includes/content/WikiTextStructure.php

   1 <?php
   2
   3 use HtmlFormatter\HtmlFormatter;
   4
   5 /**
   6  * Class allowing to explore structure of parsed wikitext.
   7  */
   8 class WikiTextStructure {
   9         /**
  10          * @var string
  11          */
  12         private $openingText;
  13         /**
  14          * @var string
  15          */
  16         private $allText;
  17         /**
  18          * @var string[]
  19          */
  20         private $auxText = [];
  21         /**
  22          * @var ParserOutput
  23          */
  24         private $parserOutput;
  25
  26         /**
  27          * @var string[] selectors to elements that are excluded entirely from search
  28          */
  29         private $excludedElementSelectors = [
  30                 // "it looks like you don't have javascript enabled..." – do not need to index
  31                 'audio', 'video',
  32                 // The [1] for references
  33                 'sup.reference',
  34                 // The ↑ next to references in the references section
  35                 '.mw-cite-backlink',
  36                 // Headings are already indexed in their own field.
  37                 'h1', 'h2', 'h3', 'h4', 'h5', 'h6',
  38                 // Collapsed fields are hidden by default so we don't want them showing up.
  39                 '.autocollapse',
  40                 // Content explicitly decided to be not searchable by editors such
  41                 // as custom navigation templates.
  42                 '.navigation-not-searchable'
  43         ];
  44
  45         /**
  46          * @var string[] selectors to elements that are considered auxiliary to article text for search
  47          */
  48         private $auxiliaryElementSelectors = [
  49                 // Thumbnail captions aren't really part of the text proper
  50                 '.thumbcaption',
  51                 // Neither are tables
  52                 'table',
  53                 // Common style for "See also:".
  54                 '.rellink',
  55                 // Common style for calling out helpful links at the top of the article.
  56                 '.dablink',
  57                 // New class users can use to mark stuff as auxiliary to searches.
  58                 '.searchaux',
  59         ];
  60
  61         /**
  62          * @param ParserOutput $parserOutput
  63          */
  64         public function __construct( ParserOutput $parserOutput ) {
  65                 $this->parserOutput = $parserOutput;
  66         }
  67
  68         /**
  69          * Get headings on the page.
  70          * @return string[]
  71          * First strip out things that look like references.  We can't use HTML filtering because
  72          * the references come back as <sup> tags without a class.  To keep from breaking stuff like
  73          *  ==Applicability of the strict mass–energy equivalence formula, ''E'' = ''mc''<sup>2</sup>==
  74          * we don't remove the whole <sup> tag.  We also don't want to strip the <sup> tag and remove
  75          * everything that looks like [2] because, I dunno, maybe there is a band named Word [2] Foo
  76          * or something.  Whatever.  So we only strip things that look like <sup> tags wrapping a
  77          * reference.  And since the data looks like:
  78          *      Reference in heading <sup>&#91;1&#93;</sup><sup>&#91;2&#93;</sup>
  79          * we can not really use HtmlFormatter as we have no suitable selector.
  80          */
  81         public function headings() {
  82                 $headings = [];
  83                 $ignoredHeadings = $this->getIgnoredHeadings();
  84                 foreach ( $this->parserOutput->getSections() as $heading ) {
  85                         $heading = $heading[ 'line' ];
  86
  87                         // Some wikis wrap the brackets in a span:
  88                         // https://en.wikipedia.org/wiki/MediaWiki:Cite_reference_link
  89                         $heading = preg_replace( '/<\/?span>/', '', $heading );
  90                         // Normalize [] so the following regexp would work.
  91                         $heading = preg_replace( [ '/&#91;/', '/&#93;/' ], [ '[', ']' ], $heading );
  92                         $heading = preg_replace( '/<sup>\s*\[\s*\d+\s*\]\s*<\/sup>/is', '', $heading );
  93
  94                         // Strip tags from the heading or else we'll display them (escaped) in search results
  95                         $heading = trim( Sanitizer::stripAllTags( $heading ) );
  96
  97                         // Note that we don't take the level of the heading into account - all headings are equal.
  98                         // Except the ones we ignore.
  99                         if ( !in_array( $heading, $ignoredHeadings ) ) {
 100                                 $headings[] = $heading;
 101                         }
 102                 }
 103                 return $headings;
 104         }
 105
 106         /**
 107          * Parse a message content into an array. This function is generally used to
 108          * parse settings stored as i18n messages (see search-ignored-headings).
 109          *
 110          * @param string $message
 111          * @return string[]
 112          */
 113         public static function parseSettingsInMessage( $message ) {
 114                 $lines = explode( "\n", $message );
 115                 $lines = preg_replace( '/#.*$/', '', $lines ); // Remove comments
 116                 $lines = array_map( 'trim', $lines );          // Remove extra spaces
 117                 $lines = array_filter( $lines );               // Remove empty lines
 118                 return $lines;
 119         }
 120
 121         /**
 122          * Get list of heading to ignore.
