Chris@0: <?php
Chris@0: namespace Masterminds\HTML5\Parser;
Chris@0: 
Chris@0: use Masterminds\HTML5\Elements;
Chris@0: 
Chris@0: /**
Chris@0:  * The HTML5 tokenizer.
Chris@0:  *
Chris@0:  * The tokenizer's role is reading data from the scanner and gathering it into
Chris@0:  * semantic units. From the tokenizer, data is emitted to an event handler,
Chris@0:  * which may (for example) create a DOM tree.
Chris@0:  *
Chris@0:  * The HTML5 specification has a detailed explanation of tokenizing HTML5. We
Chris@0:  * follow that specification to the maximum extent that we can. If you find
Chris@0:  * a discrepancy that is not documented, please file a bug and/or submit a
Chris@0:  * patch.
Chris@0:  *
Chris@0:  * This tokenizer is implemented as a recursive descent parser.
Chris@0:  *
Chris@0:  * Within the API documentation, you may see references to the specific section
Chris@0:  * of the HTML5 spec that the code attempts to reproduce. Example: 8.2.4.1.
Chris@0:  * This refers to section 8.2.4.1 of the HTML5 CR specification.
Chris@0:  *
Chris@0:  * @see http://www.w3.org/TR/2012/CR-html5-20121217/
Chris@0:  */
Chris@0: class Tokenizer
Chris@0: {
Chris@0: 
Chris@0:     protected $scanner;
Chris@0: 
Chris@0:     protected $events;
Chris@0: 
Chris@0:     protected $tok;
Chris@0: 
Chris@0:     /**
Chris@0:      * Buffer for text.
Chris@0:      */
Chris@0:     protected $text = '';
Chris@0: 
Chris@0:     // When this goes to false, the parser stops.
Chris@0:     protected $carryOn = true;
Chris@0: 
Chris@0:     protected $textMode = 0; // TEXTMODE_NORMAL;
Chris@0:     protected $untilTag = null;
Chris@0: 
Chris@0:     const CONFORMANT_XML = 'xml';
Chris@0:     const CONFORMANT_HTML = 'html';
Chris@0:     protected $mode = self::CONFORMANT_HTML;
Chris@0: 
Chris@0:     const WHITE = "\t\n\f ";
Chris@0: 
Chris@0:     /**
Chris@0:      * Create a new tokenizer.
Chris@0:      *
Chris@0:      * Typically, parsing a document involves creating a new tokenizer, giving
Chris@0:      * it a scanner (input) and an event handler (output), and then calling
Chris@0:      * the Tokenizer::parse() method.`
Chris@0:      *
Chris@0:      * @param \Masterminds\HTML5\Parser\Scanner $scanner
Chris@0:      *            A scanner initialized with an input stream.
Chris@0:      * @param \Masterminds\HTML5\Parser\EventHandler $eventHandler
Chris@0:      *            An event handler, initialized and ready to receive
Chris@0:      *            events.
Chris@0:      * @param string $mode
Chris@0:      */
Chris@0:     public function __construct($scanner, $eventHandler, $mode = self::CONFORMANT_HTML)
Chris@0:     {
Chris@0:         $this->scanner = $scanner;
Chris@0:         $this->events = $eventHandler;
Chris@0:         $this->mode = $mode;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Begin parsing.
Chris@0:      *
Chris@0:      * This will begin scanning the document, tokenizing as it goes.
Chris@0:      * Tokens are emitted into the event handler.
Chris@0:      *
Chris@0:      * Tokenizing will continue until the document is completely
Chris@0:      * read. Errors are emitted into the event handler, but
Chris@0:      * the parser will attempt to continue parsing until the
Chris@0:      * entire input stream is read.
Chris@0:      */
Chris@0:     public function parse()
Chris@0:     {
Chris@0:         do {
Chris@0:             $this->consumeData();
Chris@0:             // FIXME: Add infinite loop protection.
Chris@0:         } while ($this->carryOn);
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Set the text mode for the character data reader.
Chris@0:      *
Chris@0:      * HTML5 defines three different modes for reading text:
Chris@0:      * - Normal: Read until a tag is encountered.
Chris@0:      * - RCDATA: Read until a tag is encountered, but skip a few otherwise-
Chris@0:      * special characters.
Chris@0:      * - Raw: Read until a special closing tag is encountered (viz. pre, script)
Chris@0:      *
Chris@0:      * This allows those modes to be set.
Chris@0:      *
Chris@0:      * Normally, setting is done by the event handler via a special return code on
Chris@0:      * startTag(), but it can also be set manually using this function.
Chris@0:      *
Chris@0:      * @param integer $textmode
Chris@0:      *            One of Elements::TEXT_*
Chris@0:      * @param string $untilTag
Chris@0:      *            The tag that should stop RAW or RCDATA mode. Normal mode does not
Chris@0:      *            use this indicator.
Chris@0:      */
Chris@0:     public function setTextMode($textmode, $untilTag = null)
Chris@0:     {
Chris@0:         $this->textMode = $textmode & (Elements::TEXT_RAW | Elements::TEXT_RCDATA);
Chris@0:         $this->untilTag = $untilTag;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Consume a character and make a move.
Chris@0:      * HTML5 8.2.4.1
Chris@0:      */
Chris@0:     protected function consumeData()
Chris@0:     {
Chris@0:         // Character Ref
Chris@0:         /*
Chris@0:          * $this->characterReference() || $this->tagOpen() || $this->eof() || $this->characterData();
Chris@0:          */
Chris@0:         $this->characterReference();
Chris@0:         $this->tagOpen();
Chris@0:         $this->eof();
Chris@0:         $this->characterData();
Chris@0: 
Chris@0:         return $this->carryOn;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Parse anything that looks like character data.
Chris@0:      *
Chris@0:      * Different rules apply based on the current text mode.
