[3.0.0] Fully implement CSS extraction and cleaning. See NEWS for more information, it is now a Filter.

- Some Lexer things were moved around

git-svn-id: http://htmlpurifier.org/svnroot/htmlpurifier/trunk@1469 48356398-32a2-884e-a903-53898d9a118a

NEWS

@@ -18,8 +18,10 @@ NEWS ( CHANGELOG and HISTORY )                                     HTMLPurifier
   + PHP4 reference/foreach cruft removed (in progress)
 ! CSS properties are no case-insensitive
 ! DefinitionCacheFactory now can register new implementations
-! <style> tags can now be extracted from input HTML using %HTML.ExtractStyleBlocks.
-  These contents can be retrieved from $context->get('StyleBlocks');
+! New HTMLPurifier_Filter_ExtractStyleBlocks for extracting <style> from
+  documents and cleaning their contents up. Requires the CSSTidy library
+  <http://csstidy.sourceforge.net/>. You can access the blocks with the
+  'StyleBlocks' Context variable ($purifier->context->get('StyleBlocks'))
 . Unit tests for Injector improved
 
 2.1.3, released 2007-11-05

library/HTMLPurifier/Filter/ExtractStyleBlocks.php

@@ -0,0 +1,104 @@
+<?php
+
+require_once 'HTMLPurifier/Filter.php';
+
+/**
+ * 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
+ * @todo Allow for selectors to be munged/checked
+ * @todo Expose CSSTidy configuration so that custom changes can be made
+ */
+class HTMLPurifier_Filter_ExtractStyleBlocks extends HTMLPurifier_Filter
+{
+    
+    public $name = 'ExtractStyleBlocks';
+    private $_styleMatches = array();
+    private $_tidy, $_disableCharacterEscaping;
+    
+    /**
+     * @param $tidy Instance of csstidy to use, false to turn off cleaning,
+     *              and null to automatically instantiate
+     * @param $disable_character_escaping Whether or not to stop munging
+     *              <, > and &. This can be set to true if the CSS will
+     *              be placed in an external style and not inline.
+     */
+    public function __construct($tidy = null, $disable_character_escaping = false) {
+        if ($tidy === null) $tidy = new csstidy();
+        $this->_tidy = $tidy;
+        $this->_disableCharacterEscaping = $disable_character_escaping;
+    }
+    
+    /**
+     * Save the contents of CSS blocks to style matches
+     * @param $matches preg_replace style $matches array
+     */
+    protected function styleCallback($matches) {
+        $this->_styleMatches[] = $matches[1];
+    }
+    
+    /**
+     * Removes inline <style> tags from HTML, saves them for later use
+     * @todo Extend to indicate non-text/css style blocks
+     */
+    public function preFilter($html, $config, &$context) {
+        $html = preg_replace_callback('#<style(?:\s.*)?>(.+)</style>#isU', array($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 $css CSS styling to clean
+     * @param $config Instance of HTMLPurifier_Config
+     * @param $context Instance of HTMLPurifier_Context
+     * @return Cleaned CSS
+     */
+    public function cleanCSS($css, $config, &$context) {
+        $this->_tidy->parse($css);
+        $css_definition = $config->getDefinition('CSS');
+        foreach ($this->_tidy->css as &$decls) {
+            // $decls are all CSS declarations inside an @ selector
+            foreach ($decls as &$style) {
+                foreach ($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;
+                }
+            }
+        }
+        // remove stuff that shouldn't be used, could be reenabled
+        // after security risks are analyzed
+        $this->_tidy->import = array();
+        $this->_tidy->charset = null;
+        $this->_tidy->namespace = null;
+        $printer = new csstidy_print($this->_tidy);
+        $css = $printer->plain();
+        // we are going to escape any special characters <>& to ensure
+        // that no funny business occurs (i.e. </style> in a font-family prop).
+        if (!$this->_disableCharacterEscaping) {
+            $css = str_replace(
+                array('<',    '>',    '&'),
+                array('\3C ', '\3E ', '\26 '),
+                $css
+            );
+        }
+        return $css;
+    }
+    
+}
+

library/HTMLPurifier/Filter/YouTube.php

@@ -5,7 +5,7 @@ require_once 'HTMLPurifier/Filter.php';
 class HTMLPurifier_Filter_YouTube extends HTMLPurifier_Filter
 {
     
