and for adding * citations to pages * * @ingroup Extensions * * Documentation * @link https://www.mediawiki.org/wiki/Extension:Cite/Cite.php * * definition in HTML * @link http://www.w3.org/TR/html4/struct/text.html#edef-CITE * * definition in XHTML 2.0 * @link http://www.w3.org/TR/2005/WD-xhtml2-20050527/mod-text.html#edef_text_cite * * @bug https://phabricator.wikimedia.org/T6579 * * @author Ævar Arnfjörð Bjarmason * @copyright Copyright © 2005, Ævar Arnfjörð Bjarmason * @license GPL-2.0-or-later */ use MediaWiki\MediaWikiServices; use Wikimedia\Rdbms\IDatabase; use Wikimedia\Rdbms\Database; class Cite { /** * @todo document */ const DEFAULT_GROUP = ''; /** * Maximum storage capacity for pp_value field of page_props table * @todo Find a way to retrieve this information from the DBAL */ const MAX_STORAGE_LENGTH = 65535; // Size of MySQL 'blob' field /** * Key used for storage in parser output's ExtensionData and ObjectCache */ const EXT_DATA_KEY = 'Cite:References'; /** * Version number in case we change the data structure in the future */ const DATA_VERSION_NUMBER = 1; /** * Cache duration set when parsing a page with references */ const CACHE_DURATION_ONPARSE = 3600; // 1 hour /** * Cache duration set when fetching references from db */ const CACHE_DURATION_ONFETCH = 18000; // 5 hours /** * Datastructure representing input, in the format of: * * [ * 'user supplied' => [ * 'text' => 'user supplied reference & key', * 'count' => 1, // occurs twice * 'number' => 1, // The first reference, we want * // all occourances of it to * // use the same number * ], * 0 => 'Anonymous reference', * 1 => 'Another anonymous reference', * 'some key' => [ * 'text' => 'this one occurs once' * 'count' => 0, * 'number' => 4 * ], * 3 => 'more stuff' * ]; * * * This works because: * * PHP's datastructures are guaranteed to be returned in the * order that things are inserted into them (unless you mess * with that) * * User supplied keys can't be integers, therefore avoiding * conflict with anonymous keys * * @var array[] */ private $mRefs = []; /** * Count for user displayed output (ref[1], ref[2], ...) * * @var int */ private $mOutCnt = 0; /** * @var int[] */ private $mGroupCnt = []; /** * Counter to track the total number of (useful) calls to either the * ref or references tag hook * * @var int */ private $mCallCnt = 0; /** * The backlinks, in order, to pass as $3 to * 'cite_references_link_many_format', defined in * 'cite_references_link_many_format_backlink_labels * * @var string[] */ private $mBacklinkLabels; /** * The links to use per group, in order. * * @var (string[]|false)[] */ private $mLinkLabels = []; /** * @var Parser */ private $mParser; /** * True when the ParserAfterParse hook has been called. * Used to avoid doing anything in ParserBeforeTidy. * * @var boolean */ private $mHaveAfterParse = false; /** * True when a tag is being processed. * Used to avoid infinite recursion * * @var boolean */ public $mInCite = false; /** * True when a tag is being processed. * Used to detect the use of to define refs * * @var boolean */ public $mInReferences = false; /** * Error stack used when defining refs in * * @var string[] */ private $mReferencesErrors = []; /** * Group used when in block * * @var string */ private $mReferencesGroup = ''; /** * call stack * Used to cleanup out of sequence ref calls created by #tag * See description of function rollbackRef. * * @var (array|false)[] */ private $mRefCallStack = []; /** * @var bool */ private $mBumpRefData = false; /** * Did we install us into $wgHooks yet? * @var Boolean */ private static $hooksInstalled = false; /** * Callback function for * * @param string|null $str Raw content of the tag. * @param string[] $argv Arguments * @param Parser $parser * @param PPFrame $frame * * @return string */ public function ref( $str, array $argv, Parser $parser, PPFrame $frame ) { if ( $this->mInCite ) { return htmlspecialchars( "$str" ); } $this->mCallCnt++; $this->mInCite = true; $ret = $this->guardedRef( $str, $argv, $parser ); $this->mInCite = false; $parserOutput = $parser->getOutput(); $parserOutput->addModules( 'ext.cite.ux-enhancements' ); $parserOutput->addModuleStyles( 'ext.cite.styles' ); $frame->setVolatile(); // new tag, we may need to bump the ref data counter // to avoid overwriting a previous group $this->mBumpRefData = true; return $ret; } /** * @param string|null $str Raw content of the tag. * @param string[] $argv Arguments * @param Parser $parser * * @throws Exception * @return string */ private function guardedRef( $str, array $argv, Parser $parser ) { $this->mParser = $parser; # The key here is the "name" attribute. list( $key, $group, $follow, $dir ) = $this->refArg( $argv ); // empty string indicate invalid dir if ( $dir === '' && $str !== '' ) { $str .= $this->plainError( 'cite_error_ref_invalid_dir', $argv['dir'] ); } # Split these into groups. if ( $group === null ) { if ( $this->mInReferences ) { $group = $this->mReferencesGroup; } else { $group = self::DEFAULT_GROUP; } } /* * This section deals with constructions of the form * * * BAR * */ if ( $this->mInReferences ) { $isSectionPreview = $parser->getOptions()->getIsSectionPreview(); if ( $group != $this->mReferencesGroup ) { # and have conflicting group attributes. $this->mReferencesErrors[] = $this->error( 'cite_error_references_group_mismatch', Sanitizer::safeEncodeAttribute( $group ) ); } elseif ( $str !== '' ) { if ( !$isSectionPreview && !isset( $this->mRefs[$group] ) ) { # Called with group attribute not defined in text. $this->mReferencesErrors[] = $this->error( 'cite_error_references_missing_group', Sanitizer::safeEncodeAttribute( $group ) ); } elseif ( $key === null || $key === '' ) { # calls inside must be named $this->mReferencesErrors[] = $this->error( 'cite_error_references_no_key' ); } elseif ( !$isSectionPreview && !isset( $this->mRefs[$group][$key] ) ) { # Called with name attribute not defined in text. $this->mReferencesErrors[] = $this->error( 'cite_error_references_missing_key', Sanitizer::safeEncodeAttribute( $key ) ); } else { if ( isset( $this->mRefs[$group][$key]['text'] ) && $str !== $this->mRefs[$group][$key]['text'] ) { // two refs with same key and different content // add error message to the original ref $this->mRefs[$group][$key]['text'] .= ' ' . $this->plainError( 'cite_error_references_duplicate_key', $key ); } else { # Assign the text to corresponding ref $this->mRefs[$group][$key]['text'] = $str; } } } else { # called in has no content. $this->mReferencesErrors[] = $this->error( 'cite_error_empty_references_define', Sanitizer::safeEncodeAttribute( $key ) ); } return ''; } if ( $str === '' ) { # . This construct is invalid if # it's a contentful ref, but OK if it's a named duplicate and should # be equivalent , for compatability with #tag. if ( is_string( $key ) && $key !== '' ) { $str = null; } else { $this->mRefCallStack[] = false; return $this->error( 'cite_error_ref_no_input' ); } } if ( $key === false ) { # TODO: Comment this case; what does this condition mean? $this->mRefCallStack[] = false; return $this->error( 'cite_error_ref_too_many_keys' ); } if ( $str === null && $key === null ) { # Something like ; this makes no sense. $this->mRefCallStack[] = false; return $this->error( 'cite_error_ref_no_key' ); } if ( is_string( $key ) && preg_match( '/^\d+$/', $key ) || is_string( $follow ) && preg_match( '/^\d+$/', $follow ) ) { # Numeric names mess up the resulting id's, potentially produ- # cing duplicate id's in the XHTML. The Right Thing To Do # would be to mangle them, but it's not really high-priority # (and would produce weird id's anyway). $this->mRefCallStack[] = false; return $this->error( 'cite_error_ref_numeric_key' ); } if ( preg_match( '//', preg_replace( '#<([^ ]+?).*?>.*?|#', '', $str ) ) ) { # (bug T8199) This most likely implies that someone left off the # closing tag, which will cause the entire article to be # eaten up until the next . So we bail out early instead. # The fancy regex above first tries chopping out anything that # looks like a comment or SGML tag, which is a crude way to avoid # false alarms for ,
, etc.