Chris@0:      *
Chris@0:      * @see Elements::TEXT_RAW Elements::TEXT_RCDATA.
Chris@0:      */
Chris@0:     protected function characterData()
Chris@0:     {
Chris@0:         $tok = $this->scanner->current();
Chris@0:         if ($tok === false) {
Chris@0:             return false;
Chris@0:         }
Chris@0:         switch ($this->textMode) {
Chris@0:             case Elements::TEXT_RAW:
Chris@0:                 return $this->rawText();
Chris@0:             case Elements::TEXT_RCDATA:
Chris@0:                 return $this->rcdata();
Chris@0:             default:
Chris@0:                 if (strspn($tok, "<&")) {
Chris@0:                     return false;
Chris@0:                 }
Chris@0:                 return $this->text();
Chris@0:         }
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * This buffers the current token as character data.
Chris@0:      */
Chris@0:     protected function text()
Chris@0:     {
Chris@0:         $tok = $this->scanner->current();
Chris@0: 
Chris@0:         // This should never happen...
Chris@0:         if ($tok === false) {
Chris@0:             return false;
Chris@0:         }
Chris@0:         // Null
Chris@0:         if ($tok === "\00") {
Chris@0:             $this->parseError("Received null character.");
Chris@0:         }
Chris@0:         // fprintf(STDOUT, "Writing '%s'", $tok);
Chris@0:         $this->buffer($tok);
Chris@0:         $this->scanner->next();
Chris@0:         return true;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Read text in RAW mode.
Chris@0:      */
Chris@0:     protected function rawText()
Chris@0:     {
Chris@0:         if (is_null($this->untilTag)) {
Chris@0:             return $this->text();
Chris@0:         }
Chris@0:         $sequence = '</' . $this->untilTag . '>';
Chris@0:         $txt = $this->readUntilSequence($sequence);
Chris@0:         $this->events->text($txt);
Chris@0:         $this->setTextMode(0);
Chris@0:         return $this->endTag();
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Read text in RCDATA mode.
Chris@0:      */
Chris@0:     protected function rcdata()
Chris@0:     {
Chris@0:         if (is_null($this->untilTag)) {
Chris@0:             return $this->text();
Chris@0:         }
Chris@0:         $sequence = '</' . $this->untilTag;
Chris@0:         $txt = '';
Chris@0:         $tok = $this->scanner->current();
Chris@0: 
Chris@0:         $caseSensitive = !Elements::isHtml5Element($this->untilTag);
Chris@0:         while ($tok !== false && ! ($tok == '<' && ($this->sequenceMatches($sequence, $caseSensitive)))) {
Chris@0:             if ($tok == '&') {
Chris@0:                 $txt .= $this->decodeCharacterReference();
Chris@0:                 $tok = $this->scanner->current();
Chris@0:             } else {
Chris@0:                 $txt .= $tok;
Chris@0:                 $tok = $this->scanner->next();
Chris@0:             }
Chris@0:         }
Chris@0:         $len = strlen($sequence);
Chris@0:         $this->scanner->consume($len);
Chris@0:         $len += strlen($this->scanner->whitespace());
Chris@0:         if ($this->scanner->current() !== '>') {
Chris@0:             $this->parseError("Unclosed RCDATA end tag");
Chris@0:         }
Chris@0:         $this->scanner->unconsume($len);
Chris@0:         $this->events->text($txt);
Chris@0:         $this->setTextMode(0);
Chris@0:         return $this->endTag();
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * If the document is read, emit an EOF event.
Chris@0:      */
Chris@0:     protected function eof()
Chris@0:     {
Chris@0:         if ($this->scanner->current() === false) {
Chris@0:             // fprintf(STDOUT, "EOF");
Chris@0:             $this->flushBuffer();
Chris@0:             $this->events->eof();
Chris@0:             $this->carryOn = false;
Chris@0:             return true;
Chris@0:         }
Chris@0:         return false;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Handle character references (aka entities).
Chris@0:      *
Chris@0:      * This version is specific to PCDATA, as it buffers data into the
Chris@0:      * text buffer. For a generic version, see decodeCharacterReference().
Chris@0:      *
Chris@0:      * HTML5 8.2.4.2
Chris@0:      */
Chris@0:     protected function characterReference()
Chris@0:     {
Chris@0:         $ref = $this->decodeCharacterReference();
Chris@0:         if ($ref !== false) {
Chris@0:             $this->buffer($ref);
Chris@0:             return true;
Chris@0:         }
Chris@0:         return false;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Emit a tagStart event on encountering a tag.
Chris@0:      *
Chris@0:      * 8.2.4.8
Chris@0:      */
Chris@0:     protected function tagOpen()
Chris@0:     {
Chris@0:         if ($this->scanner->current() != '<') {
Chris@0:             return false;
Chris@0:         }
Chris@0: 
Chris@0:         // Any buffered text data can go out now.
Chris@0:         $this->flushBuffer();
Chris@0: 
Chris@0:         $this->scanner->next();
Chris@0: 
Chris@0:         return $this->markupDeclaration() || $this->endTag() || $this->processingInstruction() || $this->tagName() ||
Chris@0:           /*  This always returns false. */
Chris@0:           $this->parseError("Illegal tag opening") || $this->characterData();
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Look for markup.
Chris@0:      */
Chris@0:     protected function markupDeclaration()
Chris@0:     {
Chris@0:         if ($this->scanner->current() != '!') {
Chris@0:             return false;
Chris@0:         }
Chris@0: 
Chris@0:         $tok = $this->scanner->next();
Chris@0: 
Chris@0:         // Comment:
Chris@0:         if ($tok == '-' && $this->scanner->peek() == '-') {
Chris@0:             $this->scanner->next(); // Consume the other '-'
Chris@0:             $this->scanner->next(); // Next char.
