better HTML decoding (and encoding) - deal with numeric entities, rem…

…ove bad entities, also don't, by default, encode regular UTF-8 chars

lib/jsdom/browser/htmlencoding.js

@@ -128,18 +128,24 @@ var htmlentities = function(string, quote_style) {
 
     string = string || "";
 
-    var hash_map = {}, symbol = '', tmp_str = '', entity = '';
+    var hash_map = {}, symbol, tmp_str = '', entity = '';
     tmp_str = string.toString();
 
-    if (false === (hash_map = get_html_translation_table('HTML_ENTITIES', quote_style))) {
+
+    if (false === (hash_map = get_html_translation_table('HTML_SPECIALCHARS', quote_style))) {
         return false;
     }
-    hash_map["'"] = ''';
-    delete hash_map["&"];
-    tmp_str = tmp_str.split("&").join('&');
+
+    // amended by ejones - get_html_translation_table no longer translates
+    //   char codes by default
+
+    //hash_map["'"] = ''';
+    delete hash_map['38']; // &
+    tmp_str = tmp_str.split('&').join('&');
     for (symbol in hash_map) {
-        entity = hash_map[symbol];
-        tmp_str = tmp_str.split(symbol).join(entity);
+        tmp_str = tmp_str
+                .split(String.fromCharCode(symbol))
+                .join(hash_map[symbol]);
     }
 
     return tmp_str;
@@ -166,24 +172,32 @@ var html_entity_decode = function(string, quote_style) {
 
     string = string || "";
 
-    var hash_map = {}, symbol = '', tmp_str = '', entity = '';
+    var hash_map = {}, symbol, tmp_str = '', 
+      entity = '', entity_to_char_code = {};
     tmp_str = string.toString();
 
     if (false === (hash_map = get_html_translation_table('HTML_ENTITIES', quote_style))) {
         return false;
     }
 
-    // fix & problem
-    // http://phpjs.org/functions/get_html_translation_table:416#comment_97660
-    delete(hash_map['&']);
-    hash_map['&'] = '&';
-
-    tmp_str = tmp_str.split(''').join("'");
-    tmp_str = tmp_str.split(''').join("'");
+
+    // amended by ejones - more flexible recognition of entities
     for (symbol in hash_map) {
-        entity = hash_map[symbol];
-        tmp_str = tmp_str.split(entity).join(symbol);
+        entity_to_char_code[hash_map[symbol]] = symbol;
     }
+    tmp_str = tmp_str.replace(/&([#a-z0-9]+);/gi, function( ent, ename ) {
+        var char = entity_to_char_code[ent];
+        if (char == null && ename === 'apos') char = 39;
+        if (char == null && ename[0] === '#') {
+            if (ename[1].toLowerCase() === 'x') {
+                char = parseInt(ename.substring(2), 16);
+            } else {
+                char = parseInt(ename.substring(1), 10);
+            };
+            if (isNaN(char)) char = null;
+        };
+        return char != null ? String.fromCharCode(char) : '';
+    });
 
     return tmp_str;
 };
@@ -339,13 +353,18 @@ var get_html_translation_table = function(table, quote_style) {
     entities['62'] = '>';
 
 
-    // ascii decimals to real symbols
-    for (decimal in entities) {
-        symbol = String.fromCharCode(decimal);
-        hash_map[symbol] = entities[decimal];
-    }
+
+    // removed by ejones - translation now done in encoding/decoding
+
+    //// ascii decimals to real symbols
+    //for (decimal in entities) {
+    //    symbol = String.fromCharCode(decimal);
+    //    hash_map[symbol] = entities[decimal];
+    //}
+
+    //return hash_map;
 
-    return hash_map;
+    return entities;
 };
 
 /* }}} */