<?php
/**
 * @file
 *
 * The CSS parser
 */

namespace QueryPath\CSS;

/**
 * Parse a CSS selector.
 *
 * In CSS, a selector is used to identify which element or elements
 * in a DOM are being selected for the application of a particular style.
 * Effectively, selectors function as a query language for a structured
 * document -- almost always HTML or XML.
 *
 * This class provides an event-based parser for CSS selectors. It can be
 * used, for example, as a basis for writing a DOM query engine based on
 * CSS.
 *
 * @ingroup querypath_css
 */
class Parser {
  protected $scanner = NULL;
  protected $buffer = '';
  protected $handler = NULL;
  protected $strict = FALSE;

  protected $DEBUG = FALSE;

  /**
   * Construct a new CSS parser object. This will attempt to
   * parse the string as a CSS selector. As it parses, it will
   * send events to the EventHandler implementation.
   */
  public function __construct($string, EventHandler $handler) {
    $this->originalString = $string;
    $is = new InputStream($string);
    $this->scanner = new Scanner($is);
    $this->handler = $handler;
  }

  /**
   * Parse the selector.
   *
   * This begins an event-based parsing process that will
   * fire events as the selector is handled. A EventHandler
   * implementation will be responsible for handling the events.
   * @throws ParseException
   */
  public function parse() {

    $this->scanner->nextToken();
    while ($this->scanner->token !== FALSE) {
      // Primitive recursion detection.
      $position = $this->scanner->position();

      if ($this->DEBUG) {
        print "PARSE " . $this->scanner->token. "\n";
      }
      $this->selector();

      $finalPosition = $this->scanner->position();

      if ($this->scanner->token !== FALSE && $finalPosition == $position) {
        // If we get here, then the scanner did not pop a single character
        // off of the input stream during a full run of the parser, which
        // means that the current input does not match any recognizable
        // pattern.
        throw new ParseException('CSS selector is not well formed.');
      }

    }

  }

  /**
   * A restricted parser that can only parse simple selectors.
   * The pseudoClass handler for this parser will throw an
   * exception if it encounters a pseudo-element or the
   * negation pseudo-class.
   *
   * @deprecated This is not used anywhere in QueryPath and
   *  may be removed.
   *//*
  public function parseSimpleSelector() {
    while ($this->scanner->token !== FALSE) {
      if ($this->DEBUG) print "SIMPLE SELECTOR\n";
      $this->allElements();
      $this->elementName();
      $this->elementClass();
      $this->elementID();
      $this->pseudoClass(TRUE); // Operate in restricted mode.
      $this->attribute();

      // TODO: Need to add failure conditions here.
    }
  }*/

  /**
   * Handle an entire CSS selector.
   */
  private function selector() {
    if ($this->DEBUG) print "SELECTOR{$this->scanner->position()}\n";
    $this->consumeWhitespace(); // Remove leading whitespace
    $this->simpleSelectors();
    $this->combinator();
  }

  /**
   * Consume whitespace and return a count of the number of whitespace consumed.
   */
  private function consumeWhitespace() {
    if ($this->DEBUG) print "CONSUME WHITESPACE\n";
    $white = 0;
    while ($this->scanner->token == Token::white) {
      $this->scanner->nextToken();
      ++$white;
    }
    return $white;
  }

  /**
   * Handle one of the five combinators: '>', '+', ' ', '~', and ','.
   * This will call the appropriate event handlers.
   * @see EventHandler::directDescendant(),
   * @see EventHandler::adjacent(),
   * @see EventHandler::anyDescendant(),
   * @see EventHandler::anotherSelector().
   */
  private function combinator() {
    if ($this->DEBUG) print "COMBINATOR\n";
    /*
     * Problem: ' ' and ' > ' are both valid combinators.
     * So we have to track whitespace consumption to see
     * if we are hitting the ' ' combinator or if the
     * selector just has whitespace padding another combinator.
     */

