Bump and prep 1.34.1
[lhc/web/wiklou.git] / languages / LanguageConverter.php
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 * @ingroup Language
20 */
21 use MediaWiki\MediaWikiServices;
22
23 use MediaWiki\Logger\LoggerFactory;
24 use MediaWiki\Revision\RevisionRecord;
25
26 /**
27 * Base class for language conversion.
28 * @ingroup Language
29 *
30 * @author Zhengzhu Feng <zhengzhu@gmail.com>
31 * @author fdcn <fdcn64@gmail.com>
32 * @author shinjiman <shinjiman@gmail.com>
33 * @author PhiLiP <philip.npc@gmail.com>
34 */
35 class LanguageConverter {
36 /**
37 * languages supporting variants
38 * @since 1.20
39 * @var array
40 */
41 public static $languagesWithVariants = [
42 'en',
43 'crh',
44 'gan',
45 'iu',
46 'kk',
47 'ku',
48 'shi',
49 'sr',
50 'tg',
51 'uz',
52 'zh',
53 ];
54
55 public $mMainLanguageCode;
56
57 /**
58 * @var string[]
59 */
60 public $mVariants;
61 public $mVariantFallbacks;
62 public $mVariantNames;
63 public $mTablesLoaded = false;
64
65 /**
66 * @var ReplacementArray[]|bool[]
67 */
68 public $mTables;
69
70 // 'bidirectional' 'unidirectional' 'disable' for each variant
71 public $mManualLevel;
72
73 public $mLangObj;
74 public $mFlags;
75 public $mDescCodeSep = ':', $mDescVarSep = ';';
76 public $mUcfirst = false;
77 public $mConvRuleTitle = false;
78 public $mURLVariant;
79 public $mUserVariant;
80 public $mHeaderVariant;
81 public $mMaxDepth = 10;
82 public $mVarSeparatorPattern;
83
84 const CACHE_VERSION_KEY = 'VERSION 7';
85
86 /**
87 * @param Language $langobj
88 * @param string $maincode The main language code of this language
89 * @param string[] $variants The supported variants of this language
90 * @param array $variantfallbacks The fallback language of each variant
91 * @param array $flags Defining the custom strings that maps to the flags
92 * @param array $manualLevel Limit for supported variants
93 */
94 public function __construct( Language $langobj, $maincode, $variants = [],
95 $variantfallbacks = [], $flags = [],
96 $manualLevel = [] ) {
97 global $wgDisabledVariants;
98 $this->mLangObj = $langobj;
99 $this->mMainLanguageCode = $maincode;
100 $this->mVariants = array_diff( $variants, $wgDisabledVariants );
101 $this->mVariantFallbacks = $variantfallbacks;
102 $this->mVariantNames = Language::fetchLanguageNames();
103 $defaultflags = [
104 // 'S' show converted text
105 // '+' add rules for alltext
106 // 'E' the gave flags is error
107 // these flags above are reserved for program
108 'A' => 'A', // add rule for convert code (all text convert)
109 'T' => 'T', // title convert
110 'R' => 'R', // raw content
111 'D' => 'D', // convert description (subclass implement)
112 '-' => '-', // remove convert (not implement)
113 'H' => 'H', // add rule for convert code (but no display in placed code)
114 'N' => 'N', // current variant name
115 ];
116 $this->mFlags = array_merge( $defaultflags, $flags );
117 foreach ( $this->mVariants as $v ) {
118 if ( array_key_exists( $v, $manualLevel ) ) {
119 $this->mManualLevel[$v] = $manualLevel[$v];
120 } else {
121 $this->mManualLevel[$v] = 'bidirectional';
122 }
123 $this->mFlags[$v] = $v;
124 }
125 }
126
127 /**
128 * Get all valid variants.
129 * Call this instead of using $this->mVariants directly.
130 *
131 * @return string[] Contains all valid variants
132 */
133 public function getVariants() {
134 return $this->mVariants;
135 }
136
137 /**
138 * In case some variant is not defined in the markup, we need
139 * to have some fallback. For example, in zh, normally people
140 * will define zh-hans and zh-hant, but less so for zh-sg or zh-hk.
141 * when zh-sg is preferred but not defined, we will pick zh-hans
142 * in this case. Right now this is only used by zh.
143 *
144 * @param string $variant The language code of the variant
145 * @return string|array The code of the fallback language or the
146 * main code if there is no fallback
147 */
148 public function getVariantFallbacks( $variant ) {
149 return $this->mVariantFallbacks[$variant] ?? $this->mMainLanguageCode;
150 }
151
152 /**
153 * Get the title produced by the conversion rule.
154 * @return string The converted title text
155 */
156 public function getConvRuleTitle() {
157 return $this->mConvRuleTitle;
158 }
159
160 /**
161 * Get preferred language variant.
162 * @return string The preferred language code
163 */
164 public function getPreferredVariant() {
165 global $wgDefaultLanguageVariant, $wgUser;
166
167 $req = $this->getURLVariant();
168
169 Hooks::run( 'GetLangPreferredVariant', [ &$req ] );
170
171 if ( $wgUser->isSafeToLoad() && $wgUser->isLoggedIn() && !$req ) {
172 $req = $this->getUserVariant();
173 } elseif ( !$req ) {
174 $req = $this->getHeaderVariant();
175 }
176
177 if ( $wgDefaultLanguageVariant && !$req ) {
178 $req = $this->validateVariant( $wgDefaultLanguageVariant );
179 }
180
181 $req = $this->validateVariant( $req );
182
183 // This function, unlike the other get*Variant functions, is
184 // not memoized (i.e. there return value is not cached) since
185 // new information might appear during processing after this
186 // is first called.
