PHPIndex

<?php

/**
 * Definition that uses different definitions depending on context.
 *
 * The del and ins tags are notable because they allow different types of
 * elements depending on whether or not they're in a block or inline context.
 * Chameleon allows this behavior to happen by using two different
 * definitions depending on context.  While this somewhat generalized,
 * it is specifically intended for those two tags.
 */
class HTMLPurifier_ChildDef_Chameleon extends HTMLPurifier_ChildDef
{

    /**
     * Instance of the definition object to use when inline. Usually stricter.
     * @type HTMLPurifier_ChildDef_Optional
     */
    public $inline;

    /**
     * Instance of the definition object to use when block.
     * @type HTMLPurifier_ChildDef_Optional
     */
    public $block;

    /**
     * @type string
     */
    public $type = 'chameleon';

    /**
     * @param array $inline List of elements to allow when inline.
     * @param array $block List of elements to allow when block.
     */
    public function __construct($inline, $block)
    {
        $this->inline = new HTMLPurifier_ChildDef_Optional($inline);
        $this->block = new HTMLPurifier_ChildDef_Optional($block);
        $this->elements = $this->block->elements;
    }

    /**
     * @param HTMLPurifier_Node[] $children
     * @param HTMLPurifier_Config $config
     * @param HTMLPurifier_Context $context
     * @return bool
     */
    public function validateChildren($children, $config, $context)
    {
        if ($context->get('IsInline') === false) {
            return $this->block->validateChildren(
                $children,
                $config,
                $context
            );
        } else {
            return $this->inline->validateChildren(
                $children,
                $config,
                $context
            );
        }
    }
}

// vim: et sw=4 sts=4

<?php

/**
 * Custom validation class, accepts DTD child definitions
 *
 * @warning Currently this class is an all or nothing proposition, that is,
 *          it will only give a bool return value.
 */
class HTMLPurifier_ChildDef_Custom extends HTMLPurifier_ChildDef
{
    /**
     * @type string
     */
    public $type = 'custom';

    /**
     * @type bool
     */
    public $allow_empty = false;

    /**
     * Allowed child pattern as defined by the DTD.
     * @type string
     */
    public $dtd_regex;

    /**
     * PCRE regex derived from $dtd_regex.
     * @type string
     */
    private $_pcre_regex;

    /**
     * @param $dtd_regex Allowed child pattern from the DTD
     */
    public function __construct($dtd_regex)
    {
        $this->dtd_regex = $dtd_regex;
        $this->_compileRegex();
    }

    /**
     * Compiles the PCRE regex from a DTD regex ($dtd_regex to $_pcre_regex)
     */
    protected function _compileRegex()
    {
        $raw = str_replace(' ', '', $this->dtd_regex);
        if ($raw[0] != '(') {
            $raw = "($raw)";
        }
        $el = '[#a-zA-Z0-9_.-]+';
        $reg = $raw;

        // COMPLICATED! AND MIGHT BE BUGGY! I HAVE NO CLUE WHAT I'M
        // DOING! Seriously: if there's problems, please report them.

        // collect all elements into the $elements array
        preg_match_all("/$el/", $reg, $matches);
        foreach ($matches[0] as $match) {
            $this->elements[$match] = true;
        }

        // setup all elements as parentheticals with leading commas
        $reg = preg_replace("/$el/", '(,\\0)', $reg);

        // remove commas when they were not solicited
        $reg = preg_replace("/([^,(|]\(+),/", '\\1', $reg);

        // remove all non-paranthetical commas: they are handled by first regex
        $reg = preg_replace("/,\(/", '(', $reg);

        $this->_pcre_regex = $reg;
    }

    /**
     * @param HTMLPurifier_Node[] $children
     * @param HTMLPurifier_Config $config
     * @param HTMLPurifier_Context $context
     * @return bool
     */
    public function validateChildren($children, $config, $context)
    {
        $list_of_children = '';
        $nesting = 0; // depth into the nest
        foreach ($children as $node) {
            if (!empty($node->is_whitespace)) {
                continue;
            }
            $list_of_children .= $node->name . ',';
        }
        // add leading comma to deal with stray comma declarations
        $list_of_children = ',' . rtrim($list_of_children, ',');
        $okay =
            preg_match(
                '/^,?' . $this->_pcre_regex . '$/',
                $list_of_children
            );
        return (bool)$okay;
    }
}