    // Flag to indicate that post-checks need doing
    $inCombinator = FALSE;
    $white = $this->consumeWhitespace();
    $t = $this->scanner->token;

    if ($t == Token::rangle) {
      $this->handler->directDescendant();
      $this->scanner->nextToken();
      $inCombinator = TRUE;
      //$this->simpleSelectors();
    }
    elseif ($t == Token::plus) {
      $this->handler->adjacent();
      $this->scanner->nextToken();
      $inCombinator = TRUE;
      //$this->simpleSelectors();
    }
    elseif ($t == Token::comma) {
      $this->handler->anotherSelector();
      $this->scanner->nextToken();
      $inCombinator = TRUE;
      //$this->scanner->selectors();
    }
    elseif ($t == Token::tilde) {
      $this->handler->sibling();
      $this->scanner->nextToken();
      $inCombinator = TRUE;
    }

    // Check that we don't get two combinators in a row.
    if ($inCombinator) {
      $white = 0;
      if ($this->DEBUG) print "COMBINATOR: " . Token::name($t) . "\n";
      $this->consumeWhitespace();
      if ($this->isCombinator($this->scanner->token)) {
        throw new ParseException("Illegal combinator: Cannot have two combinators in sequence.");
      }
    }
    // Check to see if we have whitespace combinator:
    elseif ($white > 0) {
      if ($this->DEBUG) print "COMBINATOR: any descendant\n";
      $inCombinator = TRUE;
      $this->handler->anyDescendant();
    }
    else {
      if ($this->DEBUG) print "COMBINATOR: no combinator found.\n";
    }
  }

  /**
   * Check if the token is a combinator.
   */
  private function isCombinator($tok) {
    $combinators = array(Token::plus, Token::rangle, Token::comma, Token::tilde);
    return in_array($tok, $combinators);
  }

  /**
   * Handle a simple selector.
   */
  private function simpleSelectors() {
    if ($this->DEBUG) print "SIMPLE SELECTOR\n";
    $this->allElements();
    $this->elementName();
    $this->elementClass();
    $this->elementID();
    $this->pseudoClass();
    $this->attribute();
  }

  /**
   * Handles CSS ID selectors.
   * This will call EventHandler::elementID().
   */
  private function elementID() {
    if ($this->DEBUG) print "ELEMENT ID\n";
    if ($this->scanner->token == Token::octo) {
      $this->scanner->nextToken();
      if ($this->scanner->token !== Token::char) {
        throw new ParseException("Expected string after #");
      }
      $id = $this->scanner->getNameString();
      $this->handler->elementID($id);
    }
  }

  /**
   * Handles CSS class selectors.
   * This will call the EventHandler::elementClass() method.
   */
  private function elementClass() {
    if ($this->DEBUG) print "ELEMENT CLASS\n";
    if ($this->scanner->token == Token::dot) {
      $this->scanner->nextToken();
      $this->consumeWhitespace(); // We're very fault tolerent. This should prob through error.
      $cssClass = $this->scanner->getNameString();
      $this->handler->elementClass($cssClass);
    }
  }

  /**
   * Handle a pseudo-class and pseudo-element.
   *
   * CSS 3 selectors support separate pseudo-elements, using :: instead
   * of : for separator. This is now supported, and calls the pseudoElement
   * handler, EventHandler::pseudoElement().
   *
   * This will call EventHandler::pseudoClass() when a
   * pseudo-class is parsed.
   */
  private function pseudoClass($restricted = FALSE) {
    if ($this->DEBUG) print "PSEUDO-CLASS\n";
    if ($this->scanner->token == Token::colon) {

      // Check for CSS 3 pseudo element:
      $isPseudoElement = FALSE;
      if ($this->scanner->nextToken() === Token::colon) {
        $isPseudoElement = TRUE;
        $this->scanner->nextToken();
      }

      $name = $this->scanner->getNameString();
      if ($restricted && $name == 'not') {
        throw new ParseException("The 'not' pseudo-class is illegal in this context.");
      }

      $value = NULL;
      if ($this->scanner->token == Token::lparen) {
        if ($isPseudoElement) {
          throw new ParseException("Illegal left paren. Pseudo-Element cannot have arguments.");
        }
        $value = $this->pseudoClassValue();
      }

      // FIXME: This should throw errors when pseudo element has values.
      if ($isPseudoElement) {
        if ($restricted) {
          throw new ParseException("Pseudo-Elements are illegal in this context.");
        }
        $this->handler->pseudoElement($name);
        $this->consumeWhitespace();