187 if ( $req ) {
188 return $req;
189 }
190 return $this->mMainLanguageCode;
191 }
192
193 /**
194 * Get default variant.
195 * This function would not be affected by user's settings
196 * @return string The default variant code
197 */
198 public function getDefaultVariant() {
199 global $wgDefaultLanguageVariant;
200
201 $req = $this->getURLVariant();
202
203 if ( !$req ) {
204 $req = $this->getHeaderVariant();
205 }
206
207 if ( $wgDefaultLanguageVariant && !$req ) {
208 $req = $this->validateVariant( $wgDefaultLanguageVariant );
209 }
210
211 if ( $req ) {
212 return $req;
213 }
214 return $this->mMainLanguageCode;
215 }
216
217 /**
218 * Validate the variant and return an appropriate strict internal
219 * variant code if one exists. Compare to Language::hasVariant()
220 * which does a strict test.
221 *
222 * @param string|null $variant The variant to validate
223 * @return mixed Returns an equivalent valid variant code if possible,
224 * null otherwise
225 */
226 public function validateVariant( $variant = null ) {
227 if ( $variant === null ) {
228 return null;
229 }
230 // Our internal variants are always lower-case; the variant we
231 // are validating may have mixed case.
232 $variant = LanguageCode::replaceDeprecatedCodes( strtolower( $variant ) );
233 if ( in_array( $variant, $this->mVariants ) ) {
234 return $variant;
235 }
236 // Browsers are supposed to use BCP 47 standard in the
237 // Accept-Language header, but not all of our internal
238 // mediawiki variant codes are BCP 47. Map BCP 47 code
239 // to our internal code.
240 foreach ( $this->mVariants as $v ) {
241 // Case-insensitive match (BCP 47 is mixed case)
242 if ( strtolower( LanguageCode::bcp47( $v ) ) === $variant ) {
243 return $v;
244 }
245 }
246 return null;
247 }
248
249 /**
250 * Get the variant specified in the URL
251 *
252 * @return mixed Variant if one found, null otherwise
253 */
254 public function getURLVariant() {
255 global $wgRequest;
256
257 if ( $this->mURLVariant ) {
258 return $this->mURLVariant;
259 }
260
261 // see if the preference is set in the request
262 $ret = $wgRequest->getText( 'variant' );
263
264 if ( !$ret ) {
265 $ret = $wgRequest->getVal( 'uselang' );
266 }
267
268 $this->mURLVariant = $this->validateVariant( $ret );
269 return $this->mURLVariant;
270 }
271
272 /**
273 * Determine if the user has a variant set.
274 *
275 * @return mixed Variant if one found, null otherwise
276 */
277 protected function getUserVariant() {
278 global $wgUser;
279
280 // memoizing this function wreaks havoc on parserTest.php
281 /*
282 if ( $this->mUserVariant ) {
283 return $this->mUserVariant;
284 }
285 */
286
287 // Get language variant preference from logged in users
288 // Don't call this on stub objects because that causes infinite
289 // recursion during initialisation
290 if ( !$wgUser->isSafeToLoad() ) {
291 return false;
292 }
293 if ( $wgUser->isLoggedIn() ) {
294 if (
295 $this->mMainLanguageCode ==
296 MediaWikiServices::getInstance()->getContentLanguage()->getCode()
297 ) {
298 $ret = $wgUser->getOption( 'variant' );
299 } else {
300 $ret = $wgUser->getOption( 'variant-' . $this->mMainLanguageCode );
301 }
302 } else {
303 // figure out user lang without constructing wgLang to avoid
304 // infinite recursion
305 $ret = $wgUser->getOption( 'language' );
306 }
307
308 $this->mUserVariant = $this->validateVariant( $ret );
309 return $this->mUserVariant;
310 }
311
312 /**
313 * Determine the language variant from the Accept-Language header.
314 *
315 * @return mixed Variant if one found, null otherwise
316 */
317 protected function getHeaderVariant() {
318 global $wgRequest;
319
320 if ( $this->mHeaderVariant ) {
321 return $this->mHeaderVariant;
322 }
323
324 // See if some supported language variant is set in the
325 // HTTP header.
326 $languages = array_keys( $wgRequest->getAcceptLang() );
327 if ( empty( $languages ) ) {
328 return null;
329 }
330
331 $fallbackLanguages = [];
332 foreach ( $languages as $language ) {
333 $this->mHeaderVariant = $this->validateVariant( $language );
334 if ( $this->mHeaderVariant ) {
335 break;
336 }
337
338 // To see if there are fallbacks of current language.
339 // We record these fallback variants, and process
340 // them later.
