1<?php 2/** 3 * This program is free software; you can redistribute it and/or modify 4 * it under the terms of the GNU General Public License as published by 5 * the Free Software Foundation; either version 2 of the License, or 6 * (at your option) any later version. 7 * 8 * This program is distributed in the hope that it will be useful, 9 * but WITHOUT ANY WARRANTY; without even the implied warranty of 10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 11 * GNU General Public License for more details. 12 * 13 * You should have received a copy of the GNU General Public License along 14 * with this program; if not, write to the Free Software Foundation, Inc., 15 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. 16 * http://www.gnu.org/copyleft/gpl.html 17 * 18 * @file 19 * @author Zhengzhu Feng <zhengzhu@gmail.com> 20 * @author fdcn <fdcn64@gmail.com> 21 * @author shinjiman <shinjiman@gmail.com> 22 * @author PhiLiP <philip.npc@gmail.com> 23 */ 24use MediaWiki\Linker\LinkTarget; 25use MediaWiki\Logger\LoggerFactory; 26use MediaWiki\MediaWikiServices; 27use MediaWiki\Revision\RevisionRecord; 28use MediaWiki\Revision\SlotRecord; 29 30/** 31 * Base class for multi-variant language conversion. 32 * 33 * @ingroup Language 34 */ 35abstract class LanguageConverter implements ILanguageConverter { 36 use DeprecationHelper; 37 38 /** 39 * languages supporting variants 40 * @since 1.20 41 * @var array 42 */ 43 public static $languagesWithVariants = [ 44 'ban', 45 'en', 46 'crh', 47 'gan', 48 'iu', 49 'kk', 50 'ku', 51 'shi', 52 'sr', 53 'tg', 54 'uz', 55 'zh', 56 ]; 57 58 private $mTablesLoaded = false; 59 60 /** 61 * @var ReplacementArray[]|bool[] 62 */ 63 protected $mTables; 64 65 /** 66 * @var Language 67 */ 68 private $mLangObj; 69 70 private $mUcfirst = false; 71 private $mConvRuleTitle = false; 72 private $mURLVariant; 73 private $mUserVariant; 74 private $mHeaderVariant; 75 private $mMaxDepth = 10; 76 private $mVarSeparatorPattern; 77 78 private const CACHE_VERSION_KEY = 'VERSION 7'; 79 80 /** 81 * @param Language $langobj 82 */ 83 public function __construct( $langobj ) { 84 $this->deprecatePublicProperty( 'mUcfirst', '1.35', __CLASS__ ); 85 $this->deprecatePublicProperty( 'mConvRuleTitle', '1.35', __CLASS__ ); 86 $this->deprecatePublicProperty( 'mUserVariant', '1.35', __CLASS__ ); 87 $this->deprecatePublicProperty( 'mHeaderVariant', '1.35', __CLASS__ ); 88 $this->deprecatePublicProperty( 'mMaxDepth = 10', '1.35', __CLASS__ ); 89 $this->deprecatePublicProperty( 'mVarSeparatorPattern', '1.35', __CLASS__ ); 90 $this->deprecatePublicProperty( 'mLangObj', '1.35', __CLASS__ ); 91 $this->deprecatePublicProperty( 'mVariantFallbacks', '1.35', __CLASS__ ); 92 $this->deprecatePublicProperty( 'mTablesLoaded', '1.35', __CLASS__ ); 93 $this->deprecatePublicProperty( 'mTables', '1.35', __CLASS__ ); 94 95 $this->mLangObj = $langobj; 96 97 $this->deprecatePublicPropertyFallback( 'mVariants', '1.36', function () { 98 return $this->getVariants(); 99 } ); 100 101 $this->deprecatePublicPropertyFallback( 'mMainLanguageCode', '1.36', function () { 102 return $this->getMainCode(); 103 } ); 104 105 $this->deprecatePublicPropertyFallback( 'mVariantFallbacks', '1.36', function () { 106 return $this->getVariantsFallbacks(); 107 } ); 108 109 $this->deprecatePublicPropertyFallback( 'mFlags', '1.36', function () { 110 return $this->getFlags(); 111 } ); 112 113 $this->deprecatePublicPropertyFallback( 'mVariantNames', '1.36', function () { 114 return $this->getVariantNames(); 115 } ); 116 117 $this->deprecatePublicPropertyFallback( 'mDescCodeSep', '1.36', function () { 118 return $this->getDescCodeSeparator(); 119 } ); 120 121 $this->deprecatePublicPropertyFallback( 'mDescVarSep', '1.36', function () { 122 return $this->getDescVarSeparator(); 123 } ); 124 } 125 126 /** 127 * Get main language code. 128 * @since 1.36 129 * 130 * @return string 131 */ 132 abstract public function getMainCode(): string; 133 134 /** 135 * Get supported variants of the language. 136 * @since 1.36 137 * 138 * @return array 139 */ 140 abstract protected function getLanguageVariants(): array; 141 142 /** 143 * Get language variants fallbacks. 144 * @since 1.36 145 * 146 * @return array 147 */ 148 abstract public function getVariantsFallbacks(): array; 149 150 /** 151 * Get strings that maps to the flags. 152 * @since 1.36 153 * 154 * @return array 155 */ 156 final public function getFlags(): array { 157 $defaultflags = [ 158 // 'S' show converted text 159 // '+' add rules for alltext 160 // 'E' the gave flags is error 161 // these flags above are reserved for program 162 'A' => 'A', // add rule for convert code (all text convert) 163 'T' => 'T', // title convert 164 'R' => 'R', // raw content 165 'D' => 'D', // convert description (subclass implement) 166 '-' => '-', // remove convert (not implement) 167 'H' => 'H', // add rule for convert code (but no display in placed code) 168 'N' => 'N', // current variant name 169 ]; 170 $flags = array_merge( $defaultflags, $this->getAdditionalFlags() ); 171 foreach ( $this->getVariants() as $v ) { 172 $flags[$v] = $v; 173 } 174 return $flags; 175 } 176 177 /** 178 * Provides additinal flags for converter. By default it return empty array and 179 * typicslly should be overridden by implementation of converter.. 