        // Per the spec, pseudo-elements must be the last items in a selector, so we
        // check to make sure that we are either at the end of the stream or that a
        // new selector is starting. Only one pseudo-element is allowed per selector.
        if ($this->scanner->token !== FALSE && $this->scanner->token !== Token::comma) {
          throw new ParseException("A Pseudo-Element must be the last item in a selector.");
        }
      }
      else {
        $this->handler->pseudoClass($name, $value);
      }
    }
  }

  /**
   * Get the value of a pseudo-classes.
   *
   * @return string
   *  Returns the value found from a pseudo-class.
   *
   * @todo Pseudoclasses can be passed pseudo-elements and
   *  other pseudo-classes as values, which means :pseudo(::pseudo)
   *  is legal.
   */
  private function pseudoClassValue() {
    if ($this->scanner->token == Token::lparen) {
      $buf = '';

      // For now, just leave pseudoClass value vague.
      /*
      // We have to peek to see if next char is a colon because
      // pseudo-classes and pseudo-elements are legal strings here.
      print $this->scanner->peek();
      if ($this->scanner->peek() == ':') {
        print "Is pseudo\n";
        $this->scanner->nextToken();

        // Pseudo class
        if ($this->scanner->token == Token::colon) {
          $buf .= ':';
          $this->scanner->nextToken();
          // Pseudo element
          if ($this->scanner->token == Token::colon) {
            $buf .= ':';
            $this->scanner->nextToken();
          }
          // Ident
          $buf .= $this->scanner->getNameString();
        }
      }
      else {
        print "fetching string.\n";
        $buf .= $this->scanner->getQuotedString();
        if ($this->scanner->token != Token::rparen) {
          $this->throwError(Token::rparen, $this->scanner->token);
        }
        $this->scanner->nextToken();
      }
      return $buf;
      */
      $buf .= $this->scanner->getQuotedString();
      return $buf;
    }
  }

  /**
   * Handle element names.
   * This will call the EventHandler::elementName().
   *
   * This handles:
   * <code>
   *  name (EventHandler::element())
   *  |name (EventHandler::element())
   *  ns|name (EventHandler::elementNS())
   *  ns|* (EventHandler::elementNS())
   * </code>
   */
  private function elementName() {
    if ($this->DEBUG) print "ELEMENT NAME\n";
    if ($this->scanner->token === Token::pipe) {
      // We have '|name', which is equiv to 'name'
      $this->scanner->nextToken();
      $this->consumeWhitespace();
      $elementName =  $this->scanner->getNameString();
      $this->handler->element($elementName);
    }
    elseif ($this->scanner->token === Token::char) {
      $elementName =  $this->scanner->getNameString();
      if ($this->scanner->token == Token::pipe) {
        // Get ns|name
        $elementNS = $elementName;
        $this->scanner->nextToken();
        $this->consumeWhitespace();
        if ($this->scanner->token === Token::star) {
          // We have ns|*
          $this->handler->anyElementInNS($elementNS);
          $this->scanner->nextToken();
        }
        elseif ($this->scanner->token !== Token::char) {
          $this->throwError(Token::char, $this->scanner->token);
        }
        else {
          $elementName = $this->scanner->getNameString();
          // We have ns|name
          $this->handler->elementNS($elementName, $elementNS);
        }

      }
      else {
        $this->handler->element($elementName);
      }
    }
  }

  /**
   * Check for all elements designators. Due to the new CSS 3 namespace
   * support, this is slightly more complicated, now, as it handles
   * the *|name and *|* cases as well as *.
   *
   * Calls EventHandler::anyElement() or EventHandler::elementName().
   */
  private function allElements() {
    if ($this->scanner->token === Token::star) {
      $this->scanner->nextToken();
      if ($this->scanner->token === Token::pipe) {
        $this->scanner->nextToken();
        if ($this->scanner->token === Token::star) {
          // We got *|*. According to spec, this requires
          // that the element has a namespace, so we pass it on
          // to the handler:
          $this->scanner->nextToken();
          $this->handler->anyElementInNS('*');
        }
        else {
          // We got *|name, which means the name MUST be in a namespce,
          // so we pass this off to elementNameNS().
          $name = $this->scanner->getNameString();
          $this->handler->elementNS($name, '*');
        }
      }
      else {
        $this->handler->anyElement();
      }
    }
  }

