fixed bug caused when instantiating selectors with responses containing empty bodies

scrapy/selector/factories.py

@@ -24,7 +24,7 @@ def body_as_utf8(response):
         
 def xmlDoc_from_html(response):
     """Return libxml2 doc for HTMLs"""
-    utf8body = body_as_utf8(response)
+    utf8body = body_as_utf8(response) or ' '
     try:
         lxdoc = libxml2.htmlReadDoc(utf8body, response.url, 'utf-8', \
             html_parser_options)
@@ -35,7 +35,7 @@ def xmlDoc_from_html(response):
 
 def xmlDoc_from_xml(response):
     """Return libxml2 doc for XMLs"""
-    utf8body = body_as_utf8(response)
+    utf8body = body_as_utf8(response) or ' '
     try:
         lxdoc = libxml2.readDoc(utf8body, response.url, 'utf-8', \
             xml_parser_options)

scrapy/tests/test_selector.py

@@ -241,6 +241,12 @@ class XPathSelectorTestCase(unittest.TestCase):
             u'\n  ',
             u'\n  pff\n'])
 
+    @libxml2debug
+    def test_empty_bodies(self):
+        r1 = TextResponse('http://www.example.com', body='')
+        hxs = HtmlXPathSelector(r1) # shouldn't raise error
+        xxs = XmlXPathSelector(r1) # shouldn't raise error
+
     @libxml2debug
     def test_weakref_slots(self):
         """Check that classes are using slots and are weak-referenceable"""