180 * 181 * @return array 182 */ 183 protected function getAdditionalFlags(): array { 184 return []; 185 } 186 187 /** 188 * Get manual level limit for supported variants. 189 * @since 1.36 190 * 191 * @return array 192 */ 193 final public function getManualLevel() { 194 $manualLevel = $this->getAdditionalManualLevel(); 195 $result = []; 196 foreach ( $this->getVariants() as $v ) { 197 if ( array_key_exists( $v, $manualLevel ) ) { 198 $result[$v] = $manualLevel[$v]; 199 } else { 200 $result[$v] = 'bidirectional'; 201 } 202 } 203 return $result; 204 } 205 206 /** 207 * Provides additinal flags for converter. By default it return empty array and 208 * typicslly should be overridden by implementation of converter. 209 * @since 1.36 210 * 211 * @return array 212 */ 213 protected function getAdditionalManualLevel(): array { 214 return []; 215 } 216 217 /** 218 * Get desc code separator. By default returns ":", can be overridden by 219 * implementation of converter. 220 * @since 1.36 221 * 222 * @return string 223 */ 224 public function getDescCodeSeparator(): string { 225 return ':'; 226 } 227 228 /** 229 * Get desc var separator. By default returns ";", can be overridden by 230 * implementation of converter. 231 * @since 1.36 232 * 233 * @return string 234 */ 235 public function getDescVarSeparator(): string { 236 return ';'; 237 } 238 239 /** 240 * Get variant names. 241 * 242 * @return array 243 */ 244 public function getVariantNames(): array { 245 return MediaWikiServices::getInstance() 246 ->getLanguageNameUtils() 247 ->getLanguageNames(); 248 } 249 250 /** 251 * Get all valid variants for current Coverter. It uses abstract 252 * 253 * @return string[] Contains all valid variants 254 */ 255 final public function getVariants() { 256 global $wgDisabledVariants; 257 return array_diff( $this->getLanguageVariants(), $wgDisabledVariants ); 258 } 259 260 /** 261 * In case some variant is not defined in the markup, we need 262 * to have some fallback. For example, in zh, normally people 263 * will define zh-hans and zh-hant, but less so for zh-sg or zh-hk. 264 * when zh-sg is preferred but not defined, we will pick zh-hans 265 * in this case. Right now this is only used by zh. 266 * 267 * @param string $variant The language code of the variant 268 * @return string|array The code of the fallback language or the 269 * main code if there is no fallback 270 */ 271 public function getVariantFallbacks( $variant ) { 272 return $this->getVariantsFallbacks()[$variant] ?? $this->getMainCode(); 273 } 274 275 /** 276 * Get the title produced by the conversion rule. 277 * @return string The converted title text 278 */ 279 public function getConvRuleTitle() { 280 return $this->mConvRuleTitle; 281 } 282 283 /** 284 * Get preferred language variant. 285 * @return string The preferred language code 286 */ 287 public function getPreferredVariant() { 288 global $wgDefaultLanguageVariant, $wgUser; 289 290 $req = $this->getURLVariant(); 291 292 Hooks::runner()->onGetLangPreferredVariant( $req ); 293 294 // NOTE: For calls from Setup.php, wgUser or the session might not be set yet (T235360) 295 // Use case: During autocreation, User::isUsableName is called which uses interface 296 // messages for reserved usernames. 297 if ( $wgUser && $wgUser->isSafeToLoad() && $wgUser->isRegistered() && !$req ) { 298 $req = $this->getUserVariant( $wgUser ); 299 } elseif ( !$req ) { 300 $req = $this->getHeaderVariant(); 301 } 302 303 if ( $wgDefaultLanguageVariant && !$req ) { 304 $req = $this->validateVariant( $wgDefaultLanguageVariant ); 305 } 306 307 $req = $this->validateVariant( $req ); 308 309 // This function, unlike the other get*Variant functions, is 310 // not memoized (i.e. there return value is not cached) since 311 // new information might appear during processing after this 312 // is first called. 313 if ( $req ) { 314 return $req; 315 } 316 return $this->getMainCode(); 317 } 318 319 /** 320 * This function would not be affected by user's settings 321 * @return string The default variant code 322 */ 323 public function getDefaultVariant() { 324 global $wgDefaultLanguageVariant; 325 326 $req = $this->getURLVariant(); 327 328 if ( !$req ) { 329 $req = $this->getHeaderVariant(); 330 } 331 332 if ( $wgDefaultLanguageVariant && !$req ) { 333 $req = $this->validateVariant( $wgDefaultLanguageVariant ); 334 } 335 336 if ( $req ) { 337 return $req; 338 } 339 return $this->getMainCode(); 340 } 341 342 /** 343 * Validate the variant and return an appropriate strict internal 344 * variant code if one exists. Compare to Language::hasVariant() 345 * which does a strict test. 346 * 347 * @param string|null $variant The variant to validate 348 * @return mixed Returns an equivalent valid variant code if possible, 349 * null otherwise 350 */ 351 public function validateVariant( $variant = null ) { 352 if ( $variant === null ) { 353 return null; 354 } 355 // Our internal variants are always lower-case; the variant we 356 // are validating may have mixed case. 357 $variant = LanguageCode::replaceDeprecatedCodes( strtolower( $variant ) ); 358 if ( in_array( $variant, $this->getVariants() ) ) { 359 return $variant; 360 } 361 // Browsers are supposed to use BCP 47 standard in the 362 // Accept-Language header, but not all of our internal 363 // mediawiki variant codes are BCP 47. Map BCP 47 code 364 // to our internal code. 365 foreach ( $this->getVariants() as $v ) { 366 // Case-insensitive match (BCP 47 is mixed case) 367 if ( strtolower( LanguageCode::bcp47( $v ) ) === $variant ) { 368 return $v; 369 } 370 } 371 return null; 372 } 373 374 /** 375 * Get the variant specified in the URL 376 * 377 * @return mixed Variant if one found, null otherwise 378 */ 379 public function getURLVariant() { 380 global $wgRequest; 381 382 if ( $this->mURLVariant ) { 383 return $this->mURLVariant; 384 } 385 386 // see if the preference is set in the request 387 $ret = $wgRequest->getText( 'variant' ); 388 389 if ( !