341 $fallbacks = $this->getVariantFallbacks( $language );
342 if ( is_string( $fallbacks ) && $fallbacks !== $this->mMainLanguageCode ) {
343 $fallbackLanguages[] = $fallbacks;
344 } elseif ( is_array( $fallbacks ) ) {
345 $fallbackLanguages =
346 array_merge( $fallbackLanguages, $fallbacks );
347 }
348 }
349
350 if ( !$this->mHeaderVariant ) {
351 // process fallback languages now
352 $fallback_languages = array_unique( $fallbackLanguages );
353 foreach ( $fallback_languages as $language ) {
354 $this->mHeaderVariant = $this->validateVariant( $language );
355 if ( $this->mHeaderVariant ) {
356 break;
357 }
358 }
359 }
360
361 return $this->mHeaderVariant;
362 }
363
364 /**
365 * Dictionary-based conversion.
366 * This function would not parse the conversion rules.
367 * If you want to parse rules, try to use convert() or
368 * convertTo().
369 *
370 * @param string $text The text to be converted
371 * @param bool|string $toVariant The target language code
372 * @return string The converted text
373 */
374 public function autoConvert( $text, $toVariant = false ) {
375 $this->loadTables();
376
377 if ( !$toVariant ) {
378 $toVariant = $this->getPreferredVariant();
379 if ( !$toVariant ) {
380 return $text;
381 }
382 }
383
384 if ( $this->guessVariant( $text, $toVariant ) ) {
385 return $text;
386 }
387 /* we convert everything except:
388 1. HTML markups (anything between < and >)
389 2. HTML entities
390 3. placeholders created by the parser
391 IMPORTANT: Beware of failure from pcre.backtrack_limit (T124404).
392 Minimize use of backtracking where possible.
393 */
394 static $reg;
395 if ( $reg === null ) {
396 $marker = '|' . Parser::MARKER_PREFIX . '[^\x7f]++\x7f';
397
398 // this one is needed when the text is inside an HTML markup
399 $htmlfix = '|<[^>\004]++(?=\004$)|^[^<>]*+>';
400
401 // Optimize for the common case where these tags have
402 // few or no children. Thus try and possesively get as much as
403 // possible, and only engage in backtracking when we hit a '<'.
404
405 // disable convert to variants between <code> tags
406 $codefix = '<code>[^<]*+(?:(?:(?!<\/code>).)[^<]*+)*+<\/code>|';
407 // disable conversion of <script> tags
408 $scriptfix = '<script[^>]*+>[^<]*+(?:(?:(?!<\/script>).)[^<]*+)*+<\/script>|';
409 // disable conversion of <pre> tags
410 $prefix = '<pre[^>]*+>[^<]*+(?:(?:(?!<\/pre>).)[^<]*+)*+<\/pre>|';
411 // The "|.*+)" at the end, is in case we missed some part of html syntax,
412 // we will fail securely (hopefully) by matching the rest of the string.
413 $htmlFullTag = '<(?:[^>=]*+(?>[^>=]*+=\s*+(?:"[^"]*"|\'[^\']*\'|[^\'">\s]*+))*+[^>=]*+>|.*+)|';
414
415 $reg = '/' . $codefix . $scriptfix . $prefix . $htmlFullTag .
416 '&[a-zA-Z#][a-z0-9]++;' . $marker . $htmlfix . '|\004$/s';
417 }
418 $startPos = 0;
419 $sourceBlob = '';
420 $literalBlob = '';
421
422 // Guard against delimiter nulls in the input
423 // (should never happen: see T159174)
424 $text = str_replace( "\000", '', $text );
425 $text = str_replace( "\004", '', $text );
426
427 $markupMatches = null;
428 $elementMatches = null;
429
430 // We add a marker (\004) at the end of text, to ensure we always match the
431 // entire text (Otherwise, pcre.backtrack_limit might cause silent failure)
432 $textWithMarker = $text . "\004";
433 while ( $startPos < strlen( $text ) ) {
434 if ( preg_match( $reg, $textWithMarker, $markupMatches, PREG_OFFSET_CAPTURE, $startPos ) ) {
435 $elementPos = $markupMatches[0][1];
436 $element = $markupMatches[0][0];
437 if ( $element === "\004" ) {
438 // We hit the end.
439 $elementPos = strlen( $text );
440 $element = '';
441 } elseif ( substr( $element, -1 ) === "\004" ) {
442 // This can sometimes happen if we have
443 // unclosed html tags (For example
444 // when converting a title attribute
445 // during a recursive call that contains
446 // a &lt; e.g. <div title="&lt;">.
447 $element = substr( $element, 0, -1 );
448 }
449 } else {
450 // If we hit here, then Language Converter could be tricked
451 // into doing an XSS, so we refuse to translate.
452 // If non-crazy input manages to reach this code path,
453 // we should consider it a bug.
454 $log = LoggerFactory::getInstance( 'languageconverter' );
455 $log->error( "Hit pcre.backtrack_limit in " . __METHOD__
456 . ". Disabling language conversion for this page.",
457 [
458 "method" => __METHOD__,
459 "variant" => $toVariant,
460 "startOfText" => substr( $text, 0, 500 )
461 ]
462 );
463 return $text;
464 }
465 // Queue the part before the markup for translation in a batch
466 $sourceBlob .= substr( $text, $startPos, $elementPos - $startPos ) . "\000";
467
468 // Advance to the next position
469 $startPos = $elementPos + strlen( $element );
470
471 // Translate any alt or title attributes inside the matched element
472 if ( $element !== ''
473 && preg_match( '/^(<[^>\s]*+)\s([^>]*+)(.*+)$/', $element, $elementMatches )
474 ) {
475 // FIXME, this decodes entities, so if you have something
476 // like <div title="foo&lt;bar"> the bar won't get
477 // translated since after entity decoding it looks like
478 // unclosed html and we call this method recursively
479 // on attributes.
