def crawl(self, args, browser):
source = args['source']
list_page_url = args['list_page_url']
try:
appstore_modual = __import__(source)
except ImportError:
util.write_log_error(self.lock, self.logger, "can not find " + source + ".py in configured appstores")
return False
appstore_class = appstore_modual.Store()
if self.url_rd.exists(util.get_MD5(list_page_url)):
util.write_log_warn(self.lock, self.logger, "list_page_url:" + list_page_url + "was crawled in past %d seconds"%(config.CRAWLED_URL_REDIS_TIMEOUT))
return False
browser.desired_capabilities["phantomjs.page.settings.userAgent"] = appstore_class.PHANTOMJS_USER_AGENT
util.browser_get_url(browser, list_page_url)
# appstore_class.scroll_down(browser)
check_more_count = appstore_class.click_checkmore(browser)
util.write_log_warn(self.lock, self.logger, "列表页: " + list_page_url + "点击checkmore" + str(check_more_count) + "次后找不到checkmore")
detail_urls = appstore_class.get_detail_urls(browser)
if len(detail_urls) == 0:
util.write_log_warn(self.lock, self.logger, "列表页: " + list_page_url + " 找到0个app")
self.push_detail_page_message(args, detail_urls)
next_list_page = appstore_class.click_nextpage(browser)
if next_list_page:
self.push_list_page_message(args, next_list_page)
util.write_log_info(self.lock, self.logger,"SUCCESS : list_page_url: %s crawled"%(list_page_url))
self.url_rd.set(util.get_MD5(list_page_url),1)
self.url_rd.expire(util.get_MD5(list_page_url), config.CRAWLED_URL_REDIS_TIMEOUT)
return True
def work_process(self, browser):
util.write_log_info(self.lock, self.logger, 'new detail_page_handler process started')
while 1:
try:
message = self.rd.brpop([config.HIGH_DETAIL_PAGE_HANDLE_QUEUE, config.LOW_DETAIL_PAGE_HANDLE_QUEUE],0)
util.write_log_info(self.lock, self.logger, 'received new nessage: %s'%(message[1]))
self.do_work(message[1], browser)
except Exception:
import traceback
util.write_log_error(self.lock, self.logger,"detail_page_handler process catched exception." + traceback.print_exc())
def work_process(self):
util.write_log_info(self.lock, self.logger, 'new tmpfile_hander process started')
while 1:
try:
message = self.rd.brpop([config.HIGH_FILE_HANDLE_QUEUE, config.LOW_FILE_HANDLE_QUEUE],0)
util.write_log_info(self.lock, self.logger, 'received new nessage: %s'%(message[1]))
self.do_work(message[1])
except Exception:
import traceback
util.write_log_error(self.lock, self.logger,"tmpfile_hander process catched exception." + traceback.print_exc())
def work_thread(self):
util.write_log_info(self.lock, self.logger, "new downloader thread started")
while 1:
try:
message = self.rd.brpop([config.HIGH_DOWNLOAD_QUEUE, config.LOW_DOWNLOAD_QUEUE], 0)
util.write_log_info(self.lock, self.logger, "received new nessage: %s" % (message[1]))
self.do_work(message[1])
except Exception:
import traceback
util.write_log_error(
self.lock, self.logger, "downloader thread catched exception." + traceback.print_exc()
)
def normal(self, args):
session = self.prepare_download_session(args["download_cookie_producer"], args["request_headers"])
download_result, download_redirect_urls = self.download(session, args)
if download_result == "success":
util.write_log_info(
self.lock,
self.logger,
"SUCCESS : download_url: %s , file_name: %s " % (args["download_url"], args["file_name"]),
)
download_result = True
else:
util.write_log_error(
self.lock, self.logger, "FAIL : download_url: %s ; reason: %s" % (args["download_url"], download_result)
)
download_result = False
return download_result, download_redirect_urls
def main(argv):
logger = None
if config.LOGGER_ENABLED:
logger = util.initLog(config.LOGGER_NAME_FILE_HANDLER)
multiprocessing.current_process().name = 'tmpfile_hander_process_main'
tmpfile_hander = Tmpfile_hander(logger = logger, host = config.REDIS_HOST, port = config.REDIS_PORT)
util.write_log_info(tmpfile_hander.lock, tmpfile_hander.logger,"[+]start tmpfile_hander...")