$ret ) { 390 $ret = $wgRequest->getVal( 'uselang' ); 391 } 392 393 $this->mURLVariant = $this->validateVariant( $ret ); 394 return $this->mURLVariant; 395 } 396 397 /** 398 * Determine if the user has a variant set. 399 * 400 * @param User $user 401 * @return mixed Variant if one found, null otherwise 402 */ 403 protected function getUserVariant( User $user ) { 404 // This should only be called within the class after the user is known to be 405 // safe to load and logged in, but check just in case. 406 if ( !$user->isSafeToLoad() ) { 407 return false; 408 } 409 410 if ( $user->isRegistered() ) { 411 // Get language variant preference from logged in users 412 if ( 413 $this->getMainCode() == 414 MediaWikiServices::getInstance()->getContentLanguage()->getCode() 415 ) { 416 $ret = $user->getOption( 'variant' ); 417 } else { 418 $ret = $user->getOption( 'variant-' . $this->getMainCode() ); 419 } 420 } else { 421 // figure out user lang without constructing wgLang to avoid 422 // infinite recursion 423 $ret = $user->getOption( 'language' ); 424 } 425 426 $this->mUserVariant = $this->validateVariant( $ret ); 427 return $this->mUserVariant; 428 } 429 430 /** 431 * Determine the language variant from the Accept-Language header. 432 * 433 * @return mixed Variant if one found, null otherwise 434 */ 435 protected function getHeaderVariant() { 436 global $wgRequest; 437 438 if ( $this->mHeaderVariant ) { 439 return $this->mHeaderVariant; 440 } 441 442 // See if some supported language variant is set in the 443 // HTTP header. 444 $languages = array_keys( $wgRequest->getAcceptLang() ); 445 if ( empty( $languages ) ) { 446 return null; 447 } 448 449 $fallbackLanguages = []; 450 foreach ( $languages as $language ) { 451 $this->mHeaderVariant = $this->validateVariant( $language ); 452 if ( $this->mHeaderVariant ) { 453 break; 454 } 455 456 // To see if there are fallbacks of current language. 457 // We record these fallback variants, and process 458 // them later. 459 $fallbacks = $this->getVariantFallbacks( $language ); 460 if ( is_string( $fallbacks ) && $fallbacks !== $this->getMainCode() ) { 461 $fallbackLanguages[] = $fallbacks; 462 } elseif ( is_array( $fallbacks ) ) { 463 $fallbackLanguages = 464 array_merge( $fallbackLanguages, $fallbacks ); 465 } 466 } 467 468 if ( !$this->mHeaderVariant ) { 469 // process fallback languages now 470 $fallback_languages = array_unique( $fallbackLanguages ); 471 foreach ( $fallback_languages as $language ) { 472 $this->mHeaderVariant = $this->validateVariant( $language ); 473 if ( $this->mHeaderVariant ) { 474 break; 475 } 476 } 477 } 478 479 return $this->mHeaderVariant; 480 } 481 482 /** 483 * Dictionary-based conversion. 484 * This function would not parse the conversion rules. 485 * If you want to parse rules, try to use convert() or 486 * convertTo(). 487 * 488 * @param string $text The text to be converted 489 * @param bool|string $toVariant The target language code 490 * @return string The converted text 491 */ 492 public function autoConvert( $text, $toVariant = false ) { 493 $this->loadTables(); 494 495 if ( !$toVariant ) { 496 $toVariant = $this->getPreferredVariant(); 497 if ( !$toVariant ) { 498 return $text; 499 } 500 } 501 502 if ( $this->guessVariant( $text, $toVariant ) ) { 503 return $text; 504 } 505 /* we convert everything except: 506 1. HTML markups (anything between < and >) 507 2. HTML entities 508 3. placeholders created by the parser 509 IMPORTANT: Beware of failure from pcre.backtrack_limit (T124404). 510 Minimize use of backtracking where possible. 511 */ 512 static $reg; 513 if ( $reg === null ) { 514 $marker = '|' . Parser::MARKER_PREFIX . '[^\x7f]++\x7f'; 515 516 // this one is needed when the text is inside an HTML markup 517 $htmlfix = '|<[^>\004]++(?=\004$)|^[^<>]*+>'; 518 519 // Optimize for the common case where these tags have 520 // few or no children. Thus try and possesively get as much as 521 // possible, and only engage in backtracking when we hit a '<'. 522 523 // disable convert to variants between <code> tags 524 $codefix = '<code>[^<]*+(?:(?:(?!<\/code>).)[^<]*+)*+<\/code>|'; 525 // disable conversion of <script> tags 526 $scriptfix = '<script[^>]*+>[^<]*+(?:(?:(?!<\/script>).)[^<]*+)*+<\/script>|'; 527 // disable conversion of <pre> tags 528 $prefix = '<pre[^>]*+>[^<]*+(?:(?:(?!<\/pre>).)[^<]*+)*+<\/pre>|'; 529 // The "|.*+)" at the end, is in case we missed some part of html syntax, 530 // we will fail securely (hopefully) by matching the rest of the string. 531 $htmlFullTag = '<(?:[^>=]*+(?>[^>=]*+=\s*+(?:"[^"]*"|\'[^\']*\'|[^\'">\s]*+))*+[^>=]*+>|.*+)|'; 532 533 $reg = '/' . $codefix . $scriptfix . $prefix . $htmlFullTag . 