tt.py 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424
  1. """
  2. @author: luojunhui
  3. """
  4. import time
  5. import aiomysql
  6. import asyncio
  7. from static.config import db_article, db_video
  8. from applications.schedule import search_videos
  9. from applications.functions.log import logging
  10. from static.config import spider_coroutines
  11. class MatchTask1(object):
  12. """
  13. 定时执行任务
  14. """
  15. def __init__(self, mysql_client):
  16. """
  17. :param mysql_client:
  18. """
  19. self.mysql_client = mysql_client
  20. async def get_task(self):
  21. """
  22. 获取任务
  23. :return:
  24. """
  25. select_sql1 = f"""
  26. SELECT DISTINCT (content_id)
  27. FROM {db_article}
  28. WHERE content_status = 0 and process_times <= 5 and account_name = '万事如意一家子'
  29. ORDER BY request_time_stamp
  30. ASC
  31. LIMIT {spider_coroutines};
  32. """
  33. content_ids = await self.mysql_client.async_select(select_sql1)
  34. cil = []
  35. for content_id in content_ids:
  36. cil.append(content_id[0])
  37. content_ids_tuple = str(cil).replace("[", "(").replace("]", ")")
  38. if len(content_ids_tuple) > 3:
  39. select_sql = f"""
  40. SELECT trace_id, content_id, gh_id, article_title, article_text, content_status, process_times
  41. FROM {db_article}
  42. WHERE content_id in {content_ids_tuple} and process_times <= 5
  43. ORDER BY request_time_stamp
  44. ASC;
  45. """
  46. print(select_sql)
  47. task_list = await self.mysql_client.async_select(sql=select_sql)
  48. task_obj_list = [
  49. {
  50. "trace_id": item[0],
  51. "content_id": item[1],
  52. "gh_id": item[2],
  53. "title": item[3],
  54. "text": item[4],
  55. "content_status": item[5],
  56. "process_times": item[6]
  57. } for item in task_list
  58. ]
  59. logging(
  60. code="9001",
  61. info="本次任务获取到 {} 条视频".format(len(task_obj_list)),
  62. data=task_obj_list
  63. )
  64. return task_obj_list
  65. else:
  66. return []
  67. async def get_history_videos(self, content_id):
  68. """
  69. check whether the contents videos exists
  70. :param content_id:
  71. :return:
  72. """
  73. select_sql = f"""
  74. SELECT video_id
  75. FROM {db_video}
  76. where content_id = '{content_id}' and video_status = 1 order by request_time DESC;
  77. """
  78. content_videos = await self.mysql_client.async_select(select_sql)
  79. videos = [vid for vid in content_videos]
  80. if len(videos) >= 3:
  81. return videos
  82. else:
  83. return None
  84. async def judge_content_processing(self, content_id):
  85. """
  86. 判断该content_id是否在处理中
  87. :param content_id:
  88. :return:
  89. """
  90. select_sql = f"""
  91. SELECT trace_id, content_status
  92. FROM {db_article}
  93. WHERE content_id = '{content_id}'
  94. ORDER BY id DESC;
  95. """
  96. result = await self.mysql_client.async_select(select_sql)
  97. if result:
  98. for item in result:
  99. trace_id, content_status = item
  100. if content_status == 1:
  101. return False
  102. return True
  103. else:
  104. return True
  105. async def use_exists_contents_videos(self, video_id_list, params):
  106. """
  107. 使用已经存在的视频id
  108. :return:
  109. """
  110. trace_id = params['trace_id']
  111. content_id = params['content_id']
  112. select_sql = f"""
  113. SELECT kimi_title
  114. FROM {db_article}
  115. WHERE content_id = '{content_id}' and kimi_title is not null limit 1;
  116. """
  117. info = await self.mysql_client.async_select(sql=select_sql)
  118. kimi_title = info[0]
  119. update_sql = f"""
  120. UPDATE {db_article}
  121. SET
  122. kimi_title=%s,
  123. recall_video_id1=%s,
  124. recall_video_id2=%s,
  125. recall_video_id3=%s,
