#!/usr/bin/env python # Copyright (c) 2011 The Chromium Authors. All rights reserved. # Use of this source code is governed by a BSD-style license that can be # found in the LICENSE file. """Downloads web pages with fillable forms after parsing through a set of links. Used for collecting web pages with forms. Used as a standalone script. This script assumes that it's run from within the same directory in which it's checked into. If this script were to be run elsewhere then the path for REGISTER_PAGE_DIR needs to be changed. This script assumes that third party modules are installed: httplib2, lxml, pycurl. Usage: webforms_aggregator.py [options] [single url or file containing urls] Options: -l LOG_LEVEL, --log_level LOG_LEVEL LOG_LEVEL: debug, info, warning or error [default: error] -h, --help show this help message and exit """ import datetime import errno import logging import optparse import os import re # Needed in Linux so that PyCurl does not throw a segmentation fault. import signal import sys import tempfile import threading import time import urlparse import httplib2 from lxml import html, etree import pycurl REGISTER_PAGE_DIR = os.path.join(os.pardir, 'test', 'data', 'autofill', 'heuristics', 'input') NOT_FOUND_REG_PAGE_SITES_FILENAME = 'notFoundRegPageSites.txt' FORM_LOCATION_COMMENT = 'Form Location: %s' HTML_FILE_PREFIX = 'grabber-' MAX_REDIRECTIONS = 10 # Strings in a webpage that are indicative of a registration link. LINK_CLUES = ['regist', 'user', 'sign', 'login', 'account'] MAX_SAME_DOMAIN_URLS_NO = 30 MAX_TOTAL_URLS_PER_DOMAIN = 300 MAX_OPEN_FILES_NO = 500 # URLs are selected for downloading with the following rules from the link # lists, giving more weight to the links that contain a link clue. CLUE_SECURE_LINKS_NO = MAX_SAME_DOMAIN_URLS_NO * 3/10 CLUE_GENERAL_LINKS_NO = MAX_SAME_DOMAIN_URLS_NO * 3/10 SECURE_LINKS_NO = MAX_SAME_DOMAIN_URLS_NO * 2/10 GENERAL_LINKS_NO = MAX_SAME_DOMAIN_URLS_NO * 2/10 MAX_ALLOWED_THREADS = MAX_OPEN_FILES_NO / MAX_SAME_DOMAIN_URLS_NO + 1 class Retriever(object): """Download, parse, and check if the web page contains a registration form. The objects of this class has a one to one relation with the web pages. For each page that is downloaded and parsed an object of this class is created. Each Retriever object creates a curl object. This object is added to the curl multi object of the crawler object so that the corresponding pages gets downloaded. """ logger = logging.getLogger(__name__) def __init__(self, url, domain, cookie_file): """Initializes a Retriever object. Args: url: url to download page from. domain: only links with this domain will be retrieved. cookie_file: the name of a cookie file, needed for pages that use session cookies to change their contents. """ self._url = url self._domain = domain self._html_content = '' # Http links without clues from LINK_CLUES. self._general_links = [] # Http links that contain a clue from LINK_CLUES. self._clues_general_links = [] # Https links that do not contain any clues from LINK_CLUES. self._secure_links = [] # Https links that contain a clue from LINK_CLUES. self._clues_secure_links = [] self._cookie_file = cookie_file self._curl_object = None def __del__(self): """Cleans up before this object is destroyed. The function closes the corresponding curl object that does the downloading. """ if self._curl_object: self._curl_object.close() def _AddLink(self, link): """Adds url |link|, if not already present, to the appropriate list. The link only gets added to the single list that is appopriate for it: _secure_links, _general_links, _clues_secure_links or _clues_general_links. Args: link: the url that is inserted to the appropriate links list. """ # Handles sites with unicode URLs. if isinstance(link, unicode): # Encode in 'utf-8' to avoid the UnicodeEncodeError exception. link = httplib2.iri2uri(link).encode('utf-8') link_parsed = urlparse.urlparse(link) link_lists = [self._clues_secure_links, self._secure_links, self._clues_general_links, self._general_links] # Checks that the registration page is within the domain. if (self._domain in link_parsed[1] and all(link not in x for x in link_lists)): for clue in LINK_CLUES: if clue in