534 '&[a-zA-Z#][a-z0-9]++;' . $marker . $htmlfix . '|\004$/s'; 535 } 536 $startPos = 0; 537 $sourceBlob = ''; 538 $literalBlob = ''; 539 540 // Guard against delimiter nulls in the input 541 // (should never happen: see T159174) 542 $text = str_replace( "\000", '', $text ); 543 $text = str_replace( "\004", '', $text ); 544 545 $markupMatches = null; 546 $elementMatches = null; 547 548 // We add a marker (\004) at the end of text, to ensure we always match the 549 // entire text (Otherwise, pcre.backtrack_limit might cause silent failure) 550 $textWithMarker = $text . "\004"; 551 while ( $startPos < strlen( $text ) ) { 552 if ( preg_match( $reg, $textWithMarker, $markupMatches, PREG_OFFSET_CAPTURE, $startPos ) ) { 553 $elementPos = $markupMatches[0][1]; 554 $element = $markupMatches[0][0]; 555 if ( $element === "\004" ) { 556 // We hit the end. 557 $elementPos = strlen( $text ); 558 $element = ''; 559 } elseif ( substr( $element, -1 ) === "\004" ) { 560 // This can sometimes happen if we have 561 // unclosed html tags (For example 562 // when converting a title attribute 563 // during a recursive call that contains 564 // a < e.g. <div title="<">. 565 $element = substr( $element, 0, -1 ); 566 } 567 } else { 568 // If we hit here, then Language Converter could be tricked 569 // into doing an XSS, so we refuse to translate. 570 // If non-crazy input manages to reach this code path, 571 // we should consider it a bug. 572 $log = LoggerFactory::getInstance( 'languageconverter' ); 573 $log->error( "Hit pcre.backtrack_limit in " . __METHOD__ 574 . ". Disabling language conversion for this page.", 575 [ 576 "method" => __METHOD__, 577 "variant" => $toVariant, 578 "startOfText" => substr( $text, 0, 500 ) 579 ] 580 ); 581 return $text; 582 } 583 // Queue the part before the markup for translation in a batch 584 $sourceBlob .= substr( $text, $startPos, $elementPos - $startPos ) . "\000"; 585 586 // Advance to the next position 587 $startPos = $elementPos + strlen( $element ); 588 589 // Translate any alt or title attributes inside the matched element 590 if ( $element !== '' 591 && preg_match( '/^(<[^>\s]*+)\s([^>]*+)(.*+)$/', $element, $elementMatches ) 592 ) { 593 // FIXME, this decodes entities, so if you have something 594 // like <div title="foo<bar"> the bar won't get 595 // translated since after entity decoding it looks like 596 // unclosed html and we call this method recursively 597 // on attributes. 598 $attrs = Sanitizer::decodeTagAttributes( $elementMatches[2] ); 599 // Ensure self-closing tags stay self-closing. 600 $close = substr( $elementMatches[2], -1 ) === '/' ? ' /' : ''; 601 $changed = false; 602 foreach ( [ 'title', 'alt' ] as $attrName ) { 603 if ( !isset( $attrs[$attrName] ) ) { 604 continue; 605 } 606 $attr = $attrs[$attrName]; 607 // Don't convert URLs 608 if ( !strpos( $attr, '://' ) ) { 609 $attr = $this->recursiveConvertTopLevel( $attr, $toVariant ); 610 } 611 612 if ( $attr !== $attrs[$attrName] ) { 613 $attrs[$attrName] = $attr; 614 $changed = true; 615 } 616 } 617 if ( $changed ) { 618 // @phan-suppress-next-line SecurityCheck-DoubleEscaped Explained above with decodeTagAttributes 619 $element = $elementMatches[1] . Html::expandAttributes( $attrs ) . 620 $close . $elementMatches[3]; 621 } 622 } 623 $literalBlob .= $element . "\000"; 624 } 625 626 // Do the main translation batch 627 $translatedBlob = $this->translate( $sourceBlob, $toVariant ); 628 629 // Put the output back together 630 $translatedIter = StringUtils::explode( "\000", $translatedBlob ); 631 $literalIter = StringUtils::explode( "\000", $literalBlob ); 632 $output = ''; 633 while ( $translatedIter->valid() && $literalIter->valid() ) { 634 $output .= $translatedIter->current(); 635 $output .= $literalIter->current(); 636 $translatedIter->next(); 637 $literalIter->next(); 638 } 639 640 return $output; 641 } 642 643 /** 644 * Translate a string to a variant. 645 * Doesn't parse rules or do any of that other stuff, for that use 646 * convert() or convertTo(). 647 * 648 * @param string $text Text to convert 649 * @param string $variant Variant language code 650 * @return string Translated text 651 */ 652 public function translate( $text, $variant ) { 653 // If $text is empty or only includes spaces, do nothing 654 // Otherwise translate it 655 if ( trim( $text ) ) { 656 $this->loadTables(); 657 $text = $this->mTables[$variant]->replace( $text ); 658 } 659 return $text; 660 } 661 662 /** 663 * Call translate() to convert text to all valid variants. 664 * 665 * @param string $text The text to be converted 666 * @return array Variant => converted text 667 */ 668 public function autoConvertToAllVariants( $text ) { 669 $this->loadTables(); 670 671 $ret = []; 672 foreach ( $this->getVariants() as $variant ) { 673 $ret[$variant] = $this->translate( $text, $variant ); 674 } 675 676 return $ret; 677 } 678 679 /** 680 * Apply manual conversion rules. 681 * 682 * @param ConverterRule $convRule 683 */ 684 protected function applyManualConv( ConverterRule $convRule ) { 685 // Use syntax -{T|zh-cn:TitleCN; zh-tw:TitleTw}- to custom 686 // title conversion. 687 // T26072: $mConvRuleTitle was overwritten by other manual 688 // rule(s) not for title, this breaks the title conversion. 