  126. content_status=%s,
  127. process_times = %s
  128. WHERE trace_id = %s
  129. """
  130. vid1, vid2, vid3 = video_id_list[0], video_id_list[1], video_id_list[2]
  131. await self.mysql_client.async_insert(
  132. sql=update_sql,
  133. params=(
  134. kimi_title,
  135. video_id_list[0],
  136. "NULL" if vid2 is None else vid2,
  137. "NULL" if vid3 is None else vid3,
  138. 2,
  139. int(params['process_times']) + 1,
  140. trace_id
  141. )
  142. )
  143. logging(
  144. code="9002",
  145. info="已从历史文章更新,文章id: {}".format(content_id),
  146. trace_id=trace_id
  147. )
  148. async def start_process(self, params):
  149. """
  150. 开始处理
  151. :param params:
  152. :return:
  153. """
  154. # 更新文章contentId为1, 说明该文章正在处理中
  155. update_sql = f"""
  156. UPDATE {db_article}
  157. SET
  158. content_status = %s
  159. WHERE
  160. trace_id = %s;
  161. """
  162. await self.mysql_client.async_insert(
  163. sql=update_sql,
  164. params=(
  165. 1, params['trace_id']
  166. )
  167. )
  168. try:
  169. video_count = await search_videos(
  170. params={
  171. "title": params['title'],
  172. "content": params['text'],
  173. "trace_id": params['trace_id'],
  174. "content_id": params['content_id']
  175. },
  176. trace_id=params['trace_id'],
  177. gh_id=params['gh_id'],
  178. mysql_client=self.mysql_client
  179. )
  180. select_sql = f"""
  181. SELECT video_id
  182. FROM {db_video}
  183. WHERE content_id = '{params['content_id']}'
  184. """
  185. result = await self.mysql_client.async_select(sql=select_sql)
  186. vid1, vid2, vid3 = result[0], result[1], result[2]
  187. if vid1 or vid2 or vid3:
  188. update_sql2 = f"""
  189. UPDATE {db_article}
  190. SET
  191. recall_video_id1 = %s,
  192. recall_video_id2 = %s,
  193. recall_video_id3 = %s,
  194. content_status = %s,
  195. process_times = %s
  196. WHERE trace_id = %s;
  197. """
  198. await self.mysql_client.async_insert(
  199. sql=update_sql2,
  200. params=(
  201. vid1 if vid1 else "NULL",
  202. vid2 if vid2 else "NULL",
  203. vid3 if vid3 else "NULL",
  204. 2,
  205. {int(params['process_times']) + 1},
  206. params['trace_id']
  207. )
  208. )
  209. logging(
  210. code="9008",
  211. info="视频搜索成功, 状态修改为2",
  212. trace_id=params['trace_id']
  213. )
  214. else:
  215. if int(params['process_times']) < 5:
  216. update_sql3 = f"""
  217. UPDATE {db_article}
  218. SET
  219. content_status = %s,
  220. process_times = %s
  221. WHERE trace_id = %s;
  222. """
  223. await self.mysql_client.async_insert(
  224. sql=update_sql3,
  225. params=(0, int(params['process_times']) + 1, params['trace_id'])
  226. )
  227. logging(
  228. code="9018",
  229. info="视频搜索失败,回退状态为0",
  230. trace_id=params['trace_id']
  231. )
  232. else:
  233. update_sql3 = f"""
  234. UPDATE {db_article}
  235. SET
  236. content_status = %s,
  237. process_times = %s
  238. WHERE trace_id = %s;
  239. """
  240. await self.mysql_client.async_insert(
  241. sql=update_sql3,
  242. params=(3, int(params['process_times']) + 1, params['trace_id'])
  243. )
  244. logging(
  245. code="9019",
  246. info="视频多次搜索失败,状态修改为3",
  247. trace_id=params['trace_id']
  248. )
  249. except Exception as e:
  250. if int(params['process_times']) < 5:
  251. logging(
  252. code="9018",
  253. info="{}异常错误:{}, 回退状态为0".format(params['trace_id'], e),
  254. trace_id=params['trace_id']
  255. )
  256. update_sql4 = f"""
  257. UPDATE {db_article}
  258. SET
  259. content_status = %s,
  260. process_times = %s
  261. WHERE trace_id = %s;
  262. """
  263. await self.mysql_client.async_insert(
  264. sql=update_sql4,
  265. params=(0, int(params['process_times']) + 1, params['trace_id'])
