def check_data_base(self):
"""数据库结构检查"""
list = [
TBNews(),
TBArticle(),
TBProgram(),
TBWenzhangInfo(),
TBDictionaryType(),
TBDictionary(),
TBNewsGroup(),
TBNlpFilter(),
TBScrapingTarget(),
TBSoap(),
TBSoapTarget(),
TBGlobalTarget(),
TBMR(),
TBSpecialTarget(),
TBProgramType(),
TBSoapBlackList(),
TBHeavyText()
]
# list = [TBSoapBlackList()]
guard = StructureGuard(Configs())
log = guard.check(list)
if len(log) > 0:
LogGo.info(str(log))
def send_to_queue(result):
"""
传送
code: 0 到达最大访问频率
1 正常结果
:param request:
:param result:
:return:
"""
global target_mutex, target_count, target_transported_count, all_target_transported
code, value = result
target, detail_page_bundle_list, content_ruler, encode = value
if code == 1:
for detail_page_bundle in detail_page_bundle_list:
target_producer.target_queue.queue.put(
(target, detail_page_bundle, content_ruler, encode))
ScrabingTarget.set_last_access_date(target.id)
else:
LogGo.error("List Page Error:" + str(target.data_key) + " Code: " +
str(code))
ScrabingTarget.set_elog(target.id, "error code: " + str(code))
if target_mutex.acquire():
if target_count == target_transported_count:
all_target_transported = True
else:
target_transported_count += 1
LogGo.debug('target_transported_count: ' +
str(target_transported_count))
target_mutex.release()
def scrape_detail(self, target, detail_page_bundle, content_ruler, encode):
try:
order = self.web_order
ulweb = UlWebRuler()
result = ulweb.scan_detail(target, detail_page_bundle, order,
content_ruler, encode)
code, detail_page_result = result
if code == 1:
if detail_page_result is not None:
return (1, detail_page_result)
elif code == -3:
self.temp_list.append(
(target, detail_page_bundle, content_ruler, encode))
except WebTargetOutOfDateException as e:
return (-1, e.args[0])
except Exception as e:
import traceback
LogGo.warning(repr(e))
return (-1, e.args[0])
return (0, None)
def scan_list(self, target, exists):
list = []
result_list = []
cap = 'data'
ruler = 'author:author;title:title;date:posttime;img:picurl;link:url;top:top;click:readnum_newest;vote_up:likenum_newest;subject:content'
url = self.url.format(target.extra0, target.wx_hao)
header = {'X-Requested-With': 'XMLHttpRequest'}
raw = RequestHelper.get(url, header=header, file_cookie=Configs.gsdata_cookie_file)
try:
self.looper_js(list, raw, exists, ruler, cap)
except Exception as e:
E.out_err(e)
return (-1, (target, None, None, None))
if len(list) > 0:
list = self.sort(list)
list.reverse()
for item in list:
if exists.count(item['title']) < 1:
result_list.append(item)
LogGo.debug('newrank list length:' + str(len(result_list)))
if len(result_list) > 0:
return (1, (target, list, None, None))
return(-1, (target, None, None, None))
def reset_shutdown_status():
try:
file_name = Configs().system_shutdown_flag_filename
FileHelper.create(file_name, "0")
except:
LogGo.error("error while reseting shutdown flag file!!!")