689 $newConvRuleTitle = $convRule->getTitle(); 690 if ( $newConvRuleTitle ) { 691 // So I add an empty check for getTitle() 692 $this->mConvRuleTitle = $newConvRuleTitle; 693 } 694 695 // merge/remove manual conversion rules to/from global table 696 $convTable = $convRule->getConvTable(); 697 $action = $convRule->getRulesAction(); 698 foreach ( $convTable as $variant => $pair ) { 699 $v = $this->validateVariant( $variant ); 700 if ( !$v ) { 701 continue; 702 } 703 704 if ( $action == 'add' ) { 705 // More efficient than array_merge(), about 2.5 times. 706 foreach ( $pair as $from => $to ) { 707 $this->mTables[$v]->setPair( $from, $to ); 708 } 709 } elseif ( $action == 'remove' ) { 710 $this->mTables[$v]->removeArray( $pair ); 711 } 712 } 713 } 714 715 /** 716 * Auto convert a LinkTarget object to a readable string in the 717 * preferred variant. 718 * 719 * @param LinkTarget $linkTarget 720 * @return string Converted title text 721 */ 722 public function convertTitle( LinkTarget $linkTarget ) { 723 $variant = $this->getPreferredVariant(); 724 $index = $linkTarget->getNamespace(); 725 if ( $index !== NS_MAIN ) { 726 $text = $this->convertNamespace( $index, $variant ) . ':'; 727 } else { 728 $text = ''; 729 } 730 $text .= $this->translate( $linkTarget->getText(), $variant ); 731 732 return $text; 733 } 734 735 /** 736 * Get the namespace display name in the preferred variant. 737 * 738 * @param int $index Namespace id 739 * @param string|null $variant Variant code or null for preferred variant 740 * @return string Namespace name for display 741 */ 742 public function convertNamespace( $index, $variant = null ) { 743 if ( $index === NS_MAIN ) { 744 return ''; 745 } 746 747 if ( $variant === null ) { 748 $variant = $this->getPreferredVariant(); 749 } 750 751 $cache = MediaWikiServices::getInstance()->getLocalServerObjectCache(); 752 $key = $cache->makeKey( 'languageconverter', 'namespace-text', $index, $variant ); 753 $nsVariantText = $cache->get( $key ); 754 if ( $nsVariantText !== false ) { 755 return $nsVariantText; 756 } 757 758 // First check if a message gives a converted name in the target variant. 759 $nsConvMsg = wfMessage( 'conversion-ns' . $index )->inLanguage( $variant ); 760 if ( $nsConvMsg->exists() ) { 761 $nsVariantText = $nsConvMsg->plain(); 762 } 763 764 // Then check if a message gives a converted name in content language 765 // which needs extra translation to the target variant. 766 if ( $nsVariantText === false ) { 767 $nsConvMsg = wfMessage( 'conversion-ns' . $index )->inContentLanguage(); 768 if ( $nsConvMsg->exists() ) { 769 $nsVariantText = $this->translate( $nsConvMsg->plain(), $variant ); 770 } 771 } 772 773 if ( $nsVariantText === false ) { 774 // No message exists, retrieve it from the target variant's namespace names. 775 $mLangObj = MediaWikiServices::getInstance() 776 ->getLanguageFactory() 777 ->getLanguage( $variant ); 778 $nsVariantText = $mLangObj->getFormattedNsText( $index ); 779 } 780 781 $cache->set( $key, $nsVariantText, 60 ); 782 783 return $nsVariantText; 784 } 785 786 /** 787 * Convert text to different variants of a language. The automatic 788 * conversion is done in autoConvert(). Here we parse the text 789 * marked with -{}-, which specifies special conversions of the 790 * text that can not be accomplished in autoConvert(). 791 * 792 * Syntax of the markup: 793 * -{code1:text1;code2:text2;...}- or 794 * -{flags|code1:text1;code2:text2;...}- or 795 * -{text}- in which case no conversion should take place for text 796 * 797 * @warning Glossary state is maintained between calls. Never feed this 798 * method input that hasn't properly been escaped as it may result in 799 * an XSS in subsequent calls, even if those subsequent calls properly 800 * escape things. 801 * @param string $text Text to be converted, already html escaped. 802 * @return string Converted text (html) 803 */ 804 public function convert( $text ) { 805 $variant = $this->getPreferredVariant(); 806 return $this->convertTo( $text, $variant ); 807 } 808 809 /** 810 * Same as convert() except a extra parameter to custom variant. 811 * 812 * @param string $text Text to be converted, already html escaped 813 * @param-taint $text exec_html 814 * @param string $variant The target variant code 815 * @return string Converted text 816 * @return-taint escaped 817 */ 818 public function convertTo( $text, $variant ) { 819 $languageConverterFactory = MediaWikiServices::getInstance()->getLanguageConverterFactory(); 820 if ( $languageConverterFactory->isConversionDisabled() ) { 821 return $text; 822 } 823 // Reset converter state for a new converter run. 824 $this->mConvRuleTitle = false; 825 return $this->recursiveConvertTopLevel( $text, $variant ); 826 } 827 828 /** 829 * Recursively convert text on the outside. Allow to use nested 830 * markups to custom rules. 831 * 832 * @param string $text Text to be converted 833 * @param string $variant The target variant code 834 * @param int $depth Depth of recursion 835 * @return string Converted text 836 */ 837 protected function recursiveConvertTopLevel( $text, $variant, $depth = 0 ) { 838 $startPos = 0; 839 $out = ''; 840 $length = strlen( $text ); 841 $shouldConvert = !$this->guessVariant( $text, $variant ); 842 $continue = true; 843 844 $noScript = '<script.*?>.*?<\/script>(*SKIP)(*FAIL)'; 845 $noStyle = '<style.*?>.*?<\/style>(*SKIP)(*FAIL)'; 846 // phpcs:ignore Generic.Files.LineLength 847 $noHtml = '<(?:[^>=]*+(?>[^>=]*+=\s*+(?:"[^"]*"|\'[^\']*\'|[^\'">\s]*+))*+[^>=]*+>|.