link.lower(): if link_parsed[0].startswith('https'): self._clues_secure_links.append(link) return else: self._clues_general_links.append(link) return if link_parsed[0].startswith('https'): # No clues found in the link. self._secure_links.append(link) else: self._general_links.append(link) def ParseAndGetLinks(self): """Parses downloaded page and gets url link for non registration page. Checks if current page contains a registration page and if not it gets the url links. If it is a registration page, it saves it in a file as 'grabber-' + domain + '.html' after it has added the FORM_LOCATION_COMMENT and it returns True. Otherwise it returns False. Returns: True if current page contains a registration form, and False otherwise. Raises: IOError: When can't write to the file. """ if not self._domain: self.logger.error('Error: self._domain was not set') sys.exit(1) match_list = re.findall(r'(?P[\'\"])(?P(?:https?:)?//.*?)\1', self._html_content) for group_list in match_list: link = group_list[1] if link.startswith('//'): link = urlparse.urljoin(self._url, link) self._AddLink(link) try: tree = html.fromstring(self._html_content, parser=html.HTMLParser()) except etree.LxmlError: self.logger.info('\t\tSkipping: not valid HTML code in this page <<< %s', self._url) return False try: body = tree.iter('body').next() except StopIteration: self.logger.info('\t\tSkipping: no "BODY" tag in this page <<< %s', self._url) return False # Get a list of all input elements with attribute type='password' password_elements = list(body.iterfind('.//input[@type="password"]')) # Check for multiple password elements to distinguish between a login form # and a registration form (Password field and Confirm Password field). if password_elements and len(password_elements) >= 2: form_elements = [] for password_elem in password_elements: form_elem = password_elem.xpath('ancestor::form[1]') if not form_elem: continue if not form_elem[0] in form_elements: form_elements.append(form_elem[0]) else: # Confirms that the page contains a registration form if two passwords # are contained in the same form for form_elem[0]. if not os.path.isdir(REGISTER_PAGE_DIR): os.makedirs(REGISTER_PAGE_DIR) # Locate the HTML tag and insert the form location comment after it. html_tag = tree.iter('html').next() comment = etree.Comment(FORM_LOCATION_COMMENT % self._url) html_tag.insert(0, comment) # Create a new file and save the HTML registration page code. f = open('%s/%s%s.html' % (REGISTER_PAGE_DIR, HTML_FILE_PREFIX, self._domain), 'w') try: f.write(html.tostring(tree, pretty_print=True)) except IOError as e: self.logger.error('Error: %s', e) raise finally: f.close() return True # Registration page found. # Indicates page is not a registration page and links must be parsed. link_elements = list(body.iter('a')) for link_elem in link_elements: link = link_elem.get('href') if not link or '#' == link[0]: continue link = urlparse.urljoin(self._url, link) link_parsed = urlparse.urlparse(link) if not link_parsed[0].startswith('http'): continue self._AddLink(link) return False # Registration page not found. def InitRequestHead(self): """Initializes curl object for a HEAD request. A HEAD request is initiated so that we can check from the headers if this is a valid HTML file. If it is not a valid HTML file, then we do not initiate a GET request, saving any unnecessary downloadings. """ self._curl_object = pycurl.Curl() self._curl_object.setopt(pycurl.URL, self._url) # The following line fixes the GnuTLS package error that pycurl depends # on for getting https pages. self._curl_object.setopt(pycurl.SSLVERSION, pycurl.SSLVERSION_SSLv3) self._curl_object.setopt(pycurl.FOLLOWLOCATION, True) self._curl_object.setopt(pycurl.NOBODY, True) self._curl_object.setopt(pycurl.SSL_VERIFYPEER, False); self._curl_object.setopt(pycurl.MAXREDIRS, MAX_REDIRECTIONS) self._curl_object.setopt(pycurl.FAILONERROR, False) self._curl_object.setopt(pycurl.COOKIEFILE, self._cookie_file) self._curl_object.setopt(pycurl.COOKIEJAR, self._cookie_file) self._curl_object.setopt(pycurl.CONNECTTIMEOUT, 30) self._curl_object.setopt(pycurl.TIMEOUT, 300) self._curl_object.setopt(pycurl.NOSIGNAL, 1) def InitRequestGet(self): """Initializes