Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
99.67% |
301 / 302 |
|
87.50% |
7 / 8 |
CRAP | |
0.00% |
0 / 1 |
| CrhExceptions | |
99.67% |
301 / 302 |
|
87.50% |
7 / 8 |
17 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| initLcUc | |
90.00% |
9 / 10 |
|
0.00% |
0 / 1 |
4.02 | |||
| myLc | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| myUc | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| myUcWord | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| addMappings | |
100.00% |
16 / 16 |
|
100.00% |
1 / 1 |
7 | |||
| loadExceptions | |
100.00% |
15 / 15 |
|
100.00% |
1 / 1 |
1 | |||
| loadRegs | |
100.00% |
257 / 257 |
|
100.00% |
1 / 1 |
1 | |||
| 1 | <?php |
| 2 | namespace MediaWiki\Languages\Data; |
| 3 | |
| 4 | use CrhConverter as Crh; |
| 5 | |
| 6 | /** |
| 7 | * Exceptions Tables for Crimean Tatar (crh / Qırımtatarca) |
| 8 | * |
| 9 | * Adapted from https://crh.wikipedia.org/wiki/Qullan%C4%B1c%C4%B1:Don_Alessandro/Translit |
| 10 | * |
| 11 | * @ingroup Language |
| 12 | */ |
| 13 | class CrhExceptions { |
| 14 | |
| 15 | private const WB = '\b'; # default word boundary; may be updated in the future |
| 16 | |
| 17 | public function __construct() { |
| 18 | $this->loadRegs(); |
| 19 | } |
| 20 | |
| 21 | public $Cyrl2LatnExceptions = []; |
| 22 | public $Latn2CyrlExceptions = []; |
| 23 | |
| 24 | public $Cyrl2LatnPatterns = []; |
| 25 | public $Latn2CyrlPatterns = []; |
| 26 | |
| 27 | private $lc2uc; |
| 28 | private $uc2lc; |
| 29 | |
| 30 | private function initLcUc( $lcChars, $ucChars, $reinit = false ) { |
| 31 | # bail if we've already done this, unless we are re-initializing |
| 32 | if ( !$reinit && $this->lc2uc && $this->uc2lc ) { |
| 33 | return; |
| 34 | } |
| 35 | |
| 36 | # split up the lc and uc lists in a unicode-friendly way |
| 37 | $myLc = []; |
| 38 | preg_match_all( '/./u', $lcChars, $myLc ); |
| 39 | $myLc = $myLc[0]; |
| 40 | |
| 41 | $myUc = []; |
| 42 | preg_match_all( '/./u', $ucChars, $myUc ); |
| 43 | $myUc = $myUc[0]; |
| 44 | |
| 45 | # map lc to uc and vice versa |
| 46 | $this->lc2uc = array_combine( array_values( $myLc ), array_values( $myUc ) ); |
| 47 | $this->uc2lc = array_combine( array_values( $myUc ), array_values( $myLc ) ); |
| 48 | } |
| 49 | |
| 50 | private function myLc( $string ) { |
| 51 | return strtr( $string, $this->uc2lc ); |
| 52 | } |
| 53 | |
| 54 | private function myUc( $string ) { |
| 55 | return strtr( $string, $this->lc2uc ); |
| 56 | } |
| 57 | |
| 58 | private function myUcWord( $string ) { |
| 59 | return $this->myUc( mb_substr( $string, 0, 1 ) ) . $this->myLc( mb_substr( $string, 1 ) ); |
| 60 | } |
| 61 | |
| 62 | private function addMappings( $mapArray, &$A2B, &$B2A, $exactCase = false, |
| 63 | $prePat = '', $postPat = '' ) { |
| 64 | foreach ( $mapArray as $WordA => $WordB ) { |
| 65 | if ( !$exactCase ) { |
| 66 | $ucA = $this->myUc( $WordA ); |
| 67 | $ucWordA = $this->myUcWord( $WordA ); |
| 68 | $ucB = $this->myUc( $WordB ); |
| 69 | $ucWordB = $this->myUcWord( $WordB ); |
| 70 | } |
| 71 | |
| 72 | # if there are regexes, only map toward backregs |
| 73 | if ( !preg_match( '/\$[1-9]/', $WordA ) ) { |
| 74 | $A2B[ $prePat . $WordA . $postPat ] = $WordB; |
| 75 | if ( !$exactCase ) { |
| 76 | // @phan-suppress-next-line PhanPossiblyUndeclaredVariable uc vars are set when used |
| 77 | $A2B[ $prePat . $ucWordA . $postPat ] = $ucWordB; |
| 78 | // @phan-suppress-next-line PhanPossiblyUndeclaredVariable uc vars are set when used |
| 79 | $A2B[ $prePat . $ucA . $postPat ] = $ucB; |
| 80 | } |
| 81 | } |
| 82 | |
| 83 | if ( !preg_match( '/\$[1-9]/', $WordB ) ) { |
| 84 | $B2A[ $prePat . $WordB . $postPat ] = $WordA; |
| 85 | if ( !$exactCase ) { |
| 86 | // @phan-suppress-next-line PhanPossiblyUndeclaredVariable uc vars are set when used |
| 87 | $B2A[ $prePat . $ucWordB . $postPat ] = $ucWordA; |
| 88 | // @phan-suppress-next-line PhanPossiblyUndeclaredVariable uc vars are set when used |
| 89 | $B2A[ $prePat . $ucB . $postPat ] = $ucA; |
| 90 | } |
| 91 | } |
| 92 | } |
| 93 | } |
| 94 | |
| 95 | public function loadExceptions( $lcChars, $ucChars ) { |
| 96 | # init lc and uc, as needed |
| 97 | $this->initLcUc( $lcChars, $ucChars ); |
| 98 | |
| 99 | # no regex prefix/suffix needed |
| 100 | $this->addMappings( $this->ManyToOneC2LMappings, |
| 101 | // reverse exception mapping order to handle many-to-one C2L mappings |
| 102 | $this->Latn2CyrlExceptions, $this->Cyrl2LatnExceptions ); |
| 103 | $this->addMappings( $this->multiCaseMappings, |
| 104 | $this->Cyrl2LatnExceptions, $this->Latn2CyrlExceptions ); |
| 105 | $this->addMappings( $this->exactCaseMappings, |
| 106 | $this->Cyrl2LatnExceptions, $this->Latn2CyrlExceptions, true ); |
| 107 | |
| 108 | # load C2L and L2C bidirectional affix mappings |
| 109 | $this->addMappings( $this->prefixMapping, |
| 110 | $this->Cyrl2LatnPatterns, $this->Latn2CyrlPatterns, false, '/' . self::WB, '/u' ); |
| 111 | $this->addMappings( $this->suffixMapping, |
| 112 | $this->Cyrl2LatnPatterns, $this->Latn2CyrlPatterns, false, '/', self::WB . '/u' ); |
| 113 | |
| 114 | # tack on one-way mappings to the ends of the prefix and suffix patterns |
| 115 | $this->Cyrl2LatnPatterns += $this->Cyrl2LatnRegexes; |
| 116 | $this->Latn2CyrlPatterns += $this->Latn2CyrlRegexes; |
| 117 | |
| 118 | return [ $this->Cyrl2LatnExceptions, $this->Latn2CyrlExceptions, $this->Cyrl2LatnPatterns, |
| 119 | $this->Latn2CyrlPatterns, $this->CyrlCleanUpRegexes ]; |
| 120 | } |
| 121 | |
| 122 | /** |
| 123 | * @var string[] map Latin to Cyrillic and back, simple string match only (no regex) |
| 124 | * variants: all lowercase, all uppercase, first letter capitalized |
| 125 | */ |
| 126 | private $ManyToOneC2LMappings = [ |