*+)(*SKIP)(*FAIL)'; 848 while ( $startPos < $length && $continue ) { 849 $continue = preg_match( 850 // Only match -{ outside of html. 851 "/$noScript|$noStyle|$noHtml|-\{/", 852 $text, 853 $m, 854 PREG_OFFSET_CAPTURE, 855 $startPos 856 ); 857 858 if ( !$continue ) { 859 // No more markup, append final segment 860 $fragment = substr( $text, $startPos ); 861 $out .= $shouldConvert ? $this->autoConvert( $fragment, $variant ) : $fragment; 862 return $out; 863 } 864 865 // Offset of the match of the regex pattern. 866 $pos = $m[0][1]; 867 868 // Append initial segment 869 $fragment = substr( $text, $startPos, $pos - $startPos ); 870 $out .= $shouldConvert ? $this->autoConvert( $fragment, $variant ) : $fragment; 871 // -{ marker found, not in attribute 872 // Advance position up to -{ marker. 873 $startPos = $pos; 874 // Do recursive conversion 875 // Note: This passes $startPos by reference, and advances it. 876 $out .= $this->recursiveConvertRule( $text, $variant, $startPos, $depth + 1 ); 877 } 878 return $out; 879 } 880 881 /** 882 * Recursively convert text on the inside. 883 * 884 * @param string $text Text to be converted 885 * @param string $variant The target variant code 886 * @param int &$startPos 887 * @param int $depth Depth of recursion 888 * 889 * @throws MWException 890 * @return string Converted text 891 */ 892 protected function recursiveConvertRule( $text, $variant, &$startPos, $depth = 0 ) { 893 // Quick sanity check (no function calls) 894 if ( $text[$startPos] !== '-' || $text[$startPos + 1] !== '{' ) { 895 throw new MWException( __METHOD__ . ': invalid input string' ); 896 } 897 898 $startPos += 2; 899 $inner = ''; 900 $warningDone = false; 901 $length = strlen( $text ); 902 903 while ( $startPos < $length ) { 904 $m = false; 905 preg_match( '/-\{|\}-/', $text, $m, PREG_OFFSET_CAPTURE, $startPos ); 906 if ( !$m ) { 907 // Unclosed rule 908 break; 909 } 910 911 $token = $m[0][0]; 912 $pos = $m[0][1]; 913 914 // Markup found 915 // Append initial segment 916 $inner .= substr( $text, $startPos, $pos - $startPos ); 917 918 // Advance position 919 $startPos = $pos; 920 921 switch ( $token ) { 922 case '-{': 923 // Check max depth 924 if ( $depth >= $this->mMaxDepth ) { 925 $inner .= '-{'; 926 if ( !$warningDone ) { 927 $inner .= '<span class="error">' . 928 wfMessage( 'language-converter-depth-warning' ) 929 ->numParams( $this->mMaxDepth )->inContentLanguage()->text() . 930 '</span>'; 931 $warningDone = true; 932 } 933 $startPos += 2; 934 break; 935 } 936 // Recursively parse another rule 937 $inner .= $this->recursiveConvertRule( $text, $variant, $startPos, $depth + 1 ); 938 break; 939 case '}-': 940 // Apply the rule 941 $startPos += 2; 942 $rule = new ConverterRule( $inner, $this ); 943 $rule->parse( $variant ); 944 $this->applyManualConv( $rule ); 945 return $rule->getDisplay(); 946 default: 947 throw new MWException( __METHOD__ . ': invalid regex match' ); 948 } 949 } 950 951 // Unclosed rule 952 if ( $startPos < $length ) { 953 $inner .= substr( $text, $startPos ); 954 } 955 $startPos = $length; 956 return '-{' . $this->autoConvert( $inner, $variant ); 957 } 958 959 /** 960 * If a language supports multiple variants, it is possible that 961 * non-existing link in one variant actually exists in another variant. 962 * This function tries to find it. See e.g. LanguageZh.php 963 * The input parameters may be modified upon return 964 * 965 * @param string &$link The name of the link 966 * @param Title &$nt The title object of the link 967 * @param bool $ignoreOtherCond To disable other conditions when 968 * we need to transclude a template or update a category's link 969 */ 970 public function findVariantLink( &$link, &$nt, $ignoreOtherCond = false ) { 971 # If the article has already existed, there is no need to 972 # check it again, otherwise it may cause a fault. 973 if ( is_object( $nt ) && $nt->exists() ) { 974 return; 975 } 976 977 global $wgRequest; 978 $isredir = $wgRequest->getText( 'redirect', 'yes' ); 979 $action = $wgRequest->getText( 'action' ); 980 if ( $action == 'edit' && $wgRequest->getBool( 'redlink' ) ) { 981 $action = 'view'; 982 } 983 $linkconvert = $wgRequest->getText( 'linkconvert', 'yes' ); 984 $disableLinkConversion = 985 MediaWikiServices::getInstance()->getLanguageConverterFactory() 986 ->isLinkConversionDisabled(); 987 $linkBatchFactory = MediaWikiServices::getInstance()->getLinkBatchFactory(); 988 $linkBatch = $linkBatchFactory->newLinkBatch(); 989 990 $ns = NS_MAIN; 991 992 if ( $disableLinkConversion || 993 ( !$ignoreOtherCond && 994 ( $isredir == 'no' 995 || $action == 'edit' 996 || $action == 'submit' 997 || $linkconvert == 'no' ) ) ) { 998 return; 999 } 1000 1001 if ( is_object( $nt ) ) { 1002 $ns = $nt->getNamespace(); 1003 } 1004 1005 $variants = $this->autoConvertToAllVariants( $link ); 1006 if ( !$variants ) { // give up 1007 return; 1008 } 1009 1010 $titles = []; 1011 1012 foreach ( $variants as $v ) { 1013 if ( $v != $link ) { 1014 $varnt = Title::newFromText( $v, $ns ); 1015 if ( $varnt !== null ) { 1016 $linkBatch->addObj( $varnt ); 1017 $titles[] = $varnt; 1018 } 1019 } 1020 } 1021 1022 // fetch all variants in single query 1023 $linkBatch->execute(); 1024 1025 foreach ( $titles as $varnt ) { 1026 if ( $varnt->getArticleID() > 0 ) { 1027 $nt = $varnt; 1028 $link = $varnt->getText(); 1029 break; 1030 } 1031 } 1032 } 1033 1034 /** 1035 * Returns language specific hash options. 