curl object for a GET request. This is called only for valid HTML files. The Pycurl makes a GET request. The page begins to download, but since not all the data of the pages comes at once. When some of the data on the page is downloaded Pycurl will put this data in the buffer. The data is appended to the end of the page until everything is downloaded. """ self._curl_object.setopt(pycurl.NOBODY, False) self._curl_object.setopt( pycurl.WRITEFUNCTION, lambda buff: setattr( self, '_html_content', self._html_content + buff)) def Download(self): """Downloads the self._url page. It first does a HEAD request and then it proceeds to a GET request. It uses a curl object for a single download. This function is called only once for the initial url of a site when we still don't have more urls from a domain. Returns: True, if the downloaded page is valid HTML code, or False otherwise. """ self.InitRequestHead() try: self._curl_object.perform() except pycurl.error as e: self.logger.error('Error: %s, url: %s', e, self._url) return False self._url = urlparse.urljoin( self._url, self._curl_object.getinfo(pycurl.EFFECTIVE_URL)) content_type = self._curl_object.getinfo(pycurl.CONTENT_TYPE) if content_type and ('text/html' in content_type.lower()): self.InitRequestGet() try: self._curl_object.perform() except pycurl.error as e: self.logger.error('Error: %s, url: %s', e, self._url) return False return True else: self.logger.info('\tSkipping: Not an HTML page <<< %s', self._url) return False def Run(self): """Called only once for the initial url when we do not have more urls. Downloads the originally-specified site url, parses it and gets the links. Returns: True, if a registration page is found, and False otherwise. """ if self.Download(): if not self._domain: url_parsed = urlparse.urlparse(self._url) self._domain = url_parsed[1] if self._domain.startswith('www'): self._domain = '.'.join(self._domain.split('.')[1:]) if self.ParseAndGetLinks(): return True return False class Crawler(object): """Crawls a site until a registration page is found or max level is reached. Creates, uses and destroys Retriever objects. Creates a cookie temp file needed for session cookies. It keeps track of 'visited links' and 'links to visit' of the site. To do this it uses the links discovered from each Retriever object. Use Run() to crawl the site. """ try: signal.signal(signal.SIGPIPE, signal.SIG_IGN) except ImportError: pass logger = logging.getLogger(__name__) def __init__(self, url, logging_level=None): """Init crawler URL, links lists, logger, and creates a cookie temp file. The cookie temp file is needed for session cookies. Args: url: the initial "seed" url of the site. logging_level: the desired verbosity level, default is None. """ if logging_level: self.logger.setLevel(logging_level) self.url_error = False url_parsed = urlparse.urlparse(url) if not url_parsed[0].startswith('http'): self.logger.error( 'Error: "%s" does not begin with http:// or https://', url) self.url_error = True return # Example: if url is 'http://www.example.com?name=john' then value [1] or # network location is 'www.example.com'. if not url_parsed[1]: self.logger.error('Error: "%s" is not a valid url', url) self.url_error = True return self._url = url self._domain = '' # Http links that contain a clue from LINK_CLUES. self._clues_general_links = [] # Http links that do not contain any clue from LINK_CLUES. self._general_links = [] # Https links that contain a clue from LINK_CLUES. self._clues_secure_links = [] # Https links that do not contain any clue from LINK_CLUES. self._secure_links = [] # All links downloaded and parsed so far. self._links_visited = [] self._retrievers_list = [] self._cookie_file = tempfile.NamedTemporaryFile( suffix='.cookie', delete=False) self._cookie_file.close() self._cookie_file = self._cookie_file.name # Keep only the filename. def __del__(self): """Deletes cookie file when Crawler instances are destroyed.""" if hasattr(self, '_cookie_file'): self.logger.info('Deleting cookie file %s ...', self._cookie_file) os.unlink(self._cookie_file) def _MultiPerform(self, curl_multi_object): """Performs concurrent downloads using a CurlMulti object. Args: