Version 1
[yaffs-website] / vendor / ezyang / htmlpurifier / library / HTMLPurifier / Filter / ExtractStyleBlocks.php
diff --git a/vendor/ezyang/htmlpurifier/library/HTMLPurifier/Filter/ExtractStyleBlocks.php b/vendor/ezyang/htmlpurifier/library/HTMLPurifier/Filter/ExtractStyleBlocks.php
new file mode 100644 (file)
index 0000000..66f70b0
--- /dev/null
@@ -0,0 +1,341 @@
+<?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