-    public $name = 'YouTube preservation';
+    public $name = 'YouTube';
     
     public function preFilter($html, $config, &$context) {
         $pre_regex = '#<object[^>]+>.+?'.

library/HTMLPurifier/Lexer.php

@@ -79,14 +79,6 @@ It is not necessary and will have no effect for PHP 4.
 This directive has been available since 2.1.0.
 ');
 
-HTMLPurifier_ConfigSchema::define(
-    'HTML', 'ExtractStyleBlocks', false, 'bool', '
-This directive enables extraction of <code>style</code> tags contents so
-that they can be incorporated in the <code>head</code> of the document,
-after sufficient validation.
-This directive has been available since 3.0.0.
-');
-
 /**
  * Forgivingly lexes HTML (SGML-style) markup into tokens.
  * 
@@ -346,34 +338,6 @@ class HTMLPurifier_Lexer
         // represent non-SGML characters (horror, horror!)
         $html = HTMLPurifier_Encoder::cleanUTF8($html);
         
-        if ($config->get('HTML', 'ExtractStyleBlocks')) {
-            // extract <style> CSS blocks
-            $html = $this->extractStyleBlocks($html, $config, $context);
-        }
-        
-        return $html;
-    }
-    
-    private $_styleMatches = array();
-    
-    /**
-     * Save the contents of CSS blocks to style matches
-     */
-    protected function styleCallback($matches) {
-        $this->_styleMatches[] = $matches[1];
-    }
-    
-    /**
-     * Removes inline <style> tags from HTML, saves them for later use
-     * @todo Extend to indicate non-text/css style blocks
-     */
-    public function extractStyleBlocks($html, $config, $context) {
-        $html = preg_replace_callback('#<style(?:\s.*)?>(.+)</style>#isU', array($this, 'styleCallback'), $html);
-        $style_blocks = $this->_styleMatches;
-        $this->_styleMatches = array(); // reset
-        // this is a persistent context, so we have to overwrite it with every call
-        if ($context->exists('StyleBlocks')) $context->destroy('StyleBlocks');
-        $context->register('StyleBlocks', $style_blocks);
         return $html;
     }
     

test-settings.sample.php

@@ -14,6 +14,9 @@ set_time_limit(0);
 // Where is SimpleTest located?
 $simpletest_location = '/path/to/simpletest/';
 
+// Where is CSSTidy located?
+$csstidy_location    = '/path/to/csstidy/';
+
 // How many times should profiling scripts iterate over the function? More runs 
 // means more accurate results, but they'll take longer to perform.
 $GLOBALS['HTMLPurifierTest']['Runs'] = 2;