480 $attrs = Sanitizer::decodeTagAttributes( $elementMatches[2] );
481 // Ensure self-closing tags stay self-closing.
482 $close = substr( $elementMatches[2], -1 ) === '/' ? ' /' : '';
483 $changed = false;
484 foreach ( [ 'title', 'alt' ] as $attrName ) {
485 if ( !isset( $attrs[$attrName] ) ) {
486 continue;
487 }
488 $attr = $attrs[$attrName];
489 // Don't convert URLs
490 if ( !strpos( $attr, '://' ) ) {
491 $attr = $this->recursiveConvertTopLevel( $attr, $toVariant );
492 }
493
494 if ( $attr !== $attrs[$attrName] ) {
495 $attrs[$attrName] = $attr;
496 $changed = true;
497 }
498 }
499 if ( $changed ) {
500 $element = $elementMatches[1] . Html::expandAttributes( $attrs ) .
501 $close . $elementMatches[3];
502 }
503 }
504 $literalBlob .= $element . "\000";
505 }
506
507 // Do the main translation batch
508 $translatedBlob = $this->translate( $sourceBlob, $toVariant );
509
510 // Put the output back together
511 $translatedIter = StringUtils::explode( "\000", $translatedBlob );
512 $literalIter = StringUtils::explode( "\000", $literalBlob );
513 $output = '';
514 while ( $translatedIter->valid() && $literalIter->valid() ) {
515 $output .= $translatedIter->current();
516 $output .= $literalIter->current();
517 $translatedIter->next();
518 $literalIter->next();
519 }
520
521 return $output;
522 }
523
524 /**
525 * Translate a string to a variant.
526 * Doesn't parse rules or do any of that other stuff, for that use
527 * convert() or convertTo().
528 *
529 * @param string $text Text to convert
530 * @param string $variant Variant language code
531 * @return string Translated text
532 */
533 public function translate( $text, $variant ) {
534 // If $text is empty or only includes spaces, do nothing
535 // Otherwise translate it
536 if ( trim( $text ) ) {
537 $this->loadTables();
538 $text = $this->mTables[$variant]->replace( $text );
539 }
540 return $text;
541 }
542
543 /**
544 * Call translate() to convert text to all valid variants.
545 *
546 * @param string $text The text to be converted
547 * @return array Variant => converted text
548 */
549 public function autoConvertToAllVariants( $text ) {
550 $this->loadTables();
551
552 $ret = [];
553 foreach ( $this->mVariants as $variant ) {
554 $ret[$variant] = $this->translate( $text, $variant );
555 }
556
557 return $ret;
558 }
559
560 /**
561 * Apply manual conversion rules.
562 *
563 * @param ConverterRule $convRule
564 */
565 protected function applyManualConv( $convRule ) {
566 // Use syntax -{T|zh-cn:TitleCN; zh-tw:TitleTw}- to custom
567 // title conversion.
568 // T26072: $mConvRuleTitle was overwritten by other manual
569 // rule(s) not for title, this breaks the title conversion.
570 $newConvRuleTitle = $convRule->getTitle();
571 if ( $newConvRuleTitle ) {
572 // So I add an empty check for getTitle()
573 $this->mConvRuleTitle = $newConvRuleTitle;
574 }
575
576 // merge/remove manual conversion rules to/from global table
577 $convTable = $convRule->getConvTable();
578 $action = $convRule->getRulesAction();
579 foreach ( $convTable as $variant => $pair ) {
580 $v = $this->validateVariant( $variant );
581 if ( !$v ) {
582 continue;
583 }
584
585 if ( $action == 'add' ) {
586 // More efficient than array_merge(), about 2.5 times.
587 foreach ( $pair as $from => $to ) {
588 $this->mTables[$v]->setPair( $from, $to );
589 }
590 } elseif ( $action == 'remove' ) {
591 $this->mTables[$v]->removeArray( $pair );
592 }
593 }
594 }
595
596 /**
597 * Auto convert a Title object to a readable string in the
598 * preferred variant.
599 *
600 * @param Title $title A object of Title
601 * @return string Converted title text
602 */
603 public function convertTitle( $title ) {
604 $variant = $this->getPreferredVariant();
605 $index = $title->getNamespace();
606 if ( $index !== NS_MAIN ) {
607 $text = $this->convertNamespace( $index, $variant ) . ':';
608 } else {
609 $text = '';
610 }
611 $text .= $this->translate( $title->getText(), $variant );
612 return $text;
613 }
614
615 /**
616 * Get the namespace display name in the preferred variant.