// vim: et sw=4 sts=4

<?php

/**
 * Definition that disallows all elements.
 * @warning validateChildren() in this class is actually never called, because
 *          empty elements are corrected in HTMLPurifier_Strategy_MakeWellFormed
 *          before child definitions are parsed in earnest by
 *          HTMLPurifier_Strategy_FixNesting.
 */
class HTMLPurifier_ChildDef_Empty extends HTMLPurifier_ChildDef
{
    /**
     * @type bool
     */
    public $allow_empty = true;

    /**
     * @type string
     */
    public $type = 'empty';

    public function __construct()
    {
    }

    /**
     * @param HTMLPurifier_Node[] $children
     * @param HTMLPurifier_Config $config
     * @param HTMLPurifier_Context $context
     * @return array
     */
    public function validateChildren($children, $config, $context)
    {
        return array();
    }
}

// vim: et sw=4 sts=4

<?php

/**
 * Definition for list containers ul and ol.
 *
 * What does this do?  The big thing is to handle ol/ul at the top
 * level of list nodes, which should be handled specially by /folding/
 * them into the previous list node.  We generally shouldn't ever
 * see other disallowed elements, because the autoclose behavior
 * in MakeWellFormed handles it.
 */
class HTMLPurifier_ChildDef_List extends HTMLPurifier_ChildDef
{
    /**
     * @type string
     */
    public $type = 'list';
    /**
     * @type array
     */
    // lying a little bit, so that we can handle ul and ol ourselves
    // XXX: This whole business with 'wrap' is all a bit unsatisfactory
    public $elements = array('li' => true, 'ul' => true, 'ol' => true);

    public $whitespace;

    /**
     * @param array $children
     * @param HTMLPurifier_Config $config
     * @param HTMLPurifier_Context $context
     * @return array
     */
    public function validateChildren($children, $config, $context)
    {
        // Flag for subclasses
        $this->whitespace = false;

        // if there are no tokens, delete parent node
        if (empty($children)) {
            return false;
        }

        // if li is not allowed, delete parent node
        if (!isset($config->getHTMLDefinition()->info['li'])) {
            trigger_error("Cannot allow ul/ol without allowing li", E_USER_WARNING);
            return false;
        }

        // the new set of children
        $result = array();

        // a little sanity check to make sure it's not ALL whitespace
        $all_whitespace = true;

        $current_li = null;

        foreach ($children as $node) {
            if (!empty($node->is_whitespace)) {
                $result[] = $node;
                continue;
            }
            $all_whitespace = false; // phew, we're not talking about whitespace

            if ($node->name === 'li') {
                // good
                $current_li = $node;
                $result[] = $node;
            } else {
                // we want to tuck this into the previous li
                // Invariant: we expect the node to be ol/ul
                // ToDo: Make this more robust in the case of not ol/ul
                // by distinguishing between existing li and li created
                // to handle non-list elements; non-list elements should
                // not be appended to an existing li; only li created
                // for non-list. This distinction is not currently made.
                if ($current_li === null) {
                    $current_li = new HTMLPurifier_Node_Element('li');
                    $result[] = $current_li;
                }
                $current_li->children[] = $node;
                $current_li->empty = false; // XXX fascinating! Check for this error elsewhere ToDo
            }
        }
        if (empty($result)) {
            return false;
        }
        if ($all_whitespace) {
            return false;
        }
        return $result;
    }
}

// vim: et sw=4 sts=4

<?php

/**
 * Definition that allows a set of elements, and allows no children.
 * @note This is a hack to reuse code from HTMLPurifier_ChildDef_Required,
 *       really, one shouldn't inherit from the other.  Only altered behavior
 *       is to overload a returned false with an array.  Thus, it will never
 *       return false.
 */
class HTMLPurifier_ChildDef_Optional extends HTMLPurifier_ChildDef_Required
{
    /**
     * @type bool
     */
    public $allow_empty = true;