1036 * 1037 * @return string 1038 */ 1039 public function getExtraHashOptions() { 1040 $variant = $this->getPreferredVariant(); 1041 1042 return '!' . $variant; 1043 } 1044 1045 /** 1046 * Guess if a text is written in a variant. This should be implemented in subclasses. 1047 * 1048 * @param string $text The text to be checked 1049 * @param string $variant Language code of the variant to be checked for 1050 * @return bool True if $text appears to be written in $variant, false if not 1051 * 1052 * @author Nikola Smolenski <smolensk@eunet.rs> 1053 * @since 1.19 1054 */ 1055 public function guessVariant( $text, $variant ) { 1056 return false; 1057 } 1058 1059 /** 1060 * Load default conversion tables. 1061 * This method must be implemented in derived class. 1062 * 1063 * @throws MWException 1064 */ 1065 protected function loadDefaultTables() { 1066 $class = static::class; 1067 throw new MWException( "Must implement loadDefaultTables() method in class $class" ); 1068 } 1069 1070 /** 1071 * Load conversion tables either from the cache or the disk. 1072 * @private 1073 * @param bool $fromCache Load from memcached? Defaults to true. 1074 */ 1075 protected function loadTables( $fromCache = true ) { 1076 global $wgLanguageConverterCacheType; 1077 1078 if ( $this->mTablesLoaded ) { 1079 return; 1080 } 1081 1082 $this->mTablesLoaded = true; 1083 // Do not use null as starting value, as that would confuse phan a lot. 1084 $this->mTables = []; 1085 $cache = ObjectCache::getInstance( $wgLanguageConverterCacheType ); 1086 $cacheKey = $cache->makeKey( 'conversiontables', $this->getMainCode() ); 1087 if ( $fromCache ) { 1088 $this->mTables = $cache->get( $cacheKey ); 1089 } 1090 if ( !$this->mTables || !array_key_exists( self::CACHE_VERSION_KEY, $this->mTables ) ) { 1091 // not in cache, or we need a fresh reload. 1092 // We will first load the default tables 1093 // then update them using things in MediaWiki:Conversiontable/* 1094 $this->loadDefaultTables(); 1095 foreach ( $this->getVariants() as $var ) { 1096 $cached = $this->parseCachedTable( $var ); 1097 // @phan-suppress-next-next-line PhanTypeArraySuspiciousNullable 1098 // FIXME: $this->mTables could theoretically be null here 1099 $this->mTables[$var]->mergeArray( $cached ); 1100 } 1101 1102 $this->postLoadTables(); 1103 $this->mTables[self::CACHE_VERSION_KEY] = true; 1104 1105 $cache->set( $cacheKey, $this->mTables, 43200 ); 1106 } 1107 } 1108 1109 /** 1110 * Hook for post processing after conversion tables are loaded. 1111 */ 1112 protected function postLoadTables() { 1113 } 1114 1115 /** 1116 * Reload the conversion tables. 1117 * 1118 * Also used by test suites which need to reset the converter state. 1119 * 1120 * @private 1121 */ 1122 private function reloadTables() { 1123 if ( $this->mTables ) { 1124 // @phan-suppress-next-line PhanTypeObjectUnsetDeclaredProperty 1125 unset( $this->mTables ); 1126 } 1127 1128 $this->mTablesLoaded = false; 1129 $this->loadTables( false ); 1130 } 1131 1132 /** 1133 * Parse the conversion table stored in the cache. 1134 * 1135 * The tables should be in blocks of the following form: 1136 * -{ 1137 * word => word ; 1138 * word => word ; 1139 * ... 1140 * }- 1141 * 1142 * To make the tables more manageable, subpages are allowed 1143 * and will be parsed recursively if $recursive == true. 1144 * 1145 * @param string $code Language code 1146 * @param string $subpage Subpage name 1147 * @param bool $recursive Parse subpages recursively? Defaults to true. 1148 * 1149 * @return array 1150 */ 1151 private function parseCachedTable( $code, $subpage = '', $recursive = true ) { 1152 static $parsed = []; 1153 1154 $key = 'Conversiontable/' . $code; 1155 if ( $subpage ) { 1156 $key .= '/' . $subpage; 1157 } 1158 if ( array_key_exists( $key, $parsed ) ) { 1159 return []; 1160 } 1161 1162 $parsed[$key] = true; 1163 1164 if ( $subpage === '' ) { 1165 $messageCache = MediaWikiServices::getInstance()->getMessageCache(); 1166 $txt = $messageCache->getMsgFromNamespace( $key, $code ); 1167 } else { 1168 $txt = false; 1169 $title = Title::makeTitleSafe( NS_MEDIAWIKI, $key ); 1170 if ( $title && $title->exists() ) { 1171 $revision = MediaWikiServices::getInstance() 1172 ->getRevisionLookup() 1173 ->getRevisionByTitle( $title ); 1174 if ( $revision ) { 1175 $model = $revision->getSlot( 1176 SlotRecord::MAIN, 1177 RevisionRecord::RAW 1178 )->getModel(); 1179 if ( $model == CONTENT_MODEL_WIKITEXT ) { 1180 // @phan-suppress-next-line PhanUndeclaredMethod 1181 $txt = $revision->getContent( 1182 SlotRecord::MAIN, 1183 RevisionRecord::RAW 1184 )->getText(); 1185 } 1186 1187 // @todo in the future, use a specialized content model, perhaps based on json! 1188 } 1189 } 1190 } 1191 1192 # Nothing to parse if there's no text 1193 if ( $txt === false || $txt === null || $txt === '' ) { 1194 return []; 1195 } 1196 1197 // get all subpage links of the form 1198 // [[MediaWiki:Conversiontable/zh-xx/...|...]] 1199 $linkhead = $this->mLangObj->getNsText( NS_MEDIAWIKI ) . 