curl_multi_object: a curl object that downloads multiple pages concurrently. The class of this object is |pycurl.CurlMulti|. """ # Following code uses the example from section for the CurlMulti object # at http://pycurl.sourceforge.net/doc/curlmultiobject.html. while True: ret, no_handles = curl_multi_object.perform() if ret != pycurl.E_CALL_MULTI_PERFORM: break while no_handles: curl_multi_object.select(1.0) while True: ret, no_handles = curl_multi_object.perform() if ret != pycurl.E_CALL_MULTI_PERFORM: break def _GetLinksPages(self, curl_multi_object): """Downloads many pages concurrently using a CurlMulti Object. Creates many Retriever objects and adds them to a list. The constant MAX_SAME_DOMAIN_URLS_NO defines the number of pages that can be downloaded concurrently from the same domain using the pycurl multi object. It's currently set to 30 URLs. These URLs are taken from the links lists, which are from csl, gcl, sl, and gl. The rules define how many URLs are taken from each list during each iteration. Example of the rules: 3/10 from csl results in 9 URLs 3/10 from cgl results in 9 URLs 2/10 from sl results in 6 URLs 2/10 from gl results in 6 URLs Adding up the above URLs gives 30 URLs that can be downloaded concurrently. If these lists have fewer items than the defined rules, such as if a site does not contain any secure links, then csl and sl lists will be of 0 length and only 15 pages would be downloaded concurrently from the same domain. Since 30 URLs can be handled concurrently, the number of links taken from other lists can be increased. This means that we can take 24 links from the cgl list so that 24 from gfl + 6 from gl = 30 URLs. If the cgl list has less than 24 links, e.g. there are only 21 links, then only 9 links may be taken from gl so ) + 21 + 0 + 9 = 30. Args: curl_multi_object: Each Retriever object has a curl object which is added to the CurlMulti Object. """ self._retrievers_list = [] csl_no = min(CLUE_SECURE_LINKS_NO, len(self._clues_secure_links)) cgl_no = min(CLUE_GENERAL_LINKS_NO, len(self._clues_general_links)) sl_no = min(SECURE_LINKS_NO, len(self._secure_links)) gl_no = min(GENERAL_LINKS_NO, len(self._general_links)) # If some links within the list have fewer items than needed, the missing # links will be taken by the following priority: csl, cgl, sl, gl. # c: clues, s: secure, g: general, l: list. spare_links = MAX_SAME_DOMAIN_URLS_NO - (csl_no + sl_no + cgl_no + gl_no) if spare_links > 0: csl_no = min(csl_no + spare_links, len(self._clues_secure_links)) spare_links = MAX_SAME_DOMAIN_URLS_NO - (csl_no + sl_no + cgl_no + gl_no) if spare_links > 0: cgl_no = min(cgl_no + spare_links, len(self._clues_general_links)) spare_links = MAX_SAME_DOMAIN_URLS_NO - (csl_no + sl_no + cgl_no + gl_no) if spare_links > 0: sl_no = min(sl_no + spare_links, len(self._secure_links)) spare_links = MAX_SAME_DOMAIN_URLS_NO - (csl_no + sl_no + cgl_no + gl_no) if spare_links > 0: gl_no = min(gl_no + spare_links, len(self._general_links)) for no_of_links, links in [ (csl_no, self._clues_secure_links), (sl_no, self._secure_links), (cgl_no, self._clues_general_links), (gl_no, self._general_links)]: for i in xrange(no_of_links): if not links: break url = links.pop(0) self._links_visited.append(url) r = Retriever(url, self._domain, self._cookie_file) r.InitRequestHead() curl_multi_object.add_handle(r._curl_object) self._retrievers_list.append(r) if self._retrievers_list: try: self._MultiPerform(curl_multi_object) except pycurl.error as e: self.logger.error('Error: %s, url: %s', e, self._url) finally: for r in self._retrievers_list: curl_multi_object.remove_handle(r._curl_object) # |_retrievers_list[:]| is a copy of |_retrievers_list| to avoid removing # items from the iterated list. for r in self._retrievers_list[:]: r._url = urlparse.urljoin(r._url, r._curl_object.getinfo( pycurl.EFFECTIVE_URL)) content_type = r._curl_object.getinfo(pycurl.CONTENT_TYPE) if content_type and ('text/html' in content_type.lower()): r.InitRequestGet() curl_multi_object.add_handle(r._curl_object) else: self._retrievers_list.remove(r) self.logger.info('\tSkipping: Not an HTML page <<< %s', r._url) if self._retrievers_list: try: self._MultiPerform(curl_multi_object) except