| 127 | # Carefully ordered many-to-one mapping. |
| 128 | # These are ordered so that the C2L is correct (the later Latin one). |
| 129 | # See also the L2C mappings below |
| 130 | 'fevqülade' => 'февкъульаде', 'fevqulade' => 'февкъульаде', |
| 131 | 'beyude' => 'бейуде', 'beyüde' => 'бейуде', |
| 132 | 'curat' => 'джурьат', 'cürat' => 'джурьат', |
| 133 | 'mesul' => 'месуль', 'mesül' => 'месуль', |
| 134 | ]; |
| 135 | |
| 136 | /** |
| 137 | * @var string[] map Cyrillic to Latin and back, simple string match only (no regex) |
| 138 | * variants: all lowercase, all uppercase, first letter capitalized |
| 139 | */ |
| 140 | private $multiCaseMappings = [ |
| 141 | |
| 142 | #### Cyrillic to Latin |
| 143 | 'аджыумер' => 'acıümer', 'аджыусеин' => 'acıüsein', 'алейкум' => 'aleyküm', |
| 144 | 'бозтюс' => 'boztüs', 'боливия' => 'boliviya', 'большевик' => 'bolşevik', 'борис' => 'boris', |
| 145 | 'борнен' => 'bornen', 'бублик' => 'bublik', 'буддизм' => 'buddizm', 'буддист' => 'buddist', |
| 146 | 'буженина' => 'bujenina', 'бузкесен' => 'buzkesen', 'букинист' => 'bukinist', |
| 147 | 'буксир' => 'buksir', 'бульбул' => 'bülbül', 'бульвар' => 'bulvar', 'бульдог' => 'buldog', |
| 148 | 'бульдозер' => 'buldozer', 'бульон' => 'bulyon', 'бумеранг' => 'bumerang', |
| 149 | 'бунен' => 'bunen', 'буннен' => 'bunnen', 'бус-бутюн' => 'büs-bütün', |
| 150 | 'бутерброд' => 'buterbrod', 'бутилен' => 'butilen', 'бутилир' => 'butilir', |
| 151 | 'буфер' => 'bufer', 'буфет' => 'bufet', 'гобелен' => 'gobelen', 'гомео' => 'gomeo', |
| 152 | 'горизонт' => 'gorizont', 'госпитал' => 'gospital', 'готтентот' => 'gottentot', |
| 153 | 'гофрир' => 'gofrir', 'губерн' => 'gubern', 'гуверн' => 'guvern', 'гугенот' => 'gugenot', |
| 154 | 'гуливер' => 'guliver', 'гуна' => 'güna', 'гунях' => 'günâh', 'гургуль' => 'gürgül', |
| 155 | 'гуя' => 'güya', 'дёрткуль' => 'dörtkül', 'джуньджу' => 'cüncü', 'ёлнен' => 'yolnen', |
| 156 | 'зумбуль' => 'zümbül', 'ильи' => 'ilyi', 'ишунь' => 'işün', 'ковер' => 'kover', 'код' => 'kod', |
| 157 | 'койлю' => 'köylü', 'кокагъач' => 'kökağaç', 'кокбаштанкъара' => 'kökbaştanqara', |
| 158 | 'кокгогерджин' => 'kökgögercin', 'кокдогъан' => 'kökdoğan', 'коккозю' => 'kökközü', |
| 159 | 'коккъузгъун' => 'kökquzğun', 'коклюш' => 'koklüş', 'кокташ' => 'köktaş', |
| 160 | 'коктогъан' => 'köktoğan', 'коктотай' => 'köktotay', 'коллег' => 'kolleg', |
| 161 | 'коллект' => 'kollekt', 'коллекц' => 'kollekts', 'колье' => 'kolye', 'кольраби' => 'kolrabi', |
| 162 | 'кольцов' => 'koltsov', 'комби' => 'kombi', 'комеди' => 'komedi', 'коменда' => 'komenda', |
| 163 | 'комета' => 'kometa', 'комив' => 'komiv', 'комис' => 'komis', 'комит' => 'komit', |
| 164 | 'комм' => 'komm', 'коммент' => 'komment', 'коммерс' => 'kommers', 'коммерц' => 'kommerts', |
| 165 | 'комп' => 'komp', 'конве' => 'konve', 'конгени' => 'kongeni', 'конденс' => 'kondens', |
| 166 | 'кондил' => 'kondil', 'кондитер' => 'konditer', 'кондиц' => 'kondits', 'коник' => 'konik', |
| 167 | 'конкис' => 'konkis', 'консерв' => 'konserv', 'конси' => 'konsi', 'контейнер' => 'konteyner', |
| 168 | 'конти' => 'konti', 'конфе' => 'konfe', 'конфи' => 'konfi', 'конце' => 'kontse', |
| 169 | 'конъю' => 'konyu', 'коньки' => 'konki', 'коньяк' => 'konyak', 'копирле' => 'kopirle', |
| 170 | 'копия' => 'kopiya', 'корде' => 'korde', 'кореиз' => 'koreiz', 'коренн' => 'korenn', |
| 171 | 'корея' => 'koreya', 'кориа' => 'koria', 'коридор' => 'koridor', 'корне' => 'korne', |
| 172 | 'корнеев' => 'korneyev', 'корни' => 'korni', 'корре' => 'korre', 'косме' => 'kosme', |
| 173 | 'космик' => 'kosmik', 'костюм' => 'kostüm', 'котельн' => 'koteln', 'котир' => 'kotir', |
| 174 | 'котлет' => 'kotlet', 'кочерг' => 'koçerg', 'коше' => 'köşe', 'куби' => 'kubi', |
| 175 | 'кудрин' => 'kudrin', 'кузнец' => 'kuznets', 'кулинар' => 'kulinar', 'кулич' => 'kuliç', |
| 176 | 'кульмин' => 'kulmin', 'культаш' => 'kültaş', 'культе' => 'külte', 'культ' => 'kult', |
| 177 | 'куркулет' => 'kürkület', 'курсив' => 'kursiv', 'кушет' => 'kuşet', 'кушку' => 'küşkü', |
| 178 | 'куюк' => 'küyük', 'къолязма' => 'qolyazma', 'къуртумер' => 'qurtümer', |
| 179 | 'къуртусеин' => 'qurtüsein', 'медьюн' => 'medyun', 'месули' => 'mesüli', |
| 180 | 'мефкуре' => 'mefküre', 'могедек' => 'mögedek', 'мумиё' => 'mumiyo', 'мумиф' => 'mumif', |
| 181 | 'муче' => 'müçe', 'муюз' => 'müyüz', 'нумюне' => 'nümüne', 'обел' => 'obel', 'обер' => 'ober', |
| 182 | 'обли' => 'obli', 'обсе' => 'obse', 'обт' => 'obt', 'огне' => 'ogne', 'одеколон' => 'odekolon', |
| 183 | 'одеса' => 'odesa', 'одесса' => 'odessa', 'озерки' => 'ozerki', 'озерн' => 'ozern', |
| 184 | 'озёрн' => 'ozörn', 'озюя' => 'özüya', 'океан' => 'okean', 'окси' => 'oksi', |
| 185 | 'октет' => 'oktet', 'олеа' => 'olea', 'олеи' => 'olei', 'оленев' => 'olenev', 'олив' => 'oliv', |
| 186 | 'олиг' => 'olig', 'олимп' => 'olimp', 'олиф' => 'olif', 'ольчер' => 'ölçer', 'омле' => 'omle', |
| 187 | 'онен' => 'onen', 'оннен' => 'onnen', 'опера' => 'opera', 'опере' => 'opere', |
| 188 | 'оптим' => 'optim', 'опци' => 'optsi', 'орби' => 'orbi', 'орден' => 'orden', |
| 189 | 'ордер' => 'order', 'ордин' => 'ordin', 'ореа' => 'orea', 'орех' => 'oreh', |
| 190 | 'ориент' => 'oriyent', 'оркестр' => 'orkestr', 'орлин' => 'orlin', 'орни' => 'orni', |
| 191 | 'орхи' => 'orhi', 'осци' => 'ostsi', 'офис' => 'ofis', 'офиц' => 'ofits', 'офсет' => 'ofset', |
| 192 | 'очерк' => 'oçerk', 'оюннен' => 'oyunnen', 'побед' => 'pobed', 'полево' => 'polevo', |
| 193 | 'поли' => 'poli', 'полюшко' => 'polüşko', 'помидор' => 'pomidor', 'пониз' => 'poniz', |
| 194 | 'порфир' => 'porfir', 'потелов' => 'potelov', 'потюк' => 'pötük', 'почетн' => 'poçetn', |
| 195 | 'почётн' => 'poçötn', 'пукле' => 'pükle', 'пуркю' => 'pürkü', 'пурумют' => 'purümüt', |
