includes/content/WikiTextStructure.php

   1 <?php
   2
   3 use HtmlFormatter\HtmlFormatter;
   4 use MediaWiki\Logger\LoggerFactory;
   5
   6 /**
   7  * Class allowing to explore structure of parsed wikitext.
   8  */
   9 class WikiTextStructure {
  10         /**
  11          * @var string
  12          */
  13         private $openingText;
  14         /**
  15          * @var string
  16          */
  17         private $allText;
  18         /**
  19          * @var string[]
  20          */
  21         private $auxText = [];
  22         /**
  23          * @var ParserOutput
  24          */
  25         private $parserOutput;
  26
  27         /**
  28          * @var string[] selectors to elements that are excluded entirely from search
  29          */
  30         private $excludedElementSelectors = [
  31                 'audio', 'video',       // "it looks like you don't have javascript enabled..."
  32                                         // do not need to index
  33                 'sup.reference',        // The [1] for references
  34                 '.mw-cite-backlink',    // The ↑ next to references in the references section
  35                 'h1', 'h2', 'h3',       // Headings are already indexed in their own field.
  36                 'h5', 'h6', 'h4',
  37                 '.autocollapse',        // Collapsed fields are hidden by default so we don't want them
  38                                                                 // showing up.
  39         ];
  40
  41         /**
  42          * @var string[] selectors to elements that are considered auxiliary to article text for search
  43          */
  44         private $auxiliaryElementSelectors = [
  45                 '.thumbcaption',        // Thumbnail captions aren't really part of the text proper
  46                 'table',                // Neither are tables
  47                 '.rellink',             // Common style for "See also:".
  48                 '.dablink',             // Common style for calling out helpful links at the top
  49                                                                 // of the article.
  50                 '.searchaux',           // New class users can use to mark stuff as auxiliary to searches.
  51         ];
  52
  53         /**
  54          * WikiTextStructure constructor.
  55          * @param ParserOutput $parserOutput
  56          */
  57         public function __construct( ParserOutput $parserOutput ) {
  58                 $this->parserOutput = $parserOutput;
  59         }
  60
  61         /**
  62          * Get headings on the page.
  63          * @return string[]
  64          * First strip out things that look like references.  We can't use HTML filtering because
  65          * the references come back as <sup> tags without a class.  To keep from breaking stuff like
  66          *  ==Applicability of the strict mass–energy equivalence formula, ''E'' = ''mc''<sup>2</sup>==
  67          * we don't remove the whole <sup> tag.  We also don't want to strip the <sup> tag and remove
  68          * everything that looks like [2] because, I dunno, maybe there is a band named Word [2] Foo
  69          * or something.  Whatever.  So we only strip things that look like <sup> tags wrapping a
  70          * reference.  And since the data looks like:
  71          *      Reference in heading <sup>&#91;1&#93;</sup><sup>&#91;2&#93;</sup>
  72          * we can not really use HtmlFormatter as we have no suitable selector.
  73          */
  74         public function headings() {
  75                 $headings = [];
  76                 $ignoredHeadings = $this->getIgnoredHeadings();
  77                 foreach ( $this->parserOutput->getSections() as $heading ) {
  78                         $heading = $heading[ 'line' ];
  79
  80                         // Some wikis wrap the brackets in a span:
  81                         // http://en.wikipedia.org/wiki/MediaWiki:Cite_reference_link
  82                         $heading = preg_replace( '/<\/?span>/', '', $heading );
  83                         // Normalize [] so the following regexp would work.
  84                         $heading = preg_replace( [ '/&#91;/', '/&#93;/' ], [ '[', ']' ], $heading );
  85                         $heading = preg_replace( '/<sup>\s*\[\s*\d+\s*\]\s*<\/sup>/is', '', $heading );
  86
  87                         // Strip tags from the heading or else we'll display them (escaped) in search results
  88                         $heading = trim( Sanitizer::stripAllTags( $heading ) );
  89
  90                         // Note that we don't take the level of the heading into account - all headings are equal.
  91                         // Except the ones we ignore.
  92                         if ( !in_array( $heading, $ignoredHeadings ) ) {
  93                                 $headings[] = $heading;
  94                         }
  95                 }
  96                 return $headings;
  97         }
  98
  99         /**
 100          * Parse a message content into an array. This function is generally used to
 101          * parse settings stored as i18n messages (see search-ignored-headings).
 102          *
 103          * @param string $message
 104          * @return string[]
 105          */
 106         public static function parseSettingsInMessage( $message ) {
 107                 $lines = explode( "\n", $message );
 108                 $lines = preg_replace( '/#.*$/', '', $lines ); // Remove comments
 109                 $lines = array_map( 'trim', $lines );          // Remove extra spaces
 110                 $lines = array_filter( $lines );               // Remove empty lines
 111                 return $lines;
 112         }
 113
 114         /**
 115          * Get list of heading to ignore.