Chris@0:             return $this->comment();
Chris@0:         }
Chris@0: 
Chris@0:         elseif ($tok == 'D' || $tok == 'd') { // Doctype
Chris@0:             return $this->doctype();
Chris@0:         }
Chris@0: 
Chris@0:         elseif ($tok == '[') { // CDATA section
Chris@0:             return $this->cdataSection();
Chris@0:         }
Chris@0: 
Chris@0:         // FINISH
Chris@0:         $this->parseError("Expected <!--, <![CDATA[, or <!DOCTYPE. Got <!%s", $tok);
Chris@0:         $this->bogusComment('<!');
Chris@0:         return true;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Consume an end tag.
Chris@0:      * 8.2.4.9
Chris@0:      */
Chris@0:     protected function endTag()
Chris@0:     {
Chris@0:         if ($this->scanner->current() != '/') {
Chris@0:             return false;
Chris@0:         }
Chris@0:         $tok = $this->scanner->next();
Chris@0: 
Chris@0:         // a-zA-Z -> tagname
Chris@0:         // > -> parse error
Chris@0:         // EOF -> parse error
Chris@0:         // -> parse error
Chris@0:         if (! ctype_alpha($tok)) {
Chris@0:             $this->parseError("Expected tag name, got '%s'", $tok);
Chris@0:             if ($tok == "\0" || $tok === false) {
Chris@0:                 return false;
Chris@0:             }
Chris@0:             return $this->bogusComment('</');
Chris@0:         }
Chris@0: 
Chris@0:         $name = $this->scanner->charsUntil("\n\f \t>");
Chris@0:         $name = $this->mode === self::CONFORMANT_XML ? $name: strtolower($name);
Chris@0:         // Trash whitespace.
Chris@0:         $this->scanner->whitespace();
Chris@0: 
Chris@0:         if ($this->scanner->current() != '>') {
Chris@0:             $this->parseError("Expected >, got '%s'", $this->scanner->current());
Chris@0:             // We just trash stuff until we get to the next tag close.
Chris@0:             $this->scanner->charsUntil('>');
Chris@0:         }
Chris@0: 
Chris@0:         $this->events->endTag($name);
Chris@0:         $this->scanner->next();
Chris@0:         return true;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Consume a tag name and body.
Chris@0:      * 8.2.4.10
Chris@0:      */
Chris@0:     protected function tagName()
Chris@0:     {
Chris@0:         $tok = $this->scanner->current();
Chris@0:         if (! ctype_alpha($tok)) {
Chris@0:             return false;
Chris@0:         }
Chris@0: 
Chris@0:         // We know this is at least one char.
Chris@0:         $name = $this->scanner->charsWhile(":_-0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz");
Chris@0:         $name = $this->mode === self::CONFORMANT_XML ? $name : strtolower($name);
Chris@0:         $attributes = array();
Chris@0:         $selfClose = false;
Chris@0: 
Chris@0:         // Handle attribute parse exceptions here so that we can
Chris@0:         // react by trying to build a sensible parse tree.
Chris@0:         try {
Chris@0:             do {
Chris@0:                 $this->scanner->whitespace();
Chris@0:                 $this->attribute($attributes);
Chris@0:             } while (! $this->isTagEnd($selfClose));
Chris@0:         } catch (ParseError $e) {
Chris@0:             $selfClose = false;
Chris@0:         }
Chris@0: 
Chris@0:         $mode = $this->events->startTag($name, $attributes, $selfClose);
Chris@0:         // Should we do this? What does this buy that selfClose doesn't?
Chris@0:         if ($selfClose) {
Chris@0:             $this->events->endTag($name);
Chris@0:         } elseif (is_int($mode)) {
Chris@0:             // fprintf(STDOUT, "Event response says move into mode %d for tag %s", $mode, $name);
Chris@0:             $this->setTextMode($mode, $name);
Chris@0:         }
Chris@0: 
Chris@0:         $this->scanner->next();
Chris@0: 
Chris@0:         return true;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Check if the scanner has reached the end of a tag.
Chris@0:      */
Chris@0:     protected function isTagEnd(&$selfClose)
Chris@0:     {
Chris@0:         $tok = $this->scanner->current();
Chris@0:         if ($tok == '/') {
Chris@0:             $this->scanner->next();
Chris@0:             $this->scanner->whitespace();
Chris@0:             $tok = $this->scanner->current();
Chris@0: 
Chris@0:             if ($tok == '>') {
Chris@0:                 $selfClose = true;
Chris@0:                 return true;
Chris@0:             }
Chris@0:             if ($tok === false) {
Chris@0:                 $this->parseError("Unexpected EOF inside of tag.");
Chris@0:                 return true;
Chris@0:             }
Chris@0:             // Basically, we skip the / token and go on.
Chris@0:             // See 8.2.4.43.
Chris@0:             $this->parseError("Unexpected '%s' inside of a tag.", $tok);
Chris@0:             return false;
Chris@0:         }
Chris@0: 
Chris@0:         if ($tok == '>') {
Chris@0:             return true;
Chris@0:         }
Chris@0:         if ($tok === false) {
Chris@0:             $this->parseError("Unexpected EOF inside of tag.");
Chris@0:             return true;
Chris@0:         }
Chris@0: 
Chris@0:         return false;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Parse attributes from inside of a tag.
Chris@0:      */
Chris@0:     protected function attribute(&$attributes)
Chris@0:     {
Chris@0:         $tok = $this->scanner->current();
Chris@0:         if ($tok == '/' || $tok == '>' || $tok === false) {
Chris@0:             return false;
Chris@0:         }
Chris@0: 
Chris@0:         if ($tok == '<') {
Chris@0:             $this->parseError("Unexepcted '<' inside of attributes list.");
Chris@0:             // Push the < back onto the stack.
Chris@0:             $this->scanner->unconsume();
Chris@0:             // Let the caller figure out how to handle this.
