def _inner_loop(self, url):
"""Runs the inner loop of the iterative process.
Note:
The loop is executed inside the threading manager. After retrieving
the urls from the queue.
Args:
url (str): the url being requested.
"""
self.visited_urls.add(url)
if len(self.visited_urls) > self.req_limit:
self.manager.stop_all_workers()
if len(self.visited_urls) % self.log_frequency == 0:
msg = "Visited pages: {}; Targets found {}; Running visits {}".format(
len(self.visited_urls), len(self.target_pages), len(self.unvisited))
self.logger.debug(msg)
try:
new_page = wp.WebPage(url, timeout=self._timeout)
if self.identify_target(new_page):
self.target_pages[new_page.url] = new_page
else:
self.other_pages[new_page.url] = new_page
self.inner_urls |= set(new_page.child_urls)
if not self.greedy:
new_page.free() # frees memory after parsing
except (urllib.error.HTTPError, wp.TimeoutException) as e:
self._on_timeout_exception(url, e)
except urllib.error.URLError:
self.invalid_urls.add(url)
def _buildPage(self, title, tabs, content):
title = unicode(title)
subject = title
if subject == '':
title = '(Home)'
web_header = webpage.WebPageHeader(self.__wiki_name + ' >>> ' + title)
web_header.append(
'<link rel="stylesheet" type="text/css" href="/?getcss" />')
tabs_content = self.buildTabs(tabs)
menu_content = self.buildMenu()
template = self.DEFAULT_TEMPLATE
if self.__template is not None:
if os.path.isfile(self.__template):
f = file(self.__template, 'r')
try:
template = f.read()
finally:
f.close()
html_content = template % {
'wiki_name': self.__wiki_name,
'title': title,
'tabs': tabs_content,
'content': content,
'copy': self.__copyright,
'menu': menu_content
}
web_content = webpage.WebPageContent(html_content)
return webpage.WebPage(web_header, web_content)
def test_invalid_target_page(self):
"""Tests if an invalid target url yields an invalid target page."""
route = "/ganhe-brindes"
url = helpers.get_url(self.domain, route)
page = wp.WebPage(url)
self.assertTrue(page.title)
self.assertTrue(page.domain)
self.assertFalse(page.valid_target)
self.assertEqual(page.target_name, page._INVALID_TARGET)
def test_valid_target_page(self):
"""Tests if a valid target url yields a full target page."""
route = "/hypnose-eau-de-toilette-lancome-perfume-feminino/p"
url = helpers.get_url(self.domain, route)
page = wp.WebPage(url)
self.assertTrue(page.title)
self.assertTrue(page.domain)
self.assertTrue(page.valid_target)
self.assertEqual(page.domain, self.domain)
def __init__(
self, domain, req_limit=1e4, greedy=False,
indentify_target=lambda page: True):
# Initialize
self.root_page = wp.WebPage(domain)
self.target_pages = {}
self.other_pages = {}
self.visited_urls = set()
self.invalid_urls = set()
self.inner_urls = set()
self.unvisited = set()
self.manager = None
# Crawl parameters
self.req_limit = req_limit
self.greedy = greedy
self.identify_target = indentify_target
self.log_frequency = req_limit/10
self.logger = helpers.get_debug_logger("mylogger")
def test_invalid_attribute(self):
"""Tests if an invalid attribute raises AttributeError"""
with self.assertRaises(AttributeError):
wp.WebPage(1)
def test_invalid_url(self):
"""Tests if an invalid url raises URLError"""
invalid_url = "invalid"
with self.assertRaises(urllib.error.URLError):
wp.WebPage(invalid_url)
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self.domain = "www.epocacosmeticos.com.br"
self.domain_page = wp.WebPage(self.domain)
self.domain_urls = self.domain_page.child_urls
def setUp(self):
base_url = 'https://www.unrealengine.com'
query_url = base_url + '/marketplace/assets?lang=&q=test'
self.wp = webpage.WebPage(query_url, base_url)
def buildErrorPage(self, title, message):
header = webpage.WebPageHeader(title)
content = webpage.WebPageContent("<pre>%s</pre>" % message)
return webpage.WebPage(header, content)