617 *
618 * @param int $index Namespace id
619 * @param string|null $variant Variant code or null for preferred variant
620 * @return string Namespace name for display
621 */
622 public function convertNamespace( $index, $variant = null ) {
623 if ( $index === NS_MAIN ) {
624 return '';
625 }
626
627 if ( $variant === null ) {
628 $variant = $this->getPreferredVariant();
629 }
630
631 $cache = MediaWikiServices::getInstance()->getLocalServerObjectCache();
632 $key = $cache->makeKey( 'languageconverter', 'namespace-text', $index, $variant );
633 $nsVariantText = $cache->get( $key );
634 if ( $nsVariantText !== false ) {
635 return $nsVariantText;
636 }
637
638 // First check if a message gives a converted name in the target variant.
639 $nsConvMsg = wfMessage( 'conversion-ns' . $index )->inLanguage( $variant );
640 if ( $nsConvMsg->exists() ) {
641 $nsVariantText = $nsConvMsg->plain();
642 }
643
644 // Then check if a message gives a converted name in content language
645 // which needs extra translation to the target variant.
646 if ( $nsVariantText === false ) {
647 $nsConvMsg = wfMessage( 'conversion-ns' . $index )->inContentLanguage();
648 if ( $nsConvMsg->exists() ) {
649 $nsVariantText = $this->translate( $nsConvMsg->plain(), $variant );
650 }
651 }
652
653 if ( $nsVariantText === false ) {
654 // No message exists, retrieve it from the target variant's namespace names.
655 $langObj = $this->mLangObj->factory( $variant );
656 $nsVariantText = $langObj->getFormattedNsText( $index );
657 }
658
659 $cache->set( $key, $nsVariantText, 60 );
660
661 return $nsVariantText;
662 }
663
664 /**
665 * Convert text to different variants of a language. The automatic
666 * conversion is done in autoConvert(). Here we parse the text
667 * marked with -{}-, which specifies special conversions of the
668 * text that can not be accomplished in autoConvert().
669 *
670 * Syntax of the markup:
671 * -{code1:text1;code2:text2;...}- or
672 * -{flags|code1:text1;code2:text2;...}- or
673 * -{text}- in which case no conversion should take place for text
674 *
675 * @warning Glossary state is maintained between calls. Never feed this
676 * method input that hasn't properly been escaped as it may result in
677 * an XSS in subsequent calls, even if those subsequent calls properly
678 * escape things.
679 * @param string $text Text to be converted, already html escaped.
680 * @return string Converted text (html)
681 */
682 public function convert( $text ) {
683 $variant = $this->getPreferredVariant();
684 return $this->convertTo( $text, $variant );
685 }
686
687 /**
688 * Same as convert() except a extra parameter to custom variant.
689 *
690 * @param string $text Text to be converted, already html escaped
691 * @param-taint $text exec_html
692 * @param string $variant The target variant code
693 * @return string Converted text
694 * @return-taint escaped
695 */
696 public function convertTo( $text, $variant ) {
697 global $wgDisableLangConversion;
698 if ( $wgDisableLangConversion ) {
699 return $text;
700 }
701 // Reset converter state for a new converter run.
702 $this->mConvRuleTitle = false;
703 return $this->recursiveConvertTopLevel( $text, $variant );
704 }
705
706 /**
707 * Recursively convert text on the outside. Allow to use nested
708 * markups to custom rules.
709 *
710 * @param string $text Text to be converted
711 * @param string $variant The target variant code
712 * @param int $depth Depth of recursion
713 * @return string Converted text
714 */
715 protected function recursiveConvertTopLevel( $text, $variant, $depth = 0 ) {
716 $startPos = 0;
717 $out = '';
718 $length = strlen( $text );
719 $shouldConvert = !$this->guessVariant( $text, $variant );
720 $continue = 1;
721
722 $noScript = '<script.*?>.*?<\/script>(*SKIP)(*FAIL)';
723 $noStyle = '<style.*?>.*?<\/style>(*SKIP)(*FAIL)';
724 // phpcs:ignore Generic.Files.LineLength
725 $noHtml = '<(?:[^>=]*+(?>[^>=]*+=\s*+(?:"[^"]*"|\'[^\']*\'|[^\'">\s]*+))*+[^>=]*+>|.*+)(*SKIP)(*FAIL)';
726 while ( $startPos < $length && $continue ) {
727 $continue = preg_match(
728 // Only match -{ outside of html.
729 "/$noScript|$noStyle|$noHtml|-\{/",
730 $text,
731 $m,
732 PREG_OFFSET_CAPTURE,
733 $startPos
734 );
735
736 if ( !$continue ) {
737 // No more markup, append final segment
738 $fragment = substr( $text, $startPos );
739 $out .= $shouldConvert ? $this->autoConvert( $fragment, $variant ) : $fragment;
740 return $out;
741 }
742
743 // Offset of the match of the regex pattern.
744 $pos = $m[0][1];
745
746 // Append initial segment
747 $fragment = substr( $text, $startPos, $pos - $startPos );
748 $out .= $shouldConvert ? $this->autoConvert( $fragment, $variant ) : $fragment;
749 // -{ marker found, not in attribute
750 // Advance position up to -{ marker.
751 $startPos = $pos;
752 // Do recursive conversion
753 // Note: This passes $startPos by reference, and advances it.
754 $out .= $this->recursiveConvertRule( $text, $variant, $startPos, $depth + 1 );
755 }
756 return $out;
757 }
758
759 /**
760 * Recursively convert text on the inside.