Chris@0:             throw new ParseError("Start tag inside of attribute.");
Chris@0:         }
Chris@0: 
Chris@0:         $name = strtolower($this->scanner->charsUntil("/>=\n\f\t "));
Chris@0: 
Chris@0:         if (strlen($name) == 0) {
Chris@0:             $this->parseError("Expected an attribute name, got %s.", $this->scanner->current());
Chris@0:             // Really, only '=' can be the char here. Everything else gets absorbed
Chris@0:             // under one rule or another.
Chris@0:             $name = $this->scanner->current();
Chris@0:             $this->scanner->next();
Chris@0:         }
Chris@0: 
Chris@0:         $isValidAttribute = true;
Chris@0:         // Attribute names can contain most Unicode characters for HTML5.
Chris@0:         // But method "DOMElement::setAttribute" is throwing exception
Chris@0:         // because of it's own internal restriction so these have to be filtered.
Chris@0:         // see issue #23: https://github.com/Masterminds/html5-php/issues/23
Chris@0:         // and http://www.w3.org/TR/2011/WD-html5-20110525/syntax.html#syntax-attribute-name
Chris@0:         if (preg_match("/[\x1-\x2C\\/\x3B-\x40\x5B-\x5E\x60\x7B-\x7F]/u", $name)) {
Chris@0:             $this->parseError("Unexpected characters in attribute name: %s", $name);
Chris@0:             $isValidAttribute = false;
Chris@0:         }         // There is no limitation for 1st character in HTML5.
Chris@0:         // But method "DOMElement::setAttribute" is throwing exception for the
Chris@0:         // characters below so they have to be filtered.
Chris@0:         // see issue #23: https://github.com/Masterminds/html5-php/issues/23
Chris@0:         // and http://www.w3.org/TR/2011/WD-html5-20110525/syntax.html#syntax-attribute-name
Chris@0:         else
Chris@0:             if (preg_match("/^[0-9.-]/u", $name)) {
Chris@0:                 $this->parseError("Unexpected character at the begining of attribute name: %s", $name);
Chris@0:                 $isValidAttribute = false;
Chris@0:             }
Chris@0:         // 8.1.2.3
Chris@0:         $this->scanner->whitespace();
Chris@0: 
Chris@0:         $val = $this->attributeValue();
Chris@0:         if ($isValidAttribute) {
Chris@0:             $attributes[$name] = $val;
Chris@0:         }
Chris@0:         return true;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Consume an attribute value.
Chris@0:      * 8.2.4.37 and after.
Chris@0:      */
Chris@0:     protected function attributeValue()
Chris@0:     {
Chris@0:         if ($this->scanner->current() != '=') {
Chris@0:             return null;
Chris@0:         }
Chris@0:         $this->scanner->next();
Chris@0:         // 8.1.2.3
Chris@0:         $this->scanner->whitespace();
Chris@0: 
Chris@0:         $tok = $this->scanner->current();
Chris@0:         switch ($tok) {
Chris@0:             case "\n":
Chris@0:             case "\f":
Chris@0:             case " ":
Chris@0:             case "\t":
Chris@0:                 // Whitespace here indicates an empty value.
Chris@0:                 return null;
Chris@0:             case '"':
Chris@0:             case "'":
Chris@0:                 $this->scanner->next();
Chris@0:                 return $this->quotedAttributeValue($tok);
Chris@0:             case '>':
Chris@0:                 // case '/': // 8.2.4.37 seems to allow foo=/ as a valid attr.
Chris@0:                 $this->parseError("Expected attribute value, got tag end.");
Chris@0:                 return null;
Chris@0:             case '=':
Chris@0:             case '`':
Chris@0:                 $this->parseError("Expecting quotes, got %s.", $tok);
Chris@0:                 return $this->unquotedAttributeValue();
Chris@0:             default:
Chris@0:                 return $this->unquotedAttributeValue();
Chris@0:         }
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Get an attribute value string.
Chris@0:      *
Chris@0:      * @param string $quote
Chris@0:      *            IMPORTANT: This is a series of chars! Any one of which will be considered
Chris@0:      *            termination of an attribute's value. E.g. "\"'" will stop at either
Chris@0:      *            ' or ".
Chris@0:      * @return string The attribute value.
Chris@0:      */
Chris@0:     protected function quotedAttributeValue($quote)
Chris@0:     {
Chris@0:         $stoplist = "\f" . $quote;
Chris@0:         $val = '';
Chris@0: 
Chris@0:         while (true) {
Chris@0:             $tokens = $this->scanner->charsUntil($stoplist.'&');
Chris@0:             if ($tokens !== false) {
Chris@0:                 $val .= $tokens;
Chris@0:             } else {
Chris@0:                 break;
Chris@0:             }
Chris@0: 
Chris@0:             $tok = $this->scanner->current();
Chris@0:             if ($tok == '&') {
Chris@0:                 $val .= $this->decodeCharacterReference(true, $tok);
Chris@0:                 continue;
Chris@0:             }
Chris@0:             break;
Chris@0:         }
Chris@0:         $this->scanner->next();
Chris@0:         return $val;
Chris@0:     }
Chris@0: 
Chris@0:     protected function unquotedAttributeValue()
Chris@0:     {
Chris@0:         $stoplist = "\t\n\f >";
Chris@0:         $val = '';
Chris@0:         $tok = $this->scanner->current();
Chris@0:         while (strspn($tok, $stoplist) == 0 && $tok !== false) {
Chris@0:             if ($tok == '&') {
Chris@0:                 $val .= $this->decodeCharacterReference(true);
Chris@0:                 $tok = $this->scanner->current();
Chris@0:             } else {
Chris@0:                 if (strspn($tok, "\"'<=`") > 0) {
Chris@0:                     $this->parseError("Unexpected chars in unquoted attribute value %s", $tok);
Chris@0:                 }
Chris@0:                 $val .= $tok;
Chris@0:                 $tok = $this->scanner->next();
Chris@0:             }
Chris@0:         }
Chris@0:         return $val;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Consume malformed markup as if it were a comment.