    /**
     * @type string
     */
    public $type = 'optional';

    /**
     * @param array $children
     * @param HTMLPurifier_Config $config
     * @param HTMLPurifier_Context $context
     * @return array
     */
    public function validateChildren($children, $config, $context)
    {
        $result = parent::validateChildren($children, $config, $context);
        // we assume that $children is not modified
        if ($result === false) {
            if (empty($children)) {
                return true;
            } elseif ($this->whitespace) {
                return $children;
            } else {
                return array();
            }
        }
        return $result;
    }
}

// vim: et sw=4 sts=4

<?php

/**
 * Definition that allows a set of elements, but disallows empty children.
 */
class HTMLPurifier_ChildDef_Required extends HTMLPurifier_ChildDef
{
    /**
     * Lookup table of allowed elements.
     * @type array
     */
    public $elements = array();

    /**
     * Whether or not the last passed node was all whitespace.
     * @type bool
     */
    protected $whitespace = false;

    /**
     * @param array|string $elements List of allowed element names (lowercase).
     */
    public function __construct($elements)
    {
        if (is_string($elements)) {
            $elements = str_replace(' ', '', $elements);
            $elements = explode('|', $elements);
        }
        $keys = array_keys($elements);
        if ($keys == array_keys($keys)) {
            $elements = array_flip($elements);
            foreach ($elements as $i => $x) {
                $elements[$i] = true;
                if (empty($i)) {
                    unset($elements[$i]);
                } // remove blank
            }
        }
        $this->elements = $elements;
    }

    /**
     * @type bool
     */
    public $allow_empty = false;

    /**
     * @type string
     */
    public $type = 'required';

    /**
     * @param array $children
     * @param HTMLPurifier_Config $config
     * @param HTMLPurifier_Context $context
     * @return array
     */
    public function validateChildren($children, $config, $context)
    {
        // Flag for subclasses
        $this->whitespace = false;

        // if there are no tokens, delete parent node
        if (empty($children)) {
            return false;
        }

        // the new set of children
        $result = array();

        // whether or not parsed character data is allowed
        // this controls whether or not we silently drop a tag
        // or generate escaped HTML from it
        $pcdata_allowed = isset($this->elements['#PCDATA']);

        // a little sanity check to make sure it's not ALL whitespace
        $all_whitespace = true;

        $stack = array_reverse($children);
        while (!empty($stack)) {
            $node = array_pop($stack);
            if (!empty($node->is_whitespace)) {
                $result[] = $node;
                continue;
            }
            $all_whitespace = false; // phew, we're not talking about whitespace

            if (!isset($this->elements[$node->name])) {
                // special case text
                // XXX One of these ought to be redundant or something
                if ($pcdata_allowed && $node instanceof HTMLPurifier_Node_Text) {
                    $result[] = $node;
                    continue;
                }
                // spill the child contents in
                // ToDo: Make configurable
                if ($node instanceof HTMLPurifier_Node_Element) {
                    for ($i = count($node->children) - 1; $i >= 0; $i--) {
                        $stack[] = $node->children[$i];
                    }
                    continue;
                }
                continue;
            }
            $result[] = $node;
        }
        if (empty($result)) {
            return false;
        }
        if ($all_whitespace) {
            $this->whitespace = true;
            return false;
        }
        return $result;
    }
}

// vim: et sw=4 sts=4

<?php

/**
 * Takes the contents of blockquote when in strict and reformats for validation.
 */
class HTMLPurifier_ChildDef_StrictBlockquote extends HTMLPurifier_ChildDef_Required
{
    /**
     * @type array
     */
    protected $real_elements;

    /**
     * @type array
     */
    protected $fake_elements;

    /**
     * @type bool
     */
    public $allow_empty = true;

    /**
     * @type string
     */
    public $type = 'strictblockquote';

    /**
     * @type bool
     */
    protected $init = false;