761 *
762 * @param string $text Text to be converted
763 * @param string $variant The target variant code
764 * @param int &$startPos
765 * @param int $depth Depth of recursion
766 *
767 * @throws MWException
768 * @return string Converted text
769 */
770 protected function recursiveConvertRule( $text, $variant, &$startPos, $depth = 0 ) {
771 // Quick sanity check (no function calls)
772 if ( $text[$startPos] !== '-' || $text[$startPos + 1] !== '{' ) {
773 throw new MWException( __METHOD__ . ': invalid input string' );
774 }
775
776 $startPos += 2;
777 $inner = '';
778 $warningDone = false;
779 $length = strlen( $text );
780
781 while ( $startPos < $length ) {
782 $m = false;
783 preg_match( '/-\{|\}-/', $text, $m, PREG_OFFSET_CAPTURE, $startPos );
784 if ( !$m ) {
785 // Unclosed rule
786 break;
787 }
788
789 $token = $m[0][0];
790 $pos = $m[0][1];
791
792 // Markup found
793 // Append initial segment
794 $inner .= substr( $text, $startPos, $pos - $startPos );
795
796 // Advance position
797 $startPos = $pos;
798
799 switch ( $token ) {
800 case '-{':
801 // Check max depth
802 if ( $depth >= $this->mMaxDepth ) {
803 $inner .= '-{';
804 if ( !$warningDone ) {
805 $inner .= '<span class="error">' .
806 wfMessage( 'language-converter-depth-warning' )
807 ->numParams( $this->mMaxDepth )->inContentLanguage()->text() .
808 '</span>';
809 $warningDone = true;
810 }
811 $startPos += 2;
812 break;
813 }
814 // Recursively parse another rule
815 $inner .= $this->recursiveConvertRule( $text, $variant, $startPos, $depth + 1 );
816 break;
817 case '}-':
818 // Apply the rule
819 $startPos += 2;
820 $rule = new ConverterRule( $inner, $this );
821 $rule->parse( $variant );
822 $this->applyManualConv( $rule );
823 return $rule->getDisplay();
824 default:
825 throw new MWException( __METHOD__ . ': invalid regex match' );
826 }
827 }
828
829 // Unclosed rule
830 if ( $startPos < $length ) {
831 $inner .= substr( $text, $startPos );
832 }
833 $startPos = $length;
834 return '-{' . $this->autoConvert( $inner, $variant );
835 }
836
837 /**
838 * If a language supports multiple variants, it is possible that
839 * non-existing link in one variant actually exists in another variant.
840 * This function tries to find it. See e.g. LanguageZh.php
841 * The input parameters may be modified upon return
842 *
843 * @param string &$link The name of the link
844 * @param Title &$nt The title object of the link
845 * @param bool $ignoreOtherCond To disable other conditions when
846 * we need to transclude a template or update a category's link
847 */
848 public function findVariantLink( &$link, &$nt, $ignoreOtherCond = false ) {
849 # If the article has already existed, there is no need to
850 # check it again, otherwise it may cause a fault.
851 if ( is_object( $nt ) && $nt->exists() ) {
852 return;
853 }
854
855 global $wgDisableLangConversion, $wgDisableTitleConversion, $wgRequest;
856 $isredir = $wgRequest->getText( 'redirect', 'yes' );
857 $action = $wgRequest->getText( 'action' );
858 if ( $action == 'edit' && $wgRequest->getBool( 'redlink' ) ) {
859 $action = 'view';
860 }
861 $linkconvert = $wgRequest->getText( 'linkconvert', 'yes' );
862 $disableLinkConversion = $wgDisableLangConversion
863 || $wgDisableTitleConversion;
864 $linkBatch = new LinkBatch();
865
866 $ns = NS_MAIN;
867
868 if ( $disableLinkConversion ||
869 ( !$ignoreOtherCond &&
870 ( $isredir == 'no'
871 || $action == 'edit'
872 || $action == 'submit'
873 || $linkconvert == 'no' ) ) ) {
874 return;
875 }
876
877 if ( is_object( $nt ) ) {
878 $ns = $nt->getNamespace();
879 }
880
881 $variants = $this->autoConvertToAllVariants( $link );
882 if ( !$variants ) { // give up
883 return;
884 }
885
886 $titles = [];
887
888 foreach ( $variants as $v ) {
889 if ( $v != $link ) {
890 $varnt = Title::newFromText( $v, $ns );
891 if ( !is_null( $varnt ) ) {
892 $linkBatch->addObj( $varnt );
893 $titles[] = $varnt;
894 }
895 }
896 }
897
898 // fetch all variants in single query
899 $linkBatch->execute();
900
901 foreach ( $titles as $varnt ) {
902 if ( $varnt->getArticleID() > 0 ) {
903 $nt = $varnt;
904 $link = $varnt->getText();
905 break;
906 }
907 }
908 }
909
910 /**
911 * Returns language specific hash options.
912 *
913 * @return string
914 */
915 public function getExtraHashOptions() {
916 $variant = $this->getPreferredVariant();
917
918 return '!' . $variant;
919 }
920
921 /**
922 * Guess if a text is written in a variant. This should be implemented in subclasses.
