From f1df650ffe4e753be31b9cb8cb6705b996fc01dd Mon Sep 17 00:00:00 2001
From: Jeff Kehler <jeffrey.kehler@gmail.com>
Date: Sat, 4 Oct 2014 22:54:23 +0700
Subject: [PATCH] Fix issue with html pages containing <?xml> declarations

---
 goose/parsers.py | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/goose/parsers.py b/goose/parsers.py
index a43e9b47..f5374976 100644
--- a/goose/parsers.py
+++ b/goose/parsers.py
@@ -26,6 +26,7 @@
 from copy import deepcopy
 from goose.text import innerTrim
 from goose.text import encodeValue
+import re
 
 
 class Parser(object):
@@ -51,7 +52,12 @@ def css_select(self, node, selector):
     @classmethod
     def fromstring(self, html):
         html = encodeValue(html)
+
+        # remove <?xml> tag because it breaks the lxml html parser
+        html = re.sub(r'<\?xml version\=[\"\'][0-9]\.[0-9][\"\'] encoding\=(.*?)\?>', '', html)
+
         self.doc = lxml.html.fromstring(html)
+
         return self.doc
 
     @classmethod