| 196 | 'пускул' => 'püskül', 'пускур' => 'püskür', 'пусюр' => 'püsür', 'пуфле' => 'püfle', |
| 197 | 'сейитумер' => 'seyitümer', 'сейитусеин' => 'seyitüsein', 'сейитягъя' => 'seyityağya', |
| 198 | 'сейитягья' => 'seyityagya', 'сейитяхья' => 'seyityahya', 'сейитяя' => 'seyityaya', |
| 199 | 'сеитумер' => 'seitümer', 'сеитусеин' => 'seitüsein', 'сеитягъя' => 'seityağya', |
| 200 | 'сеитягья' => 'seityagya', 'сеитяхья' => 'seityahya', 'сеитяя' => 'seityaya', |
| 201 | 'сурет' => 'süret', 'увертюра' => 'uvertüra', 'угле' => 'ugle', 'узвий' => 'uzviy', |
| 202 | 'улица' => 'ulitsa', 'ультимат' => 'ultimat', 'ультра' => 'ultra', 'ульянов' => 'ulyanov', |
| 203 | 'универ' => 'univer', 'уник' => 'unik', 'унис' => 'unis', 'унит' => 'unit', 'униф' => 'unif', |
| 204 | 'унтер' => 'unter', 'урьян' => 'uryan', 'утил' => 'util', 'уткин' => 'utkin', |
| 205 | 'учебн' => 'uçebn', 'шовини' => 'şovini', 'шоссе' => 'şosse', 'шубин' => 'şubin', |
| 206 | 'шунен' => 'şunen', 'шуннен' => 'şunnen', 'шунчюн' => 'şunçün', 'щёлкино' => 'şçolkino', |
| 207 | 'эмирусеин' => 'emirüsein', 'юзбашы' => 'yüzbaşı', 'юзйыл' => 'yüzyıl', 'юртер' => 'yurter', |
| 208 | 'ющенко' => 'yuşçenko', |
| 209 | |
| 210 | ### Carefully ordered many-to-one mappings |
| 211 | # these are ordered so L2C is correct (the later Cyrillic one) |
| 212 | # see also $ManyToOneC2LMappings above for C2L |
| 213 | 'шофер' => 'şoför', 'шофёр' => 'şoför', |
| 214 | 'бугун' => 'bugün', 'бугунь' => 'bugün', |
| 215 | 'демирёл' => 'demiryol', 'демиръёл' => 'demiryol', |
| 216 | 'гонъюл' => 'göñül', 'гонъюль' => 'göñül', |
| 217 | 'коккоз' => 'kökköz', 'коккозь' => 'kökköz', |
| 218 | 'корбекул' => 'körbekül', 'корьбекул' => 'körbekül', 'корьбекуль' => 'körbekül', |
| 219 | 'муур' => 'müür', 'муурь' => 'müür', |
| 220 | 'оригинал' => 'original', 'оригиналь' => 'original', |
| 221 | 'пускю' => 'püskü', 'пуськю' => 'püskü', |
| 222 | 'къарагоз' => 'qaragöz', 'къарагозь' => 'qaragöz', |
| 223 | |
| 224 | #### Latin to Cyrillic (deduped from above) |
| 225 | |
| 226 | # слова на -аль |
| 227 | # words in -аль |
| 228 | 'актуаль' => 'aktual', 'диагональ' => 'diagonal', 'документаль' => 'dokumental', |
| 229 | 'эмсаль' => 'emsal', 'фааль' => 'faal', 'феодаль' => 'feodal', 'фестиваль' => 'festival', |
| 230 | 'горизонталь' => 'gorizontal', 'хроникаль' => 'hronikal', 'идеаль' => 'ideal', |
| 231 | 'инструменталь' => 'instrumental', 'икъмаль' => 'iqmal', 'икъбаль' => 'iqbal', |
| 232 | 'истикъбаль' => 'istiqbal', 'истикъляль' => 'istiqlâl', 'италия' => 'italiya', |
| 233 | 'италья' => 'italya', 'ишгъаль' => 'işğal', 'кафедраль' => 'kafedral', 'казуаль' => 'kazual', |
| 234 | 'коллегиаль' => 'kollegial', 'колоссаль' => 'kolossal', 'коммуналь' => 'kommunal', |
| 235 | 'кординаль' => 'kordinal', 'криминаль' => 'kriminal', 'легаль' => 'legal', |
| 236 | 'леталь' => 'letal', 'либераль' => 'liberal', 'локаль' => 'lokal', |
| 237 | 'магистраль' => 'magistral', 'материаль' => 'material', 'машиналь' => 'maşinal', |
| 238 | 'меаль' => 'meal', 'медальон' => 'medalyon', 'медаль' => 'medal', |
| 239 | 'меридиональ' => 'meridional', 'мешъаль' => 'meşal', 'минераль' => 'mineral', |
| 240 | 'минималь' => 'minimal', 'мисаль' => 'misal', 'модаль' => 'modal', 'музыкаль' => 'muzıkal', |
| 241 | 'номиналь' => 'nominal', 'нормаль' => 'normal', 'оптималь' => 'optimal', |
| 242 | 'орбиталь' => 'orbital', 'педаль' => 'pedal', 'пропорциональ' => 'proportsional', |
| 243 | 'профессиональ' => 'professional', 'радикаль' => 'radikal', 'рациональ' => 'ratsional', |
| 244 | 'реаль' => 'real', 'региональ' => 'regional', 'суаль' => 'sual', 'шималь' => 'şimal', |
| 245 | 'территориаль' => 'territorial', 'тимсаль' => 'timsal', 'тоталь' => 'total', |
| 246 | 'уникаль' => 'unikal', 'универсаль' => 'universal', 'вертикаль' => 'vertikal', |
| 247 | 'виртуаль' => 'virtual', 'визуаль' => 'vizual', 'вуаль' => 'vual', 'зональ' => 'zonal', |
| 248 | 'зуаль' => 'zual', 'италь' => 'ital', |
| 249 | |
| 250 | # слова с мягким знаком перед а, о, у, э |
| 251 | # Words with a soft sign before а, о, у, э |
| 252 | 'бильакис' => 'bilakis', 'маальэсеф' => 'maalesef', 'мельун' => 'melun', 'озьара' => 'özara', |
| 253 | 'вельасыл' => 'velasıl', 'ельаякъ' => 'yelayaq', |
| 254 | |
| 255 | # другие слова с мягким знаком |
| 256 | # Other words with a soft sign |
| 257 | 'альбатрос' => 'albatros', 'альбинос' => 'albinos', 'альбом' => 'albom', |
| 258 | 'альбумин' => 'albumin', 'алфавит' => 'alfavit', 'альфа' => 'alfa', 'альманах' => 'almanah', |
| 259 | 'альпинист' => 'alpinist', 'альтерн' => 'altern', 'альтру' => 'altru', |
| 260 | 'альвеола' => 'alveola', 'ансамбль' => 'ansambl', 'аньане' => 'anane', 'асфальт' => 'asfalt', |
| 261 | 'бальнео' => 'balneo', 'баарь' => 'baar', 'базальт' => 'bazalt', 'бинокль' => 'binokl', |
| 262 | 'девальв' => 'devalv', 'факульт' => 'fakult', 'фальсиф' => 'falsif', 'фольклор' => 'folklor', |
| 263 | 'гальван' => 'galvan', 'геральд' => 'gerald', 'женьшень' => 'jenşen', |
| 264 | 'инвентарь' => 'inventar', 'кальк' => 'kalk', 'кальмар' => 'kalmar', 'консульт' => 'konsult', |
| 265 | 'контроль' => 'kontrol', 'культур' => 'kultur', 'лагерь' => 'lager', 'макъбуль' => 'maqbul', |
| 266 | 'макъуль' => 'maqul', 'мальт' => 'malt', 'мальземе' => 'malzeme', 'меджуль' => 'mecul', |
| 267 | 'мешгуль' => 'meşgül', 'мешгъуль' => 'meşğul', 'мульти' => 'multi', |
| 268 | 'мусульман' => 'musulman', 'нефть' => 'neft', 'пальто' => 'palto', 'пароль' => 'parol', |
| 269 | 'патруль' => 'patrul', 'пенальти' => 'penalti', 'къальби' => 'qalbi', 'къальпке' => 'qalpke', |
| 270 | 'къальплер' => 'qalpler', 'къальпни' => 'qalpni', 'къальпте' => 'qalpte', 'къаарь' => 'qaar', |
| 271 | 'ресуль' => 'resul', 'рыцарь' => 'rıtsar', 'рояль' => 'royal', 'саарь' => 'saar', |