923 *
924 * @param string $text The text to be checked
925 * @param string $variant Language code of the variant to be checked for
926 * @return bool True if $text appears to be written in $variant, false if not
927 *
928 * @author Nikola Smolenski <smolensk@eunet.rs>
929 * @since 1.19
930 */
931 public function guessVariant( $text, $variant ) {
932 return false;
933 }
934
935 /**
936 * Load default conversion tables.
937 * This method must be implemented in derived class.
938 *
939 * @private
940 * @throws MWException
941 */
942 function loadDefaultTables() {
943 $class = static::class;
944 throw new MWException( "Must implement loadDefaultTables() method in class $class" );
945 }
946
947 /**
948 * Load conversion tables either from the cache or the disk.
949 * @private
950 * @param bool $fromCache Load from memcached? Defaults to true.
951 */
952 function loadTables( $fromCache = true ) {
953 global $wgLanguageConverterCacheType;
954
955 if ( $this->mTablesLoaded ) {
956 return;
957 }
958
959 $this->mTablesLoaded = true;
960 $this->mTables = null;
961 $cache = ObjectCache::getInstance( $wgLanguageConverterCacheType );
962 $cacheKey = $cache->makeKey( 'conversiontables', $this->mMainLanguageCode );
963 if ( $fromCache ) {
964 $this->mTables = $cache->get( $cacheKey );
965 }
966 if ( !$this->mTables || !array_key_exists( self::CACHE_VERSION_KEY, $this->mTables ) ) {
967 // not in cache, or we need a fresh reload.
968 // We will first load the default tables
969 // then update them using things in MediaWiki:Conversiontable/*
970 $this->loadDefaultTables();
971 foreach ( $this->mVariants as $var ) {
972 $cached = $this->parseCachedTable( $var );
973 $this->mTables[$var]->mergeArray( $cached );
974 }
975
976 $this->postLoadTables();
977 $this->mTables[self::CACHE_VERSION_KEY] = true;
978
979 $cache->set( $cacheKey, $this->mTables, 43200 );
980 }
981 }
982
983 /**
984 * Hook for post processing after conversion tables are loaded.
985 */
986 function postLoadTables() {
987 }
988
989 /**
990 * Reload the conversion tables.
991 *
992 * Also used by test suites which need to reset the converter state.
993 *
994 * @private
995 */
996 private function reloadTables() {
997 if ( $this->mTables ) {
998 // @phan-suppress-next-line PhanTypeObjectUnsetDeclaredProperty
999 unset( $this->mTables );
1000 }
1001
1002 $this->mTablesLoaded = false;
1003 $this->loadTables( false );
1004 }
1005
1006 /**
1007 * Parse the conversion table stored in the cache.
1008 *
1009 * The tables should be in blocks of the following form:
1010 * -{
1011 * word => word ;
1012 * word => word ;
1013 * ...
1014 * }-
1015 *
1016 * To make the tables more manageable, subpages are allowed
1017 * and will be parsed recursively if $recursive == true.
1018 *
1019 * @param string $code Language code
1020 * @param string $subpage Subpage name
1021 * @param bool $recursive Parse subpages recursively? Defaults to true.
1022 *
1023 * @return array
1024 */
1025 function parseCachedTable( $code, $subpage = '', $recursive = true ) {
1026 static $parsed = [];
1027
1028 $key = 'Conversiontable/' . $code;
1029 if ( $subpage ) {
1030 $key .= '/' . $subpage;
1031 }
1032 if ( array_key_exists( $key, $parsed ) ) {
1033 return [];
1034 }
1035
1036 $parsed[$key] = true;
1037
1038 if ( $subpage === '' ) {
1039 $txt = MessageCache::singleton()->getMsgFromNamespace( $key, $code );
1040 } else {
1041 $txt = false;
1042 $title = Title::makeTitleSafe( NS_MEDIAWIKI, $key );
1043 if ( $title && $title->exists() ) {
1044 $revision = Revision::newFromTitle( $title );
1045 if ( $revision ) {
1046 if ( $revision->getContentModel() == CONTENT_MODEL_WIKITEXT ) {
1047 // @phan-suppress-next-line PhanUndeclaredMethod
1048 $txt = $revision->getContent( RevisionRecord::RAW )->getText();
1049 }
1050
1051 // @todo in the future, use a specialized content model, perhaps based on json!
1052 }
1053 }
1054 }
1055
1056 # Nothing to parse if there's no text
1057 if ( $txt === false || $txt === null || $txt === '' ) {
1058 return [];
1059 }
1060
1061 // get all subpage links of the form
1062 // [[MediaWiki:Conversiontable/zh-xx/...|...]]
1063 $linkhead = $this->mLangObj->getNsText( NS_MEDIAWIKI ) .