			# Possible improvement: print the warning, followed by the contents
			# of the  tag.  This way no part of the article will be eaten
			# even temporarily.

			$this->mRefCallStack[] = false;
			return $this->error( 'cite_error_included_ref' );
		}

		if ( is_string( $key ) || is_string( $str ) ) {
			# We don't care about the content: if the key exists, the ref
			# is presumptively valid.  Either it stores a new ref, or re-
			# fers to an existing one.  If it refers to a nonexistent ref,
			# we'll figure that out later.  Likewise it's definitely valid
			# if there's any content, regardless of key.

			return $this->stack( $str, $key, $group, $follow, $argv, $dir, $parser );
		}

		# Not clear how we could get here, but something is probably
		# wrong with the types.  Let's fail fast.
		throw new Exception( 'Invalid $str and/or $key: ' . serialize( [ $str, $key ] ) );
	}

	/**
	 * Parse the arguments to the  tag
	 *
	 *  "name" : Key of the reference.
	 *  "group" : Group to which it belongs. Needs to be passed to  too.
	 *  "follow" : If the current reference is the continuation of another, key of that reference.
	 *  "dir" : set direction of text (ltr/rtl)
	 *
	 * @param string[] $argv The argument vector
	 * @return (string|false|null)[] An array with exactly four elements, where each is a string on
	 *  valid input, false on invalid input, or null on no input.
	 * @return-taint tainted
	 */
	private function refArg( array $argv ) {
		$group = null;
		$key = null;
		$follow = null;
		$dir = null;

		if ( isset( $argv['dir'] ) ) {
			// compare the dir attribute value against an explicit whitelist.
			$dir = '';
			$isValidDir = in_array( strtolower( $argv['dir'] ), [ 'ltr', 'rtl' ] );
			if ( $isValidDir ) {
				$dir = Html::expandAttributes( [ 'class' => 'mw-cite-dir-' . strtolower( $argv['dir'] ) ] );
			}

