def test_grab_parse_defensedxml(self):
with temp_dir() as tmp_dir:
injection_path = os.path.join(tmp_dir, 'injection')
with open(injection_path, 'w') as out:
out.write('Hey there!')
# Prepare file:// URL valid for both linux and windows
injection_url = 'file:///%s' % (injection_path.lstrip('/')
.replace('\\', '/'))
bad_xml = (
'<!DOCTYPE external ['
'<!ENTITY ee SYSTEM "' + injection_url + '">'
']>'
'<root>ⅇ</root>'
).encode()
xml_file = os.path.join(tmp_dir, 'bad.xml')
# On windows, use slashed instead of backslashes to avoid error:
# Invalid file://hostname/, expected localhost or 127.0.0.1 or none
if '\\' in xml_file:
xml_file = xml_file.replace('\\', '/')
with open(xml_file, 'wb') as out:
out.write(bad_xml)
grab = build_grab(content_type='xml')
file_url = 'file://%s' % xml_file
grab.go(file_url)
self.assertRaises(EntitiesForbidden, grab.doc, '//title')
def test_transport_option_as_string_fake(self):
with temp_dir() as dir_:
sys.path.insert(0, dir_)
with open(os.path.join(dir_, 'foo.py'), 'w') as out:
out.write(FAKE_TRANSPORT_CODE)
self.assert_transport_response('foo.FakeTransport', b'XYZ')
sys.path.remove(dir_)
def test_body_inmemory_false(self):
with temp_dir() as tmp_dir:
grab = build_grab()
grab.setup(body_inmemory=False)
with self.assertRaises(GrabMisuseError):
grab.go(self.server.get_url())
self.server.response['get.data'] = b'foo'
grab = build_grab()
grab.setup(body_inmemory=False)
grab.setup(body_storage_dir=tmp_dir)
grab.go(self.server.get_url())
self.assertTrue(os.path.exists(grab.doc.body_path))
self.assertTrue(tmp_dir in grab.doc.body_path)
with open(grab.doc.body_path, 'rb') as inp:
self.assertEqual(b'foo', inp.read())
# pylint: disable=protected-access
self.assertEqual(grab.doc._bytes_body, None)
# pylint: enable=protected-access
old_path = grab.doc.body_path
grab.go(self.server.get_url())
self.assertTrue(old_path != grab.doc.body_path)
with temp_dir() as tmp_dir:
self.server.response['get.data'] = 'foo'
grab = build_grab()
grab.setup(body_inmemory=False)
grab.setup(body_storage_dir=tmp_dir)
grab.setup(body_storage_filename='music.mp3')
grab.go(self.server.get_url())
self.assertTrue(os.path.exists(grab.doc.body_path))
self.assertTrue(tmp_dir in grab.doc.body_path)
with open(grab.doc.body_path, 'rb') as inp:
self.assertEqual(b'foo', inp.read())
self.assertEqual(os.path.join(tmp_dir, 'music.mp3'),
grab.doc.body_path)
self.assertEqual(grab.doc.body, b'foo')
# pylint: disable=protected-access
self.assertEqual(grab.doc._bytes_body, None)
def test_body_inmemory_false(self):
with temp_dir() as tmp_dir:
grab = build_grab()
grab.setup(body_inmemory=False)
with self.assertRaises(GrabMisuseError):
grab.go(self.server.get_url())
self.server.response['get.data'] = b'foo'
grab = build_grab()
grab.setup(body_inmemory=False)
grab.setup(body_storage_dir=tmp_dir)
grab.go(self.server.get_url())
self.assertTrue(os.path.exists(grab.doc.body_path))
self.assertTrue(tmp_dir in grab.doc.body_path)
with open(grab.doc.body_path, 'rb') as inp:
self.assertEqual(b'foo', inp.read())
# pylint: disable=protected-access
self.assertEqual(grab.doc._bytes_body, None)
# pylint: enable=protected-access
old_path = grab.doc.body_path
grab.go(self.server.get_url())
self.assertTrue(old_path != grab.doc.body_path)
with temp_dir() as tmp_dir:
self.server.response['get.data'] = 'foo'
grab = build_grab()
grab.setup(body_inmemory=False)
grab.setup(body_storage_dir=tmp_dir)
grab.setup(body_storage_filename='music.mp3')
grab.go(self.server.get_url())
self.assertTrue(os.path.exists(grab.doc.body_path))
self.assertTrue(tmp_dir in grab.doc.body_path)
with open(grab.doc.body_path, 'rb') as inp:
self.assertEqual(b'foo', inp.read())
self.assertEqual(os.path.join(tmp_dir, 'music.mp3'),
grab.doc.body_path)
self.assertEqual(grab.doc.body, b'foo')
# pylint: disable=protected-access
self.assertEqual(grab.doc._bytes_body, None)
def test_save_hash(self):
"Test `Response.save_hash` method."