Chris@0:      * 8.2.4.44
Chris@0:      *
Chris@0:      * The spec requires that the ENTIRE tag-like thing be enclosed inside of
Chris@0:      * the comment. So this will generate comments like:
Chris@0:      *
Chris@0:      * &lt;!--&lt/+foo&gt;--&gt;
Chris@0:      *
Chris@0:      * @param string $leading
Chris@0:      *            Prepend any leading characters. This essentially
Chris@0:      *            negates the need to backtrack, but it's sort of
Chris@0:      *            a hack.
Chris@0:      */
Chris@0:     protected function bogusComment($leading = '')
Chris@0:     {
Chris@0:         $comment = $leading;
Chris@0:         $tokens = $this->scanner->charsUntil('>');
Chris@0:         if ($tokens !== false) {
Chris@0:             $comment .= $tokens;
Chris@0:         }
Chris@0:         $tok = $this->scanner->current();
Chris@0:         if ($tok !== false) {
Chris@0:             $comment .= $tok;
Chris@0:         }
Chris@0: 
Chris@0:         $this->flushBuffer();
Chris@0:         $this->events->comment($comment);
Chris@0:         $this->scanner->next();
Chris@0: 
Chris@0:         return true;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Read a comment.
Chris@0:      *
Chris@0:      * Expects the first tok to be inside of the comment.
Chris@0:      */
Chris@0:     protected function comment()
Chris@0:     {
Chris@0:         $tok = $this->scanner->current();
Chris@0:         $comment = '';
Chris@0: 
Chris@0:         // <!-->. Emit an empty comment because 8.2.4.46 says to.
Chris@0:         if ($tok == '>') {
Chris@0:             // Parse error. Emit the comment token.
Chris@0:             $this->parseError("Expected comment data, got '>'");
Chris@0:             $this->events->comment('');
Chris@0:             $this->scanner->next();
Chris@0:             return true;
Chris@0:         }
Chris@0: 
Chris@0:         // Replace NULL with the replacement char.
Chris@0:         if ($tok == "\0") {
Chris@0:             $tok = UTF8Utils::FFFD;
Chris@0:         }
Chris@0:         while (! $this->isCommentEnd()) {
Chris@0:             $comment .= $tok;
Chris@0:             $tok = $this->scanner->next();
Chris@0:         }
Chris@0: 
Chris@0:         $this->events->comment($comment);
Chris@0:         $this->scanner->next();
Chris@0:         return true;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Check if the scanner has reached the end of a comment.
Chris@0:      */
Chris@0:     protected function isCommentEnd()
Chris@0:     {
Chris@0:         $tok = $this->scanner->current();
Chris@0: 
Chris@0:         // EOF
Chris@0:         if ($tok === false) {
Chris@0:             // Hit the end.
Chris@0:             $this->parseError("Unexpected EOF in a comment.");
Chris@0:             return true;
Chris@0:         }
Chris@0: 
Chris@0:         // If it doesn't start with -, not the end.
Chris@0:         if ($tok != '-') {
Chris@0:             return false;
Chris@0:         }
Chris@0: 
Chris@0:         // Advance one, and test for '->'
Chris@0:         if ($this->scanner->next() == '-' && $this->scanner->peek() == '>') {
Chris@0:             $this->scanner->next(); // Consume the last '>'
Chris@0:             return true;
Chris@0:         }
Chris@0:         // Unread '-';
Chris@0:         $this->scanner->unconsume(1);
Chris@0:         return false;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Parse a DOCTYPE.
Chris@0:      *
Chris@0:      * Parse a DOCTYPE declaration. This method has strong bearing on whether or
Chris@0:      * not Quirksmode is enabled on the event handler.
Chris@0:      *
Chris@0:      * @todo This method is a little long. Should probably refactor.
Chris@0:      */
Chris@0:     protected function doctype()
Chris@0:     {
Chris@0:         if (strcasecmp($this->scanner->current(), 'D')) {
Chris@0:             return false;
Chris@0:         }
Chris@0:         // Check that string is DOCTYPE.
Chris@0:         $chars = $this->scanner->charsWhile("DOCTYPEdoctype");
Chris@0:         if (strcasecmp($chars, 'DOCTYPE')) {
Chris@0:             $this->parseError('Expected DOCTYPE, got %s', $chars);
Chris@0:             return $this->bogusComment('<!' . $chars);
Chris@0:         }
Chris@0: 
Chris@0:         $this->scanner->whitespace();
Chris@0:         $tok = $this->scanner->current();
Chris@0: 
Chris@0:         // EOF: die.
Chris@0:         if ($tok === false) {
Chris@0:             $this->events->doctype('html5', EventHandler::DOCTYPE_NONE, '', true);
Chris@0:             return $this->eof();
Chris@0:         }
Chris@0: 
Chris@0:         $doctypeName = '';
Chris@0: 
Chris@0:         // NULL char: convert.
Chris@0:         if ($tok === "\0") {
Chris@0:             $this->parseError("Unexpected null character in DOCTYPE.");
Chris@0:             $doctypeName .= UTF8::FFFD;
Chris@0:             $tok = $this->scanner->next();
Chris@0:         }
Chris@0: 
Chris@0:         $stop = " \n\f>";
Chris@0:         $doctypeName = $this->scanner->charsUntil($stop);
Chris@0:         // Lowercase ASCII, replace \0 with FFFD
Chris@0:         $doctypeName = strtolower(strtr($doctypeName, "\0", UTF8Utils::FFFD));
Chris@0: 
Chris@0:         $tok = $this->scanner->current();
Chris@0: 
Chris@0:         // If false, emit a parse error, DOCTYPE, and return.