tests/HTMLPurifier/Filter/ExtractStyleBlocksTest.php

@@ -0,0 +1,112 @@
+<?php
+
+require_once 'HTMLPurifier/Filter/ExtractStyleBlocks.php';
+
+/**
+ * @todo Assimilate CSSTidy into our library
+ */
+class HTMLPurifier_Filter_ExtractStyleBlocksTest extends HTMLPurifier_Harness
+{
+    
+    // usual use case:
+    function test_tokenizeHTML_extractStyleBlocks() {
+        $purifier = new HTMLPurifier($this->config);
+        $purifier->addFilter(new HTMLPurifier_Filter_ExtractStyleBlocks());
+        $result = $purifier->purify('<style type="text/css">.foo {text-align:center;bogus:remove-me;}</style>Test<style>* {font-size:12pt;}</style>');
+        $this->assertIdentical($result, 'Test');
+        $this->assertIdentical($purifier->context->get('StyleBlocks'),
+            array(
+                ".foo {\ntext-align:center;\n}",
+                "* {\nfont-size:12pt;\n}"
+            )
+        );
+    }
+    
+    function assertExtractStyleBlocks($html, $expect = true, $styles = array()) {
+        $filter = new HTMLPurifier_Filter_ExtractStyleBlocks(false); // disable cleaning
+        if ($expect === true) $expect = $html;
+        $result = $filter->preFilter($html, $this->config, $this->context);
+        $this->assertIdentical($result, $expect);
+        $this->assertIdentical($this->context->get('StyleBlocks'), $styles);
+    }
+    
+    function test_extractStyleBlocks_preserve() {
+        $this->assertExtractStyleBlocks('Foobar');
+    }
+    
+    function test_extractStyleBlocks_allStyle() {
+        $this->assertExtractStyleBlocks('<style>foo</style>', '', array('foo'));
+    }
+    
+    function test_extractStyleBlocks_multipleBlocks() {
+        $this->assertExtractStyleBlocks(
+          "<style>1</style><style>2</style>NOP<style>4</style>",
+          "NOP",
+          array('1', '2', '4')
+        );
+    }
+    
+    function test_extractStyleBlocks_blockWithAttributes() {
+        $this->assertExtractStyleBlocks(
+          '<style type="text/css">css</style>',
+          '',
+          array('css')
+        );
+    }
+    
+    function test_extractStyleBlocks_styleWithPadding() {
+        $this->assertExtractStyleBlocks(
+          "Alas<styled>Awesome</styled>\n<style>foo</style> Trendy!",
+          "Alas<styled>Awesome</styled>\n Trendy!",
+          array('foo')
+        );
+    }
+    
+    function assertCleanCSS($input, $expect = true) {
+        $filter = new HTMLPurifier_Filter_ExtractStyleBlocks();
+        if ($expect === true) $expect = $input;
+        $result = $filter->cleanCSS($input, $this->config, $this->context);
+        $this->assertIdentical($result, $expect);
+    }
+    
+    function test_cleanCSS_malformed() {
+        $this->assertCleanCSS('</style>', '');
+    }
+    
+    function test_cleanCSS_selector() {
+        $this->assertCleanCSS("a .foo #id div.cl#foo {\nfont-weight:700;\n}");
+    }
+    
+    function test_cleanCSS_angledBrackets() {
+        $this->assertCleanCSS(
+            ".class {\nfont-family:'</style>';\n}",
+            ".class {\nfont-family:'\\3C /style\\3E ';\n}"
+        );
+    }
+    
+    function test_cleanCSS_angledBrackets2() {
+        // CSSTidy's behavior in this case is wrong, and should be fixed
+        //$this->assertCleanCSS(
+        //    "span[title=\"</style>\"] {\nfont-size:12pt;\n}",
+        //    "span[title=\"\\3C /style\\3E \"] {\nfont-size:12pt;\n}"
+        //);
+    }
+    
+    function test_cleanCSS_bogus() {
+        $this->assertCleanCSS("div {bogus:tree;}", "div {\n}");
+    }
+    
+    function test_cleanCSS_escapeCodes() {
+        $this->assertCleanCSS(
+            ".class {\nfont-family:'\\3C /style\\3E ';\n}"
+        );
+    }
+    
+    function test_cleanCSS_noEscapeCodes() {
+        $filter = new HTMLPurifier_Filter_ExtractStyleBlocks(null, true);
+        $input = ".class {\nfont-family:'</style>';\n}";
+        $result = $filter->cleanCSS($input, $this->config, $this->context);
+        $this->assertIdentical($result, $input);
+    }
+    
+}

tests/HTMLPurifier/LexerTest.php

@@ -31,48 +31,6 @@ class HTMLPurifier_LexerTest extends HTMLPurifier_Harness
         $this->assertIsA($lexer, 'HTMLPurifier_Lexer_DirectLex');
     }
     