| 272 | 'спираль' => 'spiral', 'сульх' => 'sulh', 'сумбуль' => 'sumbul', 'суньий' => 'suniy', |
| 273 | 'темаюль' => 'temayul', 'шампунь' => 'şampun', 'вальс' => 'vals', 'вальц' => 'valts', |
| 274 | 'ведомость' => 'vedomost', 'зулькъарнейн' => 'zulqarneyn', 'январь' => 'yanvar', |
| 275 | 'февраль' => 'fevral', 'июнь' => 'iyün', 'сентябрь' => 'sentâbr', 'октябрь' => 'oktâbr', |
| 276 | 'ноябрь' => 'noyabr', 'декабрь' => 'dekabr', |
| 277 | |
| 278 | # слова с твёрдым знаком |
| 279 | # Words with a solid sign |
| 280 | 'бидъат' => 'bidat', 'бузъюрек' => 'buzyürek', 'атешъюрек' => 'ateşyürek', |
| 281 | 'алъянакъ' => 'alyanaq', 'инъекц' => 'inyekts', 'мефъум' => 'mefum', 'мешъум' => 'meşum', |
| 282 | 'объект' => 'obyekt', 'разъезд' => 'razyezd', 'субъект' => 'subyekt', 'хавъяр' => 'havyar', |
| 283 | 'ямъям' => 'yamyam', |
| 284 | |
| 285 | # слова с буквой щ |
| 286 | # words with щ |
| 287 | 'ящик' => 'yaşçik', 'мещан' => 'meşçan', |
| 288 | |
| 289 | # слова с ц |
| 290 | # words with ц |
| 291 | 'акциз' => 'aktsiz', 'ацет' => 'atset', 'блиц' => 'blits', 'бруцеллёз' => 'brutsellöz', |
| 292 | 'доцент' => 'dotsent', 'фармацевт' => 'farmatsevt', 'глицер' => 'glitser', |
| 293 | 'люцерна' => 'lütserna', 'лицей' => 'litsey', 'меццо' => 'metstso', 'наци' => 'natsi', |
| 294 | 'проце' => 'protse', 'рецеп' => 'retsep', 'реценз' => 'retsenz', 'теплица' => 'teplitsa', |
| 295 | 'вице' => 'vitse', 'швейцар' => 'şveytsar', 'богородиц' => 'bogorodits', |
| 296 | 'бруцел' => 'brutsel', 'дацюк' => 'datsük', 'доницетти' => 'donitsetti', |
| 297 | 'драцена' => 'dratsena', 'контрацеп' => 'kontratsep', 'коцюб' => 'kotsüb', |
| 298 | 'меценат' => 'metsenat', 'мицел' => 'mitsel', 'моцарт' => 'motsart', 'плац' => 'plats', |
| 299 | 'плацен' => 'platsen', 'прецедент' => 'pretsedent', 'прецес' => 'pretses', |
| 300 | 'прицеп' => 'pritsep', 'спец' => 'spets', 'троиц' => 'troits', 'шприц' => 'şprits', |
| 301 | 'эпицентр' => 'epitsentr', 'яценюк' => 'yatsenük', |
| 302 | |
| 303 | # слова с тс |
| 304 | # words with тс |
| 305 | 'агъартс' => 'ağarts', 'агъыртс' => 'ağırts', 'бильдиртс' => 'bildirts', 'битсин' => 'bitsin', |
| 306 | 'буюльтс' => 'büyülts', 'буютс' => 'büyüts', 'гебертс' => 'geberts', 'делиртс' => 'delirts', |
| 307 | 'эгрильтс' => 'egrilts', 'эксильтс' => 'eksilts', 'эшитс' => 'eşits', 'иритс' => 'irits', |
| 308 | 'иситс' => 'isits', 'ичиртс' => 'içirts', 'кертсин' => 'kertsin', 'кенишлетс' => 'kenişlets', |
| 309 | 'кийсетс' => 'kiysets', 'копюртс' => 'köpürts', 'косьтертс' => 'kösterts', |
| 310 | 'кучертс' => 'küçerts', 'кучюльтс' => 'küçülts', 'пертсин' => 'pertsin', 'къайтс' => 'qayts', |
| 311 | 'къутсуз' => 'qutsuz', 'орьтс' => 'örts', 'отьс' => 'öts', 'тартс' => 'tarts', |
| 312 | 'тутсун' => 'tutsun', 'тюнъюльтс' => 'tüñülts', 'тюртс' => 'türts', 'янъартс' => 'yañarts', |
| 313 | 'ебертс' => 'yeberts', 'ешертс' => 'yeşerts', 'йиритс' => 'yirits', |
| 314 | |
| 315 | # разные исключения |
| 316 | # different exceptions |
| 317 | 'бюджет' => 'bücet', 'бюллет' => 'büllet', 'бюро' => 'büro', 'бюст' => 'büst', |
| 318 | 'диалог' => 'dialog', 'ханымэфенди' => 'hanımefendi', 'каньон' => 'kanyon', |
| 319 | 'кирил' => 'kiril', 'кирилл' => 'kirill', 'кёрджа' => 'körca', 'коy' => 'köy', |
| 320 | 'кулеръюзь' => 'küleryüz', 'маалле' => 'маальle', 'майор' => 'mayor', 'маниал' => 'manиаль', |
| 321 | 'нормала' => 'нормальa', 'проект' => 'proekt', 'район' => 'rayon', 'сойады' => 'soyadı', |
| 322 | 'спортсмен' => 'sportsmen', 'услюп' => 'üslüp', 'услюб' => 'üslüb', 'вакъиал' => 'vaqиаль', |
| 323 | 'юзйыллыкъ' => 'yüzyıllıq', 'койот' => 'koyot', |
| 324 | |
| 325 | # имена собственные |
| 326 | # proper names |
| 327 | 'адольф' => 'adolf', 'альберт' => 'albert', 'бешуй' => 'beşüy', 'флотск' => 'flotsk', |
| 328 | 'гайана' => 'gayana', 'грэсовский' => 'gresovskiy', 'гриц' => 'grits', 'гурджи' => 'gürci', |
| 329 | 'игорь' => 'igor', 'ильич' => 'ilyiç', 'ильин' => 'ilyin', 'исмаил' => 'ismail', |
| 330 | 'киттс' => 'kitts', 'комсомольск' => 'komsomolsk', 'корьбекулю' => 'körbekülü', |
| 331 | 'куницын' => 'kunitsın', 'львив' => 'lviv', 'львов' => 'lvov', 'марьино' => 'maryino', |
| 332 | 'махульдюр' => 'mahuldür', 'павел' => 'pavel', 'пантикапейон' => 'pantikapeyon', |
| 333 | 'къуртсейит' => 'qurtseyit', 'къуртсеит' => 'qurtseit', 'смаил' => 'smail', |
| 334 | 'советск' => 'sovetsk', 'шемьи-заде' => 'şemi-zade', 'тсвана' => 'tsvana', |
| 335 | 'учьэвли' => 'üçevli', 'йохан' => 'yohan', 'йорк' => 'york', 'винныця' => 'vinnıtsâ', |
| 336 | 'винница' => 'vinnitsa', 'хмельницк' => 'hmelnitsk', 'хмельныцк' => 'hmelnıtsk', |
| 337 | 'зайце' => 'zaytse', 'чистеньк' => 'çistenk', 'кольчуг' => 'kolçug', 'ручьи' => 'ruçyi', |
| 338 | 'ботсвана' => 'botsvana', 'большой' => 'bolşoy', 'большое' => 'bolşoye', |
| 339 | 'большая' => 'bolşaya', 'ущелье' => 'uşçelye', 'ущельное' => 'uşçelnoye', |
| 340 | 'предущельное' => 'preduşçelnoye', 'новенькое' => 'novenkoye', 'новосельц' => 'novoselts', |
| 341 | 'мелко' => 'melko', 'овощ' => 'ovoşç', 'перепёлк' => 'perepölk', 'рощин' => 'roşçin', |
| 342 | 'братск' => 'bratsk', 'краснофлотск' => 'krasnoflotsk', 'синицин' => 'sinitsin', |
| 343 | 'синицын' => 'sinitsın', 'льгов' => 'lgov', 'желто' => 'jelto', 'жёлт' => 'jölt', |
| 344 | 'пермь' => 'perm', 'солдатск' => 'soldatsk', 'кольцо' => 'koltso', 'шелко' => 'şelko', |
| 345 | 'охотск' => 'ohotsk', 'марий эл' => 'mariy el', 'мариуполь' => 'mariupol', |
| 346 | 'белгород' => 'belgorod', 'иркутск' => 'irkutsk', 'Иркутск' => 'İrkutsk', 'орёл' => 'oröl', |
| 347 | 'рязанск' => 'râzansk', 'рязань' => 'râzan', 'тверск' => 'tversk', 'тверь' => 'tver', |
| 348 | 'ярославль' => 'yaroslavl', 'благовеще' => 'blagoveşçe', 'мальдив' => 'maldiv', |