			unset( $argv['dir'] );
		}

		if ( $argv === [] ) {
			// No key
			return [ null, null, false, $dir ];
		}

		if ( isset( $argv['name'] ) && isset( $argv['follow'] ) ) {
			return [ false, false, false, false ];
		}

		if ( isset( $argv['name'] ) ) {
			// Key given.
			$key = trim( $argv['name'] );
			unset( $argv['name'] );
		}
		if ( isset( $argv['follow'] ) ) {
			// Follow given.
			$follow = trim( $argv['follow'] );
			unset( $argv['follow'] );
		}
		if ( isset( $argv['group'] ) ) {
			// Group given.
			$group = $argv['group'];
			unset( $argv['group'] );
		}

		if ( $argv !== [] ) {
			// Invalid key
			return [ false, false, false, false ];
		}

		return [ $key, $group, $follow, $dir ];
	}

	/**
	 * Populate $this->mRefs based on input and arguments to 
	 *
	 * @param string $str Input from the  tag
	 * @param string|null $key Argument to the  tag as returned by $this->refArg()
	 * @param string $group
	 * @param string|null $follow
	 * @param string[] $call
	 * @param string $dir ref direction
	 * @param Parser $parser
	 *
	 * @throws Exception
	 * @return string
	 */
	private function stack( $str, $key, $group, $follow, array $call, $dir, Parser $parser ) {
		if ( !isset( $this->mRefs[$group] ) ) {
			$this->mRefs[$group] = [];
		}
		if ( !isset( $this->mGroupCnt[$group] ) ) {
			$this->mGroupCnt[$group] = 0;
		}
		if ( $follow != null ) {
			if ( isset( $this->mRefs[$group][$follow] ) && is_array( $this->mRefs[$group][$follow] ) ) {
				// add text to the note that is being followed
				$this->mRefs[$group][$follow]['text'] .= ' ' . $str;
			} else {
				// insert part of note at the beginning of the group
				$groupsCount = count( $this->mRefs[$group] );
				for ( $k = 0; $k < $groupsCount; $k++ ) {
					if ( !isset( $this->mRefs[$group][$k]['follow'] ) ) {
						break;
					}
				}
				array_splice( $this->mRefs[$group], $k, 0, [ [
					'count' => -1,
					'text' => $str,
					'key' => ++$this->mOutCnt,
					'follow' => $follow,
					'dir' => $dir
				] ] );
				array_splice( $this->mRefCallStack, $k, 0,
					[ [ 'new', $call, $str, $key, $group, $this->mOutCnt ] ] );
			}
			// return an empty string : this is not a reference
			return '';
		}

		if ( $key === null ) {
			// No key
			// $this->mRefs[$group][] = $str;

			$this->mRefs[$group][] = [
				'count' => -1,
				'text' => $str,
				'key' => ++$this->mOutCnt,
				'dir' => $dir
			];
			$this->mRefCallStack[] = [ 'new', $call, $str, $key, $group, $this->mOutCnt ];

			return $this->linkRef( $group, $this->mOutCnt );
		}
		if ( !is_string( $key ) ) {
			throw new Exception( 'Invalid stack key: ' . serialize( $key ) );
		}

		// Valid key
		if ( !isset( $this->mRefs[$group][$key] ) || !is_array( $this->mRefs[$group][$key] ) ) {
			// First occurrence
			$this->mRefs[$group][$key] = [
				'text' => $str,
				'count' => 0,
				'key' => ++$this->mOutCnt,
				'number' => ++$this->mGroupCnt[$group],
				'dir' => $dir
			];
			$this->mRefCallStack[] = [ 'new', $call, $str, $key, $group, $this->mOutCnt ];

			return $this->linkRef(
				$group,
				$key,
				$this->mRefs[$group][$key]['key'] . "-" . $this->mRefs[$group][$key]['count'],
				$this->mRefs[$group][$key]['number'],
				"-" . $this->mRefs[$group][$key]['key']
			);
		}

		// We've been here before
		if ( $this->mRefs[$group][$key]['text'] === null && $str !== '' ) {
			// If no text found before, use this text
			$this->mRefs[$group][$key]['text'] = $str;
			// Use the dir parameter only from the full definition of a named ref tag
			$this->mRefs[$group][$key]['dir'] = $dir;
			$this->mRefCallStack[] = [ 'assign', $call, $str, $key, $group,
				$this->mRefs[$group][$key]['key'] ];
		} else {
			if ( $str != null && $str !== ''
				// T205803 different strip markers might hide the same text
				&& $parser->mStripState->unstripBoth( $str )
					!== $parser->mStripState->unstripBoth( $this->mRefs[$group][$key]['text'] )
			) {
				// two refs with same key and different content
				// add error message to the original ref
				$this->mRefs[$group][$key]['text'] .= ' ' . $this->plainError(
					'cite_error_references_duplicate_key', $key
				);
			}
			$this->mRefCallStack[] = [ 'increment', $call, $str, $key, $group,
				$this->mRefs[$group][$key]['key'] ];
		}
		return $this->linkRef(
			$group,
			$key,
			$this->mRefs[$group][$key]['key'] . "-" . ++$this->mRefs[$group][$key]['count'],
			$this->mRefs[$group][$key]['number'],
			"-" . $this->mRefs[$group][$key]['key']
		);
	}