 116          * @return string[]
 117          */
 118         private function getIgnoredHeadings() {
 119                 static $ignoredHeadings = null;
 120                 if ( $ignoredHeadings === null ) {
 121                         $ignoredHeadings = [];
 122                         $source = wfMessage( 'search-ignored-headings' )->inContentLanguage();
 123                         if ( $source->isBlank() ) {
 124                                 // Try old version too, just in case
 125                                 $source = wfMessage( 'cirrussearch-ignored-headings' )->inContentLanguage();
 126                         }
 127                         if ( !$source->isDisabled() ) {
 128                                 $lines = self::parseSettingsInMessage( $source->plain() );
 129                                 $ignoredHeadings = $lines;               // Now we just have headings!
 130                         }
 131                 }
 132                 return $ignoredHeadings;
 133         }
 134
 135         /**
 136          * Extract parts of the text - opening, main and auxiliary.
 137          */
 138         private function extractWikitextParts() {
 139                 if ( !is_null( $this->allText ) ) {
 140                         return;
 141                 }
 142                 $this->parserOutput->setEditSectionTokens( false );
 143                 $this->parserOutput->setTOCEnabled( false );
 144                 $text = $this->parserOutput->getText();
 145                 if ( strlen( $text ) == 0 ) {
 146                         $this->allText = "";
 147                         // empty text - nothing to seek here
 148                         return;
 149                 }
 150                 $opening = null;
 151
 152                 $this->openingText = $this->extractHeadingBeforeFirstHeading( $text );
 153
 154                 // Add extra spacing around break tags so text crammed together like<br>this
 155                 // doesn't make one word.
 156                 $text = str_replace( '<br', "\n<br", $text );
 157
 158                 $formatter = new HtmlFormatter( $text );
 159
 160                 // Strip elements from the page that we never want in the search text.
 161                 $formatter->remove( $this->excludedElementSelectors );
 162                 $formatter->filterContent();
 163
 164                 // Strip elements from the page that are auxiliary text.  These will still be
 165                 // searched but matches will be ranked lower and non-auxiliary matches will be
 166                 // preferred in highlighting.
 167                 $formatter->remove( $this->auxiliaryElementSelectors );
 168                 $auxiliaryElements = $formatter->filterContent();
 169                 $this->allText = trim( Sanitizer::stripAllTags( $formatter->getText() ) );
 170                 foreach ( $auxiliaryElements as $auxiliaryElement ) {
 171                         $this->auxText[] =
 172                                 trim( Sanitizer::stripAllTags( $formatter->getText( $auxiliaryElement ) ) );
 173                 }
 174         }
 175
 176         /**
 177          * Get text before first heading.
 178          * @param string $text
 179          * @return string|null
 180          */
 181         private function extractHeadingBeforeFirstHeading( $text ) {
 182                 $matches = [];
 183                 if ( !preg_match( '/<h[123456]>/', $text, $matches, PREG_OFFSET_CAPTURE ) ) {
 184                         // There isn't a first heading so we interpret this as the article
 185                         // being entirely without heading.
 186                         return null;
 187                 }
 188                 $text = substr( $text, 0, $matches[ 0 ][ 1 ] );
 189                 if ( !$text ) {
 190                         // There isn't any text before the first heading so we declare there isn't
 191                         // a first heading.
 192                         return null;
 193                 }
 194
 195                 $formatter = new HtmlFormatter( $text );
 196                 $formatter->remove( $this->excludedElementSelectors );
 197                 $formatter->remove( $this->auxiliaryElementSelectors );
 198                 $formatter->filterContent();
 199                 $text = trim( Sanitizer::stripAllTags( $formatter->getText() ) );
 200
 201                 if ( !$text ) {
 202                         // There isn't any text after filtering before the first heading so we declare
 203                         // that there isn't a first heading.
 204                         return null;
 205                 }
 206
 207                 return $text;
 208         }
 209
 210         /**
 211          * Get opening text
 212          * @return string
 213          */
 214         public function getOpeningText() {
 215                 $this->extractWikitextParts();
 216                 return $this->openingText;
 217         }
 218
 219         /**
 220          * Get main text
 221          * @return string
 222          */
 223         public function getMainText() {
 224                 $this->extractWikitextParts();
 225                 return $this->allText;
 226         }
 227
 228         /**
 229          * Get auxiliary text
 230          * @return string[]
 231          */
 232         public function getAuxiliaryText() {
 233                 $this->extractWikitextParts();
 234                 return $this->auxText;
 235         }
 236
 237         /**
 238          * Get the defaultsort property
 239          * @return string|null
 240          */
 241         public function getDefaultSort() {
 242                 return $this->parserOutput->getProperty( 'defaultsort' );
 243         }
 244 }