    /**
     * @param HTMLPurifier_Config $config
     * @return array
     * @note We don't want MakeWellFormed to auto-close inline elements since
     *       they might be allowed.
     */
    public function getAllowedElements($config)
    {
        $this->init($config);
        return $this->fake_elements;
    }

    /**
     * @param array $children
     * @param HTMLPurifier_Config $config
     * @param HTMLPurifier_Context $context
     * @return array
     */
    public function validateChildren($children, $config, $context)
    {
        $this->init($config);

        // trick the parent class into thinking it allows more
        $this->elements = $this->fake_elements;
        $result = parent::validateChildren($children, $config, $context);
        $this->elements = $this->real_elements;

        if ($result === false) {
            return array();
        }
        if ($result === true) {
            $result = $children;
        }

        $def = $config->getHTMLDefinition();
        $block_wrap_name = $def->info_block_wrapper;
        $block_wrap = false;
        $ret = array();

        foreach ($result as $node) {
            if ($block_wrap === false) {
                if (($node instanceof HTMLPurifier_Node_Text && !$node->is_whitespace) ||
                    ($node instanceof HTMLPurifier_Node_Element && !isset($this->elements[$node->name]))) {
                        $block_wrap = new HTMLPurifier_Node_Element($def->info_block_wrapper);
                        $ret[] = $block_wrap;
                }
            } else {
                if ($node instanceof HTMLPurifier_Node_Element && isset($this->elements[$node->name])) {
                    $block_wrap = false;

                }
            }
            if ($block_wrap) {
                $block_wrap->children[] = $node;
            } else {
                $ret[] = $node;
            }
        }
        return $ret;
    }

    /**
     * @param HTMLPurifier_Config $config
     */
    private function init($config)
    {
        if (!$this->init) {
            $def = $config->getHTMLDefinition();
            // allow all inline elements
            $this->real_elements = $this->elements;
            $this->fake_elements = $def->info_content_sets['Flow'];
            $this->fake_elements['#PCDATA'] = true;
            $this->init = true;
        }
    }
}

// vim: et sw=4 sts=4

<?php

/**
 * Definition for tables.  The general idea is to extract out all of the
 * essential bits, and then reconstruct it later.
 *
 * This is a bit confusing, because the DTDs and the W3C
 * validators seem to disagree on the appropriate definition. The
 * DTD claims:
 *
 *      (CAPTION?, (COL*|COLGROUP*), THEAD?, TFOOT?, TBODY+)
 *
 * But actually, the HTML4 spec then has this to say:
 *
 *      The TBODY start tag is always required except when the table
 *      contains only one table body and no table head or foot sections.
 *      The TBODY end tag may always be safely omitted.
 *
 * So the DTD is kind of wrong.  The validator is, unfortunately, kind
 * of on crack.
 *
 * The definition changed again in XHTML1.1; and in my opinion, this
 * formulation makes the most sense.
 *
 *      caption?, ( col* | colgroup* ), (( thead?, tfoot?, tbody+ ) | ( tr+ ))
 *
 * Essentially, we have two modes: thead/tfoot/tbody mode, and tr mode.
 * If we encounter a thead, tfoot or tbody, we are placed in the former
 * mode, and we *must* wrap any stray tr segments with a tbody. But if
 * we don't run into any of them, just have tr tags is OK.
 */
class HTMLPurifier_ChildDef_Table extends HTMLPurifier_ChildDef
{
    /**
     * @type bool
     */
    public $allow_empty = false;

    /**
     * @type string
     */
    public $type = 'table';

    /**
     * @type array
     */
    public $elements = array(
        'tr' => true,
        'tbody' => true,
        'thead' => true,
        'tfoot' => true,
        'caption' => true,
        'colgroup' => true,
        'col' => true
    );

    public function __construct()
    {
    }

    /**
     * @param array $children
     * @param HTMLPurifier_Config $config
     * @param HTMLPurifier_Context $context
     * @return array
     */
    public function validateChildren($children, $config, $context)
    {
        if (empty($children)) {
            return false;
        }

        // only one of these elements is allowed in a table
        $caption = false;
        $thead = false;
        $tfoot = false;

        // whitespace
        $initial_ws = array();
        $after_caption_ws = array();
        $after_thead_ws = array();
        $after_tfoot_ws = array();

        // as many of these as you want
        $cols = array();
        $content = array();

        $tbody_mode = false; // if true, then we need to wrap any stray
                             // <tr>s with a <tbody>.