Chris@0:         if ($tok === false) {
Chris@0:             $this->parseError('Unexpected EOF in DOCTYPE declaration.');
Chris@0:             $this->events->doctype($doctypeName, EventHandler::DOCTYPE_NONE, null, true);
Chris@0:             return true;
Chris@0:         }
Chris@0: 
Chris@0:         // Short DOCTYPE, like <!DOCTYPE html>
Chris@0:         if ($tok == '>') {
Chris@0:             // DOCTYPE without a name.
Chris@0:             if (strlen($doctypeName) == 0) {
Chris@0:                 $this->parseError("Expected a DOCTYPE name. Got nothing.");
Chris@0:                 $this->events->doctype($doctypeName, 0, null, true);
Chris@0:                 $this->scanner->next();
Chris@0:                 return true;
Chris@0:             }
Chris@0:             $this->events->doctype($doctypeName);
Chris@0:             $this->scanner->next();
Chris@0:             return true;
Chris@0:         }
Chris@0:         $this->scanner->whitespace();
Chris@0: 
Chris@0:         $pub = strtoupper($this->scanner->getAsciiAlpha());
Chris@0:         $white = strlen($this->scanner->whitespace());
Chris@0: 
Chris@0:         // Get ID, and flag it as pub or system.
Chris@0:         if (($pub == 'PUBLIC' || $pub == 'SYSTEM') && $white > 0) {
Chris@0:             // Get the sys ID.
Chris@0:             $type = $pub == 'PUBLIC' ? EventHandler::DOCTYPE_PUBLIC : EventHandler::DOCTYPE_SYSTEM;
Chris@0:             $id = $this->quotedString("\0>");
Chris@0:             if ($id === false) {
Chris@0:                 $this->events->doctype($doctypeName, $type, $pub, false);
Chris@0:                 return false;
Chris@0:             }
Chris@0: 
Chris@0:             // Premature EOF.
Chris@0:             if ($this->scanner->current() === false) {
Chris@0:                 $this->parseError("Unexpected EOF in DOCTYPE");
Chris@0:                 $this->events->doctype($doctypeName, $type, $id, true);
Chris@0:                 return true;
Chris@0:             }
Chris@0: 
Chris@0:             // Well-formed complete DOCTYPE.
Chris@0:             $this->scanner->whitespace();
Chris@0:             if ($this->scanner->current() == '>') {
Chris@0:                 $this->events->doctype($doctypeName, $type, $id, false);
Chris@0:                 $this->scanner->next();
Chris@0:                 return true;
Chris@0:             }
Chris@0: 
Chris@0:             // If we get here, we have <!DOCTYPE foo PUBLIC "bar" SOME_JUNK
Chris@0:             // Throw away the junk, parse error, quirks mode, return true.
Chris@0:             $this->scanner->charsUntil(">");
Chris@0:             $this->parseError("Malformed DOCTYPE.");
Chris@0:             $this->events->doctype($doctypeName, $type, $id, true);
Chris@0:             $this->scanner->next();
Chris@0:             return true;
Chris@0:         }
Chris@0: 
Chris@0:         // Else it's a bogus DOCTYPE.
Chris@0:         // Consume to > and trash.
Chris@0:         $this->scanner->charsUntil('>');
Chris@0: 
Chris@0:         $this->parseError("Expected PUBLIC or SYSTEM. Got %s.", $pub);
Chris@0:         $this->events->doctype($doctypeName, 0, null, true);
Chris@0:         $this->scanner->next();
Chris@0:         return true;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Utility for reading a quoted string.
Chris@0:      *
Chris@0:      * @param string $stopchars
Chris@0:      *            Characters (in addition to a close-quote) that should stop the string.
Chris@0:      *            E.g. sometimes '>' is higher precedence than '"' or "'".
Chris@0:      * @return mixed String if one is found (quotations omitted)
Chris@0:      */
Chris@0:     protected function quotedString($stopchars)
Chris@0:     {
Chris@0:         $tok = $this->scanner->current();
Chris@0:         if ($tok == '"' || $tok == "'") {
Chris@0:             $this->scanner->next();
Chris@0:             $ret = $this->scanner->charsUntil($tok . $stopchars);
Chris@0:             if ($this->scanner->current() == $tok) {
Chris@0:                 $this->scanner->next();
Chris@0:             } else {
Chris@0:                 // Parse error because no close quote.
Chris@0:                 $this->parseError("Expected %s, got %s", $tok, $this->scanner->current());
Chris@0:             }
Chris@0:             return $ret;
Chris@0:         }
Chris@0:         return false;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Handle a CDATA section.
Chris@0:      */
Chris@0:     protected function cdataSection()
Chris@0:     {
Chris@0:         if ($this->scanner->current() != '[') {
Chris@0:             return false;
Chris@0:         }
Chris@0:         $cdata = '';
Chris@0:         $this->scanner->next();
Chris@0: 
Chris@0:         $chars = $this->scanner->charsWhile('CDAT');
Chris@0:         if ($chars != 'CDATA' || $this->scanner->current() != '[') {
Chris@0:             $this->parseError('Expected [CDATA[, got %s', $chars);
Chris@0:             return $this->bogusComment('<![' . $chars);
Chris@0:         }
Chris@0: 
Chris@0:         $tok = $this->scanner->next();
Chris@0:         do {
Chris@0:             if ($tok === false) {
Chris@0:                 $this->parseError('Unexpected EOF inside CDATA.');
Chris@0:                 $this->bogusComment('<![CDATA[' . $cdata);
Chris@0:                 return true;
Chris@0:             }
Chris@0:             $cdata .= $tok;
Chris@0:             $tok = $this->scanner->next();
Chris@0:         } while (! $this->sequenceMatches(']]>'));
Chris@0: 
Chris@0:         // Consume ]]>
Chris@0:         $this->scanner->consume(3);
Chris@0: 
Chris@0:         $this->events->cdata($cdata);
Chris@0:         return true;
Chris@0:     }
Chris@0: 
Chris@0:     // ================================================================
Chris@0:     // Non-HTML5
Chris@0:     // ================================================================
Chris@0:     /**
Chris@0:      * Handle a processing instruction.