	/**
	 * Partially undoes the effect of calls to stack()
	 *
	 * Called by guardedReferences()
	 *
	 * The option to define  within  makes the
	 * behavior of  context dependent.  This is normally fine
	 * but certain operations (especially #tag) lead to out-of-order
	 * parser evaluation with the  tags being processed before
	 * their containing  element is read.  This leads to
	 * stack corruption that this function works to fix.
	 *
	 * This function is not a total rollback since some internal
	 * counters remain incremented.  Doing so prevents accidentally
	 * corrupting certain links.
	 *
	 * @param string $type
	 * @param string|null $key
	 * @param string $group
	 * @param int $index
	 */
	private function rollbackRef( $type, $key, $group, $index ) {
		if ( !isset( $this->mRefs[$group] ) ) {
			return;
		}

		if ( $key === null ) {
			foreach ( $this->mRefs[$group] as $k => $v ) {
				if ( $this->mRefs[$group][$k]['key'] === $index ) {
					$key = $k;
					break;
				}
			}
		}

		// Sanity checks that specified element exists.
		if ( $key === null ) {
			return;
		}
		if ( !isset( $this->mRefs[$group][$key] ) ) {
			return;
		}
		if ( $this->mRefs[$group][$key]['key'] != $index ) {
			return;
		}

		switch ( $type ) {
		case 'new':
			# Rollback the addition of new elements to the stack.
			unset( $this->mRefs[$group][$key] );
			if ( $this->mRefs[$group] === [] ) {
				unset( $this->mRefs[$group] );
				unset( $this->mGroupCnt[$group] );
			}
			break;
		case 'assign':
			# Rollback assignment of text to pre-existing elements.
			$this->mRefs[$group][$key]['text'] = null;
			# continue without break
		case 'increment':
			# Rollback increase in named ref occurrences.
			$this->mRefs[$group][$key]['count']--;
			break;
		}
	}

	/**
	 * Callback function for 
	 *
	 * @param string|null $str Raw content of the  tag.
	 * @param string[] $argv Arguments
	 * @param Parser $parser
	 * @param PPFrame $frame
	 *
	 * @return string
	 */
	public function references( $str, array $argv, Parser $parser, PPFrame $frame ) {
		if ( $this->mInCite || $this->mInReferences ) {
			if ( $str === null ) {
				return htmlspecialchars( "" );
			}
			return htmlspecialchars( "$str" );
		}
		$this->mCallCnt++;
		$this->mInReferences = true;
		$ret = $this->guardedReferences( $str, $argv, $parser );
		$this->mInReferences = false;
		$frame->setVolatile();
		return $ret;
	}

	/**
	 * Must only be called from references(). Use that to prevent recursion.
	 *
	 * @param string|null $str Raw content of the  tag.
	 * @param string[] $argv
	 * @param Parser $parser
	 *
	 * @return string
	 */
	private function guardedReferences(
		$str,
		array $argv,
		Parser $parser
	) {
		global $wgCiteResponsiveReferences;

		$this->mParser = $parser;

		if ( isset( $argv['group'] ) ) {
			$group = $argv['group'];
			unset( $argv['group'] );
		} else {
			$group = self::DEFAULT_GROUP;
		}

		if ( strval( $str ) !== '' ) {
			$this->mReferencesGroup = $group;

			# Detect whether we were sent already rendered s.
			# Mostly a side effect of using #tag to call references.
			# The following assumes that the parsed s sent within
			# the  block were the most recent calls to
			# .  This assumption is true for all known use cases,
			# but not strictly enforced by the parser.  It is possible
			# that some unusual combination of #tag,  and
			# conditional parser functions could be created that would
			# lead to malformed references here.
			$count = substr_count( $str, Parser::MARKER_PREFIX . "-ref-" );
			$redoStack = [];

			# Undo effects of calling  while unaware of containing 
			for ( $i = 1; $i <= $count; $i++ ) {
				if ( !$this->mRefCallStack ) {
					break;
				}

				$call = array_pop( $this->mRefCallStack );
				$redoStack[] = $call;
				if ( $call !== false ) {
					list( $type, $ref_argv, $ref_str,
						$ref_key, $ref_group, $ref_index ) = $call;
					$this->rollbackRef( $type, $ref_key, $ref_group, $ref_index );
				}
			}

			# Rerun  call now that mInReferences is set.
			for ( $i = count( $redoStack ) - 1; $i >= 0; $i-- ) {
				$call = $redoStack[$i];
				if ( $call !== false ) {
					list( $type, $ref_argv, $ref_str,
						$ref_key, $ref_group, $ref_index ) = $call;
					$this->guardedRef( $ref_str, $ref_argv, $parser );
				}
			}

			# Parse $str to process any unparsed  tags.
			$parser->recursiveTagParse( $str );

			# Reset call stack
			$this->mRefCallStack = [];
		}

		if ( isset( $argv['responsive'] ) ) {
			$responsive = $argv['responsive'] !== '0';
			unset( $argv['responsive'] );
		} else {
			$responsive = $wgCiteResponsiveReferences;
		}

		// There are remaining parameters we don't recognise
		if ( $argv ) {
			return $this->error( 'cite_error_references_invalid_parameters' );
		}

		$s = $this->referencesFormat( $group, $responsive );

