<?php
// why is this a top level function? Because PHP 5.2.0 doesn't seem to
// understand how to interpret this filter if it's a static method.
// It's all really silly, but if we go this route it might be reasonable
// to coalesce all of these methods into one.
function htmlpurifier_filter_extractstyleblocks_muteerrorhandler()
{
}
/**
* This filter extracts <style> blocks from input HTML, cleans them up
* using CSSTidy, and then places them in $purifier->context->get('StyleBlocks')
* so they can be used elsewhere in the document.
*
* @note
* See tests/HTMLPurifier/Filter/ExtractStyleBlocksTest.php for
* sample usage.
*
* @note
* This filter can also be used on stylesheets not included in the
* document--something purists would probably prefer. Just directly
* call HTMLPurifier_Filter_ExtractStyleBlocks->cleanCSS()
*/
class HTMLPurifier_Filter_ExtractStyleBlocks extends HTMLPurifier_Filter
{
/**
* @type string
*/
public $name = 'ExtractStyleBlocks';
/**
* @type array
*/
private $_styleMatches = array();
/**
* @type csstidy
*/
private $_tidy;
/**
* @type HTMLPurifier_AttrDef_HTML_ID
*/
private $_id_attrdef;
/**
* @type HTMLPurifier_AttrDef_CSS_Ident
*/
private $_class_attrdef;
/**
* @type HTMLPurifier_AttrDef_Enum
*/
private $_enum_attrdef;
public function __construct()
{
$this->_tidy = new csstidy();
$this->_tidy->set_cfg('lowercase_s', false);
$this->_id_attrdef = new HTMLPurifier_AttrDef_HTML_ID(true);
$this->_class_attrdef = new HTMLPurifier_AttrDef_CSS_Ident();
$this->_enum_attrdef = new HTMLPurifier_AttrDef_Enum(
array(
'first-child',
'link',
'visited',
'active',
'hover',
'focus'
)
);
}
/**
* Save the contents of CSS blocks to style matches
* @param array $matches preg_replace style $matches array
*/
protected function styleCallback($matches)
{
$this->_styleMatches[] = $matches[1];
}
/**
* Removes inline <style> tags from HTML, saves them for later use
* @param string $html
* @param HTMLPurifier_Config $config
* @param HTMLPurifier_Context $context
* @return string
* @todo Extend to indicate non-text/css style blocks
*/
public function preFilter($html, $config, $context)
{
$tidy = $config->get('Filter.ExtractStyleBlocks.TidyImpl');
if ($tidy !== null) {
$this->_tidy = $tidy;
}
// NB: this must be NON-greedy because if we have
// <style>foo</style> <style>bar</style>
// we must not grab foo</style> <style>bar
$html = preg_replace_callback('#<style(?:\s.*)?>(.*)<\/style>#isU', array($this, 'styleCallback'), $html);
$style_blocks = $this->_styleMatches;
$this->_styleMatches = array(); // reset
$context->register('StyleBlocks', $style_blocks); // $context must not be reused
if ($this->_tidy) {
foreach ($style_blocks as &$style) {
$style = $this->cleanCSS($style, $config, $context);
}
}
return $html;
}
/**
* Takes CSS (the stuff found in <style>) and cleans it.
* @warning Requires CSSTidy <http://csstidy.sourceforge.net/>
* @param string $css CSS styling to clean
* @param HTMLPurifier_Config $config
* @param HTMLPurifier_Context $context
* @throws HTMLPurifier_Exception
* @return string Cleaned CSS
*/
public function cleanCSS($css, $config, $context)
{
// prepare scope
$scope = $config->get('Filter.ExtractStyleBlocks.Scope');
if ($scope !== null) {
$scopes = array_map('trim', explode(',', $scope));
} else {
$scopes = array();
}
// remove comments from CSS
$css = trim($css);
if (strncmp('<!--', $css, 4) === 0) {
$css = substr($css, 4);
}
if (strlen($css) > 3 && substr($css, -3) == '-->') {
$css = substr($css, 0, -3);
}
$css = trim($css);
set_error_handler('htmlpurifier_filter_extractstyleblocks_muteerrorhandler');
$this->_tidy->parse($css);
restore_error_handler();
$css_definition = $config->getDefinition('CSS');
$html_definition = $config->getDefinition('HTML');
$new_css = array();
foreach ($this->_tidy->css as $k => $decls) {
// $decls are all CSS declarations inside an @ selector
$new_decls = array();
foreach ($decls as $selector => $style) {
$selector = trim($selector);
if ($selector === '') {
continue;
} // should not happen
// Parse the selector
// Here is the relevant part of the CSS grammar:
//
// ruleset
// : selector [ ',' S* selector ]* '{' ...
// selector
// : simple_selector [ combinator selector | S+ [ combinator? selector ]? ]?