with temp_dir() as tmp_dir:
with open(IMG_FILE, 'rb') as inp:
img_data = inp.read()
self.server.response['get.data'] = img_data
grab = build_grab()
grab.go(self.server.get_url())
path = grab.doc.save_hash(self.server.get_url(), tmp_dir)
with open(os.path.join(tmp_dir, path), 'rb') as inp:
test_data = inp.read()
self.assertEqual(test_data, img_data)
def test_save(self):
"Test `Response.save` method."
with temp_dir() as tmp_dir:
with open(IMG_FILE, 'rb') as inp:
img_data = inp.read()
tmp_file = os.path.join(tmp_dir, 'file.bin')
self.server.response['get.data'] = img_data
grab = build_grab()
grab.go(self.server.get_url())
grab.doc.save(tmp_file)
with open(tmp_file, 'rb') as inp:
self.assertEqual(inp.read(), img_data)
def test_log_option(self):
with temp_dir() as tmp_dir:
reset_request_counter()
log_file_path = os.path.join(tmp_dir, 'lograb.html')
grab = build_grab()
grab.setup(log_file=log_file_path)
self.server.response['get.data'] = 'omsk'
self.assertEqual(os.listdir(tmp_dir), [])
grab.go(self.server.get_url())
self.assertEqual(os.listdir(tmp_dir), ['lograb.html'])
with open(log_file_path) as inp:
self.assertEqual(inp.read(), 'omsk')
def test_log_option(self):
with temp_dir() as tmp_dir:
reset_request_counter()
log_file_path = os.path.join(tmp_dir, 'lograb.html')
grab = build_grab()
grab.setup(log_file=log_file_path)
self.server.response['get.data'] = 'omsk'
self.assertEqual(os.listdir(tmp_dir), [])
grab.go(self.server.get_url())
self.assertEqual(os.listdir(tmp_dir), ['lograb.html'])
with open(log_file_path) as inp:
self.assertEqual(inp.read(), 'omsk')
def test_lxml_security_bug(self):
with temp_dir() as tmp_dir:
injection_path = os.path.join(tmp_dir, 'injection')
with open(injection_path, 'w') as out:
out.write('Hey there!')
# Prepare file:// URL valid for both linux and windows
injection_url = 'file:///%s' % (injection_path.lstrip('/').replace(
'\\', '/'))
bad_xml = ('<!DOCTYPE external ['
'<!ENTITY ee SYSTEM "' + injection_url + '">'
']>'
'<root>ⅇ</root>').encode()
tree = parse(BytesIO(bad_xml))
self.assertEqual(tree.xpath('//root/text()')[0], 'Hey there!')
def test_log_dir_response_content(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir)
self.server.response['get.data'] = 'omsk'
self.server.response['headers'] = [('X-Engine', 'PHP')]
self.assertEqual(os.listdir(tmp_dir), [])
grab.go(self.server.get_url())
self.assertEqual(sorted(os.listdir(tmp_dir)),
['01.html', '01.log'])
with open(os.path.join(tmp_dir, '01.log')) as inp:
log_file_content = inp.read()
self.assertTrue('x-engine' in log_file_content.lower())
def test_log_dir_response_content(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir)
self.server.response['get.data'] = 'omsk'
self.server.response['headers'] = [('X-Engine', 'PHP')]
self.assertEqual(os.listdir(tmp_dir), [])
grab.go(self.server.get_url())
self.assertEqual(sorted(os.listdir(tmp_dir)),
['01.html', '01.log'])
with open(os.path.join(tmp_dir, '01.log')) as inp:
log_file_content = inp.read()
self.assertTrue('x-engine' in log_file_content.lower())
def test_log_dir_request_content_is_empty(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir)
grab.setup(headers={'X-Name': 'spider'}, post='xxxPost')
self.assertEqual(os.listdir(tmp_dir), [])
grab.go(self.server.get_url())
self.assertEqual(sorted(os.listdir(tmp_dir)),
['01.html', '01.log'])
with open(os.path.join(tmp_dir, '01.log')) as inp:
log_file_content = inp.read()
self.assertFalse('X-Name' in log_file_content)
self.assertFalse('xxxPost' in log_file_content)
def test_lxml_security_bug(self):
with temp_dir() as tmp_dir:
injection_path = os.path.join(tmp_dir, 'injection')
with open(injection_path, 'w') as out:
out.write('Hey there!')