tmpfile_hander.launch()
try:
content=raw_input()
while content != 'exit':
content=raw_input()
except KeyboardInterrupt:
pass
util.write_log_info(tmpfile_hander.lock, tmpfile_hander.logger,"user determined tmpfile_hander")
tmpfile_hander.quit()
def work_thread(self):
util.write_log_info(self.lock, self.logger,
'new downloader thread started')
while 1:
try:
message = self.rd.brpop(
[config.HIGH_DOWNLOAD_QUEUE, config.LOW_DOWNLOAD_QUEUE], 0)
util.write_log_info(self.lock, self.logger,
'received new nessage: %s' % (message[1]))
self.do_work(message[1])
except Exception:
import traceback
util.write_log_error(
self.lock, self.logger,
'downloader thread catched exception.' +
traceback.print_exc())
def work_process(self):
util.write_log_info(self.lock, self.logger,
'new tmpfile_hander process started')
while 1:
try:
message = self.rd.brpop([
config.HIGH_FILE_HANDLE_QUEUE, config.LOW_FILE_HANDLE_QUEUE
], 0)
util.write_log_info(self.lock, self.logger,
'received new nessage: %s' % (message[1]))
self.do_work(message[1])
except Exception:
import traceback
util.write_log_error(
self.lock, self.logger,
"tmpfile_hander process catched exception." +
traceback.print_exc())
def normal(self, args):
session = self.prepare_download_session(
args['download_cookie_producer'], args['request_headers'])
download_result, download_redirect_urls = self.download(session, args)
if download_result == 'success':
util.write_log_info(
self.lock, self.logger,
"SUCCESS : download_url: %s , file_name: %s " %
(args['download_url'], args['file_name']))
download_result = True
else:
util.write_log_error(
self.lock, self.logger,
"FAIL : download_url: %s ; reason: %s" %
(args['download_url'], download_result))
download_result = False
return download_result, download_redirect_urls
def main(argv):
import socket
socket.setdefaulttimeout(config.SOCKET_TIMEOUT)
logger = None
if config.LOGGER_ENABLED:
logger = util.initLog(config.LOGGER_NAME_DETAIL_PAGE_HANDLER)
multiprocessing.current_process().name = 'detail_page_handler_process_main'
detail_page_handler = Detail_page_handler(logger = logger, host = config.REDIS_HOST, port = config.REDIS_PORT)
util.write_log_info(detail_page_handler.lock, detail_page_handler.logger,"[+]start detail_page_handler...")
detail_page_handler.launch()
try:
content=raw_input()
while content != 'exit':
content=raw_input()
except KeyboardInterrupt:
pass
util.write_log_info(detail_page_handler.lock, detail_page_handler.logger,"user determined detail_page_handler")
detail_page_handler.quit()
def main(argv):
import socket
socket.setdefaulttimeout(config.SOCKET_TIMEOUT)
logger = None
if config.LOGGER_ENABLED:
logger = util.initLog(config.LOGGER_NAME_DOWNLOADER)
threading.current_thread().setName("downloader_thread_main")
downloader = Downloader(logger=logger, host=config.REDIS_HOST, port=config.REDIS_PORT)
util.write_log_info(downloader.lock, downloader.logger, "[+]start downloader...")
downloader.launch()
try:
content = raw_input()
while content != "exit":
content = raw_input()
except KeyboardInterrupt:
pass
util.write_log_info(downloader.lock, downloader.logger, "user determined downloader")
downloader.quit()
def main(argv):
logger = None
if config.LOGGER_ENABLED:
logger = util.initLog(config.LOGGER_NAME_FILE_HANDLER)
multiprocessing.current_process().name = 'tmpfile_hander_process_main'
tmpfile_hander = Tmpfile_hander(logger=logger,
host=config.REDIS_HOST,
port=config.REDIS_PORT)
util.write_log_info(tmpfile_hander.lock, tmpfile_hander.logger,
"[+]start tmpfile_hander...")