		# Append errors generated while processing 
		if ( $this->mReferencesErrors ) {
			$s .= "\n" . implode( "
\n", $this->mReferencesErrors ); $this->mReferencesErrors = []; } return $s; } /** * Make output to be returned from the references() function. * * If called outside of references(), caller is responsible for ensuring * `mInReferences` is enabled before the call and disabled after call. * * @param string $group * @param bool $responsive * @return string HTML ready for output */ private function referencesFormat( $group, $responsive ) { if ( !$this->mRefs || !isset( $this->mRefs[$group] ) ) { return ''; } $ent = []; foreach ( $this->mRefs[$group] as $k => $v ) { $ent[] = $this->referencesFormatEntry( $k, $v ); } // Add new lines between the list items (ref entires) to avoid confusing tidy (T15073). // Note: This builds a string of wikitext, not html. $parserInput = Html::rawElement( 'ol', [ 'class' => [ 'references' ] ], "\n" . implode( "\n", $ent ) . "\n" ); // Live hack: parse() adds two newlines on WM, can't reproduce it locally -ævar $ret = rtrim( $this->mParser->recursiveTagParse( $parserInput ), "\n" ); if ( $responsive ) { // Use a DIV wrap because column-count on a list directly is broken in Chrome. // See https://bugs.chromium.org/p/chromium/issues/detail?id=498730. $wrapClasses = [ 'mw-references-wrap' ]; if ( count( $this->mRefs[$group] ) > 10 ) { $wrapClasses[] = 'mw-references-columns'; } $ret = Html::rawElement( 'div', [ 'class' => $wrapClasses ], $ret ); } if ( !$this->mParser->getOptions()->getIsPreview() ) { // save references data for later use by LinksUpdate hooks $this->saveReferencesData( $group ); } // done, clean up so we can reuse the group unset( $this->mRefs[$group] ); unset( $this->mGroupCnt[$group] ); return $ret; } /** * Format a single entry for the referencesFormat() function * * @param string $key The key of the reference * @param mixed $val The value of the reference, string for anonymous * references, array for user-suppplied * @return string Wikitext */ private function referencesFormatEntry( $key, $val ) { // Anonymous reference if ( !is_array( $val ) ) { return wfMessage( 'cite_references_link_one', $this->normalizeKey( self::getReferencesKey( $key ) ), $this->normalizeKey( $this->refKey( $key ) ), $this->referenceText( $key, $val ), $val['dir'] )->inContentLanguage()->plain(); } $text = $this->referenceText( $key, $val['text'] ); if ( isset( $val['follow'] ) ) { return wfMessage( 'cite_references_no_link', $this->normalizeKey( self::getReferencesKey( $val['follow'] ) ), $text )->inContentLanguage()->plain(); } if ( !isset( $val['count'] ) ) { // this handles the case of section preview for list-defined references return wfMessage( 'cite_references_link_many', $this->normalizeKey( self::getReferencesKey( $key . "-" . ( $val['key'] ?? '' ) ) ), '', $text )->inContentLanguage()->plain(); } if ( $val['count'] < 0 ) { return wfMessage( 'cite_references_link_one', $this->normalizeKey( self::getReferencesKey( $val['key'] ) ), $this->normalizeKey( # $this->refKey( $val['key'], $val['count'] ) $this->refKey( $val['key'] ) ), $text, $val['dir'] )->inContentLanguage()->plain(); // Standalone named reference, I want to format this like an // anonymous reference because displaying "1. 1.1 Ref text" is // overkill and users frequently use named references when they // don't need them for convenience } if ( $val['count'] === 0 ) { return wfMessage( 'cite_references_link_one', $this->normalizeKey( self::getReferencesKey( $key . "-" . $val['key'] ) ), $this->normalizeKey( # $this->refKey( $key, $val['count'] ), $this->refKey( $key, $val['key'] . "-" . $val['count'] ) ), $text, $val['dir'] )->inContentLanguage()->plain(); // Named references with >1 occurrences } $links = []; // for group handling, we have an extra key here. for ( $i = 0; $i <= $val['count']; ++$i ) { $links[] = wfMessage( 'cite_references_link_many_format', $this->normalizeKey( $this->refKey( $key, $val['key'] . "-$i" ) ), $this->referencesFormatEntryNumericBacklinkLabel( $val['number'], $i, $val['count'] ), $this->referencesFormatEntryAlternateBacklinkLabel( $i ) )->inContentLanguage()->plain(); } $list = $this->listToText( $links ); return wfMessage( 'cite_references_link_many', $this->normalizeKey( self::getReferencesKey( $key . "-" . $val['key'] ) ), $list, $text, $val['dir'] )->inContentLanguage()->plain(); } /** * Returns formatted reference text * @param string $key * @param string|null $text * @return string */ private function referenceText( $key, $text ) { if ( $text === null || $text === '' ) { if ( $this->mParser->getOptions()->getIsSectionPreview() ) { return $this->warning( 'cite_warning_sectionpreview_no_text', $key, 'noparse' ); } return $this->plainError( 'cite_error_references_no_text', $key ); } return '' . rtrim( $text, "\n" ) . "\n"; } /** * Generate a numeric backlink given a base number and an * offset, e.g. $base = 1, $offset = 2; = 1.2 * Since bug #5525, it correctly does 1.9 -> 1.10 as well as 1.099 -> 1.100 * * @param int $base * @param int $offset * @param int $max Maximum value expected. * @return string */ private function referencesFormatEntryNumericBacklinkLabel( $base, $offset, $max ) { $scope = strlen( $max ); $ret = MediaWikiServices::getInstance()->getContentLanguage()->formatNum( sprintf( "%s.%0{$scope}s", $base, $offset ) ); return $ret; } /** * Generate a custom format backlink given an offset, e.g. * $offset = 2; = c if $this->mBacklinkLabels = [ 'a', * 'b', 'c', ...]. Return an error if the offset > the # of * array items * * @param int $offset * * @return string */ private function referencesFormatEntryAlternateBacklinkLabel( $offset ) { if ( !isset( $this->mBacklinkLabels ) ) { $this->genBacklinkLabels(); } if ( isset( $this->mBacklinkLabels[$offset] ) ) { return $this->mBacklinkLabels[$offset]; } else { // Feed me! return $this->plainError( 'cite_error_references_no_backlink_label', null ); } } /** * Generate a custom format link for a group given an offset, e.g. * the second is b if $this->mLinkLabels["foo"] = * [ 'a', 'b', 'c', ...]. * Return an error if the offset > the # of array items * * @param int $offset * @param string $group The group name * @param string $label The text to use if there's no message for them. * * @return string */ private function getLinkLabel( $offset, $group, $label ) { $message = "cite_link_label_group-$group"; if ( !isset( $this->mLinkLabels[$group] ) ) { $this->genLinkLabels( $group, $message ); } if ( $this->mLinkLabels[$group] === false ) { // Use normal representation, ie. "$group 1", "$group 2"... return $label; } if ( isset( $this->mLinkLabels[$group][$offset - 1] ) ) { return $this->mLinkLabels[$group][$offset - 1]; } else { // Feed me! return $this->plainError( 'cite_error_no_link_label_group', [ $group, $message ] ); } } /** * Return an id for use in wikitext output based on a key and * optionally the number of it, used in , not * (since otherwise it would link to itself) * * @param string $key * @param int|null $num The number of the key * @return string A key for use in wikitext */ private function refKey( $key, $num = null ) { $prefix = wfMessage( 'cite_reference_link_prefix' )->inContentLanguage()->text(); $suffix = wfMessage( 'cite_reference_link_suffix' )->inContentLanguage()->text(); if ( $num !== null ) { $key = wfMessage( 'cite_reference_link_key_with_num', $key, $num ) ->inContentLanguage()->plain(); } return "$prefix$key$suffix"; } /** * Return an id for use in wikitext output based on a key and * optionally the number of it, used in , not * (since otherwise it would link to itself) * * @param string $key * @return string A key for use in wikitext */ public static function getReferencesKey( $key ) { $prefix = wfMessage( 'cite_references_link_prefix' )->inContentLanguage()->text(); $suffix = wfMessage( 'cite_references_link_suffix' )->inContentLanguage()->text(); return "$prefix$key$suffix"; } /** * Generate a link ( element from a key * and return XHTML ready for output * * @suppress SecurityCheck-DoubleEscaped * @param string $group * @param string $key The key for the link * @param int|null $count The index of the key, used for distinguishing * multiple occurrences of the same key * @param int|null $label The label to use for the link, I want to * use the same label for all occourances of * the same named reference. * @param string $subkey * * @return string */ private function linkRef( $group, $key, $count = null, $label = null, $subkey = '' ) { $contLang = MediaWikiServices::getInstance()->getContentLanguage(); if ( $label === null ) { $label = ++$this->mGroupCnt[$group]; } return $this->mParser->recursiveTagParse( wfMessage( 'cite_reference_link', $this->normalizeKey( $this->refKey( $key, $count ) ), $this->normalizeKey( self::getReferencesKey( $key . $subkey ) ), Sanitizer::safeEncodeAttribute( $this->getLinkLabel( $label, $group, ( ( $group === self::DEFAULT_GROUP ) ? '' : "$group " ) . $contLang->formatNum( $label ) ) ) )->inContentLanguage()->plain() ); } /** * Normalizes and sanitizes a reference key * * @param string $key * @return string */ private function normalizeKey( $key ) { $ret = Sanitizer::escapeIdForAttribute( $key ); $ret = preg_replace( '/__+/', '_', $ret ); $ret = Sanitizer::safeEncodeAttribute( $ret ); return $ret; } /** * This does approximately the same thing as * Language::listToText() but due to this being used for a * slightly different purpose (people might not want , as the * first separator and not 'and' as the second, and this has to * use messages from the content language) I'm rolling my own. * * @param string[] $arr The array to format * @return string */ private function listToText( $arr ) { $cnt = count( $arr ); if ( $cnt === 1 ) { // Enforce always returning a string return (string)$arr[0]; } $sep = wfMessage( 'cite_references_link_many_sep' )->inContentLanguage()->plain(); $and = wfMessage( 'cite_references_link_many_and' )->inContentLanguage()->plain(); $t = array_slice( $arr, 0, $cnt - 1 ); return implode( $sep, $t ) . $and . $arr[$cnt - 1]; } /** * Generate the labels to pass to the * 'cite_references_link_many_format' message, the format is an * arbitrary number of tokens separated by whitespace. */ private function genBacklinkLabels() { $text = wfMessage( 'cite_references_link_many_format_backlink_labels' ) ->inContentLanguage()->plain(); $this->mBacklinkLabels = preg_split( '/\s+/', $text ); } /** * Generate the labels to pass to the * 'cite_reference_link' message instead of numbers, the format is an * arbitrary number of tokens separated by whitespace. * * @param string $group * @param string $message */ private function genLinkLabels( $group, $message ) { $text = false; $msg = wfMessage( $message )->inContentLanguage(); if ( $msg->exists() ) { $text = $msg->plain(); } $this->mLinkLabels[$group] = $text ? preg_split( '/\s+/', $text ) : false; } /** * Gets run when Parser::clearState() gets run, since we don't * want the counts to transcend pages and other instances * * @param Parser $parser */ public function clearState( Parser $parser ) { if ( $parser->extCite !== $this ) { $parser->extCite->clearState( $parser ); return; } # Don't clear state when we're in the middle of parsing # a tag if ( $this->mInCite || $this->mInReferences ) { return; } $this->mGroupCnt = []; $this->mOutCnt = 0; $this->mCallCnt = 0; $this->mRefs = []; $this->mReferencesErrors = []; $this->mRefCallStack = []; } /** * Gets run when the parser is cloned. * * @param Parser $parser */ public function cloneState( Parser $parser ) { if ( $parser->extCite !== $this ) { $parser->extCite->cloneState( $parser ); return; } $parser->extCite = clone $this; $parser->setHook( 'ref', [ $parser->extCite, 'ref' ] ); $parser->setHook( 'references', [ $parser->extCite, 'references' ] ); // Clear the state, making sure it will actually work. $parser->extCite->mInCite = false; $parser->extCite->mInReferences = false; $parser->extCite->clearState( $parser ); } /** * Called at the end of page processing to append a default references * section, if refs were used without a main references tag. If there are references * in a custom group, and there is no references tag for it, show an error * message for that group. * If we are processing a section preview, this adds the missing * references tags and does not add the errors. * * @param bool $afterParse True if called from the ParserAfterParse hook * @param Parser $parser * @param string &$text */ public function checkRefsNoReferences( $afterParse, $parser, &$text ) { global $wgCiteResponsiveReferences; if ( $parser->extCite === null ) { return; } if ( $parser->extCite !== $this ) { $parser->extCite->checkRefsNoReferences( $afterParse, $parser, $text ); return; } if ( $afterParse ) { $this->mHaveAfterParse = true; } elseif ( $this->mHaveAfterParse ) { return; } if ( !$parser->getOptions()->getIsPreview() ) { // save references data for later use by LinksUpdate hooks if ( $this->mRefs && isset( $this->mRefs[self::DEFAULT_GROUP] ) ) { $this->saveReferencesData(); } $isSectionPreview = false; } else { $isSectionPreview = $parser->getOptions()->getIsSectionPreview(); } $s = ''; foreach ( $this->mRefs as $group => $refs ) { if ( !$refs ) { continue; } if ( $group === self::DEFAULT_GROUP || $isSectionPreview ) { $this->mInReferences = true; $s .= $this->referencesFormat( $group, $wgCiteResponsiveReferences ); $this->mInReferences = false; } else { $s .= "\n
" . $this->error( 'cite_error_group_refs_without_references', Sanitizer::safeEncodeAttribute( $group ) ); } } if ( $isSectionPreview && $s !== '' ) { // provide a preview of references in its own section $text .= "\n" . '
'; $headerMsg = wfMessage( 'cite_section_preview_references' ); if ( !$headerMsg->isDisabled() ) { $text .= '