Chris@0:      *
Chris@0:      * XML processing instructions are supposed to be ignored in HTML5,
Chris@0:      * treated as "bogus comments". However, since we're not a user
Chris@0:      * agent, we allow them. We consume until ?> and then issue a
Chris@0:      * EventListener::processingInstruction() event.
Chris@0:      */
Chris@0:     protected function processingInstruction()
Chris@0:     {
Chris@0:         if ($this->scanner->current() != '?') {
Chris@0:             return false;
Chris@0:         }
Chris@0: 
Chris@0:         $tok = $this->scanner->next();
Chris@0:         $procName = $this->scanner->getAsciiAlpha();
Chris@0:         $white = strlen($this->scanner->whitespace());
Chris@0: 
Chris@0:         // If not a PI, send to bogusComment.
Chris@0:         if (strlen($procName) == 0 || $white == 0 || $this->scanner->current() == false) {
Chris@0:             $this->parseError("Expected processing instruction name, got $tok");
Chris@0:             $this->bogusComment('<?' . $tok . $procName);
Chris@0:             return true;
Chris@0:         }
Chris@0: 
Chris@0:         $data = '';
Chris@0:         // As long as it's not the case that the next two chars are ? and >.
Chris@0:         while (! ($this->scanner->current() == '?' && $this->scanner->peek() == '>')) {
Chris@0:             $data .= $this->scanner->current();
Chris@0: 
Chris@0:             $tok = $this->scanner->next();
Chris@0:             if ($tok === false) {
Chris@0:                 $this->parseError("Unexpected EOF in processing instruction.");
Chris@0:                 $this->events->processingInstruction($procName, $data);
Chris@0:                 return true;
Chris@0:             }
Chris@0:         }
Chris@0: 
Chris@0:         $this->scanner->next(); // >
Chris@0:         $this->scanner->next(); // Next token.
Chris@0:         $this->events->processingInstruction($procName, $data);
Chris@0:         return true;
Chris@0:     }
Chris@0: 
Chris@0:     // ================================================================
Chris@0:     // UTILITY FUNCTIONS
Chris@0:     // ================================================================
Chris@0: 
Chris@0:     /**
Chris@0:      * Read from the input stream until we get to the desired sequene
Chris@0:      * or hit the end of the input stream.
Chris@0:      */
Chris@0:     protected function readUntilSequence($sequence)
Chris@0:     {
Chris@0:         $buffer = '';
Chris@0: 
Chris@0:         // Optimization for reading larger blocks faster.
Chris@0:         $first = substr($sequence, 0, 1);
Chris@0:         while ($this->scanner->current() !== false) {
Chris@0:             $buffer .= $this->scanner->charsUntil($first);
Chris@0: 
Chris@0:             // Stop as soon as we hit the stopping condition.
Chris@0:             if ($this->sequenceMatches($sequence, false)) {
Chris@0:                 return $buffer;
Chris@0:             }
Chris@0:             $buffer .= $this->scanner->current();
Chris@0:             $this->scanner->next();
Chris@0:         }
Chris@0: 
Chris@0:         // If we get here, we hit the EOF.
Chris@0:         $this->parseError("Unexpected EOF during text read.");
Chris@0:         return $buffer;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Check if upcomming chars match the given sequence.
Chris@0:      *
Chris@0:      * This will read the stream for the $sequence. If it's
Chris@0:      * found, this will return true. If not, return false.
Chris@0:      * Since this unconsumes any chars it reads, the caller
Chris@0:      * will still need to read the next sequence, even if
Chris@0:      * this returns true.
Chris@0:      *
Chris@0:      * Example: $this->sequenceMatches('</script>') will
Chris@0:      * see if the input stream is at the start of a
Chris@0:      * '</script>' string.
Chris@0:      */
Chris@0:     protected function sequenceMatches($sequence, $caseSensitive = true)
Chris@0:     {
Chris@0:         $len = strlen($sequence);
Chris@0:         $buffer = '';
Chris@0:         for ($i = 0; $i < $len; ++ $i) {
Chris@0:             $tok = $this->scanner->current();
Chris@0:             $buffer .= $tok;
Chris@0: 
Chris@0:             // EOF. Rewind and let the caller handle it.
Chris@0:             if ($tok === false) {
Chris@0:                 $this->scanner->unconsume($i);
Chris@0:                 return false;
Chris@0:             }
Chris@0:             $this->scanner->next();
Chris@0:         }
Chris@0: 
Chris@0:         $this->scanner->unconsume($len);
Chris@0:         return $caseSensitive ? $buffer == $sequence : strcasecmp($buffer, $sequence) === 0;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Send a TEXT event with the contents of the text buffer.
Chris@0:      *
Chris@0:      * This emits an EventHandler::text() event with the current contents of the
Chris@0:      * temporary text buffer. (The buffer is used to group as much PCDATA
Chris@0:      * as we can instead of emitting lots and lots of TEXT events.)
Chris@0:      */
Chris@0:     protected function flushBuffer()
Chris@0:     {
Chris@0:         if ($this->text === '') {
Chris@0:             return;
Chris@0:         }
Chris@0:         $this->events->text($this->text);
Chris@0:         $this->text = '';
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Add text to the temporary buffer.
Chris@0:      *
Chris@0:      * @see flushBuffer()
Chris@0:      */
Chris@0:     protected function buffer($str)
Chris@0:     {
Chris@0:         $this->text .= $str;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Emit a parse error.