1200 ':Conversiontable'; 1201 $subs = StringUtils::explode( '[[', $txt ); 1202 $sublinks = []; 1203 foreach ( $subs as $sub ) { 1204 $link = explode( ']]', $sub, 2 ); 1205 if ( count( $link ) != 2 ) { 1206 continue; 1207 } 1208 $b = explode( '|', $link[0], 2 ); 1209 $b = explode( '/', trim( $b[0] ), 3 ); 1210 if ( count( $b ) == 3 ) { 1211 $sublink = $b[2]; 1212 } else { 1213 $sublink = ''; 1214 } 1215 1216 if ( $b[0] == $linkhead && $b[1] == $code ) { 1217 $sublinks[] = $sublink; 1218 } 1219 } 1220 1221 // parse the mappings in this page 1222 $blocks = StringUtils::explode( '-{', $txt ); 1223 $ret = []; 1224 $first = true; 1225 foreach ( $blocks as $block ) { 1226 if ( $first ) { 1227 // Skip the part before the first -{ 1228 $first = false; 1229 continue; 1230 } 1231 $mappings = explode( '}-', $block, 2 )[0]; 1232 $stripped = str_replace( [ "'", '"', '*', '#' ], '', $mappings ); 1233 $table = StringUtils::explode( ';', $stripped ); 1234 foreach ( $table as $t ) { 1235 $m = explode( '=>', $t, 3 ); 1236 if ( count( $m ) != 2 ) { 1237 continue; 1238 } 1239 // trim any trailling comments starting with '//' 1240 $tt = explode( '//', $m[1], 2 ); 1241 $ret[trim( $m[0] )] = trim( $tt[0] ); 1242 } 1243 } 1244 1245 // recursively parse the subpages 1246 if ( $recursive ) { 1247 foreach ( $sublinks as $link ) { 1248 $s = $this->parseCachedTable( $code, $link, $recursive ); 1249 $ret = $s + $ret; 1250 } 1251 } 1252 1253 if ( $this->mUcfirst ) { 1254 foreach ( $ret as $k => $v ) { 1255 $ret[$this->mLangObj->ucfirst( $k )] = $this->mLangObj->ucfirst( $v ); 1256 } 1257 } 1258 return $ret; 1259 } 1260 1261 /** 1262 * Enclose a string with the "no conversion" tag. This is used by 1263 * various functions in the Parser. 1264 * 1265 * @param string $text Text to be tagged for no conversion 1266 * @param bool $noParse Unused 1267 * @return string The tagged text 1268 */ 1269 public function markNoConversion( $text, $noParse = false ) { 1270 # don't mark if already marked 1271 if ( strpos( $text, '-{' ) || strpos( $text, '}-' ) ) { 1272 return $text; 1273 } 1274 1275 $ret = "-{R|$text}-"; 1276 return $ret; 1277 } 1278 1279 /** 1280 * Convert the sorting key for category links. This should make different 1281 * keys that are variants of each other map to the same key. 1282 * 1283 * @param string $key 1284 * 1285 * @return string 1286 */ 1287 public function convertCategoryKey( $key ) { 1288 return $key; 1289 } 1290 1291 /** 1292 * Refresh the cache of conversion tables when 1293 * MediaWiki:Conversiontable* is updated. 1294 * 1295 * @param LinkTarget $linkTarget The LinkTarget of the page being updated 1296 */ 1297 public function updateConversionTable( LinkTarget $linkTarget ) { 1298 if ( $linkTarget->getNamespace() === NS_MEDIAWIKI ) { 1299 $t = explode( '/', $linkTarget->getDBkey(), 3 ); 1300 $c = count( $t ); 1301 if ( $c > 1 && $t[0] == 'Conversiontable' ) { 1302 if ( $this->validateVariant( $t[1] ) ) { 1303 $this->reloadTables(); 1304 } 1305 } 1306 } 1307 } 1308 1309 /** 1310 * Get the cached separator pattern for ConverterRule::parseRules() 1311 * @return string 1312 */ 1313 public function getVarSeparatorPattern() { 1314 if ( $this->mVarSeparatorPattern === null ) { 1315 // varsep_pattern for preg_split: 1316 // text should be splited by ";" only if a valid variant 1317 // name exist after the markup, for example: 1318 // -{zh-hans:<span style="font-size:120%;">xxx</span>;zh-hant:\ 1319 // <span style="font-size:120%;">yyy</span>;}- 1320 // we should split it as: 1321 // [ 1322 // [0] => 'zh-hans:<span style="font-size:120%;">xxx</span>' 1323 // [1] => 'zh-hant:<span style="font-size:120%;">yyy</span>' 1324 // [2] => '' 1325 // ] 1326 $expandedVariants = []; 1327 foreach ( $this->getVariants() as $variant ) { 1328 $expandedVariants[ $variant ] = 1; 1329 // Accept standard BCP 47 names for variants as well. 1330 $expandedVariants[ LanguageCode::bcp47( $variant ) ] = 1; 1331 } 1332 // Accept old deprecated names for variants 1333 foreach ( LanguageCode::getDeprecatedCodeMapping() as $old => $new ) { 1334 if ( isset( $expandedVariants[ $new ] ) ) { 1335 $expandedVariants[ $old ] = 1; 1336 } 1337 } 1338 1339 $pat = '/;\s*(?='; 1340 foreach ( $expandedVariants as $variant => $ignore ) { 1341 // zh-hans:xxx;zh-hant:yyy 1342 $pat .= $variant . '\s*:|'; 1343 // xxx=>zh-hans:yyy; xxx=>zh-hant:zzz 1344 $pat .= '[^;]*?=>\s*' . $variant . '\s*:|'; 1345 } 1346 $pat .= '\s*$)/'; 1347 $this->mVarSeparatorPattern = $pat; 1348 } 1349 return $this->mVarSeparatorPattern; 1350 } 1351 1352 /** 1353 * Check if this is a language with variants 1354 * 1355 * @since 1.35 1356 * 1357 * @return bool 1358 */ 1359 public function hasVariants() { 1360 return count( $this->getVariants() ) > 1; 1361 } 1362 1363 /** 1364 * Strict check if the language has the specific variant. 1365 * 1366 * Compare to LanguageConverter::validateVariant() which does a more 1367 * lenient check and attempts to coerce the given code to a valid one. 1368 * 1369 * @since 1.35 1370 * @param string $variant 1371 * @return bool 1372 */ 1373 public function hasVariant( $variant ) { 1374 return $variant && ( $variant === $this->validateVariant( $variant ) ); 1375 } 1376 1377 /** 1378 * Perform output conversion on a string, and encode for safe HTML output. 1379 * 1380 * @since 1.35 1381 * 1382 * @param string $text Text to be converted 1383 * @return string 1384 */ 1385 public function convertHtml( $text ) { 1386 // @phan-suppress-next-line SecurityCheck-DoubleEscaped convert() is documented to return html 1387 return htmlspecialchars( $this->convert( $text ) ); 1388 } 1389} 1390