| 349 | 'бальбек' => 'balbek', 'альчик' => 'alçik', 'харьков' => 'harkov', 'волынск' => 'volınsk', |
| 350 | 'волынь' => 'volın', |
| 351 | |
| 352 | ]; |
| 353 | |
| 354 | /** |
| 355 | * @var string[] map Cyrillic to Latin and back, simple string match only (no regex) |
| 356 | * no variants: map exactly as is |
| 357 | */ |
| 358 | private $exactCaseMappings = [ |
| 359 | # аббревиатуры |
| 360 | # abbreviations |
| 361 | 'ОБСЕ' => 'OBSE', 'КъМДж' => 'QMC', 'КъДж' => 'QC', 'КъАЭ' => 'QAE', 'ГъСМК' => 'ĞSMK', |
| 362 | 'ШСДжБ' => 'ŞSCB', 'КъМШСДж' => 'QMŞSC', 'КъАССР' => 'QASSR', 'КъДМПУ' => 'QDMPU', |
| 363 | 'КъМПУ' => 'QMPU', |
| 364 | ]; |
| 365 | |
| 366 | /** |
| 367 | * @var string[] map Cyrillic to Latin and back, match end of word |
| 368 | * variants: all lowercase, all uppercase, first letter capitalized |
| 369 | * "first letter capitalized" variant was in the source |
| 370 | * items with capture group refs (e.g., $1) are only mapped from the |
| 371 | * regex to the reference |
| 372 | */ |
| 373 | private $suffixMapping = [ |
| 374 | # originally C2L |
| 375 | 'иаль' => 'ial', 'нуль' => 'nul', 'кой' => 'köy', 'койнинъ' => 'köyniñ', 'койни' => 'köyni', |
| 376 | 'койге' => 'köyge', 'койде' => 'köyde', 'койдеки' => 'köydeki', 'койден' => 'köyden', |
| 377 | 'козь' => 'köz', '-юнджи' => '-ünci', '-юнджиде' => '-üncide', '-юнджиден' => '-ünciden', |
| 378 | |
| 379 | # originally L2C, here swapped |
| 380 | 'льная' => 'lnaya', 'льное' => 'lnoye', 'льный' => 'lnıy', 'льний' => 'lniy', |
| 381 | 'льская' => 'lskaya', 'льский' => 'lskiy', 'льское' => 'lskoye', 'ополь' => 'opol', |
| 382 | 'щее' => 'şçeye', 'щий' => 'şçiy', 'щая' => 'şçaya', 'цепс' => 'tseps', |
| 383 | |
| 384 | ]; |
| 385 | |
| 386 | /** |
| 387 | * @var string[] map Cyrillic to Latin and back, match beginning of word |
| 388 | * variants: all lowercase letters, all uppercase letters, first letter capitalized |
| 389 | * items with capture group refs (e.g., $1) are only mapped from the |
| 390 | * regex to the reference |
| 391 | */ |
| 392 | private $prefixMapping = [ |
| 393 | # originally C2L |
| 394 | 'буюк([^ъ])' => 'büyük$1', 'бую([гдйлмнпрстчшc])(и)' => 'büyü$1$2', |
| 395 | 'буют([^ыа])' => 'büyüt$1', 'джонк([^ъ])' => 'cönk$1', 'коюм' => 'köyüm', 'коюнъ' => 'köyüñ', |
| 396 | 'коюн([ди])' => 'köyün$1', 'куе' => 'küye', 'куркке' => 'kürkke', 'куркни' => 'kürkni', |
| 397 | 'куркте' => 'kürkte', 'куркчю' => 'kürkçü', 'кою' => 'köyü', |
| 398 | 'жизнь' => 'jizn', |
| 399 | |
| 400 | # арабизмы на муи- муэ- / Arabic муи- муэ- |
| 401 | 'му([иэИЭ])' => 'mü$1', |
| 402 | |
| 403 | # originally L2C, here swapped |
| 404 | 'роль$1' => 'rol([^ü]|' . self::WB . ')', |
| 405 | 'усть$1' => 'üst([^ü]|' . self::WB . ')', |
| 406 | |
| 407 | # more prefixes |
| 408 | 'ком-кок' => 'köm-kök', |
| 409 | |
| 410 | ]; |
| 411 | |
| 412 | private $Cyrl2LatnRegexes = []; |
| 413 | private $Latn2CyrlRegexes = []; |
| 414 | |
| 415 | private function loadRegs() { |
| 416 | // Regexes as keys need to be declared in a function. |
| 417 | $this->Cyrl2LatnRegexes = [ |
| 418 | # относятся ко всему слову |
| 419 | # whole words |
| 420 | |
| 421 | // TODO: refactor upper/lower/first capital whole words without |
| 422 | // regexes into simpler list |
| 423 | |
| 424 | '/' . self::WB . 'КъЮШ' . self::WB . '/u' => 'QYŞ', |
| 425 | '/' . self::WB . 'ЮШ' . self::WB . '/u' => 'YŞ', |
| 426 | |
| 427 | '/' . self::WB . 'кок' . self::WB . '/u' => 'kök', |
| 428 | '/' . self::WB . 'Кок' . self::WB . '/u' => 'Kök', |
| 429 | '/' . self::WB . 'КОК' . self::WB . '/u' => 'KÖK', |
| 430 | '/' . self::WB . 'ком-кок' . self::WB . '/u' => 'köm-kök', |
| 431 | '/' . self::WB . 'Ком-кок' . self::WB . '/u' => 'Köm-kök', |
| 432 | '/' . self::WB . 'КОМ-КОК' . self::WB . '/u' => 'KÖM-KÖK', |
| 433 | |
| 434 | '/' . self::WB . 'коп' . self::WB . '/u' => 'köp', |
| 435 | '/' . self::WB . 'Коп' . self::WB . '/u' => 'Köp', |
| 436 | '/' . self::WB . 'КОП' . self::WB . '/u' => 'KÖP', |
| 437 | |
| 438 | '/' . self::WB . 'курк' . self::WB . '/u' => 'kürk', |
| 439 | '/' . self::WB . 'Курк' . self::WB . '/u' => 'Kürk', |
| 440 | '/' . self::WB . 'КУРК' . self::WB . '/u' => 'KÜRK', |
| 441 | |
| 442 | '/' . self::WB . 'ог' . self::WB . '/u' => 'ög', |
| 443 | '/' . self::WB . 'Ог' . self::WB . '/u' => 'Ög', |
| 444 | '/' . self::WB . 'ОГ' . self::WB . '/u' => 'ÖG', |
| 445 | |
| 446 | '/' . self::WB . 'юрип' . self::WB . '/u' => 'yürip', |
| 447 | '/' . self::WB . 'Юрип' . self::WB . '/u' => 'Yürip', |
| 448 | '/' . self::WB . 'ЮРИП' . self::WB . '/u' => 'YÜRİP', |
| 449 | |
| 450 | '/' . self::WB . 'юз' . self::WB . '/u' => 'yüz', |
| 451 | '/' . self::WB . 'Юз' . self::WB . '/u' => 'Yüz', |
| 452 | '/' . self::WB . 'ЮЗ' . self::WB . '/u' => 'YÜZ', |
| 453 | |
| 454 | '/' . self::WB . 'юк' . self::WB . '/u' => 'yük', |
| 455 | '/' . self::WB . 'Юк' . self::WB . '/u' => 'Yük', |
| 456 | '/' . self::WB . 'ЮК' . self::WB . '/u' => 'YÜK', |
| 457 | |
| 458 | '/' . self::WB . 'буюп' . self::WB . '/u' => 'büyüp', |
| 459 | '/' . self::WB . 'Буюп' . self::WB . '/u' => 'Büyüp', |
| 460 | '/' . self::WB . 'БУЮП' . self::WB . '/u' => 'BÜYÜP', |
| 461 | |
| 462 | '/' . self::WB . 'буюк' . self::WB . '/u' => 'büyük', |
| 463 | '/' . self::WB . 'Буюк' . self::WB . '/u' => 'Büyük', |
| 464 | '/' . self::WB . 'БУЮК' . self::WB . '/u' => 'BÜYÜK', |
| 465 | |
| 466 | '/' . self::WB . 'джонк' . self::WB . '/u' => 'cönk', |
| 467 | '/' . self::WB . 'Джонк' . self::WB . '/u' => 'Cönk', |
| 468 | '/' . self::WB . 'ДЖОНК' . self::WB . '/u' => 'CÖNK', |
| 469 | '/' . self::WB . 'джонкю' . self::WB . '/u' => 'cönkü', |
| 470 | '/' . self::WB . 'Джонкю' . self::WB . '/u' => 'Cönkü', |