Chris@0:      *
Chris@0:      * A parse error always returns false because it never consumes any
Chris@0:      * characters.
Chris@0:      */
Chris@0:     protected function parseError($msg)
Chris@0:     {
Chris@0:         $args = func_get_args();
Chris@0: 
Chris@0:         if (count($args) > 1) {
Chris@0:             array_shift($args);
Chris@0:             $msg = vsprintf($msg, $args);
Chris@0:         }
Chris@0: 
Chris@0:         $line = $this->scanner->currentLine();
Chris@0:         $col = $this->scanner->columnOffset();
Chris@0:         $this->events->parseError($msg, $line, $col);
Chris@0:         return false;
Chris@0:     }
Chris@0: 
Chris@0:     /**
Chris@0:      * Decode a character reference and return the string.
Chris@0:      *
Chris@0:      * Returns false if the entity could not be found. If $inAttribute is set
Chris@0:      * to true, a bare & will be returned as-is.
Chris@0:      *
Chris@0:      * @param boolean $inAttribute
Chris@0:      *            Set to true if the text is inside of an attribute value.
Chris@0:      *            false otherwise.
Chris@0:      */
Chris@0:     protected function decodeCharacterReference($inAttribute = false)
Chris@0:     {
Chris@0: 
Chris@0:         // If it fails this, it's definitely not an entity.
Chris@0:         if ($this->scanner->current() != '&') {
Chris@0:             return false;
Chris@0:         }
Chris@0: 
Chris@0:         // Next char after &.
Chris@0:         $tok = $this->scanner->next();
Chris@0:         $entity = '';
Chris@0:         $start = $this->scanner->position();
Chris@0: 
Chris@0:         if ($tok == false) {
Chris@0:             return '&';
Chris@0:         }
Chris@0: 
Chris@0:         // These indicate not an entity. We return just
Chris@0:         // the &.
Chris@0:         if (strspn($tok, static::WHITE . "&<") == 1) {
Chris@0:             // $this->scanner->next();
Chris@0:             return '&';
Chris@0:         }
Chris@0: 
Chris@0:         // Numeric entity
Chris@0:         if ($tok == '#') {
Chris@0:             $tok = $this->scanner->next();
Chris@0: 
Chris@0:             // Hexidecimal encoding.
Chris@0:             // X[0-9a-fA-F]+;
Chris@0:             // x[0-9a-fA-F]+;
Chris@0:             if ($tok == 'x' || $tok == 'X') {
Chris@0:                 $tok = $this->scanner->next(); // Consume x
Chris@0: 
Chris@0:                 // Convert from hex code to char.
Chris@0:                 $hex = $this->scanner->getHex();
Chris@0:                 if (empty($hex)) {
Chris@0:                     $this->parseError("Expected &#xHEX;, got &#x%s", $tok);
Chris@0:                     // We unconsume because we don't know what parser rules might
Chris@0:                     // be in effect for the remaining chars. For example. '&#>'
Chris@0:                     // might result in a specific parsing rule inside of tag
Chris@0:                     // contexts, while not inside of pcdata context.
Chris@0:                     $this->scanner->unconsume(2);
Chris@0:                     return '&';
Chris@0:                 }
Chris@0:                 $entity = CharacterReference::lookupHex($hex);
Chris@0:             }             // Decimal encoding.
Chris@0:             // [0-9]+;
Chris@0:             else {
Chris@0:                 // Convert from decimal to char.
Chris@0:                 $numeric = $this->scanner->getNumeric();
Chris@0:                 if ($numeric === false) {
Chris@0:                     $this->parseError("Expected &#DIGITS;, got &#%s", $tok);
Chris@0:                     $this->scanner->unconsume(2);
Chris@0:                     return '&';
Chris@0:                 }
Chris@0:                 $entity = CharacterReference::lookupDecimal($numeric);
Chris@0:             }
Chris@0:         } elseif ($tok === '=' && $inAttribute) {
Chris@0:             return '&';
Chris@0:         } else { // String entity.
Chris@0: 
Chris@0:             // Attempt to consume a string up to a ';'.
Chris@0:             // [a-zA-Z0-9]+;
Chris@0:             $cname = $this->scanner->getAsciiAlphaNum();
Chris@0:             $entity = CharacterReference::lookupName($cname);
Chris@0: 
Chris@0:             // When no entity is found provide the name of the unmatched string
Chris@0:             // and continue on as the & is not part of an entity. The & will
Chris@0:             // be converted to &amp; elsewhere.
Chris@0:             if ($entity == null) {
Chris@0:                 if (!$inAttribute || strlen($cname) === 0) {
Chris@0:                     $this->parseError("No match in entity table for '%s'", $cname);
Chris@0:                 }
Chris@0:                 $this->scanner->unconsume($this->scanner->position() - $start);
Chris@0:                 return '&';
Chris@0:             }
Chris@0:         }
Chris@0: 
Chris@0:         // The scanner has advanced the cursor for us.
Chris@0:         $tok = $this->scanner->current();
Chris@0: 
Chris@0:         // We have an entity. We're done here.
Chris@0:         if ($tok == ';') {
Chris@0:             $this->scanner->next();
Chris@0:             return $entity;
Chris@0:         }
Chris@0: 
Chris@0:         // If in an attribute, then failing to match ; means unconsume the
Chris@0:         // entire string. Otherwise, failure to match is an error.
Chris@0:         if ($inAttribute) {
Chris@0:             $this->scanner->unconsume($this->scanner->position() - $start);
Chris@0:             return '&';
Chris@0:         }
Chris@0: 
Chris@0:         $this->parseError("Expected &ENTITY;, got &ENTITY%s (no trailing ;) ", $tok);
Chris@0:         return '&' . $entity;
Chris@0:     }
Chris@0: }