ENHANCEMENT Allowing custom HTMLCleaner implementations via Diff:: and new TidyHTMLCleaner and PurifierHTMLCleaner classes

core/Diff.php

@@ -677,8 +677,11 @@ class Diff
 		if (!$cleaner) {
 			if (class_exists(self::$html_cleaner_class)) {
 				$cleaner = new self::$html_cleaner_class;
+			} else {
+				$cleaner = HTMLCleaner::inst();    //load cleaner if the dependent class is available
 			}
 		}
+
 		if ($cleaner) {
 			$content = $cleaner->cleanHTML($content);
 		} else {

core/HTMLCleaner.php

@@ -0,0 +1,98 @@
+<?php
+/**
+ * @package sapphire
+ * @subpackage misc
+ */
+
+/**
+ * Base class for HTML cleaning implementations.
+ */
+abstract class HTMLCleaner extends Object {
+
+	/**
+	 * @var array
+	 */
+	protected $defaultConfig = array();
+	
+	/**
+	 * @var $config Array configuration variables for HTMLCleaners that support configuration (like Tidy)
+	 */
+	public $config;
+
+	/**
+	 * @param Array The configuration for the cleaner, if necessary
+	 */
+	public function __construct($config = null) {
+		if ($config) $this->config = array_merge($this->defaultConfig, $config);
+		else $this->config = $this->defaultConfig;
+	}
+	
+	/**
+	 * @param Array
+	 */
+	public function setConfig($config) {
+		$this->config = $config;
+	}
+	
+	/**
+	 * @return Array
+	 */
+	public function getConfig() {
+		return $this->config;
+	}
+
+	/**
+	 * Passed a string, return HTML that has been tidied.
+	 * 
+	 * @param String HTML
+	 * @return String HTML, tidied
+	 */
+	public abstract function cleanHTML($content);
+
+	/**
+	 * Experimental inst class to create a default html cleaner class
+	 * 
+	 * @return PurifierHTMLCleaner|TidyHTMLCleaner
+	 */
+	public static function inst() {
+		if (class_exists('HTMLPurifier')) return new PurifierHTMLCleaner();
+		elseif (class_exists('tidy')) return new TidyHTMLCleaner();
+	}
+}
+
+
+/**
+ * Cleans HTML using the HTMLPurifier package
+ * http://htmlpurifier.org/
+ */
+class PurifierHTMLCleaner extends HTMLCleaner {
+	
+	public function cleanHTML($content) {
+		$html = new HTMLPurifier();
+		$doc = new SS_HTMLValue($html->purify($content));
+		return $doc->getContent();
+	}
+}
+
+/**
+ * Cleans HTML using the Tidy package
+ * http://php.net/manual/en/book.tidy.php
+ */
+class TidyHTMLCleaner extends HTMLCleaner {
+
+	protected $defaultConfig = array(
+		'clean' => true,
+		'output-xhtml' => true,
+		'show-body-only' => true,
+		'wrap' => 0,
+		'doctype' => 'omit',
+		'input-encoding' => 'utf8',
+		'output-encoding' => 'utf8'
+	);
+
+	public function cleanHTML($content) {
+		$tidy = new tidy();
+		$output = $tidy->repairString($content, $this->config);
+		return $output;
+	}
+}

tests/core/HTMLCleanerTest.php

@@ -0,0 +1,28 @@
+<?php
+/**
+ * @package sapphire
+ * @subpackage tests
+ */
+
+class HTMLCleanerTest extends SapphireTest {
+	
+	function testHTMLClean() {
+		$cleaner = HTMLCleaner::inst();
+
+		if ($cleaner) {
+			$this->assertEquals(
+				$cleaner->cleanHTML('<p>wrong <b>nesting</i></p>' . "\n"),
+				'<p>wrong <b>nesting</b></p>' . "\n",
+				"HTML cleaned properly"
+			);
+			$this->assertEquals(
+				$cleaner->cleanHTML('<p>unclosed paragraph' . "\n"),
+				'<p>unclosed paragraph</p>' . "\n",
+				"HTML cleaned properly"
+			);
+		} else {
+			$this->markTestSkipped('No HTMLCleaner library available (tidy or HTMLBeautifier)');
+		}
+	}
+
+}