<?php
namespace HTML5\Parser;

use HTML5\Elements;
/**
 * Create an HTML5 DOM tree from events.
 *
 * This attempts to create a DOM from events emitted by a parser. This
 * attempts (but does not guarantee) to up-convert older HTML documents
 * to HTML5. It does this by applying HTML5's rules, but it will not
 * change the architecture of the document itself.
 *
 * Many of the error correction and quirks features suggested in the specification
 * are implemented herein; however, not all of them are. Since we do not
 * assume a graphical user agent, no presentation-specific logic is conducted
 * during tree building.
 *
 * FIXME: The present tree builder does not exactly follow the state machine rules
 * for insert modes as outlined in the HTML5 spec. The processor needs to be
 * re-written to accomodate this. See, for example, the Go language HTML5
 * parser.
 */
class DOMTreeBuilder implements EventHandler {


  /**
   * Defined in 8.2.5.
   */
  const IM_INITIAL = 0;
  const IM_BEFORE_HTML = 1;
  const IM_BEFORE_HEAD = 2;
  const IM_IN_HEAD = 3;
  const IM_IN_HEAD_NOSCRIPT = 4;
  const IM_AFTER_HEAD = 5;
  const IM_IN_BODY = 6;
  const IM_TEXT = 7;
  const IM_IN_TABLE = 8;
  const IM_IN_TABLE_TEXT = 9;
  const IM_IN_CAPTION = 10;
  const IM_IN_COLUMN_GROUP = 11;
  const IM_IN_TABLE_BODY = 12;
  const IM_IN_ROW = 13;
  const IM_IN_CELL = 14;
  const IM_IN_SELECT = 15;
  const IM_IN_SELECT_IN_TABLE = 16;
  const IM_AFTER_BODY = 17;
  const IM_IN_FRAMESET = 18;
  const IM_AFTER_FRAMESET = 19;
  const IM_AFTER_AFTER_BODY = 20;
  const IM_AFTER_AFTER_FRAMESET = 21;

  const IM_IN_SVG = 22;
  const IM_IN_MATHML = 23;

  protected $stack = array();
  protected $current; // Pointer in the tag hierarchy.
  protected $doc;

  protected $processor;

  protected $insertMode = 0;

  /**
   * Quirks mode is enabled by default. Any document that is missing the
   * DT will be considered to be in quirks mode.
   */
  protected $quirks = TRUE;

  public $isFragment = FALSE;

  public function __construct($isFragment = FALSE) {
    $impl = new \DOMImplementation();
    // XXX:
    // Create the doctype. For now, we are always creating HTML5
    // documents, and attempting to up-convert any older DTDs to HTML5.
    $dt = $impl->createDocumentType('html');
    //$this->doc = \DOMImplementation::createDocument(NULL, 'html', $dt);
    $this->doc = $impl->createDocument(NULL, NULL, $dt);
    $this->doc->errors = array();

    // $this->current = $this->doc->documentElement;
    $this->current = $this->doc; //->documentElement;

    // Create a rules engine for tags.
    $this->rules = new TreeBuildingRules($this->doc);

    if ($isFragment) {
      $this->isFragment = TRUE;
      $this->insertMode = static::IM_IN_BODY;
      $ele = $this->doc->createElement('html');
      $this->doc->appendChild($ele);
      $this->current = $ele;
    }
  }

  /**
   * Get the document.
   */
  public function document() {
    return $this->doc;
  }

  /**
   * Get the DOM fragment for the body.
   *
   * This returns a DOMNodeList because a fragment may have zero or more
   * DOMNodes at its root.
   *
   * @see http://www.w3.org/TR/2012/CR-html5-20121217/syntax.html#concept-frag-parse-context
   *
   * @return \DOMFragmentDocumentFragment
   */
  public function fragment() {
    $append = $this->doc->documentElement->childNodes;
    $frag = $this->doc->createDocumentFragment();

    // appendChild() modifies the DOMNodeList, so we
    // have to buffer up the items first, then use the
    // array buffer and loop twice.
    $buffer = array();
    foreach ($append as $node) {
      $buffer[] = $node;
    }

    foreach ($buffer as $node) {
      $frag->appendChild($node);
    }

    $frag->errors = $this->doc->errors;
    return $frag;
  }

  /**
   * Provide an instruction processor.
   *
   * This is used for handling Processor Instructions as they are
   * inserted. If omitted, PI's are inserted directly into the DOM tree.
   */
  public function setInstructionProcessor(\HTML5\InstructionProcessor $proc) {
    $this->processor = $proc;
  }

  public function doctype($name, $idType = 0, $id = NULL, $quirks = FALSE) {
    // This is used solely for setting quirks mode. Currently we don't
    // try to preserve the inbound DT. We convert it to HTML5.
    $this->quirks = $quirks;

    if ($this->insertMode > static::IM_INITIAL) {
      $this->parseError("Illegal placement of DOCTYPE tag. Ignoring: " . $name);
      return;
    }