        $ws_accum =& $initial_ws;

        foreach ($children as $node) {
            if ($node instanceof HTMLPurifier_Node_Comment) {
                $ws_accum[] = $node;
                continue;
            }
            switch ($node->name) {
            case 'tbody':
                $tbody_mode = true;
                // fall through
            case 'tr':
                $content[] = $node;
                $ws_accum =& $content;
                break;
            case 'caption':
                // there can only be one caption!
                if ($caption !== false)  break;
                $caption = $node;
                $ws_accum =& $after_caption_ws;
                break;
            case 'thead':
                $tbody_mode = true;
                // XXX This breaks rendering properties with
                // Firefox, which never floats a <thead> to
                // the top. Ever. (Our scheme will float the
                // first <thead> to the top.)  So maybe
                // <thead>s that are not first should be
                // turned into <tbody>? Very tricky, indeed.
                if ($thead === false) {
                    $thead = $node;
                    $ws_accum =& $after_thead_ws;
                } else {
                    // Oops, there's a second one! What
                    // should we do?  Current behavior is to
                    // transmutate the first and last entries into
                    // tbody tags, and then put into content.
                    // Maybe a better idea is to *attach
                    // it* to the existing thead or tfoot?
                    // We don't do this, because Firefox
                    // doesn't float an extra tfoot to the
                    // bottom like it does for the first one.
                    $node->name = 'tbody';
                    $content[] = $node;
                    $ws_accum =& $content;
                }
                break;
            case 'tfoot':
                // see above for some aveats
                $tbody_mode = true;
                if ($tfoot === false) {
                    $tfoot = $node;
                    $ws_accum =& $after_tfoot_ws;
                } else {
                    $node->name = 'tbody';
                    $content[] = $node;
                    $ws_accum =& $content;
                }
                break;
            case 'colgroup':
            case 'col':
                $cols[] = $node;
                $ws_accum =& $cols;
                break;
            case '#PCDATA':
                // How is whitespace handled? We treat is as sticky to
                // the *end* of the previous element. So all of the
                // nonsense we have worked on is to keep things
                // together.
                if (!empty($node->is_whitespace)) {
                    $ws_accum[] = $node;
                }
                break;
            }
        }

        if (empty($content) && $thead === false && $tfoot === false) {
            return false;
        }

        $ret = $initial_ws;
        if ($caption !== false) {
            $ret[] = $caption;
            $ret = array_merge($ret, $after_caption_ws);
        }
        if ($cols !== false) {
            $ret = array_merge($ret, $cols);
        }
        if ($thead !== false) {
            $ret[] = $thead;
            $ret = array_merge($ret, $after_thead_ws);
        }
        if ($tfoot !== false) {
            $ret[] = $tfoot;
            $ret = array_merge($ret, $after_tfoot_ws);
        }

        if ($tbody_mode) {
            // we have to shuffle tr into tbody
            $current_tr_tbody = null;

            foreach($content as $node) {
                switch ($node->name) {
                case 'tbody':
                    $current_tr_tbody = null;
                    $ret[] = $node;
                    break;
                case 'tr':
                    if ($current_tr_tbody === null) {
                        $current_tr_tbody = new HTMLPurifier_Node_Element('tbody');
                        $ret[] = $current_tr_tbody;
                    }
                    $current_tr_tbody->children[] = $node;
                    break;
                case '#PCDATA':
                    //assert($node->is_whitespace);
                    if ($current_tr_tbody === null) {
                        $ret[] = $node;
                    } else {
                        $current_tr_tbody->children[] = $node;
                    }
                    break;
                }
            }
        } else {
            $ret = array_merge($ret, $content);
        }

        return $ret;

    }
}

// vim: et sw=4 sts=4