task1.py 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356
  1. """
  2. @author: luojunhui
  3. """
  4. import asyncio
  5. from static.config import db_article, db_video
  6. from applications.schedule import search_videos
  7. from applications.functions.log import logging
  8. from static.config import spider_coroutines
  9. # Temporary solution for task dead-lock
  10. g_values = {'row_offset': 0}
  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_limit = spider_coroutines * 5
  26. select_sql1 = f"""
  27. SELECT content_id, id
  28. FROM {db_article}
  29. WHERE content_status = 0 and process_times <= 3
  30. AND id >= {g_values['row_offset']}
  31. ORDER BY id
  32. LIMIT {select_limit};
  33. """
  34. content_ids = await self.mysql_client.async_select(select_sql1)
  35. unique_content_ids = set()
  36. for content_id in content_ids:
  37. unique_content_ids.add(content_id[0])
  38. if not unique_content_ids:
  39. return []
  40. g_values['row_offset'] = content_ids[-1][1]
  41. print(f"update row offset to: {g_values['row_offset']}")
  42. unique_content_ids = list(unique_content_ids)[0:spider_coroutines]
  43. content_ids_tuple = str(unique_content_ids).replace("[", "(").replace("]", ")")
  44. if len(content_ids_tuple) > 3:
  45. select_sql = f"""
  46. SELECT trace_id, content_id, gh_id, article_title, article_text, content_status, process_times
  47. FROM {db_article}
  48. WHERE content_id in {content_ids_tuple} and process_times <= 3
  49. ORDER BY request_time_stamp
  50. ASC;
  51. """
  52. task_list = await self.mysql_client.async_select(sql=select_sql)
  53. task_obj_list = [
  54. {
  55. "trace_id": item[0],
  56. "content_id": item[1],
  57. "gh_id": item[2],
  58. "title": item[3],
  59. "text": item[4],
  60. "content_status": item[5],
  61. "process_times": item[6]
  62. } for item in task_list
  63. ]
  64. logging(
  65. code="9001",
  66. function="task1.get_task",
  67. info="本次任务获取到 {} 条视频".format(len(task_obj_list)),
  68. data=[x['content_id'] for x in task_obj_list]
  69. )
  70. return task_obj_list
  71. else:
  72. return []
  73. async def get_history_videos(self, content_id):
  74. """
  75. check whether the contents videos exists
  76. :param content_id:
  77. :return:
  78. """
  79. select_sql = f"""
  80. SELECT video_id
  81. FROM {db_video}
  82. where content_id = '{content_id}' and video_status = 1 order by request_time DESC;
  83. """
  84. content_videos = await self.mysql_client.async_select(select_sql)
  85. videos = [vid for vid in content_videos]
  86. if len(videos) >= 3:
  87. return videos
  88. else:
  89. return None
  90. async def judge_content_processing(self, content_id):
  91. """
  92. 判断该content_id是否在处理中
  93. :param content_id:
  94. :return:
  95. """
  96. select_sql = f"""
  97. SELECT trace_id, content_status
  98. FROM {db_article}
  99. WHERE content_id = '{content_id}'
  100. ORDER BY id DESC;
  101. """
  102. result = await self.mysql_client.async_select(select_sql)
  103. if result:
  104. for item in result:
  105. trace_id, content_status = item
  106. if content_status == 1:
  107. return False
  108. return True
  109. else:
  110. return True
  111. async def use_exists_contents_videos(self, video_id_list, params):
  112. """
  113. 使用已经存在的视频id
  114. :return:
  115. """
  116. trace_id = params['trace_id']
  117. content_id = params['content_id']
  118. select_sql = f"""
  119. SELECT kimi_title
  120. FROM {db_article}
  121. WHERE content_id = '{content_id}' and kimi_title is not null limit 1;
  122. """
  123. info = await self.mysql_client.async_select(sql=select_sql)
  124. kimi_title = info[0]
  125. update_sql = f"""
  126. UPDATE {db_article}
  127. SET
  128. kimi_title=%s,
  129. recall_video_id1=%s,
  130. recall_video_id2=%s,
  131. recall_video_id3=%s,
  132. content_status=%s,
  133. process_times = %s
  134. WHERE trace_id = %s
  135. """
  136. vid1, vid2, vid3 = video_id_list[0], video_id_list[1], video_id_list[2]
  137. await self.mysql_client.async_insert(
  138. sql=update_sql,
  139. params=(
  140. kimi_title,
  141. video_id_list[0],
  142. "NULL" if vid2 is None else vid2,
  143. "NULL" if vid3 is None else vid3,
  144. 2,
  145. int(params['process_times']) + 1,
  146. trace_id
  147. )
  148. )
  149. logging(
  150. code="9002",
  151. info="已从历史文章更新,文章id: {}".format(content_id),
  152. trace_id=trace_id
  153. )
  154. async def start_process(self, params):
  155. """
  156. 开始处理
  157. :param params:
  158. :return:
  159. """
  160. # 更新文章contentId为1, 说明该文章正在处理中
  161. update_sql = f"""
  162. UPDATE {db_article}
  163. SET
  164. content_status = %s
  165. WHERE
  166. trace_id = %s;
  167. """
  168. await self.mysql_client.async_insert(
  169. sql=update_sql,