tmpfile_hander.launch()
try:
content = raw_input()
while content != 'exit':
content = raw_input()
except KeyboardInterrupt:
pass
util.write_log_info(tmpfile_hander.lock, tmpfile_hander.logger,
"user determined tmpfile_hander")
tmpfile_hander.quit()
def crawl(self, args):
file_name = args['file_name']
target_folder = args['target_folder']
apk_info = self.prepare_apk_info(args)
util.makeDirs(self.lock, target_folder)
if not util.verify_apk(file_name):
util.write_log_warn(self.lock, self.logger, "FAILED : %s jarsigner faild."%(file_name))
os.remove(file_name)
return False
md5 = util.get_file_MD5(file_name)
apk_info['MD5'] = md5
target_name = target_folder + 'cell-' + md5[:2] + '/' + md5 + '.apk'
if os.path.exists(target_name):
os.remove(file_name)
else:
apk_info['new'] = 'True'
os.rename(args['file_name'], target_name)
util.writeApkInfoFile(self.lock, target_folder,apk_info)
util.writeRecord(self.lock, target_folder,apk_info)
util.write_log_info(self.lock, self.logger, "SUCCESS : %s handle success"%(args['file_name']))
return True
def main(argv):
import socket
socket.setdefaulttimeout(config.SOCKET_TIMEOUT)
logger = None
if config.LOGGER_ENABLED:
logger = util.initLog(config.LOGGER_NAME_DOWNLOADER)
threading.current_thread().setName('downloader_thread_main')
downloader = Downloader(logger=logger,
host=config.REDIS_HOST,
port=config.REDIS_PORT)
util.write_log_info(downloader.lock, downloader.logger,
"[+]start downloader...")
downloader.launch()
try:
content = raw_input()
while content != 'exit':
content = raw_input()
except KeyboardInterrupt:
pass
util.write_log_info(downloader.lock, downloader.logger,
"user determined downloader")
downloader.quit()
def crawl(self, args):
file_name = args['file_name']
target_folder = args['target_folder']
apk_info = self.prepare_apk_info(args)
util.makeDirs(self.lock, target_folder)
if not util.verify_apk(file_name):
util.write_log_warn(self.lock, self.logger,
"FAILED : %s jarsigner faild." % (file_name))
os.remove(file_name)
return False
md5 = util.get_file_MD5(file_name)
apk_info['MD5'] = md5
target_name = target_folder + 'cell-' + md5[:2] + '/' + md5 + '.apk'
if os.path.exists(target_name):
os.remove(file_name)
else:
apk_info['new'] = 'True'
os.rename(args['file_name'], target_name)
util.writeApkInfoFile(self.lock, target_folder, apk_info)
util.writeRecord(self.lock, target_folder, apk_info)
util.write_log_info(
self.lock, self.logger,
"SUCCESS : %s handle success" % (args['file_name']))
return True
browser.desired_capabilities["phantomjs.page.settings.userAgent"] = appstore_class.PHANTOMJS_USER_AGENT
util.browser_get_url(browser, detail_page_url)
download_url = appstore_class.get_download_url(browser)
if download_url is None:
util.write_log_error(self.lock, self.logger, "详情页: " + detail_page_url + " 找不到download_url")
return False
out_json_message = appstore_class.make_detail2download_json_message(browser)
out_json_message['detail_url'] = detail_page_url
out_json_message['download_url'] = download_url
if args['level'] = 'high':
out_json_message['next_queue'] = config.HIGH_DOWNLOAD_QUEUE
else:
out_json_message['next_queue'] = config.LOW_DOWNLOAD_QUEUE
out_json_message['level'] = args['level']
out_json_message['message_type'] = args['message_type']
util.write_log_info(self.lock, self.logger,"SUCCESS : detail_page_url: %s crawled"%(detail_page_url))
self.url_rd.set(util.get_MD5(detail_page_url),1)
self.url_rd.expire(util.get_MD5(detail_page_url), config.CRAWLED_URL_REDIS_TIMEOUT)
return self.push_message(out_json_message)
def normal(self, args, browser):
util.write_log_warn(self.lock, self.logger, '"normal" message type did nothing.')
return True
def do_work(self, message, browser):
util.random_sleep()
args = self.prepare_args(message)
if args is None:
return
return eval('self.'+args['message_type'])(args, browser)