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                 // CSS stylesheets aren't content
  33                 'style',
  34                 // The [1] for references from Cite
  35                 'sup.reference',
  36                 // The ↑ next to references in the references section from Cite
  37                 '.mw-cite-backlink',
  38                 // Headings are already indexed in their own field.
  39                 'h1', 'h2', 'h3', 'h4', 'h5', 'h6',
  40                 // Collapsed fields are hidden by default so we don't want them showing up.
  41                 '.autocollapse',
  42                 // Content explicitly decided to be not searchable by editors such
  43                 // as custom navigation templates.
  44                 '.navigation-not-searchable',
  45                 // User-facing interface code prompting the user to act from WikibaseMediaInfo
  46                 '.wbmi-entityview-emptyCaption',
  47         ];
  48
  49         /**
  50          * @var string[] selectors to elements that are considered auxiliary to article text for search
  51          */
  52         private $auxiliaryElementSelectors = [
  53                 // Thumbnail captions aren't really part of the text proper
  54                 '.thumbcaption',
  55                 // Neither are tables
  56                 'table',
  57                 // Common style for "See also:".
  58                 '.rellink',
  59                 // Common style for calling out helpful links at the top of the article.
  60                 '.dablink',
  61                 // New class users can use to mark stuff as auxiliary to searches.
  62                 '.searchaux',
  63         ];
  64
  65         /**
  66          * @param ParserOutput $parserOutput
  67          */
  68         public function __construct( ParserOutput $parserOutput ) {
  69                 $this->parserOutput = $parserOutput;
  70         }
  71
  72         /**
  73          * Get headings on the page.
  74          * @return string[]
  75          * First strip out things that look like references.  We can't use HTML filtering because
  76          * the references come back as <sup> tags without a class.  To keep from breaking stuff like
  77          *  ==Applicability of the strict mass–energy equivalence formula, ''E'' = ''mc''<sup>2</sup>==
  78          * we don't remove the whole <sup> tag.  We also don't want to strip the <sup> tag and remove
  79          * everything that looks like [2] because, I dunno, maybe there is a band named Word [2] Foo
  80          * or something.  Whatever.  So we only strip things that look like <sup> tags wrapping a
  81          * reference.  And since the data looks like:
  82          *      Reference in heading <sup>&#91;1&#93;</sup><sup>&#91;2&#93;</sup>
  83          * we can not really use HtmlFormatter as we have no suitable selector.
  84          */
  85         public function headings() {
  86                 $headings = [];
  87                 $ignoredHeadings = $this->getIgnoredHeadings();
  88                 foreach ( $this->parserOutput->getSections() as $heading ) {
  89                         $heading = $heading[ 'line' ];
  90
  91                         // Some wikis wrap the brackets in a span:
  92                         // https://en.wikipedia.org/wiki/MediaWiki:Cite_reference_link
  93                         $heading = preg_replace( '/<\/?span>/', '', $heading );
  94                         // Normalize [] so the following regexp would work.
  95                         $heading = preg_replace( [ '/&#91;/', '/&#93;/' ], [ '[', ']' ], $heading );
  96                         $heading = preg_replace( '/<sup>\s*\[\s*\d+\s*\]\s*<\/sup>/is', '', $heading );
  97
  98                         // Strip tags from the heading or else we'll display them (escaped) in search results
  99                         $heading = trim( Sanitizer::stripAllTags( $heading ) );
 100
 101                         // Note that we don't take the level of the heading into account - all headings are equal.
 102                         // Except the ones we ignore.
 103                         if ( !in_array( $heading, $ignoredHeadings ) ) {
 104                                 $headings[] = $heading;
 105                         }
 106                 }
 107                 return $headings;
 108         }
 109
 110         /**
 111          * Parse a message content into an array. This function is generally used to
 112          * parse settings stored as i18n messages (see search-ignored-headings).
 113          *
 114          * @param string $message
 115          * @return string[]
 116          */
 117         public static function parseSettingsInMessage( $message ) {
 118                 $lines = explode( "\n", $message );
 119                 $lines = preg_replace( '/#.*$/', '', $lines ); // Remove comments
 120                 $lines = array_map( 'trim', $lines );          // Remove extra spaces
 121                 $lines = array_filter( $lines );               // Remove empty lines
 122                 return $lines;
 123         }
 124
 125         /**
 126          * Get list of heading to ignore.