-    // HTMLPurifier_Lexer->extractStyleBlocks() --------------------------------
-    
-    function assertExtractStyleBlocks($html, $expect = true, $styles = array()) {
-        $lexer = HTMLPurifier_Lexer::create($this->config);
-        if ($expect === true) $expect = $html;
-        $result = $lexer->extractStyleBlocks($html, $this->config, $this->context);
-        $this->assertIdentical($result, $expect);
-        $this->assertIdentical($this->context->get('StyleBlocks'), $styles);
-    }
-    
-    function test_extractStyleBlocks_preserve() {
-        $this->assertExtractStyleBlocks('Foobar');
-    }
-    
-    function test_extractStyleBlocks_allStyle() {
-        $this->assertExtractStyleBlocks('<style>foo</style>', '', array('foo'));
-    }
-    
-    function test_extractStyleBlocks_multipleBlocks() {
-        $this->assertExtractStyleBlocks(
-          "<style>1</style><style>2</style>NOP<style>4</style>",
-          "NOP",
-          array('1', '2', '4')
-        );
-    }
-    
-    function test_extractStyleBlocks_blockWithAttributes() {
-        $this->assertExtractStyleBlocks(
-          '<style type="text/css">css</style>',
-          '',
-          array('css')
-        );
-    }
-    
-    function test_extractStyleBlocks_styleWithPadding() {
-        $this->assertExtractStyleBlocks(
-          "Alas<styled>Awesome</styled>\n<style>foo</style> Trendy!",
-          "Alas<styled>Awesome</styled>\n Trendy!",
-          array('foo')
-        );
-    }
-    
     // HTMLPurifier_Lexer->parseData() -----------------------------------------
     
     function assertParseData($input, $expect = true) {
@@ -553,17 +511,6 @@ class HTMLPurifier_LexerTest extends HTMLPurifier_Harness
         );
     }
     
-    function test_tokenizeHTML_extractStyleBlocks() {
-        $this->config->set('HTML', 'ExtractStyleBlocks', true);
-        $this->assertTokenization(
-            '<style type="text/css">.foo {text-align:center;}</style>Test',
-            array(
-                new HTMLPurifier_Token_Text('Test')
-            )
-        );
-        $this->assertIdentical($this->context->get('StyleBlocks'), array('.foo {text-align:center;}'));
-    }
-    
     /*
     
     function test_tokenizeHTML_() {

tests/index.php

@@ -16,6 +16,7 @@ $GLOBALS['HTMLPurifierTest'] = array();
 $GLOBALS['HTMLPurifierTest']['PEAR'] = false; // do PEAR tests
 $GLOBALS['HTMLPurifierTest']['PH5P'] = version_compare(PHP_VERSION, "5", ">=") && class_exists('DOMDocument');
 $simpletest_location = 'simpletest/'; // reasonable guess
+$csstidy_location = false;
 
 // load SimpleTest
 if (file_exists('../conf/test-settings.php')) include '../conf/test-settings.php';
@@ -24,6 +25,11 @@ require_once $simpletest_location . 'unit_tester.php';
 require_once $simpletest_location . 'reporter.php';
 require_once $simpletest_location . 'mock_objects.php';
 
+if ($csstidy_location !== false) {
+    require_once $csstidy_location . 'class.csstidy.php';
+    require_once $csstidy_location . 'class.csstidy_print.php';
+}
+
 error_reporting(E_ALL | E_STRICT); // after SimpleTest is loaded, turn on compile time errors
 
 // load Debugger

tests/test_files.php

@@ -75,6 +75,7 @@ $test_files[] = 'HTMLPurifier/ErrorCollectorTest.php';
 $test_files[] = 'HTMLPurifier/EncoderTest.php';
 $test_files[] = 'HTMLPurifier/EntityLookupTest.php';
 $test_files[] = 'HTMLPurifier/EntityParserTest.php';
+$test_files[] = 'HTMLPurifier/Filter/ExtractStyleBlocksTest.php';
 $test_files[] = 'HTMLPurifier/GeneratorTest.php';
 $test_files[] = 'HTMLPurifier/HTMLDefinitionTest.php';
 $test_files[] = 'HTMLPurifier/HTMLModuleManagerTest.php';