Ignore attributes with illegal chars in name (fixes #23)

This is neccesary because method "DOMElement::setAttribute"
throws exception for wrong names so DOM elements
can't contain these attributes.

src/HTML5/Parser/DOMTreeBuilder.php

@@ -234,7 +234,13 @@ public function startTag($name, $attributes = array(), $selfClosing = FALSE) {
         $aName = Elements::normalizeMathMlAttribute($aName);
       }
 
-      $ele->setAttribute($aName, $aVal);
+      try {
+        $ele->setAttribute($aName, $aVal);
+      }
+      catch(\DOMException $e) {
+        $this->parseError("Illegal attribute name for tag $name. Ignoring: $aName");
+        continue;
+      }
 
       // This is necessary on a non-DTD schema, like HTML5.
       if ($aName == 'id') {

src/HTML5/Parser/Tokenizer.php

@@ -414,16 +414,33 @@ protected function attribute(&$attributes) {
       $name = $this->scanner->current();
       $this->scanner->next();
     }
-    if (preg_match('/[\'\"]/', $name)) {
-    //if (strspn($name, '\'\"')) {
+
+    $isValidAttribute = TRUE;
+    // Attribute names can contain most Unicode characters for HTML5.
+    // But method "DOMElement::setAttribute" is throwing exception
+    // because of it's own internal restriction so these have to be filtered.
+    // see issue #23: https://github.com/Masterminds/html5-php/issues/23
+    // and http://www.w3.org/TR/2011/WD-html5-20110525/syntax.html#syntax-attribute-name
+    if (preg_match("/[\x1-\x2C\\/\x3B-\x40\x5B-\x5E\x60\x7B-\x7F]/u", $name)) {
       $this->parseError("Unexpected characters in attribute name: %s", $name);
+      $isValidAttribute = FALSE;
+    }
+    // There is no limitation for 1st character in HTML5.
+    // But method "DOMElement::setAttribute" is throwing exception for the
+    // characters below so they have to be filtered.
+    // see issue #23: https://github.com/Masterminds/html5-php/issues/23
+    // and http://www.w3.org/TR/2011/WD-html5-20110525/syntax.html#syntax-attribute-name
+    else if (preg_match("/^[0-9.-]/u", $name)) {
+      $this->parseError("Unexpected character at the begining of attribute name: %s", $name);
+      $isValidAttribute = FALSE;
     }
     // 8.1.2.3
     $this->scanner->whitespace();
 
     $val = $this->attributeValue();
-    //return array($name, $val);
-    $attributes[$name] = $val;
+    if($isValidAttribute) {
+      $attributes[$name] = $val;
+    }
     return TRUE;
   }
 

test/HTML5/Parser/TokenizerTest.php

@@ -363,11 +363,18 @@ public function testTagAttributes() {
       // This will emit an entity lookup failure for &red.
       "<foo a='blue&red'>" => array('foo', array('a' => 'blue&red'), FALSE),
       "<foo a='blue&&amp;&red'>" => array('foo', array('a' => 'blue&&&red'), FALSE),
-      '<foo b"="baz">' => array('foo', array('b"' => 'baz'), FALSE),
       '<foo bar=>' => array('foo', array('bar' => NULL), FALSE),
       '<foo bar="oh' => array('foo', array('bar' => 'oh'), FALSE),
       '<foo bar=oh">' => array('foo', array('bar' => 'oh"'), FALSE),
 
+      // these attributes are ignored because of current implementation
+      // of method "DOMElement::setAttribute"
+      // see issue #23: https://github.com/Masterminds/html5-php/issues/23
+      '<foo b"="baz">' => array('foo', array(), FALSE),
+      '<foo 2abc="baz">' => array('foo', array(), FALSE),
+      '<foo ?="baz">' => array('foo', array(), FALSE),
+      '<foo foo?bar="baz">' => array('foo', array(), FALSE),
+
     );
     foreach ($bad as $test => $expects) {
       $events = $this->parse($test);