pycurl.error as e: self.logger.error('Error: %s, url: %s', e, self._url) finally: for r in self._retrievers_list: curl_multi_object.remove_handle(r._curl_object) self.logger.info('Downloaded: %s', r._url) def _LogRegPageFound(self, retriever): """Display logging for registration page found. Args: retriever: The object that has retrieved the page. """ self.logger.info('\t##############################################') self.logger.info('\t### %s ###', retriever._domain) self.logger.info('\t##############################################') self.logger.info('\t!!!!!!!!! registration page FOUND !!!!!!!!!!!') self.logger.info('\t%s', retriever._url) self.logger.info('\t##############################################') def _GetNewLinks(self, retriever): """Appends new links discovered by each retriever to the appropriate lists. Links are copied to the links list of the crawler object, which holds all the links found from all retrievers that the crawler object created. The Crawler object exists as far as a specific site is examined and the Retriever object exists as far as a page of this site is examined. Args: retriever: a temporary object that downloads a specific page, parses the content and gets the page's href link. """ for link in retriever._clues_secure_links: if (not link in self._clues_secure_links and not link in self._links_visited): self._clues_secure_links.append(link) for link in retriever._secure_links: if (not link in self._secure_links and not link in self._links_visited): self._secure_links.append(link) for link in retriever._clues_general_links: if (not link in self._clues_general_links and not link in self._links_visited): self._clues_general_links.append(link) for link in retriever._general_links: if (not link in self._general_links and not link in self._links_visited): self._general_links.append(link) def Run(self): """Runs the Crawler. Creates a Retriever object and calls its run method to get the first links, and then uses CurlMulti object and creates many Retriever objects to get the subsequent pages. The number of pages (=Retriever objs) created each time is restricted by MAX_SAME_DOMAIN_URLS_NO. After this number of Retriever objects download and parse their pages, we do the same again. The number of total pages visited is kept in urls_visited. If no registration page is found, the Crawler object will give up its try after MAX_TOTAL_URLS_PER_DOMAIN is reached. Returns: True is returned if registration page is found, or False otherwise. """ reg_page_found = False if self.url_error: return False r = Retriever(self._url, self._domain, self._cookie_file) if r.Run(): self._LogRegPageFound(r) reg_page_found = True else: self._url = r._url self._domain = r._domain self.logger.info('url to crawl: %s', self._url) self.logger.info('domain: %s', self._domain) self._links_visited.append(r._url) self._GetNewLinks(r) urls_visited = 1 while True: if (not (self._clues_secure_links or self._secure_links or self._clues_general_links or self._general_links) or urls_visited >= MAX_TOTAL_URLS_PER_DOMAIN): break # Registration page not found. m = pycurl.CurlMulti() self._GetLinksPages(m) urls_visited += len(self._retrievers_list) self.logger.info('\t<----- URLs visited for domain "%s": %d ----->', self._domain, urls_visited) for r in self._retrievers_list: if r.ParseAndGetLinks(): self._LogRegPageFound(r) reg_page_found = True break else: self.logger.info('parsed: %s', r._url) self._GetNewLinks(r) m.close() if reg_page_found: break while self._retrievers_list: r = self._retrievers_list.pop() return reg_page_found class WorkerThread(threading.Thread): """Creates a new thread of execution.""" def __init__(self, url): """Creates _url and page_found attri to populate urls_with_no_reg_page file. Used after thread's termination for the creation of a file with a list of the urls for which a registration page wasn't found. Args: url: will be used as an argument to create a Crawler object later. """ threading.Thread.