def start_mormal_mission(self):
global all_target_transported
target_list = get_target_list()
self.target_producer = target_producer(target_list,
self.config.target_pool_size,
self.config.target_queue_size)
self.target_consumer = target_consumer()
self.upload_consumer = upload_consumer(self.config.uploader_queue_size)
self.upload_consumer.start()
self.target_consumer.start()
self.target_producer.start()
# self.target_producer.pool.close()
# self.target_producer.pool.join()
while True: # LogGo.debug(">>> target queue unfinishd count: " + str(self.target_producer.target_queue.queue.unfinished_tasks))
time.sleep(5)
LogGo.debug("target_transported_over: " +
str(target_producer.is_all_target_transported()))
# self.target_consumer.queue.queue.join()
# time.sleep(6000)
LogGo.info('Loop Done! task count: ' + str(len(target_list)))
SMTPServer.launch_mission_report()
def newrank_detail(self, target, detail_page_bundle, content_ruler,
encode):
LogGo.info("On newrank detail: " + str(target.data_key))
try:
order = self.wechat_order
newrank = NewrankRuler()
result = newrank.scan_detail(target, detail_page_bundle, order,
content_ruler, encode)
code, detail_page_result = result
if code == 1:
if detail_page_result is not None:
return (1, detail_page_result)
elif code == -3:
self.temp_list.append(
(target, detail_page_bundle, content_ruler, encode))
except Exception:
LogGo.warning('error in newrank detail')
return (-1, None)
return (0, None)
def sogou_transfor(self):
LogGo.info('搜狗转移')
result = 0
# sogou = SougouTransforRuler()
#
# existsUrls = self.news.get_all_title()
# order = self.news.get_max_order_code('wechat')
#
# news, article = sogou.ExtraList(existsUrls, order)
#
# content = '搜狗转移任务(' + ')' + '\r\n'
# content += '此次采集数量: ' + str(len(news)) + '\r\n'
#
# if self.store(news, article):
# result = 1
# else:
# result = -1
#
# if result == 1:
# content += '存储成功!'
#
# SMTPServer.build_mission_report(content)
# print(content)
return result
def store_soap_target(self, targets: [], banned: bool = False):
status = False
element = 0
for target in targets:
try:
if banned:
self.banned_program.save(target)
else:
self.soap_target.save(target)
element += 1
except Exception as e:
import traceback
msg = traceback.format_exc()
LogGo.warning(msg)
LogGo.info('Total :' + str(element) + ' / ' + str(len(targets)) +
' elements Saved!')
if element == 0:
status = False
else:
status = True
return status
def gs_detail(self, target, detail_page_bundle, content_ruler, encode):
LogGo.info("On gs detail: " + str(target.data_key))
try:
order = self.wechat_order
gs = GsdataRuler()
result = gs.scan_detail(target, detail_page_bundle, order,
content_ruler, encode)
code, detail_page_result = result
if code == 1:
if detail_page_result is not None:
return (1, detail_page_result)
elif code == 3:
self.temp_list.append(
(target, detail_page_bundle, content_ruler, encode))
except Exception as e:
import traceback
LogGo.warning(repr(e))
return (-1, e.args[0])
return (0, None)
def scan_list(self, target, exists):
self.limited_forward_count = target.limited_forward_count
self.limited_attitude_count = target.limited_attitude_count
list = []
result_list = []
"""模拟登陆"""
status = 'you got it'
"""如果登陆成功"""
if status != '':
self.loops(target,exists,list)
if len(list) < 1:
return (0, (target, None, None, None))
else:
LogGo.warning("Weibo: Loop scan faild!")
return (-1, (target, None, None, None))
if len(list) > 0:
list = self.purify(list)
list.reverse()
for item in list:
if exists.count(item['id']) < 1:
result_list.append(item)
LogGo.debug('newrank list length:' + str(len(result_list)))
if len(result_list) > 0:
return (1, (target, list, None, None))
return(-1, (target, None, None, None))
def set_shutdown_status():
try:
file_name = Configs().system_shutdown_flag_filename
FileHelper.create(file_name, "1")
except:
LogGo.error("something wrong at setting shutdown flag file !!!")