    $this->insertMode = static::IM_BEFORE_HTML;
  }

  /**
   * Process the start tag.
   *
   * @todo
   *   - XMLNS namespace handling (we need to parse, even if it's not valid)
   *   - XLink, MathML and SVG namespace handling
   *   - Omission rules: 8.1.2.4 Optional tags
   */
  public function startTag($name, $attributes = array(), $selfClosing = FALSE) {
    // fprintf(STDOUT, $name);
    $lname = $this->normalizeTagName($name);

    // Make sure we have an html element.
    if (!$this->doc->documentElement && $name !== 'html') {
      $this->startTag('html');
    }

    // Set quirks mode if we're at IM_INITIAL with no doctype.
    if ($this->insertMode == static::IM_INITIAL) {
      $this->quirks = TRUE;
      $this->parseError("No DOCTYPE specified.");
    }

    // SPECIAL TAG HANDLING:
    // Spec says do this, and "don't ask."
    if ($name == 'image') {
      $name = 'img';
    }


    // Autoclose p tags where appropriate.
    if ($this->insertMode >= static::IM_IN_BODY && Elements::isA($name, Elements::AUTOCLOSE_P)) {
      $this->autoclose('p');
    }

    // Set insert mode:
    switch ($name) {
    case 'html':
      $this->insertMode = static::IM_BEFORE_HEAD;
      break;
    case 'head':
      if ($this->insertMode > static::IM_BEFORE_HEAD) {
        $this->parseError("Unexpected head tag outside of head context.");
      }
      else {
        $this->insertMode = static::IM_IN_HEAD;
      }
      break;
    case 'body':
      $this->insertMode = static::IM_IN_BODY;
      break;
    case 'svg':
      $this->insertMode = static::IM_IN_SVG;
      break;
    case 'math':
      $this->insertMode = static::IM_IN_MATHML;
      break;
    case 'noscript':
      if ($this->insertMode == static::IM_IN_HEAD) {
        $this->insertMode = static::IM_IN_HEAD_NOSCRIPT;
      }
      break;

    }

    // Special case handling for SVG.
    if ($this->insertMode == static::IM_IN_SVG) {
      $lname = Elements::normalizeSvgElement($lname);
    }

    try {
      $ele = $this->doc->createElement($lname);
    }
    catch(\DOMException $e) {
      $this->parseError("Illegal tag name: <$lname>. Replaced with <invalid>.");
      $ele = $this->doc->createElement('invalid');
    }

    foreach ($attributes as $aName => $aVal) {

      if ($this->insertMode == static::IM_IN_SVG) {
        $aName = Elements::normalizeSvgAttribute($aName);
      }
      elseif ($this->insertMode == static::IM_IN_MATHML) {
        $aName = Elements::normalizeMathMlAttribute($aName);
      }

      try {
        $ele->setAttribute($aName, $aVal);
      }
      catch(\DOMException $e) {
        $this->parseError("Illegal attribute name for tag $name. Ignoring: $aName");
        continue;
      }

      // This is necessary on a non-DTD schema, like HTML5.
      if ($aName == 'id') {
        $ele->setIdAttribute('id', TRUE);
      }
    }

    // Some elements have special processing rules. Handle those separately.
    if ($this->rules->hasRules($name)) {
      $this->current = $this->rules->evaluate($ele, $this->current);
    }
    // Otherwise, it's a standard element.
    else {
      $this->current->appendChild($ele);

      // XXX: Need to handle self-closing tags and unary tags.
      if (!Elements::isA($name, Elements::VOID_TAG)) {
        $this->current = $ele;
      }
    }

    // This is sort of a last-ditch attempt to correct for cases where no head/body
    // elements are provided.
    if ($this->insertMode <= static::IM_BEFORE_HEAD && $name != 'head' && $name != 'html') {
      $this->insertMode = static::IM_IN_BODY;
    }

    // Return the element mask, which the tokenizer can then use to set
    // various processing rules.
    return Elements::element($name);
  }

  public function endTag($name) {
    $lname = $this->normalizeTagName($name);

    // Ignore closing tags for unary elements.
    if (Elements::isA($name, Elements::VOID_TAG)) {
      return;
    }

    if ($this->insertMode <= static::IM_BEFORE_HTML) {
      // 8.2.5.4.2
      if (in_array($name, array('html', 'br', 'head', 'title'))) {
        $this->startTag('html');
        $this->endTag($name);
        $this->insertMode = static::IM_BEFORE_HEAD;
        return;
      }

      // Ignore the tag.
      $this->parseError("Illegal closing tag at global scope.");
      return;
    }