| 471 | '/' . self::WB . 'ДЖОНКЮ' . self::WB . '/u' => 'CÖNKÜ', |
| 472 | |
| 473 | '/' . self::WB . 'куркчи/u' => 'kürkçi', |
| 474 | '/' . self::WB . 'Куркчи/u' => 'Kürkçi', |
| 475 | '/' . self::WB . 'КУРКЧИ/u' => 'KÜRKÇI', |
| 476 | |
| 477 | '/' . self::WB . 'устке' . self::WB . '/u' => 'üstke', |
| 478 | '/' . self::WB . 'Устке' . self::WB . '/u' => 'Üstke', |
| 479 | '/' . self::WB . 'УСТКЕ' . self::WB . '/u' => 'ÜSTKE', |
| 480 | '/' . self::WB . 'устте' . self::WB . '/u' => 'üstte', |
| 481 | '/' . self::WB . 'Устте' . self::WB . '/u' => 'Üstte', |
| 482 | '/' . self::WB . 'УСТТЕ' . self::WB . '/u' => 'ÜSTTE', |
| 483 | '/' . self::WB . 'усттен' . self::WB . '/u' => 'üstten', |
| 484 | '/' . self::WB . 'Усттен' . self::WB . '/u' => 'Üstten', |
| 485 | '/' . self::WB . 'УСТТЕН' . self::WB . '/u' => 'ÜSTTEN', |
| 486 | |
| 487 | # отдельно стоящие Ё и Я |
| 488 | # stand-alone Ё and Я |
| 489 | '/' . self::WB . 'Я' . self::WB . '/u' => 'Ya', |
| 490 | '/' . self::WB . 'Ё' . self::WB . '/u' => 'Yo', |
| 491 | |
| 492 | # относятся к началу слова |
| 493 | # word prefixes |
| 494 | '/' . self::WB . 'КъЮШн/u' => 'QYŞn', |
| 495 | '/' . self::WB . 'ЮШн/u' => 'YŞn', |
| 496 | |
| 497 | # need to convert digraphs (гъ, къ, нъ, дж) now to match patterns |
| 498 | '/гъ/u' => 'ğ', |
| 499 | '/Г[ъЪ]/u' => 'Ğ', |
| 500 | '/къ/u' => 'q', |
| 501 | '/К[ъЪ]/u' => 'Q', |
| 502 | '/нъ/u' => 'ñ', |
| 503 | '/Н[ъЪ]/u' => 'Ñ', |
| 504 | '/дж/u' => 'c', |
| 505 | '/Д[жЖ]/u' => 'C', |
| 506 | |
| 507 | # о => ö |
| 508 | '/' . self::WB . '([' . Crh::C_M_CONS . '])о([' . Crh::C_CONS . '])([' . Crh::C_CONS . |
| 509 | '])([еиэюьü])/u' => '$1ö$2$3$4', |
| 510 | '/' . self::WB . 'о([' . Crh::C_CONS . '])([' . Crh::C_CONS . '])([еиэюьü])/u' => 'ö$1$2$3', |
| 511 | '/' . self::WB . '([' . Crh::C_M_CONS . '])О([' . Crh::C_CONS . '])([' . Crh::C_CONS . |
| 512 | '])([еиэюьüЕИЭЮЬÜ])/u' => '$1Ö$2$3$4', |
| 513 | '/' . self::WB . 'О([' . Crh::C_CONS . '])([' . Crh::C_CONS . '])([еиэюьüЕИЭЮЬÜ])/u' |
| 514 | => 'Ö$1$2$3', |
| 515 | |
| 516 | '/' . self::WB . '([' . Crh::C_M_CONS . '])о([' . Crh::C_CONS . '])([еиэюьü])/u' => '$1ö$2$3', |
| 517 | '/' . self::WB . 'о([' . Crh::C_CONS . '])([еиэюьü])/u' => 'ö$1$2', |
| 518 | '/' . self::WB . '([' . Crh::C_M_CONS . '])О([' . Crh::C_CONS . '])([еиэюьüЕИЭЮЬÜ])/u' |
| 519 | => '$1Ö$2$3', |
| 520 | '/' . self::WB . 'О([' . Crh::C_CONS . '])([еиэюьüЕИЭЮЬÜ])/u' => 'Ö$1$2', |
| 521 | |
| 522 | # ё => yö |
| 523 | '/' . self::WB . 'ё([' . Crh::C_CONS . '])([' . Crh::C_CONS . '])([ьеюü])/u' => 'yö$1$2$3', |
| 524 | '/' . self::WB . 'Ё([' . Crh::C_CONS_LC . '])([' . Crh::C_CONS_LC . '])([ьеюü])/u' => 'Yö$1$2$3', |
| 525 | '/' . self::WB . 'Ё([' . Crh::C_CONS_UC . '])([' . Crh::C_CONS_UC . '])([ЬЕЮÜ])/u' => 'YÖ$1$2$3', |
| 526 | '/' . self::WB . 'ё([' . Crh::C_CONS . '])([ьеюü])/u' => 'yö$1$2', |
| 527 | '/' . self::WB . 'Ё([' . Crh::C_CONS_LC . '])([ьеюü])/u' => 'Yö$1$2', |
| 528 | '/' . self::WB . 'Ё([' . Crh::C_CONS_UC . '])([ЬЕЮÜ])/u' => 'YÖ$1$2', |
| 529 | |
| 530 | # у => ü, ую => üyü |
| 531 | '/' . self::WB . '([' . Crh::C_M_CONS . '])у([' . Crh::C_CONS . '])([' . Crh::C_CONS . |
| 532 | '])([еиэюьü])/u' => '$1ü$2$3$4', |
| 533 | '/' . self::WB . 'у([' . Crh::C_CONS . '])([' . Crh::C_CONS . '])([еиэюьü])/u' => 'ü$1$2$3', |
| 534 | '/' . self::WB . 'ую([' . Crh::C_CONS . '])([' . Crh::C_CONS . '])([еиэюьü])/u' => 'üyü$1$2$3', |
| 535 | '/' . self::WB . '([' . Crh::C_M_CONS . '])У([' . Crh::C_CONS . '])([' . Crh::C_CONS . |
| 536 | '])([еиэюьüЕИЭЮЬÜ])/u' => '$1Ü$2$3$4', |
| 537 | '/' . self::WB . 'У([' . Crh::C_CONS . '])([' . Crh::C_CONS . '])([еиэюьüЕИЭЮЬÜ])/u' |
| 538 | => 'Ü$1$2$3', |
| 539 | '/' . self::WB . 'Ую([' . Crh::C_CONS . '])([' . Crh::C_CONS . '])([еиэюьü])/u' => 'Üyü$1$2$3', |
| 540 | '/' . self::WB . 'УЮ([' . Crh::C_CONS . '])([' . Crh::C_CONS . '])([еиэюьü])/u' => 'ÜYÜ$1$2$3', |
| 541 | |
| 542 | '/' . self::WB . '([' . Crh::C_M_CONS . '])у([' . Crh::C_CONS . '])([еиэюьü])/u' => '$1ü$2$3', |
| 543 | '/' . self::WB . 'у([' . Crh::C_CONS . '])([еиэюьü])/u' => 'ü$1$2', |
| 544 | '/' . self::WB . 'ую([' . Crh::C_CONS . '])([еиэюьü])/u' => 'üyü$1$2', |
| 545 | '/' . self::WB . '([' . Crh::C_M_CONS . '])У([' . Crh::C_CONS . '])([еиэюьüЕИЭЮЬÜ])/u' |
| 546 | => '$1Ü$2$3', |
| 547 | '/' . self::WB . 'У([' . Crh::C_CONS . '])([еиэюьüЕИЭЮЬÜ])/u' => 'Ü$1$2', |
| 548 | '/' . self::WB . 'Ую([' . Crh::C_CONS . '])([еиэюьü])/u' => 'Üyü$1$2', |
| 549 | '/' . self::WB . 'УЮ([' . Crh::C_CONS . '])([еиэюьü])/u' => 'ÜYÜ$1$2', |
| 550 | |
| 551 | # ю => yü |
| 552 | '/' . self::WB . '([аыоуеиёюАЫОУЕИЁЮ]?)ю([' . Crh::C_CONS . '])([' . Crh::C_CONS . '])([ьеюü])/u' |
| 553 | => '$1yü$2$3$4', |
| 554 | '/' . self::WB . '([АЫОУЕИЁЮ]?)Ю([' . Crh::C_CONS_LC . '])([' . Crh::C_CONS_LC . '])([ьеюü])/u' |
| 555 | => '$1Yü$2$3$4', |
| 556 | '/' . self::WB . '([АЫОУЕИЁЮ]?)Ю([' . Crh::C_CONS_UC . '])([' . Crh::C_CONS_UC . '])([ЬЕЮÜ])/u' |
| 557 | => '$1YÜ$2$3$4', |
| 558 | '/' . self::WB . '([аыоуеиёюАЫОУЕИЁЮ]?)ю([' . Crh::C_CONS . '])([ьеюü])/u' => '$1yü$2$3', |
| 559 | '/' . self::WB . '([АЫОУЕИЁЮ]?)Ю([' . Crh::C_CONS_LC . '])([ьеюü])/u' => '$1Yü$2$3', |
| 560 | '/' . self::WB . '([АЫОУЕИЁЮ]?)Ю([' . Crh::C_CONS_UC . '])([ЬЕЮÜ])/u' => '$1YÜ$2$3', |
| 561 | |
| 562 | # e => ye, я => ya |
| 563 | '/' . self::WB . 'е/u' => 'ye', |
| 564 | '/' . self::WB . 'Е([' . Crh::C_LC . 'cğñqöü])/u' => 'Ye$1', |
| 565 | '/' . self::WB . 'Е([' . Crh::C_UC . 'CĞÑQÖÜ])/u' => 'YE$1', |
| 566 | '/' . self::WB . 'я/u' => 'ya', |
| 567 | '/' . self::WB . 'Я([' . Crh::C_LC . 'cğñqöü])/u' => 'Ya$1', |
| 568 | '/' . self::WB . 'Я([' . Crh::C_UC . 'CĞÑQÖÜ])/u' => 'YA$1', |
| 569 | '/([аеёиоуыэюяйьъaeöüАЕЁИОУЫЭЮЯЙЬЪAEÖÜ])е/u' => '$1ye', |