def purify(self,list):
if len(list) < 1:
return []
result = []
for i in list:
try:
flag = True
id = i['id']
text = i['text']
limited_attitude_count = i['attitudes_count']
limited_forward_count = i['reposts_count']
program_count = 0
# 基础过滤重复id
for seq in result[::-1]:
sid = seq['id']
if id == sid:
flag = False
break
# 第一次节目名过滤（有可能会包含到非节目）
if flag and text.count('《') < 1:
flag = False
# 条二次参数过滤
if flag and self.limited_attitude_count is not None and limited_attitude_count is not None:
if limited_attitude_count < self.limited_attitude_count:
flag = False
if flag and self.limited_forward_count is not None and limited_forward_count is not None:
if limited_forward_count < self.limited_forward_count:
flag = False
# 第三次依据节目名过滤
if flag:
for program in self.exist_program:
if text.count(program) >= 1:
program_count = program_count + 1
if program_count > 3:
flag = False
break
if flag:
result.append(i)
except Exception as e:
import traceback
msg = traceback.format_exc()
LogGo.warning(msg)
return result
def base_init():
LogGo.init(Configs())
RequestHelper.init(Configs())
SMTPServer.init(Configs())
Download(Configs())
RequestHelperClassVer.init(Configs())
ProxyHelper.init(Configs())
MysqlHelper.init(Configs())
BaseStrategy.init()
def ExtraList(self, target, existsUrls, order):
# order = ScrappdeDataDao.get_max_order_code() # 数据库中排序代码
result = []
url = str(target.extra0)
next_index = ""
"""抓取地址"""
raw = WechatRuler.req._get(url)
try:
trup = ExtraJSON.extraWechatList(raw, 'msgList', WechatRuler.keys)
list = trup[0]
next_index = str(trup[1])
except Exception as e:
print(e)
print("ERROR")
return result
while True:
try:
print('>>> scaning id: ' + next_index)
LogGo.info('>>> scaning id: ' + next_index)
tup = self.loopToFail(url, next_index)
re_list = tup[0]
next_index = str(tup[1])
is_continue = tup[2]
if len(re_list) > 0:
for item in re_list:
list.append(item)
# break
else:
break
if is_continue != 1:
break
except Exception as e:
print(e)
break
print('>>> list scaning completed')
print('>>>')
list.reverse()
print('>>> Start Build SQL')
result = self.build_base_dic(target,list,existsUrls,order)
print('>>> Build SQL Success')
print('>>>')
return result
def start(self):
"""
开启线程池，把所有 Target 放入池中，单独获取抓取列表
:return:
"""
while True:
LogGo.info("Start target pool")
[
self.pool.apply_async(target_producer.worker, (target, ))
for target in self.targets
]
time.sleep(Configs().work_interval)
def check_shutdown_status():
try:
file_name = Configs().system_shutdown_flag_filename
status = int(StringHelper.trim(FileHelper.read(file_name)))
if status == 1:
somebody_help.reset_shutdown_status()
return True
else:
return False
except:
LogGo.error("system_shutdown_flag_file unavailable!")
return False
def store_program(self, ids, programs):
status = False
if len(ids) != len(programs):
LogGo.error("ids count unmatch programs count")
return False
count = len(ids)
element = 0
if count > 0:
for id, program in zip(ids, programs):
try:
self.program.save(program, id)
element += 1
except Exception as e:
import traceback
msg = traceback.format_exc()
LogGo.warning(msg)
LogGo.info('Total :' + str(count) + ' / ' + str(element) +
' elements Saved!')
if element == 0:
status = False
else:
status = True
else:
LogGo.info("0 Element!")
return status
def start(self):
try:
if self.config.check_table:
self.check_data_base()
# RequestHelperClassVer.init(self.config)
# ProxyHelper.init(self.config)
# MysqlHelper.init(self.config)
self.start_mormal_mission()
except Exception as e:
import traceback
msg = traceback.format_exc()
LogGo.warning(msg)
def newrank_list(self, target):
LogGo.info("On newrank list: " + str(target.data_key))
try:
exists = self.exists_title
newrank = NewrankRuler()
code, value = newrank.scan_list(target, exists)
return (code, value)
except Exception as e:
import traceback
LogGo.warning(repr(e))
return (0, None)
def gs_list(self, target):
LogGo.info("On gs list: " + str(target.data_key))
try:
exists = self.exists_signature
gs = GsdataRuler()
code, value = gs.scan_list(target, exists)
return (code, value)
except Exception as e:
import traceback
LogGo.warning(repr(e))
return (0, None)
def weibo_list(self, target):
LogGo.info("On weibo list: " + str(target.data_key))
try:
exists = self.exists_identifier
weibo = WeiboRuler()
code, value = weibo.scan_list(target, exists)
return (code, value)
except Exception as e:
import traceback
LogGo.warning(repr(e))
return (0, None)
def store(self, news_list: [], article_list: [], heavy_list: [] = None):
status = False
if len(article_list) != len(news_list):
LogGo.error("news count unmatch article count")
return False
count = len(news_list)
element = 0
if count > 0:
if heavy_list is not None:
for news, article, heavy in zip(news_list, article_list,
heavy_list):
try:
id = genUUID()
self.news.save(news, id)
self.article.save(article, id)
self.heavy.save_with_news_id(heavy, id)
element += 1
except Exception as e:
LogGo.warning(repr(e))
else:
for news, article in zip(news_list, article_list):
try:
id = genUUID()
self.news.save(news, id)
self.article.save(article, id)
element += 1
except Exception as e:
LogGo.warning(repr(e))
LogGo.info('Total :' + str(count) + ' / ' + str(element) +
' elements Saved!')
if element == 0:
status = False
else:
status = True
# return True
else:
LogGo.info("0 Element!")