  /**
   * Handler an attribute.
   * An attribute can be in one of two forms:
   * <code>[attrName]</code>
   * or
   * <code>[attrName="AttrValue"]</code>
   *
   * This may call the following event handlers: EventHandler::attribute().
   */
  private function attribute() {
    if($this->scanner->token == Token::lsquare) {
      $attrVal = $op = $ns = NULL;

      $this->scanner->nextToken();
      $this->consumeWhitespace();

      if ($this->scanner->token === Token::at) {
        if ($this->strict) {
          throw new ParseException('The @ is illegal in attributes.');
        }
        else {
          $this->scanner->nextToken();
          $this->consumeWhitespace();
        }
      }

      if ($this->scanner->token === Token::star) {
        // Global namespace... requires that attr be prefixed,
        // so we pass this on to a namespace handler.
        $ns = '*';
        $this->scanner->nextToken();
      }
      if ($this->scanner->token === Token::pipe) {
        // Skip this. It's a global namespace.
        $this->scanner->nextToken();
        $this->consumeWhitespace();
      }

      $attrName = $this->scanner->getNameString();
      $this->consumeWhitespace();

      // Check for namespace attribute: ns|attr. We have to peek() to make
      // sure that we haven't hit the |= operator, which looks the same.
      if ($this->scanner->token === Token::pipe && $this->scanner->peek() !== '=') {
        // We have a namespaced attribute.
        $ns = $attrName;
        $this->scanner->nextToken();
        $attrName = $this->scanner->getNameString();
        $this->consumeWhitespace();
      }

      // Note: We require that operators do not have spaces
      // between characters, e.g. ~= , not ~ =.

      // Get the operator:
      switch ($this->scanner->token) {
        case Token::eq:
          $this->consumeWhitespace();
          $op = EventHandler::isExactly;
          break;
        case Token::tilde:
          if ($this->scanner->nextToken() !== Token::eq) {
            $this->throwError(Token::eq, $this->scanner->token);
          }
          $op = EventHandler::containsWithSpace;
          break;
        case Token::pipe:
          if ($this->scanner->nextToken() !== Token::eq) {
            $this->throwError(Token::eq, $this->scanner->token);
          }
          $op = EventHandler::containsWithHyphen;
          break;
        case Token::star:
          if ($this->scanner->nextToken() !== Token::eq) {
            $this->throwError(Token::eq, $this->scanner->token);
          }
          $op = EventHandler::containsInString;
          break;
        case Token::dollar;
          if ($this->scanner->nextToken() !== Token::eq) {
            $this->throwError(Token::eq, $this->scanner->token);
          }
          $op = EventHandler::endsWith;
          break;
        case Token::carat:
          if ($this->scanner->nextToken() !== Token::eq) {
            $this->throwError(Token::eq, $this->scanner->token);
          }
          $op = EventHandler::beginsWith;
          break;
      }

      if (isset($op)) {
        // Consume '=' and go on.
        $this->scanner->nextToken();
        $this->consumeWhitespace();

        // So... here we have a problem. The grammer suggests that the
        // value here is String1 or String2, both of which are enclosed
        // in quotes of some sort, and both of which allow lots of special
        // characters. But the spec itself includes examples like this:
        //   [lang=fr]
        // So some bareword support is assumed. To get around this, we assume
        // that bare words follow the NAME rules, while quoted strings follow
        // the String1/String2 rules.

        if ($this->scanner->token === Token::quote || $this->scanner->token === Token::squote) {
          $attrVal = $this->scanner->getQuotedString();
        }
        else {
          $attrVal = $this->scanner->getNameString();
        }

        if ($this->DEBUG) {
          print "ATTR: $attrVal AND OP: $op\n";
        }
      }

      $this->consumeWhitespace();

      if ($this->scanner->token != Token::rsquare) {
        $this->throwError(Token::rsquare, $this->scanner->token);
      }

      if (isset($ns)) {
        $this->handler->attributeNS($attrName, $ns, $attrVal, $op);
      }
      elseif (isset($attrVal)) {
        $this->handler->attribute($attrName, $attrVal, $op);
      }
      else {
        $this->handler->attribute($attrName);
      }
      $this->scanner->nextToken();
    }
  }

  /**
   * Utility for throwing a consistantly-formatted parse error.
   */
  private function throwError($expected, $got) {
    $filter = sprintf('Expected %s, got %s', Token::name($expected), Token::name($got));
    throw new ParseException($filter);
  }

}