def fetch_domains(self):
self.domains = []
for n in self.links:
url = Link(n, self.url)
if url.is_valid():
self.domains.append(url.domain)
return self.domains
def fetch_domains_id(self):
self.domain_ids = []
for n in self.links:
url = Link(n, self.url)
if url.is_valid():
self.domain_ids.append(url.netloc)
return self.domain_ids
def fetch_domains_id(self): self.domain_ids = [] for n in self.links: url = Link(n, self.url) if url.is_valid(): self.domain_ids.append(url.netloc) return self.domain_ids
def fetch_domains(self): self.domains = [] for n in self.links: url = Link(n, self.url) if url.is_valid(): self.domains.append(url.domain) return self.domains
def is_valid(self):
#logging.info("Valid url?")
url = Link(self.url, self.source_url, self.debug)
if url.is_valid():
#logging.info("Yes")
return True
else:
self.msg = url.msg
self.code = url.code
self.step = "Validating page"
self.status = False
return False
def is_valid(self):
#logging.info("Valid url?")
url = Link(self.url, self.source_url, self.debug)
if url.is_valid():
#logging.info("Yes")
return True
else:
self.msg = url.msg
self.code = url.code
self.step = "Validating page"
self.status = False
return False
def fetch_links(self):
''' extract raw_links and domains '''
self.domains = []
self.links = []
self.domain_ids = []
links = [n.get('href') for n in self.doc.find_all("a")]
links = [n for n in links if n is not None and n != "" and n != "/" and n[0] !="#"]
for url in links:
if url.startswith('mailto'):
pass
if url.startswith('javascript'):
pass
else:
l = Link(url)
if l.is_valid():
url, domain, domain_id = l.clean_url(url, self.url)
self.domains.append(domain)
self.links.append(url)
self.domain_ids.append(domain_id)
return (self.links, self.domains, self.domain_ids)
def fetch_links(self):
''' extract raw_links and domains '''
self.domains = []
self.links = []
self.domain_ids = []
links = [n.get('href') for n in self.doc.find_all("a")]
links = [
n for n in links
if n is not None and n != "" and n != "/" and n[0] != "#"
]
for url in links:
if url.startswith('mailto'):
pass
if url.startswith('javascript'):
pass
else:
l = Link(url)
if l.is_valid():
url, domain, domain_id = l.clean_url(url, self.url)
self.domains.append(domain)
self.links.append(url)
self.domain_ids.append(domain_id)
return (self.links, self.domains, self.domain_ids)
def check_link(self, url, source_url):
url = Link(url, source_url)
if url.is_valid():
return url.url
else:
return None
def check_link(self, url, source_url): url = Link(url, source_url) if url.is_valid(): return url.url else: return None