| 570 | '/([аеёиоуыэюяйьъaeöüАЕЁИОУЫЭЮЯЙЬЪAEÖÜ])Е([' . Crh::C_LC . 'cğñqöü])/u' => '$1Ye$2', |
| 571 | '/([аеёиоуыэюяйьъaeöüАЕЁИОУЫЭЮЯЙЬЪAEÖÜ])Е([' . Crh::C_UC . 'CĞÑQÖÜ])/u' => '$1YE$2', |
| 572 | '/([аеёиоуыэюяйьъaeöüğqАЕЁИОУЫЭЮЯЙЬЪAEÖÜĞQ])я/u' => '$1ya', |
| 573 | '/([аеёиоуыэюяйьъaeöüğqАЕЁИОУЫЭЮЯЙЬЪAEÖÜĞQ])Я([' . Crh::C_LC . 'cğñqöü])/u' => '$1Ya$2', |
| 574 | '/([аеёиоуыэюяйьъaeöüğqАЕЁИОУЫЭЮЯЙЬЪAEÖÜĞQ])Я([' . Crh::C_UC . 'CĞÑQÖÜ])/u' => '$1YA$2', |
| 575 | |
| 576 | # не зависят от места в слове |
| 577 | # position independent |
| 578 | |
| 579 | # слова на -льон |
| 580 | # words with -льон |
| 581 | '/льон/u' => 'lyon', |
| 582 | '/ЛЬОН/u' => 'LYON', |
| 583 | |
| 584 | '/козь([^я])/u' => 'köz$1', |
| 585 | '/Козь([^я])/u' => 'Köz$1', |
| 586 | '/КОЗЬ([^Я])/u' => 'KÖZ$1', |
| 587 | |
| 588 | # Ö, Ü 1-й заход: ё, ю после согласных > ö, ü |
| 589 | # Ö, Ü 1st instance: ё, ю after consonants > ö, ü |
| 590 | '/([' . Crh::C_CONS . '])ю/u' => '$1ü', |
| 591 | '/([' . Crh::C_CONS . '])Ю/u' => '$1Ü', |
| 592 | '/([' . Crh::C_CONS . '])ё/u' => '$1ö', |
| 593 | '/([' . Crh::C_CONS . '])Ё/u' => '$1Ö', |
| 594 | |
| 595 | # остальные вхождения о, у, ё, ю |
| 596 | # other occurrences of о, у, ё, ю |
| 597 | '/Ё([' . Crh::C_UC . 'CĞÑQÖÜ])/u' => 'YO$1', |
| 598 | '/Ю([' . Crh::C_UC . 'CĞÑQÖÜ])/u' => 'YU$1', |
| 599 | |
| 600 | # Ц & Щ |
| 601 | '/Ц([' . Crh::C_UC . 'CĞÑQÖÜ])/u' => 'TS$1', |
| 602 | '/Щ([' . Crh::C_UC . 'CĞÑQÖÜ])/u' => 'ŞÇ$1', |
| 603 | ]; |
| 604 | |
| 605 | $this->Latn2CyrlRegexes = [ |
| 606 | |
| 607 | // TODO: refactor upper/lower/first capital whole words without |
| 608 | // regexes into simpler list |
| 609 | |
| 610 | '/' . self::WB . 'an' . self::WB . '/u' => 'ань', |
| 611 | '/' . self::WB . 'An' . self::WB . '/u' => 'Ань', |
| 612 | '/' . self::WB . 'AN' . self::WB . '/u' => 'АНЬ', |
| 613 | '/' . self::WB . 'ange' . self::WB . '/u' => 'аньге', |
| 614 | '/' . self::WB . 'Ange' . self::WB . '/u' => 'Аньге', |
| 615 | '/' . self::WB . 'ANGE' . self::WB . '/u' => 'АНЬГЕ', |
| 616 | '/' . self::WB . 'ande' . self::WB . '/u' => 'аньде', |
| 617 | '/' . self::WB . 'Ande' . self::WB . '/u' => 'Аньде', |
| 618 | '/' . self::WB . 'ANDE' . self::WB . '/u' => 'АНЬДЕ', |
| 619 | '/' . self::WB . 'anki' . self::WB . '/u' => 'аньки', |
| 620 | '/' . self::WB . 'Anki' . self::WB . '/u' => 'Аньки', |
| 621 | '/' . self::WB . 'ANKİ' . self::WB . '/u' => 'АНЬКИ', |
| 622 | '/' . self::WB . 'deral' . self::WB . '/u' => 'деръал', |
| 623 | '/' . self::WB . 'Deral' . self::WB . '/u' => 'Деръал', |
| 624 | '/' . self::WB . 'DERAL' . self::WB . '/u' => 'ДЕРЪАЛ', |
| 625 | '/' . self::WB . 'kör' . self::WB . '/u' => 'кёр', |
| 626 | '/' . self::WB . 'Kör' . self::WB . '/u' => 'Кёр', |
| 627 | '/' . self::WB . 'KÖR' . self::WB . '/u' => 'КЁР', |
| 628 | '/' . self::WB . 'mer' . self::WB . '/u' => 'мэр', |
| 629 | '/' . self::WB . 'Mer' . self::WB . '/u' => 'Мэр', |
| 630 | '/' . self::WB . 'MER' . self::WB . '/u' => 'МЭР', |
| 631 | |
| 632 | '/' . self::WB . 'cönk/u' => 'джонк', |
| 633 | '/' . self::WB . 'Cönk/u' => 'Джонк', |
| 634 | '/' . self::WB . 'CÖNK/u' => 'ДЖОНК', |
| 635 | |
| 636 | # (y)etsin -> етсин/этсин |
| 637 | # note that target starts with CYRILLIC е/Е! |
| 638 | '/yetsin/u' => 'етсин', |
| 639 | '/Yetsin/u' => 'Етсин', |
| 640 | '/YETSİN/u' => 'ЕТСИН', |
| 641 | |
| 642 | # note that target starts with LATIN e/E! |
| 643 | # (other transformations will determine CYRILLIC е/э as needed) |
| 644 | '/etsin/u' => 'eтсин', |
| 645 | '/Etsin/u' => 'Eтсин', |
| 646 | '/ETSİN/u' => 'EТСИН', |
| 647 | |
| 648 | # буква Ё - первый заход |
| 649 | # расставляем Ь после согласных |
| 650 | '/' . self::WB . '([yY])ö([' . Crh::L_N_CONS . '])([aAuU' . Crh::L_CONS . ']|' . self::WB . ')/u' |
| 651 | => '$1ö$2ь$3', |
| 652 | '/' . self::WB . '([yY])Ö([' . Crh::L_N_CONS . '])([aAuU' . Crh::L_CONS . ']|' . self::WB . ')/u' |
| 653 | => '$1Ö$2Ь$3', |
| 654 | '/' . self::WB . 'AQŞ([^AEI]|' . self::WB . ')/u' => 'АКъШ$1', |
| 655 | |
| 656 | # буква Ю - первый заход |
| 657 | # расставляем Ь после согласных |
| 658 | '/' . self::WB . '([yY])ü([' . Crh::L_N_CONS . '])([aAuU' . Crh::L_CONS . ']|' . self::WB . ')/u' |
| 659 | => '$1ü$2ь$3', |
| 660 | '/' . self::WB . '([yY])Ü([' . Crh::L_N_CONS . '])([aAuU' . Crh::L_CONS . ']|' . self::WB . ')/u' |
| 661 | => '$1Ü$2Ь$3', |
| 662 | |
| 663 | '/' . self::WB . '([bcgkpşBCGKPŞ])ö([' . Crh::L_N_CONS . '])([' . Crh::L_CONS . ']|' . |
| 664 | self::WB . ')/u' => '$1ö$2ь$3', |
| 665 | '/' . self::WB . '([bcgkpşBCGKPŞ])Ö([' . Crh::L_N_CONS . '])([' . Crh::L_CONS . ']|' . |
| 666 | self::WB . ')/u' => '$1Ö$2Ь$3', |
| 667 | '/' . self::WB . '([bcgkpşBCGKPŞ])ü([' . Crh::L_N_CONS . '])([' . Crh::L_CONS . ']|' . |
| 668 | self::WB . ')/u' => '$1ü$2ь$3', |
| 669 | '/' . self::WB . '([bcgkpşBCGKPŞ])Ü([' . Crh::L_N_CONS . '])([' . Crh::L_CONS . ']|' . |
| 670 | self::WB . ')/u' => '$1Ü$2Ь$3', |
| 671 | |
| 672 | # ö и ü в начале слова |
| 673 | # случаи, когда нужен Ь |
| 674 | '/' . self::WB . 'ö([' . Crh::L_N_CONS . 'pP])([' . Crh::L_CONS . ']|' . self::WB . ')/u' |
| 675 | => 'ö$1ь$2', |
| 676 | '/' . self::WB . 'Ö([' . Crh::L_N_CONS_LC . 'p])([' . Crh::L_CONS . ']|' . self::WB . ')/u' |
| 677 | => 'Ö$1ь$2', |
| 678 | '/' . self::WB . 'Ö([' . Crh::L_N_CONS_UC . 'P])([' . Crh::L_CONS . ']|' . self::WB . ')/u' |
| 679 | => 'Ö$1Ь$2', |
| 680 | '/' . self::WB . 'ü([' . Crh::L_N_CONS . '])([' . Crh::L_CONS . ']|' . self::WB . ')/u' |
| 681 | => 'ü$1ь$2', |
| 682 | '/' . self::WB . 'Ü([' . Crh::L_N_CONS_LC . '])([' . Crh::L_CONS . ']|' . self::WB . ')/u' |
| 683 | => 'Ü$1ь$2', |
| 684 | '/' . self::WB . 'Ü([' . Crh::L_N_CONS_UC . '])([' . Crh::L_CONS . ']|' . self::WB . ')/u' |
| 685 | => 'Ü$1Ь$2', |
| 686 | |