// combinator
// : '+' S*
// : '>' S*
// simple_selector
// : element_name [ HASH | class | attrib | pseudo ]*
// | [ HASH | class | attrib | pseudo ]+
// element_name
// : IDENT | '*'
// ;
// class
// : '.' IDENT
// ;
// attrib
// : '[' S* IDENT S* [ [ '=' | INCLUDES | DASHMATCH ] S*
// [ IDENT | STRING ] S* ]? ']'
// ;
// pseudo
// : ':' [ IDENT | FUNCTION S* [IDENT S*]? ')' ]
// ;
//
// For reference, here are the relevant tokens:
//
// HASH #{name}
// IDENT {ident}
// INCLUDES ==
// DASHMATCH |=
// STRING {string}
// FUNCTION {ident}\(
//
// And the lexical scanner tokens
//
// name {nmchar}+
// nmchar [_a-z0-9-]|{nonascii}|{escape}
// nonascii [\240-\377]
// escape {unicode}|\\[^\r\n\f0-9a-f]
// unicode \\{h}}{1,6}(\r\n|[ \t\r\n\f])?
// ident -?{nmstart}{nmchar*}
// nmstart [_a-z]|{nonascii}|{escape}
// string {string1}|{string2}
// string1 \"([^\n\r\f\\"]|\\{nl}|{escape})*\"
// string2 \'([^\n\r\f\\"]|\\{nl}|{escape})*\'
//
// We'll implement a subset (in order to reduce attack
// surface); in particular:
//
// - No Unicode support
// - No escapes support
// - No string support (by proxy no attrib support)
// - element_name is matched against allowed
// elements (some people might find this
// annoying...)
// - Pseudo-elements one of :first-child, :link,
// :visited, :active, :hover, :focus
// handle ruleset
$selectors = array_map('trim', explode(',', $selector));
$new_selectors = array();
foreach ($selectors as $sel) {
// split on +, > and spaces
$basic_selectors = preg_split('/\s*([+> ])\s*/', $sel, -1, PREG_SPLIT_DELIM_CAPTURE);
// even indices are chunks, odd indices are
// delimiters
$nsel = null;
$delim = null; // guaranteed to be non-null after
// two loop iterations
for ($i = 0, $c = count($basic_selectors); $i < $c; $i++) {
$x = $basic_selectors[$i];
if ($i % 2) {
// delimiter
if ($x === ' ') {
$delim = ' ';
} else {
$delim = ' ' . $x . ' ';
}
} else {
// simple selector
$components = preg_split('/([#.:])/', $x, -1, PREG_SPLIT_DELIM_CAPTURE);
$sdelim = null;
$nx = null;
for ($j = 0, $cc = count($components); $j < $cc; $j++) {
$y = $components[$j];
if ($j === 0) {
if ($y === '*' || isset($html_definition->info[$y = strtolower($y)])) {
$nx = $y;
} else {
// $nx stays null; this matters
// if we don't manage to find
// any valid selector content,
// in which case we ignore the
// outer $delim
}
} elseif ($j % 2) {
// set delimiter
$sdelim = $y;
} else {
$attrdef = null;
if ($sdelim === '#') {
$attrdef = $this->_id_attrdef;
} elseif ($sdelim === '.') {
$attrdef = $this->_class_attrdef;
} elseif ($sdelim === ':') {
$attrdef = $this->_enum_attrdef;
} else {
throw new HTMLPurifier_Exception('broken invariant sdelim and preg_split');
}
$r = $attrdef->validate($y, $config, $context);
if ($r !== false) {
if ($r !== true) {
$y = $r;
}
if ($nx === null) {
$nx = '';
}
$nx .= $sdelim . $y;
}
}
}
if ($nx !== null) {
if ($nsel === null) {
$nsel = $nx;
} else {
$nsel .= $delim . $nx;
}
} else {
// delimiters to the left of invalid
// basic selector ignored
}
}
}
if ($nsel !== null) {
if (!empty($scopes)) {
foreach ($scopes as $s) {
$new_selectors[] = "$s $nsel";
}
} else {
$new_selectors[] = $nsel;
}
}
}
if (empty($new_selectors)) {
continue;
}
$selector = implode(', ', $new_selectors);
foreach ($style as $name => $value) {
if (!isset($css_definition->info[$name])) {
unset($style[$name]);
continue;
}
$def = $css_definition->info[$name];
$ret = $def->validate($value, $config, $context);
if ($ret === false) {
unset($style[$name]);
} else {
$style[$name] = $ret;
}
}
$new_decls[$selector] = $style;
}
$new_css[$k] = $new_decls;
}
// remove stuff that shouldn't be used, could be reenabled
// after security risks are analyzed
$this->_tidy->css = $new_css;
$this->_tidy->import = array();
$this->_tidy->charset = null;
$this->_tidy->namespace = null;
$css = $this->_tidy->print->plain();
// we are going to escape any special characters <>& to ensure
// that no funny business occurs (i.e. </style> in a font-family prop).
if ($config->get('Filter.ExtractStyleBlocks.Escaping')) {
$css = str_replace(
array('<', '>', '&'),
array('\3C ', '\3E ', '\26 '),
$css
);
}
return $css;
}
}
// vim: et sw=4 sts=4