  266. )
  267. else:
  268. logging(
  269. code="9019",
  270. info="{}异常错误:{}, 状态修改为3".format(params['trace_id'], e),
  271. trace_id=params['trace_id']
  272. )
  273. update_sql4 = f"""
  274. UPDATE {db_article}
  275. SET
  276. content_status = %s,
  277. process_times = %s
  278. WHERE trace_id = %s;
  279. """
  280. await self.mysql_client.async_insert(
  281. sql=update_sql4,
  282. params=(3, int(params['process_times']) + 1, params['trace_id'])
  283. )
  284. async def process_task(self, params):
  285. """
  286. 异步执行
  287. :param params:
  288. :return:
  289. """
  290. content_id = params['content_id']
  291. trace_id = params['trace_id']
  292. # 判断该篇文章是否存在未下架的视频,且判断是否有3条, 如果没有三条,则启动新抓取任务,后续优化点
  293. video_id_list = await self.get_history_videos(content_id=content_id)
  294. if video_id_list:
  295. # 说明已经存在了结果, 将该条记录下的video_id拿出来
  296. logging(
  297. code="9001",
  298. info="存在历史文章",
  299. trace_id=trace_id
  300. )
  301. await self.use_exists_contents_videos(video_id_list=video_id_list, params=params)
  302. else:
  303. flag = await self.judge_content_processing(content_id)
  304. if flag:
  305. logging(
  306. code="9004",
  307. info="无正在处理的文章ID, 开始处理",
  308. trace_id=trace_id
  309. )
  310. await self.start_process(params=params)
  311. else:
  312. logging(
  313. code="9003",
  314. info="该文章ID正在请求--文章ID {}".format(content_id),
  315. trace_id=trace_id
  316. )
  317. async def deal(self):
  318. """
  319. 处理
  320. :return:
  321. """
  322. task_list = await self.get_task()
  323. task_dict = {}
  324. for task in task_list:
  325. key = task['content_id']
  326. task_dict[key] = task
  327. process_list = []
  328. for item in task_dict:
  329. process_list.append(task_dict[item])
  330. if process_list:
  331. # for task in task_list:
  332. # await self.process_task(task)
  333. tasks = [self.process_task(params) for params in process_list]
  334. await asyncio.gather(*tasks)
  335. else:
  336. logging(
  337. code="9008",
  338. info="没有要处理的请求"
  339. )
  340. class TaskMySQLClient(object):
  341. """
  342. Async MySQL
  343. """
  344. def __init__(self):
  345. self.mysql_pool = None
  346. async def init_pool(self):
  347. """
  348. 初始化连接
  349. :return:
  350. """
  351. self.mysql_pool = await aiomysql.create_pool(
  352. host='rm-bp1159bu17li9hi94.mysql.rds.aliyuncs.com',
  353. port=3306,
  354. user='crawler',
  355. password='crawler123456@',
  356. db='piaoquan-crawler',
  357. charset='utf8mb4',
  358. connect_timeout=120,
  359. )
  360. print("mysql init successfully")
  361. async def close_pool(self):
  362. """
  363. 关闭 mysql 连接
  364. :return:
  365. """
  366. self.mysql_pool.close()
  367. await self.mysql_pool.wait_closed()
  368. async def async_select(self, sql):
  369. """
  370. select method
  371. :param sql:
  372. :return:
  373. """
  374. async with self.mysql_pool.acquire() as conn:
  375. async with conn.cursor() as cursor:
  376. await cursor.execute(sql)
  377. result = await cursor.fetchall()
  378. return result
  379. async def async_insert(self, sql, params):
  380. """
  381. insert and update method
  382. :param params:
  383. :param sql:
  384. :return:
  385. """
  386. async with self.mysql_pool.acquire() as coon:
  387. async with coon.cursor() as cursor:
  388. await cursor.execute(sql, params)
  389. await coon.commit()
  390. async def main():
  391. """
  392. main job
  393. :return:
  394. """
  395. TMC = TaskMySQLClient()
  396. await TMC.init_pool()
  397. PD = MatchTask1(TMC)
  398. await PD.deal()
  399. if __name__ == '__main__':
  400. while True:
  401. asyncio.run(main())
  402. time.sleep(10)