add parser_cls argument, changes default html parser to html.HTMLParser

This changes the default HTML parser to html.HTMLParser, and also
introduces a parameter in Selector to specify another parser class
if desired.

The parser parameter will enable users that care a big deal about
performance to use a custom parser if desired.

This will affect Scrapy because it just uses the default here, but
doesn't seem to have a perceived impact on performance, as per @kmike
benchmark shared here:

https://gist.github.com/kmike/af647777cef39c3d01071905d176c006

parsel/selector.py

@@ -5,7 +5,7 @@
 import sys
 
 import six
-from lxml import etree
+from lxml import etree, html
 
 from .utils import flatten, iflatten, extract_regex
 from .csstranslator import HTMLTranslator, GenericTranslator
@@ -17,7 +17,7 @@ def __init__(self, *args, **kwargs):
         super(SafeXMLParser, self).__init__(*args, **kwargs)
 
 _ctgroup = {
-    'html': {'_parser': etree.HTMLParser,
+    'html': {'_parser': html.HTMLParser,
              '_csstranslator': HTMLTranslator(),
              '_tostring_method': 'html'},
     'xml': {'_parser': SafeXMLParser,
@@ -139,9 +139,9 @@ class Selector(object):
     selectorlist_cls = SelectorList
 
     def __init__(self, text=None, type=None, namespaces=None, root=None,
-                 base_url=None, _expr=None):
+                 base_url=None, _expr=None, parser_cls=None):
         self.type = st = _st(type or self._default_type)
-        self._parser = _ctgroup[st]['_parser']
+        self._parser = parser_cls or _ctgroup[st]['_parser']
         self._csstranslator = _ctgroup[st]['_csstranslator']
         self._tostring_method = _ctgroup[st]['_tostring_method']
 

tests/test_selector.py

@@ -404,6 +404,15 @@ def test_configure_base_url(self):
         sel = self.sscls(text=u'nothing', base_url='http://example.com')
         self.assertEquals(u'http://example.com', sel.root.base)
 
+    def test_custom_parser_cls(self):
+        from lxml import etree, html
+        text = u"<a>example</a>"
+        with self.assertRaisesRegexp(AttributeError, "has no attribute 'make_links_absolute'"):
+            sel = Selector(text, parser_cls=etree.HTMLParser)
+            sel.xpath('//a')[0].root.make_links_absolute
+
+        sel = Selector(text, parser_cls=html.HTMLParser)
+        self.assertIsNotNone(sel.xpath('//a')[0].root.make_links_absolute)
 
     def test_extending_selector(self):
         class MySelectorList(Selector.selectorlist_cls):
@@ -418,6 +427,7 @@ class MySelector(Selector):
         self.assertIsInstance(sel.css('div'), MySelectorList)
         self.assertIsInstance(sel.css('div')[0], MySelector)
 
+
 class ExsltTestCase(unittest.TestCase):
 
     sscls = Selector