__init__(self) self._url = url self.page_found = False def run(self): """Execution of thread creates a Crawler object and runs it. Caution: this function name should not be changed to 'Run' or any other names because it is overriding the 'run' method of the 'threading.Thread' class. Otherwise it will never be called. """ self.page_found = Crawler(self._url).Run() class ThreadedCrawler(object): """Calls the Run function of WorkerThread which creates & runs a Crawler obj. The crawler object runs concurrently, examining one site each. """ logger = logging.getLogger(__name__) def __init__(self, urls_file, logging_level=None): """Creates threaded Crawler objects. Args: urls_file: a text file containing a URL in each line. logging_level: verbosity level, default is None. Raises: IOError: If cannot find URLs from the list. """ if logging_level: self.logger.setLevel(logging_level) self._urls_list = [] f = open(urls_file) try: for url in f.readlines(): url = url.strip() if not urlparse.urlparse(url)[0].startswith('http'): self.logger.info( '%s: skipping this (does not begin with "http://")', url) continue self._urls_list.append(url) except IOError as e: self.logger.error('Error: %s', e) raise finally: f.close() if not self._urls_list: error_msg = 'No URLs were found.' self.logger.error('ERROR: %s', error_msg) raise IOError(error_msg) def Run(self): """Runs Crawler objects using python threads. Number of concurrent threads is restricted to MAX_ALLOWED_THREADS. Returns: The number of registration pages found. -1 if no URLs are given. Raises: OSError: When creating the same directory that already exists. """ if self._urls_list: allThreads = [] # originalNumThreads is the number of threads just before the # ThreadedCrawler starts creating new threads. As a standalone script it # will be 1. originalNumThreads = threading.active_count() for url in self._urls_list: self.logger.info('URL fed to a crawler thread: %s', url) t = WorkerThread(url) t.start() allThreads.append(t) while threading.active_count() >= ( MAX_ALLOWED_THREADS + originalNumThreads): time.sleep(.4) while threading.active_count() > originalNumThreads: time.sleep(.4) self.logger.info('----------------') self.logger.info('--- FINISHED ---') self.logger.info('----------------') urls_no = 0 urls_not_found_no = 0 not_file_name = os.path.join( REGISTER_PAGE_DIR, NOT_FOUND_REG_PAGE_SITES_FILENAME) not_file_dir = os.path.dirname(not_file_name) try: os.makedirs(not_file_dir) except OSError as e: if e.errno != errno.EEXIST: raise fnot = open(not_file_name, 'wb') try: for t in sorted(allThreads, key=lambda t: t._url): urls_no += 1 if not t.page_found: urls_not_found_no += 1 fnot.write('%s' % t._url) fnot.write(os.linesep) except IOError as e: self.logger.error('Error: %s', e) finally: fnot.close() self.logger.info('Total number of URLs given: %d\n', urls_no) self.logger.info( 'Registration pages found: %d\n', (urls_no - urls_not_found_no)) self.logger.info( 'URLs that did not return a registration page: %d\n', urls_not_found_no) return urls_no - urls_not_found_no else: self.logger.error('Error: no URLs were found.') return -1 def main(): usage = 'usage: %prog [options] single_url_or_urls_filename' parser = optparse.OptionParser(usage) parser.add_option( '-l', '--log_level', metavar='LOG_LEVEL', default='error', help='LOG_LEVEL: debug, info, warning or error [default: %default]') (options, args) = parser.parse_args() options.log_level = options.log_level.upper() if options.log_level not in ['DEBUG', 'INFO', 'WARNING', 'ERROR']: print 'Wrong log_level argument.' parser.print_help() return 1 options.log_level = getattr(logging, options.log_level) if len(args) != 1: parser.error('Wrong number of arguments.') logger = logging.getLogger(__name__) if options.log_level: console = logging.StreamHandler() logger.addHandler(console) logger.setLevel(options.log_level) arg_is_a_file = os.path.isfile(args[0]) if arg_is_a_file: CrawlerClass = ThreadedCrawler else: CrawlerClass = Crawler t0 = datetime.datetime.now() c = CrawlerClass(args[0], options.log_level) c.Run() if not arg_is_a_file and c.url_error: logger.error( 'ERROR: "%s" is neither a valid filename nor a valid URL' % args[0]) t1 = datetime.datetime.now() delta_t = t1 - t0 logger.info('Started at: %s\n', t0) logger.info('Ended at: %s\n', t1) logger.info('Total execution time: %s\n', delta_t) return 0 if __name__ == "__main__": sys.exit(main())