# Prepare file:// URL valid for both linux and windows
injection_url = 'file:///%s' % (injection_path.lstrip('/')
.replace('\\', '/'))
bad_xml = (
'<!DOCTYPE external ['
'<!ENTITY ee SYSTEM "' + injection_url + '">'
']>'
'<root>ⅇ</root>'
).encode()
tree = parse(BytesIO(bad_xml))
self.assertEqual(tree.xpath('//root/text()')[0], 'Hey there!')
def test_log_dir_request_content_is_empty(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir)
grab.setup(headers={'X-Name': 'spider'}, post='xxxPost')
self.assertEqual(os.listdir(tmp_dir), [])
grab.go(self.server.get_url())
self.assertEqual(sorted(os.listdir(tmp_dir)),
['01.html', '01.log'])
with open(os.path.join(tmp_dir, '01.log')) as inp:
log_file_content = inp.read()
self.assertFalse('X-Name' in log_file_content)
self.assertFalse('xxxPost' in log_file_content)
def test_grab_parse_defensedxml(self):
with temp_dir() as tmp_dir:
injection_path = os.path.join(tmp_dir, 'injection')
with open(injection_path, 'w') as out:
out.write('Hey there!')
# Prepare file:// URL valid for both linux and windows
injection_url = 'file:///%s' % (injection_path.lstrip('/').replace(
'\\', '/'))
bad_xml = ('<!DOCTYPE external ['
'<!ENTITY ee SYSTEM "' + injection_url + '">'
']>'
'<root>ⅇ</root>').encode()
xml_file = os.path.join(tmp_dir, 'bad.xml')
with open(xml_file, 'wb') as out:
out.write(bad_xml)
grab = build_grab(content_type='xml')
grab.go('file://%s' % xml_file)
self.assertRaises(EntitiesForbidden, grab.doc, '//title')
def test_log_dir_option(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir)
self.server.response_once['get.data'] = 'omsk1'
self.server.response['get.data'] = 'omsk2'
self.assertEqual(os.listdir(tmp_dir), [])
grab.go(self.server.get_url())
grab.go(self.server.get_url())
self.assertEqual(sorted(os.listdir(tmp_dir)),
['01.html', '01.log', '02.html', '02.log'])
with open(os.path.join(tmp_dir, '01.html')) as inp:
self.assertEqual(inp.read(), 'omsk1')
with open(os.path.join(tmp_dir, '02.html')) as inp:
self.assertEqual(inp.read(), 'omsk2')
def test_log_dir_option(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir)
self.server.response_once['get.data'] = 'omsk1'
self.server.response['get.data'] = 'omsk2'
self.assertEqual(os.listdir(tmp_dir), [])
grab.go(self.server.get_url())
grab.go(self.server.get_url())
self.assertEqual(sorted(os.listdir(tmp_dir)),
['01.html', '01.log', '02.html', '02.log'])
with open(os.path.join(tmp_dir, '01.html')) as inp:
self.assertEqual(inp.read(), 'omsk1')
with open(os.path.join(tmp_dir, '02.html')) as inp:
self.assertEqual(inp.read(), 'omsk2')
def test_log_dir_request_content_headers_and_post(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir, debug=True)
grab.setup(headers={'X-Name': 'spider'}, post={'xxx': 'Post'})
self.assertEqual(os.listdir(tmp_dir), [])
grab.go(self.server.get_url())
self.assertEqual(sorted(os.listdir(tmp_dir)),
['01.html', '01.log'])
with open(os.path.join(tmp_dir, '01.log')) as inp:
log_file_content = inp.read()
#if not 'x-name' in log_file_content.lower():
# print('CONTENT OF 01.log:')
# print(log_file_content)
self.assertTrue('x-name' in log_file_content.lower())
self.assertTrue('xxx=post' in log_file_content.lower())
def test_log_dir_request_content_headers_and_post(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir, debug=True)
grab.setup(headers={'X-Name': 'spider'}, post={'xxx': 'Post'})
self.assertEqual(os.listdir(tmp_dir), [])
grab.go(self.server.get_url())
self.assertEqual(sorted(os.listdir(tmp_dir)),
['01.html', '01.log'])
with open(os.path.join(tmp_dir, '01.log')) as inp:
log_file_content = inp.read()
