranslationServer::TranslationServer() :
locale("en"),
enabled(true) {
singleton = this;
for (int i = 0; locale_list[i]; ++i) {
locale_name_map.insert(locale_list[i], String::utf8(locale_names[i]));
}
}