' . $headerMsg->escaped() . '

'; } $text .= $s . '
'; } else { $text .= $s; } } /** * Saves references in parser extension data * This is called by each tag, and by checkRefsNoReferences * Assumes $this->mRefs[$group] is set * * @param string $group */ private function saveReferencesData( $group = self::DEFAULT_GROUP ) { global $wgCiteStoreReferencesData; if ( !$wgCiteStoreReferencesData ) { return; } $savedRefs = $this->mParser->getOutput()->getExtensionData( self::EXT_DATA_KEY ); if ( $savedRefs === null ) { // Initialize array structure $savedRefs = [ 'refs' => [], 'version' => self::DATA_VERSION_NUMBER, ]; } if ( $this->mBumpRefData ) { // This handles pages with multiple tags with tags in between. // On those, a group can appear several times, so we need to avoid overwriting // a previous appearance. $savedRefs['refs'][] = []; $this->mBumpRefData = false; } $n = count( $savedRefs['refs'] ) - 1; // save group $savedRefs['refs'][$n][$group] = $this->mRefs[$group]; $this->mParser->getOutput()->setExtensionData( self::EXT_DATA_KEY, $savedRefs ); } /** * Initialize the parser hooks * * @param Parser $parser */ public static function setHooks( Parser $parser ) { global $wgHooks; $parser->extCite = new self(); if ( !self::$hooksInstalled ) { $wgHooks['ParserClearState'][] = [ $parser->extCite, 'clearState' ]; $wgHooks['ParserCloned'][] = [ $parser->extCite, 'cloneState' ]; $wgHooks['ParserAfterParse'][] = [ $parser->extCite, 'checkRefsNoReferences', true ]; $wgHooks['ParserBeforeTidy'][] = [ $parser->extCite, 'checkRefsNoReferences', false ]; self::$hooksInstalled = true; } $parser->setHook( 'ref', [ $parser->extCite, 'ref' ] ); $parser->setHook( 'references', [ $parser->extCite, 'references' ] ); } /** * Return an error message based on an error ID and parses it * * @param string $key Message name for the error * @param string[]|string|null $param Parameter to pass to the message * @return string HTML ready for output */ private function error( $key, $param = null ) { $error = $this->plainError( $key, $param ); return $this->mParser->recursiveTagParse( $error ); } /** * Return an error message based on an error ID as unescaped plaintext. * * @param string $key Message name for the error * @param string[]|string|null $param Parameter to pass to the message * @return string wikitext ready for output * @return-taint tainted */ private function plainError( $key, $param = null ) { # For ease of debugging and because errors are rare, we # use the user language and split the parser cache. $lang = $this->mParser->getOptions()->getUserLangObj(); $dir = $lang->getDir(); # We rely on the fact that PHP is okay with passing unused argu- # ments to functions. If $1 is not used in the message, wfMessage will # just ignore the extra parameter. $msg = wfMessage( 'cite_error', wfMessage( $key, $param )->inLanguage( $lang )->plain() ) ->inLanguage( $lang ) ->plain(); $this->mParser->addTrackingCategory( 'cite-tracking-category-cite-error' ); $ret = Html::rawElement( 'span', [ 'class' => 'error mw-ext-cite-error', 'lang' => $lang->getHtmlCode(), 'dir' => $dir, ], $msg ); return $ret; } /** * Return a warning message based on a warning ID * * @param string $key Message name for the warning. Name should start with cite_warning_ * @param string|null $param Parameter to pass to the message * @param string $parse Whether to parse the message ('parse') or not ('noparse') * @return string XHTML or wikitext ready for output */ private function warning( $key, $param = null, $parse = 'parse' ) { # For ease of debugging and because errors are rare, we # use the user language and split the parser cache. $lang = $this->mParser->getOptions()->getUserLangObj(); $dir = $lang->getDir(); # We rely on the fact that PHP is okay with passing unused argu- # ments to functions. If $1 is not used in the message, wfMessage will # just ignore the extra parameter. $msg = wfMessage( 'cite_warning', wfMessage( $key, $param )->inLanguage( $lang )->plain() ) ->inLanguage( $lang ) ->plain(); $key = preg_replace( '/^cite_warning_/', '', $key ) . ''; $ret = Html::rawElement( 'span', [ 'class' => 'warning mw-ext-cite-warning mw-ext-cite-warning-' . Sanitizer::escapeClass( $key ), 'lang' => $lang->getHtmlCode(), 'dir' => $dir, ], $msg ); if ( $parse === 'parse' ) { $ret = $this->mParser->recursiveTagParse( $ret ); } return $ret; } /** * Fetch references stored for the given title in page_props * For performance, results are cached * * @param Title $title * @return array|false */ public static function getStoredReferences( Title $title ) { global $wgCiteStoreReferencesData; if ( !$wgCiteStoreReferencesData ) { return false; } $cache = MediaWikiServices::getInstance()->getMainWANObjectCache(); $key = $cache->makeKey( self::EXT_DATA_KEY, $title->getArticleID() ); return $cache->getWithSetCallback( $key, self::CACHE_DURATION_ONFETCH, function ( $oldValue, &$ttl, array &$setOpts ) use ( $title ) { $dbr = wfGetDB( DB_REPLICA ); $setOpts += Database::getCacheSetOptions( $dbr ); return self::recursiveFetchRefsFromDB( $title, $dbr ); }, [ 'checkKeys' => [ $key ], 'lockTSE' => 30, ] ); } /** * Reconstructs compressed json by successively retrieving the properties references-1, -2, etc * It attempts the next step when a decoding error occurs. * Returns json_decoded uncompressed string, with validation of json * * @param Title $title * @param IDatabase $dbr * @param string $string * @param int $i * @return array|false */ private static function recursiveFetchRefsFromDB( Title $title, IDatabase $dbr, $string = '', $i = 1 ) { $id = $title->getArticleID(); $result = $dbr->selectField( 'page_props', 'pp_value', [ 'pp_page' => $id, 'pp_propname' => 'references-' . $i ], __METHOD__ ); if ( $result !== false ) { $string .= $result; $decodedString = gzdecode( $string ); if ( $decodedString !== false ) { $json = json_decode( $decodedString, true ); if ( json_last_error() === JSON_ERROR_NONE ) { return $json; } // corrupted json ? // shouldn't happen since when string is truncated, gzdecode should fail wfDebug( "Corrupted json detected when retrieving stored references for title id $id" ); } // if gzdecode fails, try to fetch next references- property value return self::recursiveFetchRefsFromDB( $title, $dbr, $string, ++$i ); } else { // no refs stored in page_props at this index if ( $i > 1 ) { // shouldn't happen wfDebug( "Failed to retrieve stored references for title id $id" ); } return false; } } }