def test_single_element(self):
source = BytesIO(b'<img src="#">')
context = iterparse(source)
elements = list(context) # consume the generator
self.assertTrue(len(elements) == 1)
el, attr, url, pos = elements.pop()
self.assertEqual(url, '#')
self.assertEqual(pos, 0)
self.assertEqual(attr, 'src')
self.assertEqual(el.tag, 'img')
self.assertEqual(el.attrib, {'src': '#'})
def test_source_is_not_bytes(self):
source = StringIO('<img src="#">')
context = iterparse(source)
lst = list(context)
self.assertTrue(len(lst) == 1)
el, attr, url, pos = lst.pop()
self.assertEqual(url, '#')
self.assertEqual(pos, 0)
self.assertEqual(attr, 'src')
self.assertEqual(el.tag, 'img')
self.assertEqual(el.attrib, {'src': '#'})
def test_meta_charset_insertion(self):
context = iterparse(BytesIO(b''), include_meta_charset_tag=True)
self.assertEqual(list(context), [])
elms = context.root.xpath('//meta')
self.assertEqual(len(elms), 1)
meta = elms.pop()
self.assertEqual(meta.tag, 'meta')
self.assertEqual(meta.attrib['charset'], 'iso-8859-1')
self.assertEqual(
tostring(meta),
tostring(lxml.html.Element('meta', charset='iso-8859-1')))
self.assertEqual(tostring(context.root),
b'<html><head>%s</head></html>' % tostring(meta))
def test_root_post_iteration(self):
source = BytesIO(b'<img src="#">')
context = iterparse(source)
list(context) # consume the generator
self.assertTrue(context.root is not None)
self.assertTrue(isinstance(context.root, lxml.etree._Element))
def test_root_pre_iteration(self):
source = BytesIO(b'<img src="#">')
context = iterparse(source)
self.assertTrue(hasattr(context, 'root'))
self.assertTrue(context.root is None)
def test_return_type(self):
source = BytesIO(b'<img src="#">')
context = iterparse(source)
self.assertTrue(isinstance(context, Iterator))
class TestFullLatin1EncodedHTMLParsing(unittest.TestCase):
# Single instance of the parser to avoid creating anew each time.
context = iterparse(BytesIO(html.encode('latin1', 'replace')),
encoding='latin1')
# NOTE: Methods are ordered in the sequence they will be parsed
def test_a_first_meta_element(self):
el, attr, url, pos = next(self.context)
self.assertEqual(el.tag, 'meta')
self.assertEqual(url, 'http://nx-domain.com/redirect')
self.assertEqual(pos, 2)
self.assertEqual(attr, 'content')
self.assertEqual(el.attrib, {
'http-equiv': 'refresh',
'content': "3;http://nx-domain.com/redirect"
})
def test_b_first_link_element(self):
el, attr, url, pos = next(self.context)
self.assertEqual(el.tag, 'link')
self.assertEqual(url, 'css/main.css')
self.assertEqual(pos, 0)
self.assertEqual(attr, 'href')
self.assertEqual(el.attrib, {
'rel': 'stylesheet',
'href': "css/main.css"
})
def test_c_second_link_element(self):
el, attr, url, pos = next(self.context)
self.assertEqual(el.tag, 'link')
self.assertEqual(url, 'http://nx-domain.com/css/style.css')
self.assertEqual(pos, 0)
self.assertEqual(attr, 'href')
self.assertEqual(el.attrib, {
'rel': 'stylesheet',
'href': 'http://nx-domain.com/css/style.css'
})
def test_d_style_tag_css_url_element(self):
el, attr, url, pos = next(self.context)
self.assertEqual(el.tag, 'style')
self.assertEqual(url, "img/background.png")
self.assertEqual(pos, 49)
self.assertEqual(attr, None)
self.assertEqual(el.attrib, {})
def test_e_style_tag_css_import_element(self):
el, attr, url, pos = next(self.context)
self.assertEqual(el.tag, 'style')
self.assertEqual(url, "css/theme.css")
self.assertEqual(pos, 10)
self.assertEqual(attr, None)
self.assertEqual(el.attrib, {})
def test_f_first_anchor_element(self):
el, attr, url, pos = next(self.context)
self.assertEqual(el.tag, 'a')
self.assertEqual(url, "#")
self.assertEqual(pos, 0)
self.assertEqual(attr, 'href')
self.assertEqual(el.attrib, {'href': '#'})
def test_g_second_anchor_element(self):
el, attr, url, pos = next(self.context)
self.assertEqual(el.tag, 'a')
self.assertEqual(url, "javascript:void(0);")
self.assertEqual(pos, 0)
self.assertEqual(attr, 'href')
self.assertEqual(el.attrib, {'href': "javascript:void(0);"})
def test_h_third_anchor_element(self):
el, attr, url, pos = next(self.context)
self.assertEqual(el.tag, 'a')
self.assertEqual(url, "http://new-site.com")
self.assertEqual(pos, 0)
self.assertEqual(attr, 'href')
self.assertEqual(el.attrib, {'href': "http://new-site.com"})
def test_i_inline_style_element(self):
el, attr, url, pos = next(self.context)
self.assertEqual(el.tag, 'div')
self.assertEqual(url, "img/background.png")
self.assertEqual(pos, 17)
self.assertEqual(attr, 'style')
self.assertEqual(el.attrib,
{'style': "background: url('img/background.png');"})
def test_j_first_img_element(self):
el, attr, url, pos = next(self.context)
self.assertEqual(el.tag, 'img')
self.assertEqual(url, "img/img1.png")
self.assertEqual(pos, 0)
self.assertEqual(attr, 'src')
self.assertEqual(el.attrib, {'src': "img/img1.png", 'alt': 'img1-alt'})
def test_k_second_img_element(self):
el, attr, url, pos = next(self.context)
self.assertEqual(el.tag, 'img')
self.assertEqual(url, "http://static-site.com/img/img3.png")
self.assertEqual(pos, 0)
self.assertEqual(attr, 'src')
self.assertEqual(el.attrib, {
'src': "http://static-site.com/img/img3.png",
'alt': ""
})
def test_y_empty_iterator(self):
with self.assertRaises(StopIteration):
next(self.context)
self.assertEqual(len(list(self.context)), 0)
def test_z_root_tree_attribute(self):
self.assertTrue(hasattr(self.context.root, 'getroottree'))
self.assertTrue(isinstance(self.context.root, lxml.etree.ElementBase))
self.assertTrue(
isinstance(self.context.root.getroottree(),
lxml.etree._ElementTree))
def test_source_is_empty(self):
source = BytesIO(b'')
context = iterparse(source)
self.assertEqual(list(context), [])
self.assertEqual(tostring(context.root), b'<html></html>')
def test_source_without_read_method(self):
with self.assertRaises(TypeError):
iterparse(object())
with self.assertRaises(TypeError):
iterparse(None)
def test_source_is_empty_string(self):
with self.assertRaises(TypeError):
iterparse('')
