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