1064 ':Conversiontable';
1065 $subs = StringUtils::explode( '[[', $txt );
1066 $sublinks = [];
1067 foreach ( $subs as $sub ) {
1068 $link = explode( ']]', $sub, 2 );
1069 if ( count( $link ) != 2 ) {
1070 continue;
1071 }
1072 $b = explode( '|', $link[0], 2 );
1073 $b = explode( '/', trim( $b[0] ), 3 );
1074 if ( count( $b ) == 3 ) {
1075 $sublink = $b[2];
1076 } else {
1077 $sublink = '';
1078 }
1079
1080 if ( $b[0] == $linkhead && $b[1] == $code ) {
1081 $sublinks[] = $sublink;
1082 }
1083 }
1084
1085 // parse the mappings in this page
1086 $blocks = StringUtils::explode( '-{', $txt );
1087 $ret = [];
1088 $first = true;
1089 foreach ( $blocks as $block ) {
1090 if ( $first ) {
1091 // Skip the part before the first -{
1092 $first = false;
1093 continue;
1094 }
1095 $mappings = explode( '}-', $block, 2 )[0];
1096 $stripped = str_replace( [ "'", '"', '*', '#' ], '', $mappings );
1097 $table = StringUtils::explode( ';', $stripped );
1098 foreach ( $table as $t ) {
1099 $m = explode( '=>', $t, 3 );
1100 if ( count( $m ) != 2 ) {
1101 continue;
1102 }
1103 // trim any trailling comments starting with '//'
1104 $tt = explode( '//', $m[1], 2 );
1105 $ret[trim( $m[0] )] = trim( $tt[0] );
1106 }
1107 }
1108
1109 // recursively parse the subpages
1110 if ( $recursive ) {
1111 foreach ( $sublinks as $link ) {
1112 $s = $this->parseCachedTable( $code, $link, $recursive );
1113 $ret = $s + $ret;
1114 }
1115 }
1116
1117 if ( $this->mUcfirst ) {
1118 foreach ( $ret as $k => $v ) {
1119 $ret[$this->mLangObj->ucfirst( $k )] = $this->mLangObj->ucfirst( $v );
1120 }
1121 }
1122 return $ret;
1123 }
1124
1125 /**
1126 * Enclose a string with the "no conversion" tag. This is used by
1127 * various functions in the Parser.
1128 *
1129 * @param string $text Text to be tagged for no conversion
1130 * @param bool $noParse Unused
1131 * @return string The tagged text
1132 */
1133 public function markNoConversion( $text, $noParse = false ) {
1134 # don't mark if already marked
1135 if ( strpos( $text, '-{' ) || strpos( $text, '}-' ) ) {
1136 return $text;
1137 }
1138
1139 $ret = "-{R|$text}-";
1140 return $ret;
1141 }
1142
1143 /**
1144 * Convert the sorting key for category links. This should make different
1145 * keys that are variants of each other map to the same key.
1146 *
1147 * @param string $key
1148 *
1149 * @return string
1150 */
1151 function convertCategoryKey( $key ) {
1152 return $key;
1153 }
1154
1155 /**
1156 * Refresh the cache of conversion tables when
1157 * MediaWiki:Conversiontable* is updated.
1158 *
1159 * @param Title $titleobj The Title of the page being updated
1160 */
1161 public function updateConversionTable( Title $titleobj ) {
1162 if ( $titleobj->getNamespace() == NS_MEDIAWIKI ) {
1163 $title = $titleobj->getDBkey();
1164 $t = explode( '/', $title, 3 );
1165 $c = count( $t );
1166 if ( $c > 1 && $t[0] == 'Conversiontable' ) {
1167 if ( $this->validateVariant( $t[1] ) ) {
1168 $this->reloadTables();
1169 }
1170 }
1171 }
1172 }
1173
1174 /**
1175 * Get the cached separator pattern for ConverterRule::parseRules()
1176 * @return string
1177 */
1178 function getVarSeparatorPattern() {
1179 if ( is_null( $this->mVarSeparatorPattern ) ) {
1180 // varsep_pattern for preg_split:
1181 // text should be splited by ";" only if a valid variant
1182 // name exist after the markup, for example:
1183 // -{zh-hans:<span style="font-size:120%;">xxx</span>;zh-hant:\
1184 // <span style="font-size:120%;">yyy</span>;}-
1185 // we should split it as:
1186 // [
1187 // [0] => 'zh-hans:<span style="font-size:120%;">xxx</span>'
1188 // [1] => 'zh-hant:<span style="font-size:120%;">yyy</span>'
1189 // [2] => ''
1190 // ]
1191 $expandedVariants = [];
1192 foreach ( $this->mVariants as $variant ) {
1193 $expandedVariants[ $variant ] = 1;
1194 // Accept standard BCP 47 names for variants as well.
1195 $expandedVariants[ LanguageCode::bcp47( $variant ) ] = 1;
1196 }
1197 // Accept old deprecated names for variants
1198 foreach ( LanguageCode::getDeprecatedCodeMapping() as $old => $new ) {
1199 if ( isset( $expandedVariants[ $new ] ) ) {
1200 $expandedVariants[ $old ] = 1;
1201 }
1202 }
1203
1204 $pat = '/;\s*(?=';
1205 foreach ( $expandedVariants as $variant => $ignore ) {
1206 // zh-hans:xxx;zh-hant:yyy
1207 $pat .= $variant . '\s*:|';
1208 // xxx=>zh-hans:yyy; xxx=>zh-hant:zzz
1209 $pat .= '[^;]*?=>\s*' . $variant . '\s*:|';
1210 }
1211 $pat .= '\s*$)/';
1212 $this->mVarSeparatorPattern = $pat;
1213 }
1214 return $this->mVarSeparatorPattern;
1215 }
1216 }