| 687 | '/ts' . self::WB . '/u' => 'ц', |
| 688 | '/şç' . self::WB . '/u' => 'щ', |
| 689 | '/Ş[çÇ]' . self::WB . '/u' => 'Щ', |
| 690 | '/T[sS]' . self::WB . '/u' => 'Ц', |
| 691 | |
| 692 | # Ь после Л |
| 693 | # add Ь after Л |
| 694 | '/([' . Crh::L_F . '])l([' . Crh::L_CONS_LC . ']|' . self::WB . ')/u' => '$1ль$2', |
| 695 | '/([' . Crh::L_F_UC . '])L([' . Crh::L_CONS . ']|' . self::WB . ')/u' => '$1ЛЬ$2', |
| 696 | |
| 697 | # относятся к началу слова |
| 698 | '/' . self::WB . 'ts/u' => 'ц', |
| 699 | '/' . self::WB . 'T[sS]/u' => 'Ц', |
| 700 | |
| 701 | '/' . self::WB . 'şç/u' => 'щ', |
| 702 | '/' . self::WB . 'Ş[çÇ]/u' => 'Щ', |
| 703 | |
| 704 | # Э |
| 705 | '/(' . self::WB . '|[' . Crh::L_VOW . 'аеэяАЕЭЯ])e/u' => '$1э', |
| 706 | '/(' . self::WB . '|[' . Crh::L_VOW_UC . 'АЕЭЯ])E/u' => '$1Э', |
| 707 | |
| 708 | '/' . self::WB . '([' . Crh::L_M_CONS . '])ö/u' => '$1о', |
| 709 | '/' . self::WB . '([' . Crh::L_M_CONS . '])Ö/u' => '$1О', |
| 710 | '/' . self::WB . '([' . Crh::L_M_CONS . '])ü/u' => '$1у', |
| 711 | '/' . self::WB . '([' . Crh::L_M_CONS . '])Ü/u' => '$1У', |
| 712 | |
| 713 | '/' . self::WB . 'ö/u' => 'о', |
| 714 | '/' . self::WB . 'Ö/u' => 'О', |
| 715 | '/' . self::WB . 'ü/u' => 'у', |
| 716 | '/' . self::WB . 'Ü/u' => 'У', |
| 717 | |
| 718 | # некоторые исключения |
| 719 | # some exceptions |
| 720 | '/maal([^e])/u' => 'мааль$1', |
| 721 | '/Maal([^e])/u' => 'Мааль$1', |
| 722 | '/MAAL([^E])/u' => 'МААЛЬ$1', |
| 723 | '/küf([^eü])/u' => 'куфь$1', |
| 724 | '/Küf([^eü])/u' => 'Куфь$1', |
| 725 | '/KÜF([^EÜ])/u' => 'КУФЬ$1', |
| 726 | '/köz([^eü])/u' => 'козь$1', |
| 727 | '/Köz([^eü])/u' => 'Козь$1', |
| 728 | '/KÖZ([^EÜ])/u' => 'КОЗЬ$1', |
| 729 | |
| 730 | # Punctuation |
| 731 | '/#|No\./u' => '№', |
| 732 | |
| 733 | # некоторые случаи употребления Ц |
| 734 | '/tsi([^zñ])/u' => 'ци$1', |
| 735 | '/T[sS][iİ]([^zZñÑ])/u' => 'ЦИ$1', |
| 736 | '/ts([ou])/u' => 'ц$1', |
| 737 | '/T[sS]([oOuU])/u' => 'Ц$1', |
| 738 | '/ts([' . Crh::L_CONS . '])/u' => 'ц$1', |
| 739 | '/T[sS]([' . Crh::L_CONS . '])/u' => 'Ц$1', |
| 740 | '/([' . Crh::L_CONS . '])ts/u' => '$1ц', |
| 741 | '/([' . Crh::L_CONS . '])T[sS]/u' => '$1Ц', |
| 742 | '/tsиал/u' => 'циал', |
| 743 | '/TSИАЛ/u' => 'ЦИАЛ', |
| 744 | |
| 745 | # убираем ьi |
| 746 | # remove ьi (note Cyrillic ь and Latin i) |
| 747 | '/[ьЬ]([iİ])/u' => '$1', |
| 748 | |
| 749 | # ya & ye |
| 750 | '/([' . Crh::L_CONS . '])ya/u' => '$1ья', |
| 751 | '/([' . Crh::L_CONS . '])Y[aA]/u' => '$1ЬЯ', |
| 752 | '/([' . Crh::L_CONS . '])ye/u' => '$1ье', |
| 753 | '/([' . Crh::L_CONS . '])Y[eE]/u' => '$1ЬЕ', |
| 754 | |
| 755 | # расставляем Ь перед Ё |
| 756 | # place Ь in front of Ё |
| 757 | '/([' . Crh::L_CONS . '])y[oö]/u' => '$1ьё', |
| 758 | '/([' . Crh::L_CONS . '])Y[oOöÖ]/u' => '$1ЬЁ', |
| 759 | # оставшиеся вхождения yo и yö |
| 760 | # remaining occurrences of yo and yö |
| 761 | '/y[oö]/u' => 'ё', |
| 762 | '/[yY][oOöÖ]/u' => 'Ё', |
| 763 | |
| 764 | # расставляем Ь перед Ю |
| 765 | # place Ь in front of Ю |
| 766 | '/([' . Crh::L_CONS . '])y[uü]/u' => '$1ью', |
| 767 | '/([' . Crh::L_CONS . '])Y[uUüÜ]/u' => '$1ЬЮ', |
| 768 | # оставшиеся вхождения yu и yü |
| 769 | # remaining occurrences of yu and yü |
| 770 | '/y[uü]/u' => 'ю', |
| 771 | '/[yY][uUüÜ]/u' => 'Ю', |
| 772 | |
| 773 | # убираем ьa |
| 774 | # remove ьa (note Cyrillic ь and Latin a) |
| 775 | '/[ьЬ]([aA])/u' => '$1', |
| 776 | |
| 777 | # дж |
| 778 | '/C([' . Crh::L_UC . Crh::C_UC . 'АЕЁЙОУЭЮЯ])/u' => 'ДЖ$1', |
| 779 | '/([' . Crh::L_UC . Crh::C_UC . 'АЕЁЙОУЭЮЯ])C/u' => '$1ДЖ', |
| 780 | |
| 781 | # гъ, къ, нъ |
| 782 | '/Ğ([' . Crh::L_UC . Crh::C_UC . '])/u' => 'ГЪ$1', |
| 783 | '/([' . Crh::L_UC . Crh::C_UC . 'Ъ])Ğ/u' => '$1ГЪ', |
| 784 | |
| 785 | '/Q([' . Crh::L_UC . Crh::C_UC . '])/u' => 'КЪ$1', |
| 786 | '/([' . Crh::L_UC . Crh::C_UC . 'Ъ])Q/u' => '$1КЪ', |
| 787 | |
| 788 | '/Ñ([' . Crh::L_UC . Crh::C_UC . '])/u' => 'НЪ$1', |
| 789 | '/([' . Crh::L_UC . Crh::C_UC . 'Ъ])Ñ/u' => '$1НЪ', |
| 790 | |
| 791 | ]; |
| 792 | } |
| 793 | |
| 794 | private $CyrlCleanUpRegexes = [ |
| 795 | '/([клнрст])ь\1/u' => '$1$1', |
| 796 | '/([КЛНРСТ])Ь\1/u' => '$1$1', |
| 797 | '/К[ьЬ]к/u' => 'Кк', |
| 798 | '/Л[ьЬ]л/u' => 'Лл', |
| 799 | '/Н[ьЬ]н/u' => 'Нн', |
| 800 | '/Р[ьЬ]р/u' => 'Рр', |
| 801 | '/С[ьЬ]с/u' => 'Сс', |
| 802 | '/Т[ьЬ]т/u' => 'Тт', |
| 803 | |
| 804 | # убираем ьы и ь..ы |
| 805 | # remove ьы и ь..ы |
| 806 | '/[ьЬ]ы/u' => 'ы', |
| 807 | '/ЬЫ/u' => 'Ы', |
| 808 | '/[ьЬ]([гдклмнпрстчшГДКЛМНПРСТЧШ])ы/u' => '$1ы', |
| 809 | '/Ь([гдклмнпрстчшГДКЛМНПРСТЧШ])Ы/u' => '$1Ы', |
| 810 | '/[ьЬ]([гкнГКН])([ъЪ])ы/u' => '$1$2ы', |
| 811 | '/Ь([ГКН])ЪЫ/u' => '$1ЪЫ', |
| 812 | |
| 813 | # убираем йь |
| 814 | # remove йь |
| 815 | '/йь/u' => 'й', |
| 816 | '/ЙЬ/u' => 'Й', |
| 817 | |
| 818 | # частичное решение проблемы слова юз - 100 |
| 819 | # Partial solution to the problem of the word юз ("100") |
| 820 | # notice that these are cross-word patterns |
| 821 | '/эки юзь/u' => 'эки юз', '/Эки юзь/u' => 'Эки юз', '/ЭКИ ЮЗЬ/u' => 'ЭКИ ЮЗ', |
| 822 | '/учь юзь/u' => 'учь юз', '/Учь юзь/u' => 'Учь юз', '/УЧЬ ЮЗЬ/u' => 'УЧЬ ЮЗ', |
| 823 | '/дёрт юзь/u' => 'дёрт юз', '/Дёрт юзь/u' => 'Дёрт юз', '/ДЁРТ ЮЗЬ/u' => 'ДЁРТ ЮЗ', |
| 824 | '/беш юзь/u' => 'беш юз', '/Беш юзь/u' => 'Беш юз', '/БЕШ ЮЗЬ/u' => 'БЕШ ЮЗ', |
| 825 | '/алты юзь/u' => 'алты юз', '/Алты юзь/u' => 'Алты юз', '/АЛТЫ ЮЗЬ/u' => 'АЛТЫ ЮЗ', |
| 826 | '/еди юзь/u' => 'еди юз', '/Еди юзь/u' => 'Еди юз', '/ЕДИ ЮЗЬ/u' => 'ЕДИ ЮЗ', |
| 827 | '/секиз юзь/u' => 'секиз юз', '/Секиз юзь/u' => 'Секиз юз', '/СЕКИЗ ЮЗЬ/u' => 'СЕКИЗ ЮЗ', |
| 828 | '/докъуз юзь/u' => 'докъуз юз', '/Докъуз юзь/u' => 'Докъуз юз', '/ДОКЪУЗ ЮЗЬ/u' => 'ДОКЪУЗ ЮЗ', |
| 829 | ]; |
| 830 | } |