includes/parser/RemexRemoveTagHandler.php

   1 <?php
   2
   3 namespace MediaWiki\Parser;
   4
   5 use Wikimedia\RemexHtml\Tokenizer\Attributes;
   6 use Wikimedia\RemexHtml\Tokenizer\PlainAttributes;
   7 use Wikimedia\RemexHtml\Tokenizer\RelayTokenHandler;
   8 use Wikimedia\RemexHtml\Tokenizer\TokenHandler;
   9
  10 /**
  11  * Helper class for Sanitizer::removeSomeTags().
  12  * @internal
  13  */
  14 class RemexRemoveTagHandler extends RelayTokenHandler {
  15         /**
  16          * @var string The original HTML source string (used for fallback text
  17          * when rejecting an HTML tag).
  18          */
  19         private $source;
  20
  21         /**
  22          * @var array<string,true> Set of HTML tags which can be self-closed.
  23          */
  24         private $htmlsingle;
  25
  26         /**
  27          * @var array<string,true> Self-closed tags which are on $htmlsingle
  28          * but not on $htmlsingleonly will be emitted as an empty element.
  29          */
  30         private $htmlsingleonly;
  31
  32         /**
  33          * @var array<string,true> Set of allowed HTML open/close tags.
  34          */
  35         private $htmlelements;
  36
  37         /**
  38          * @var ?callable(Attributes,mixed...):Attributes Callback to mutate or
  39          * sanitize attributes.
  40          */
  41         private $attrCallback;
  42
  43         /**
  44          * @var ?array $args Optional extra arguments to provide to the
  45          * $attrCallback.
  46          */
  47         private $callbackArgs;
  48
  49         /**
  50          * @param TokenHandler $nextHandler Handler to relay accepted tokens.
  51          * @param string $source Input source string.
  52          * @param array $tagData Information about allowed/rejected tags.
  53          * @param ?callable $attrCallback Attribute handler callback.
  54          *   The full signature is ?callable(Attributes,mixed...):Attributes
  55          * @param ?array $callbackArgs Optional arguments to attribute handler.
  56          */
  57         public function __construct(
  58                 TokenHandler $nextHandler,
  59                 string $source,
  60                 array $tagData,
  61                 ?callable $attrCallback,
  62                 ?array $callbackArgs
  63         ) {
  64                 parent::__construct( $nextHandler );
  65                 $this->source = $source;
  66                 $this->htmlsingle = $tagData['htmlsingle'];
  67                 $this->htmlsingleonly = $tagData['htmlsingleonly'];
  68                 $this->htmlelements = $tagData['htmlelements'];
  69                 $this->attrCallback = $attrCallback;
  70                 $this->callbackArgs = $callbackArgs ?? [];
  71         }
  72
  73         /**
  74          * @inheritDoc
  75          */
  76         public function comment( $text, $sourceStart, $sourceLength ) {
  77                 // Don't relay comments.
  78         }
  79
  80         /**
  81          * Takes attribute names and values for a tag and the tag name and
  82          * validates that the tag is allowed to be present.
  83          * This DOES NOT validate the attributes, nor does it validate the
  84          * tags themselves. This method only handles the special circumstances
  85          * where we may want to allow a tag within content but ONLY when it has
  86          * specific attributes set.
  87          *
  88          * @param string $element
  89          * @param Attributes $attrs
  90          * @return bool
  91          *
  92          * @see Sanitizer::validateTag()
  93          */
  94         private static function validateTag( string $element, Attributes $attrs ): bool {
  95                 if ( $element == 'meta' || $element == 'link' ) {
  96                         $params = $attrs->getValues();
  97                         if ( !isset( $params['itemprop'] ) ) {
  98                                 // <meta> and <link> must have an itemprop="" otherwise they are not valid or safe in content
  99                                 return false;
 100                         }
 101                         if ( $element == 'meta' && !isset( $params['content'] ) ) {
 102                                 // <meta> must have a content="" for the itemprop
 103                                 return false;
 104                         }
 105                         if ( $element == 'link' && !isset( $params['href'] ) ) {
 106                                 // <link> must have an associated href=""
 107                                 return false;
 108                         }
 109                 }
 110
 111                 return true;
 112         }
 113
 114         /**
 115          * @inheritDoc
 116          */
 117         public function startTag( $name, Attributes $attrs, $selfClose, $sourceStart, $sourceLength ) {
 118                 // Handle a start tag from the tokenizer: either relay it to the
 119                 // next stage, or re-emit it as raw text.
 120
 121                 $badtag = false;
 122                 $t = strtolower( $name );
 123                 if ( isset( $this->htmlelements[$t] ) ) {
 124                         if ( $this->attrCallback ) {
 125                                 $attrs = ( $this->attrCallback )( $attrs, ...$this->callbackArgs );
 126                         }
 127                         if ( $selfClose && !( isset( $this->htmlsingle[$t] ) || isset( $this->htmlsingleonly[$t] ) ) ) {
 128                                 // Remove the self-closing slash, to be consistent with
 129                                 // HTML5 semantics. T134423
 130                                 $selfClose = false;
 131                         }
 132                         if ( !self::validateTag( $t, $attrs ) ) {
 133                                 $badtag = true;
 134                         }
 135                         $fixedAttrs = Sanitizer::validateTagAttributes( $attrs->getValues(), $t );
 136                         $attrs = new PlainAttributes( $fixedAttrs );
 137                         if ( !$badtag ) {
 138                                 if ( $selfClose && !isset( $this->htmlsingleonly[$t] ) ) {
 139                                         // Interpret self-closing tags as empty tags even when
 140                                         // HTML5 would interpret them as start tags.  Such input
 141                                         // is commonly seen on Wikimedia wikis with this intention.
 142                                         $this->nextHandler->startTag( $name, $attrs, false, $sourceStart, $sourceLength );
 143                                         $this->nextHandler->endTag( $name, $sourceStart + $sourceLength, 0 );
 144                                 } else {
 145                                         $this->nextHandler->startTag( $name, $attrs, $selfClose, $sourceStart, $sourceLength );
 146                                 }
 147                                 return;
 148                         }
 149                 }
 150                 // Emit this as a text node instead.
 151                 $this->nextHandler->characters( $this->source, $sourceStart, $sourceLength, $sourceStart, $sourceLength );
 152         }
 153
 154         /**
 155          * @inheritDoc
 156          */
 157         public function endTag( $name, $sourceStart, $sourceLength ) {
 158                 // Handle an end tag from the tokenizer: either relay it to the
 159                 // next stage, or re-emit it as raw text.
 160
 161                 $t = strtolower( $name );
 162                 if ( isset( $this->htmlelements[$t] ) ) {
 163                         // This is a good tag, relay it.
 164                         $this->nextHandler->endTag( $name, $sourceStart, $sourceLength );
 165                 } else {
 166                         // Emit this as a text node instead.
 167                         $this->nextHandler->characters( $this->source, $sourceStart, $sourceLength, $sourceStart, $sourceLength );
 168                 }
 169         }
 170
 171 }