Convert named HTML entities before processing as XML

Author: clue

src/Io/Loader.php

@@ -6,6 +6,19 @@
 
 class Loader
 {
+    private $entities;
+
+    public function __construct(array $entities = null)
+    {
+        if ($entities === null) {
+            // get all HTML entities (minus those for XML parsing)
+            $entities = get_html_translation_table(HTML_ENTITIES, ENT_NOQUOTES, 'UTF-8');
+            unset($entities['<'], $entities['>'], $entities['&']);
+        }
+
+        $this->entities = $entities;
+    }
+
     public function loadXmlFile($path)
     {
         return $this->loadXmlString(file_get_contents($path));
@@ -16,8 +29,8 @@ public function loadXmlString($html)
         // fix invalid markup of help link in footer of outdated ViewVC versions
         $html = str_replace('Help</strong></td>', 'Help</a></strong></td>', $html);
 
-        // replace unneeded HTML entities
-        $html = str_replace('&nbsp;', ' ', $html);
+        // replace named HTML entities with their UTF-8 value
+        $html = str_replace(array_values($this->entities), array_keys($this->entities), $html);
 
         // clean up namespace declaration
         $html = str_replace('xmlns="', 'ns="', $html);

tests/Io/LoaderTest.php

@@ -29,4 +29,13 @@ function ($path) {
             scandir(__DIR__ . '/../fixtures/')
         ));
     }
+
+    public function testHtmlEntities()
+    {
+        $str = '<p>&auml;&hellip;&nbsp;&copy;</p>';
+        $xml = $this->loader->loadXmlString($str);
+
+        // c3 a4 e2 80 a6 c2 a0 c2 a9
+        $this->assertEquals('ä… ©', (string)$xml);
+    }
 }