#if not 'x-name' in log_file_content.lower():
# print('CONTENT OF 01.log:')
# print(log_file_content)
self.assertTrue('x-name' in log_file_content.lower())
self.assertTrue('xxx=post' in log_file_content.lower())
def test_log_dir_response_network_error(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir, timeout=1, user_agent='Perl',
debug=True)
self.server.response['get.data'] = 'omsk'
self.server.response['headers'] = [('X-Engine', 'PHP')]
self.server.response['sleep'] = 2
self.assertEqual(os.listdir(tmp_dir), [])
try:
grab.go(self.server.get_url())
except GrabTimeoutError:
pass
self.assertEqual(sorted(os.listdir(tmp_dir)),
['01.html', '01.log'])
with open(os.path.join(tmp_dir, '01.log')) as inp:
log_file_content = inp.read()
self.assertTrue('user-agent: perl' in log_file_content.lower())
def test_grab_parse_defensedxml(self):
with temp_dir() as tmp_dir:
injection_path = os.path.join(tmp_dir, 'injection')
with open(injection_path, 'w') as out:
out.write('Hey there!')
# Prepare file:// URL valid for both linux and windows
injection_url = 'file:///%s' % (injection_path.lstrip('/').replace(
'\\', '/'))
bad_xml = ('<!DOCTYPE external ['
'<!ENTITY ee SYSTEM "' + injection_url + '">'
']>'
'<root>ⅇ</root>').encode()
xml_file = os.path.join(tmp_dir, 'bad.xml')
# On windows, use slashed instead of backslashes to avoid error:
# Invalid file://hostname/, expected localhost or 127.0.0.1 or none
if '\\' in xml_file:
xml_file = xml_file.replace('\\', '/')
with open(xml_file, 'wb') as out:
out.write(bad_xml)
grab = build_grab(content_type='xml')
file_url = 'file://%s' % xml_file
grab.go(file_url)
self.assertRaises(EntitiesForbidden, grab.doc, '//title')
def test_log_dir_response_content_thread(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir)
self.server.response['get.data'] = 'omsk'
self.server.response['headers'] = [('X-Engine', 'PHP')]
self.assertEqual(os.listdir(tmp_dir), [])
def func():
grab.go(self.server.get_url())
thread = threading.Thread(target=func)
thread.start()
thread.join()
files = os.listdir(tmp_dir)
self.assertEqual(2, len([x for x in files if '01-thread' in x]))
fname = [x for x in files if x.endswith('.log')][0]
with open(os.path.join(tmp_dir, fname)) as inp:
log_file_content = inp.read()
self.assertTrue('x-engine' in log_file_content.lower())
def test_log_dir_response_content_thread(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir)
self.server.response['get.data'] = 'omsk'
self.server.response['headers'] = [('X-Engine', 'PHP')]
self.assertEqual(os.listdir(tmp_dir), [])
def func():
grab.go(self.server.get_url())
thread = threading.Thread(target=func)
thread.start()
thread.join()
files = os.listdir(tmp_dir)
self.assertEqual(2, len([x for x in files if '01-thread' in x]))
fname = [x for x in files if x.endswith('.log')][0]
with open(os.path.join(tmp_dir, fname)) as inp:
log_file_content = inp.read()
self.assertTrue('x-engine' in log_file_content.lower())
def test_log_dir_response_network_error(self):
with temp_dir() as tmp_dir:
reset_request_counter()
grab = build_grab()
grab.setup(log_dir=tmp_dir,
timeout=1,
user_agent='Perl',
debug=True)
self.server.response['get.data'] = 'omsk'
self.server.response['headers'] = [('X-Engine', 'PHP')]
self.server.response['sleep'] = 2
self.assertEqual(os.listdir(tmp_dir), [])
try:
grab.go(self.server.get_url())
except GrabTimeoutError:
pass
self.assertEqual(sorted(os.listdir(tmp_dir)),
['01.html', '01.log'])
with open(os.path.join(tmp_dir, '01.log')) as inp:
log_file_content = inp.read()
self.assertTrue('user-agent: perl' in log_file_content.lower())