def test_create_links(self):
# no pages should exist
self.assertEqual(Page.select().count(), 0)
# create links
links = create_links(link1, [link2])
self.assertTrue(len(links), 1)
link = links[0]
self.assertEqual(link.from_page.url, link1)
self.assertEqual(link.to_page.url, link2)
# 2 pages should exist
self.assertEqual(Page.select().count(), 2)
def test_create_links(self):
# no pages should exist
self.assertEqual(Page.select().count(), 0)
# create links
links = create_links(link1, [link2])
self.assertTrue(len(links), 1)
link = links[0]
self.assertEqual(link.from_page.url, link1)
self.assertEqual(link.to_page.url, link2)
# 2 pages should exist
self.assertEqual(Page.select().count(), 2)
def test_page_twice(self):
# make sure database is cleared out between tests
url = 'http://www.example.com/foo'
Page.create(url=url, content='hi world', status_code=200)
self.assertEqual(Page.select().count(), 1)
self.assertRaises(IntegrityError, Page.create, url=url, content='hi world', status_code=200)
def test_page_twice(self):
# make sure database is cleared out between tests
url = 'http://www.example.com/foo'
Page.create(url=url, content='hi world', status_code=200)
self.assertEqual(Page.select().count(), 1)
self.assertRaises(IntegrityError,
Page.create,
url=url,
content='hi world',
status_code=200)
def test_permanent_redirect(self, requests_get, requests_head):
url = "http://www.example.com/foo"
redirect_url = "http://www.example.com/bar"
headers = {'location': redirect_url}
requests_head.return_value = MagicMock(status_code=301,
headers=headers)
page = Page.create(url=url, content='', status_code=0)
add_page_info_to_page(page)
to_page = Page.select().where(Page.url == redirect_url).first()
self.assertTrue(to_page)
url_redirect_link = Link.select().where(Link.from_page == page,
Link.to_page == to_page)
self.assertTrue(url_redirect_link.exists())
self.assertEqual(requests_head.call_count, 1)
self.assertFalse(requests_get.called)
self.assertEqual(page.content, redirect_url)
def test_permanent_redirect(self, requests_get, requests_head):
url = "http://www.example.com/foo"
redirect_url = "http://www.example.com/bar"
headers = { 'location': redirect_url }
requests_head.return_value = MagicMock(status_code=301, headers=headers)
page = Page.create(url=url, content='', status_code=0)
add_page_info_to_page(page)
to_page = Page.select().where(Page.url == redirect_url).first()
self.assertTrue(to_page)
url_redirect_link = Link.select().where(
Link.from_page == page,
Link.to_page == to_page)
self.assertTrue(url_redirect_link.exists())
self.assertEqual(requests_head.call_count, 1)
self.assertFalse(requests_get.called)
self.assertEqual(page.content, redirect_url)
def sizeof_fmt(num, suffix='B'):
""" print formatted file size
http://stackoverflow.com/a/1094933
"""
for unit in ['', 'Ki', 'Mi', 'Gi', 'Ti', 'Pi', 'Ei', 'Zi']:
if abs(num) < 1024.0:
return "%3.1f%s%s" % (num, unit, suffix)
num /= 1024.0
return "%.1f%s%s" % (num, 'Yi', suffix)
if __name__ == "__main__":
initialize('corpus.db')
page_count = Page.select().count()
crawled_count = Page.select().where(
(Page.status_code == 200) &
((Page.content_type == 'text/html') |
(Page.content_type == 'text/plain')))\
.count()
redirect_count = Page.select().where(Page.status_code == 301).count()
to_crawl_count = Page.select().where(Page.status_code == 0).count()
other_count = page_count - crawled_count - redirect_count - to_crawl_count
link_count = Link.select().count()
corpus_size = os.stat('corpus.db').st_size
corpus_size = sizeof_fmt(corpus_size)
print('crawled pages: {}'.format(crawled_count))
def sizeof_fmt(num, suffix='B'):
""" print formatted file size
http://stackoverflow.com/a/1094933
"""
for unit in ['','Ki','Mi','Gi','Ti','Pi','Ei','Zi']:
if abs(num) < 1024.0:
return "%3.1f%s%s" % (num, unit, suffix)
num /= 1024.0
return "%.1f%s%s" % (num, 'Yi', suffix)
if __name__ == "__main__":
initialize('corpus.db')
page_count = Page.select().count()
crawled_count = Page.select().where(
(Page.status_code == 200) &
((Page.content_type == 'text/html') |
(Page.content_type == 'text/plain')))\
.count()
redirect_count = Page.select().where(Page.status_code == 301).count()
to_crawl_count = Page.select().where(Page.status_code == 0).count()
other_count = page_count - crawled_count - redirect_count - to_crawl_count
link_count = Link.select().count()
corpus_size = os.stat('corpus.db').st_size
corpus_size = sizeof_fmt(corpus_size)
print('crawled pages: {}'.format(crawled_count))