    // Special case handling for SVG.
    if ($this->insertMode == static::IM_IN_SVG) {
      $lname = Elements::normalizeSvgElement($lname);
    }

    // XXX: Not sure whether we need this anymore.
    // if ($name != $lname) {
    //  return $this->quirksTreeResolver($lname);
    //}

    // XXX: HTML has no parent. What do we do, though,
    // if this element appears in the wrong place?
    if ($lname == 'html') {
      return;
    }

    //$this->current = $this->current->parentNode;
    if (!$this->autoclose($lname)) {
      $this->parseError('Could not find closing tag for ' . $lname);
    }

    //switch ($this->insertMode) {
    switch ($lname) {
    case "head":
      $this->insertMode = static::IM_AFTER_HEAD;
      break;
    case "body":
      $this->insertMode = static::IM_AFTER_BODY;
      break;
    case "svg":
    case "mathml":
      $this->insertMode = static::IM_IN_BODY;
      break;
    }
  }

  public function comment($cdata) {
    // TODO: Need to handle case where comment appears outside of the HTML tag.
    $node = $this->doc->createComment($cdata);
    $this->current->appendChild($node);
  }

  public function text($data) {
    // XXX: Hmmm.... should we really be this strict?
    if ($this->insertMode < static::IM_IN_HEAD) {
      // Per '8.2.5.4.3 The "before head" insertion mode' the characters
      // " \t\n\r\f" should be ignored but no mention of a parse error. This is
      // practical as most documents contain these characters. Other text is not
      // expected here so recording a parse error is necessary.
      $dataTmp = trim($data, " \t\n\r\f");
      if (!empty($dataTmp)) {
        //fprintf(STDOUT, "Unexpected insert mode: %d", $this->insertMode);
        $this->parseError("Unexpected text. Ignoring: " . $dataTmp);
      }
      return;
    }
    //fprintf(STDOUT, "Appending text %s.", $data);
    $node = $this->doc->createTextNode($data);
    $this->current->appendChild($node);
  }

  public function eof() {
    // If the $current isn't the $root, do we need to do anything?
  }

  public function parseError($msg, $line = 0, $col = 0) {
    $this->doc->errors[] = sprintf("Line %d, Col %d: %s", $line, $col, $msg);
  }

  public function cdata($data) {
    $node = $this->doc->createCDATASection($data);
    $this->current->appendChild($node);
  }

  public function processingInstruction($name, $data = NULL) {
    // XXX: Ignore initial XML declaration, per the spec.
    if ($this->insertMode == static::IM_INITIAL && 'xml' == strtolower($name)) {
      return;
    }

    // Important: The processor may modify the current DOM tree however
    // it sees fit.
    if (isset($this->processor)) {
      $res = $this->processor->process($this->current, $name, $data);
      if (!empty($res)) {
        $this->current = $res;
      }
      return;
    }

    // Otherwise, this is just a dumb PI element.
    $node = $this->doc->createProcessingInstruction($name, $data);

    $this->current->appendChild($node);
  }

  // ==========================================================================
  // UTILITIES
  // ==========================================================================

  /**
   * Apply normalization rules to a tag name.
   *
   * See sections 2.9 and 8.1.2.
   *
   * @param string $name
   *   The tag name.
   * @return string
   *   The normalized tag name.
   */
  protected function normalizeTagName($name) {
    /* Section 2.9 suggests that we should not do this.
    if (strpos($name, ':') !== FALSE) {
      // We know from the grammar that there must be at least one other
      // char besides :, since : is not a legal tag start.
      $parts = explode(':', $name);
      return array_pop($parts);
    }
     */

    return $name;
  }

  protected function quirksTreeResolver($name) {
    throw new \Exception("Not implemented.");

  }

  /**
   * Automatically climb the tree and close the closest node with the matching $tag.
   */
  protected function autoclose($tag) {
    $working = $this->current;
    do {
      if ($working->nodeType != XML_ELEMENT_NODE) {
        return FALSE;
      }
      if ($working->tagName == $tag) {
        $this->current = $working->parentNode;
        return TRUE;
      }
    } while ($working = $working->parentNode);
    return FALSE;

  }

  /**
   * Checks if the given tagname is an ancestor of the present candidate.
   *
   * If $this->current or anything above $this->current matches the given tag
   * name, this returns TRUE.
   */
  protected function isAncestor($tagname) {
    $candidate = $this->current;
    while ($candidate->nodeType === XML_ELEMENT_NODE) {
      if ($candidate->tagName == $tagname) {
        return TRUE;
      }
      $candidate = $candidate->parentNode;
    }
    return FALSE;
  }

  /**
   * Returns TRUE if the immediate parent element is of the given tagname.
   */
  protected function isParent($tagname) {
    return $this->current->tagName == $tagname;
  }


}