# return False
return status
def start(self):
result = 0
try:
LogGo.info('搜狗爬虫')
# Updatemp.loot()
return result
except Exception as e:
import traceback
msg = traceback.format_exc()
print(msg)
LogGo.warning(e)
result = -1
return result
def weibo_detail(self, target, detail_page_bundle, content_ruler, encode):
LogGo.info("On weibo detail: " + str(target.data_key))
try:
order = self.weibo_order
weibo = WeiboRuler()
detail_page_result_dic = weibo.scan_detail(target,
detail_page_bundle,
order, content_ruler,
encode)
if detail_page_result_dic is not None:
return (1, detail_page_result_dic)
except Exception as e:
import traceback
LogGo.warning(repr(e))
return (-1, e.args[0])
return (0, None)
def scan_list(self, target, exists):
"""请求参数"""
par = (['flag', 'true'], ['uuid', target.extra0])
"""抓取关键字"""
keys = [
'title', 'author', 'publicTime', 'url', 'clicksCount', 'likeCount',
'publicTime', 'summary'
]
list = []
result_list = []
try:
raw = RequestHelper.post(NewrankRuler.url,
par,
file_cookie=Configs.newrank_cookie_file)
except Exception as e:
import traceback
msg = traceback.format_exc()
# print(msg)
LogGo.warning(msg)
return (-1, (target, None, None, None))
try:
list = ExtraJSON.extra_newrank_wechat_list(raw, keys)
except:
return (-1, (target, None, None, None))
if len(list) > 0:
list.reverse()
for item in list:
if exists.count(item['title']) < 1:
result_list.append(item)
LogGo.debug('newrank list length:' + str(len(result_list)))
if len(result_list) > 0:
return (1, (target, list, None, None))
return (-1, (target, None, None, None))
def store_program_type(self, pair):
status = False
count = len(pair)
element = 0
if count > 0:
for item in pair:
try:
program, type = item
self.program_type.save_by_program_type(program, type)
element += 1
except Exception as e:
import traceback
msg = traceback.format_exc()
LogGo.warning(msg)
LogGo.info('Total :' + str(count) + ' / ' + str(element) +
' elements Saved!')
if element == 0:
status = False
else:
status = True
else:
LogGo.info("0 Element!")
return status
def store_count(self, list):
status = False
count = len(list)
element = 0
update = 0
if count > 0:
for program in list:
try:
if self.check_for_exists(program) == 1:
update += 1
else:
id = genUUID()
self.pc.save(program, id)
element += 1
except Exception as e:
import traceback
msg = traceback.format_exc()
LogGo.warning(msg)
LogGo.info('Total :' + str(count) + ' (' + str(element) +
' Saved, ' + str(update) + ' Updated)')
if element == 0 and update == 0:
status = False
else:
status = True
else:
LogGo.info("0 Element!")
# return False
return status
def gensto(self, dao, list):
status = False
count = len(list)
element = 0
if count > 0:
for item in list:
try:
id = genUUID()
dao.save_or_update(item, id)
element += 1
except Exception as e:
import traceback
msg = traceback.format_exc()
# print(msg)
LogGo.warning(repr(e))
LogGo.info('Total :' + str(count) + ' / ' + str(element) +
' elements Saved!')
if element == 0:
status = False
else:
status = True
else:
LogGo.info("0 Element!")
stauts = False
return status
def genup(self, dao, list):
status = False
count = len(list)
element = 0
if count > 0:
for (update, where) in list:
try:
dao.update(update, where)
element += 1
except Exception as e:
import traceback
msg = traceback.format_exc()
# print(msg)
LogGo.warning(repr(e))
LogGo.info('Total :' + str(count) + ' / ' + str(element) +
' elements Updated!')
if element == 0:
status = False
else:
status = True
else:
LogGo.info("0 Element!")
stauts = False
return status