 123          * @return string[]
 124          */
 125         private function getIgnoredHeadings() {
 126                 static $ignoredHeadings = null;
 127                 if ( $ignoredHeadings === null ) {
 128                         $ignoredHeadings = [];
 129                         $source = wfMessage( 'search-ignored-headings' )->inContentLanguage();
 130                         if ( $source->isBlank() ) {
 131                                 // Try old version too, just in case
 132                                 $source = wfMessage( 'cirrussearch-ignored-headings' )->inContentLanguage();
 133                         }
 134                         if ( !$source->isDisabled() ) {
 135                                 $lines = self::parseSettingsInMessage( $source->plain() );
 136                                 $ignoredHeadings = $lines;               // Now we just have headings!
 137                         }
 138                 }
 139                 return $ignoredHeadings;
 140         }
 141
 142         /**
 143          * Extract parts of the text - opening, main and auxiliary.
 144          */
 145         private function extractWikitextParts() {
 146                 if ( !is_null( $this->allText ) ) {
 147                         return;
 148                 }
 149                 $this->parserOutput->setEditSectionTokens( false );
 150                 $this->parserOutput->setTOCEnabled( false );
 151                 $text = $this->parserOutput->getText();
 152                 if ( strlen( $text ) == 0 ) {
 153                         $this->allText = "";
 154                         // empty text - nothing to seek here
 155                         return;
 156                 }
 157                 $opening = null;
 158
 159                 $this->openingText = $this->extractHeadingBeforeFirstHeading( $text );
 160
 161                 // Add extra spacing around break tags so text crammed together like<br>this
 162                 // doesn't make one word.
 163                 $text = str_replace( '<br', "\n<br", $text );
 164
 165                 $formatter = new HtmlFormatter( $text );
 166
 167                 // Strip elements from the page that we never want in the search text.
 168                 $formatter->remove( $this->excludedElementSelectors );
 169                 $formatter->filterContent();
 170
 171                 // Strip elements from the page that are auxiliary text.  These will still be
 172                 // searched but matches will be ranked lower and non-auxiliary matches will be
 173                 // preferred in highlighting.
 174                 $formatter->remove( $this->auxiliaryElementSelectors );
 175                 $auxiliaryElements = $formatter->filterContent();
 176                 $this->allText = trim( Sanitizer::stripAllTags( $formatter->getText() ) );
 177                 foreach ( $auxiliaryElements as $auxiliaryElement ) {
 178                         $this->auxText[] =
 179                                 trim( Sanitizer::stripAllTags( $formatter->getText( $auxiliaryElement ) ) );
 180                 }
 181         }
 182
 183         /**
 184          * Get text before first heading.
 185          * @param string $text
 186          * @return string|null
 187          */
 188         private function extractHeadingBeforeFirstHeading( $text ) {
 189                 $matches = [];
 190                 if ( !preg_match( '/<h[123456]>/', $text, $matches, PREG_OFFSET_CAPTURE ) ) {
 191                         // There isn't a first heading so we interpret this as the article
 192                         // being entirely without heading.
 193                         return null;
 194                 }
 195                 $text = substr( $text, 0, $matches[ 0 ][ 1 ] );
 196                 if ( !$text ) {
 197                         // There isn't any text before the first heading so we declare there isn't
 198                         // a first heading.
 199                         return null;
 200                 }
 201
 202                 $formatter = new HtmlFormatter( $text );
 203                 $formatter->remove( $this->excludedElementSelectors );
 204                 $formatter->remove( $this->auxiliaryElementSelectors );
 205                 $formatter->filterContent();
 206                 $text = trim( Sanitizer::stripAllTags( $formatter->getText() ) );
 207
 208                 if ( !$text ) {
 209                         // There isn't any text after filtering before the first heading so we declare
 210                         // that there isn't a first heading.
 211                         return null;
 212                 }
 213
 214                 return $text;
 215         }
 216
 217         /**
 218          * Get opening text
 219          * @return string
 220          */
 221         public function getOpeningText() {
 222                 $this->extractWikitextParts();
 223                 return $this->openingText;
 224         }
 225
 226         /**
 227          * Get main text
 228          * @return string
 229          */
 230         public function getMainText() {
 231                 $this->extractWikitextParts();
 232                 return $this->allText;
 233         }
 234
 235         /**
 236          * Get auxiliary text
 237          * @return string[]
 238          */
 239         public function getAuxiliaryText() {
 240                 $this->extractWikitextParts();
 241                 return $this->auxText;
 242         }
 243
 244         /**
 245          * Get the defaultsort property
 246          * @return string|null
 247          */
 248         public function getDefaultSort() {
 249                 return $this->parserOutput->getProperty( 'defaultsort' );
 250         }
 251 }