 127          * @return string[]
 128          */
 129         private function getIgnoredHeadings() {
 130                 static $ignoredHeadings = null;
 131                 if ( $ignoredHeadings === null ) {
 132                         $ignoredHeadings = [];
 133                         $source = wfMessage( 'search-ignored-headings' )->inContentLanguage();
 134                         if ( $source->isBlank() ) {
 135                                 // Try old version too, just in case
 136                                 $source = wfMessage( 'cirrussearch-ignored-headings' )->inContentLanguage();
 137                         }
 138                         if ( !$source->isDisabled() ) {
 139                                 $lines = self::parseSettingsInMessage( $source->plain() );
 140                                 $ignoredHeadings = $lines;               // Now we just have headings!
 141                         }
 142                 }
 143                 return $ignoredHeadings;
 144         }
 145
 146         /**
 147          * Extract parts of the text - opening, main and auxiliary.
 148          */
 149         private function extractWikitextParts() {
 150                 if ( !is_null( $this->allText ) ) {
 151                         return;
 152                 }
 153                 $text = $this->parserOutput->getText( [
 154                         'enableSectionEditTokens' => false,
 155                         'allowTOC' => false,
 156                 ] );
 157                 if ( $text === '' ) {
 158                         $this->allText = "";
 159                         // empty text - nothing to seek here
 160                         return;
 161                 }
 162                 $opening = null;
 163
 164                 $this->openingText = $this->extractHeadingBeforeFirstHeading( $text );
 165
 166                 $formatter = new HtmlFormatter( $text );
 167
 168                 // Strip elements from the page that we never want in the search text.
 169                 $formatter->remove( $this->excludedElementSelectors );
 170                 $formatter->filterContent();
 171
 172                 // Strip elements from the page that are auxiliary text.  These will still be
 173                 // searched but matches will be ranked lower and non-auxiliary matches will be
 174                 // preferred in highlighting.
 175                 $formatter->remove( $this->auxiliaryElementSelectors );
 176                 $auxiliaryElements = $formatter->filterContent();
 177                 $this->allText = trim( Sanitizer::stripAllTags( $formatter->getText() ) );
 178                 foreach ( $auxiliaryElements as $auxiliaryElement ) {
 179                         $this->auxText[] =
 180                                 trim( Sanitizer::stripAllTags( $formatter->getText( $auxiliaryElement ) ) );
 181                 }
 182         }
 183
 184         /**
 185          * Get text before first heading.
 186          * @param string $text
 187          * @return string|null
 188          */
 189         private function extractHeadingBeforeFirstHeading( $text ) {
 190                 $matches = [];
 191                 if ( !preg_match( '/<h[123456]>/', $text, $matches, PREG_OFFSET_CAPTURE ) ) {
 192                         // There isn't a first heading so we interpret this as the article
 193                         // being entirely without heading.
 194                         return null;
 195                 }
 196                 $text = substr( $text, 0, $matches[ 0 ][ 1 ] );
 197                 if ( !$text ) {
 198                         // There isn't any text before the first heading so we declare there isn't
 199                         // a first heading.
 200                         return null;
 201                 }
 202
 203                 $formatter = new HtmlFormatter( $text );
 204                 $formatter->remove( $this->excludedElementSelectors );
 205                 $formatter->remove( $this->auxiliaryElementSelectors );
 206                 $formatter->filterContent();
 207                 $text = trim( Sanitizer::stripAllTags( $formatter->getText() ) );
 208
 209                 if ( !$text ) {
 210                         // There isn't any text after filtering before the first heading so we declare
 211                         // that there isn't a first heading.
 212                         return null;
 213                 }
 214
 215                 return $text;
 216         }
 217
 218         /**
 219          * Get opening text
 220          * @return string
 221          */
 222         public function getOpeningText() {
 223                 $this->extractWikitextParts();
 224                 return $this->openingText;
 225         }
 226
 227         /**
 228          * Get main text
 229          * @return string
 230          */
 231         public function getMainText() {
 232                 $this->extractWikitextParts();
 233                 return $this->allText;
 234         }
 235
 236         /**
 237          * Get auxiliary text
 238          * @return string[]
 239          */
 240         public function getAuxiliaryText() {
 241                 $this->extractWikitextParts();
 242                 return $this->auxText;
 243         }
 244
 245         /**
 246          * Get the defaultsort property
 247          * @return string|null
 248          */
 249         public function getDefaultSort() {
 250                 return $this->parserOutput->getProperty( 'defaultsort' );
 251         }
 252 }