  170. params=(
  171. 1, params['trace_id']
  172. )
  173. )
  174. try:
  175. video_count = await search_videos(
  176. params={
  177. "title": params['title'],
  178. "content": params['text'],
  179. "trace_id": params['trace_id'],
  180. "content_id": params['content_id']
  181. },
  182. trace_id=params['trace_id'],
  183. gh_id=params['gh_id'],
  184. mysql_client=self.mysql_client
  185. )
  186. select_sql = f"""
  187. SELECT video_id
  188. FROM {db_video}
  189. WHERE content_id = '{params['content_id']}'
  190. """
  191. result = await self.mysql_client.async_select(sql=select_sql)
  192. vid1, vid2, vid3 = result[0], result[1], result[2]
  193. if vid1 or vid2 or vid3:
  194. update_sql2 = f"""
  195. UPDATE {db_article}
  196. SET
  197. recall_video_id1 = %s,
  198. recall_video_id2 = %s,
  199. recall_video_id3 = %s,
  200. content_status = %s,
  201. process_times = %s
  202. WHERE trace_id = %s;
  203. """
  204. await self.mysql_client.async_insert(
  205. sql=update_sql2,
  206. params=(
  207. vid1 if vid1 else "NULL",
  208. vid2 if vid2 else "NULL",
  209. vid3 if vid3 else "NULL",
  210. 2,
  211. {int(params['process_times']) + 1},
  212. params['trace_id']
  213. )
  214. )
  215. logging(
  216. code="9008",
  217. info="视频搜索成功, 状态修改为2",
  218. trace_id=params['trace_id']
  219. )
  220. else:
  221. if int(params['process_times']) < 3:
  222. update_sql3 = f"""
  223. UPDATE {db_article}
  224. SET
  225. content_status = %s,
  226. process_times = %s
  227. WHERE trace_id = %s;
  228. """
  229. await self.mysql_client.async_insert(
  230. sql=update_sql3,
  231. params=(0, int(params['process_times']) + 1, params['trace_id'])
  232. )
  233. logging(
  234. code="9018",
  235. info="视频搜索失败,回退状态为0",
  236. trace_id=params['trace_id']
  237. )
  238. else:
  239. update_sql3 = f"""
  240. UPDATE {db_article}
  241. SET
  242. content_status = %s,
  243. process_times = %s
  244. WHERE trace_id = %s;
  245. """
  246. await self.mysql_client.async_insert(
  247. sql=update_sql3,
  248. params=(3, int(params['process_times']) + 1, params['trace_id'])
  249. )
  250. logging(
  251. code="9019",
  252. info="视频多次搜索失败,状态修改为3",
  253. trace_id=params['trace_id']
  254. )
  255. except Exception as e:
  256. if int(params['process_times']) < 3:
  257. logging(
  258. code="9018",
  259. info="{}异常错误:{}, 回退状态为0".format(params['trace_id'], e),
  260. trace_id=params['trace_id']
  261. )
  262. update_sql4 = f"""
  263. UPDATE {db_article}
  264. SET
  265. content_status = %s,
  266. process_times = %s
  267. WHERE trace_id = %s;
  268. """
  269. await self.mysql_client.async_insert(
  270. sql=update_sql4,
  271. params=(0, int(params['process_times']) + 1, params['trace_id'])
  272. )
  273. else:
  274. logging(
  275. code="9019",
  276. info="{}异常错误:{}, 状态修改为3".format(params['trace_id'], e),
  277. trace_id=params['trace_id']
  278. )
  279. update_sql4 = f"""
  280. UPDATE {db_article}
  281. SET
  282. content_status = %s,
  283. process_times = %s
  284. WHERE trace_id = %s;
  285. """
  286. await self.mysql_client.async_insert(
  287. sql=update_sql4,
  288. params=(3, int(params['process_times']) + 1, params['trace_id'])
  289. )
  290. async def process_task(self, params):
  291. """
  292. 异步执行
  293. :param params:
  294. :return:
  295. """
  296. content_id = params['content_id']
  297. trace_id = params['trace_id']
  298. # 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
  299. video_id_list = await self.get_history_videos(content_id=content_id)
  300. if video_id_list:
  301. # 说明已经存在了结果, 将该条记录下的video_id拿出来
  302. logging(
  303. code="9001",
  304. info="存在历史文章",
  305. trace_id=trace_id
  306. )
  307. # await self.use_exists_contents_videos(video_id_list=video_id_list, params=params)
  308. else:
  309. flag = await self.judge_content_processing(content_id)
  310. if flag:
  311. logging(
  312. code="9004",
  313. info="无正在处理的文章ID, 开始处理",
  314. trace_id=trace_id
  315. )
  316. await self.start_process(params=params)
  317. else:
  318. logging(
  319. code="9003",
  320. info="该文章ID正在请求--文章ID {}".format(content_id),
  321. trace_id=trace_id
  322. )
  323. async def deal(self):
  324. """
  325. 处理
  326. :return:
  327. """
  328. task_list = await self.get_task()
  329. task_dict = {}
  330. for task in task_list:
  331. key = task['content_id']
  332. task_dict[key] = task
  333. process_list = []
  334. for item in task_dict:
  335. process_list.append(task_dict[item])
  336. if process_list:
  337. # for task in task_list:
  338. # await self.process_task(task)
  339. tasks = [self.process_task(params) for params in process_list]
  340. await asyncio.gather(*tasks)
  341. else:
  342. logging(
  343. code="9008",
  344. info="没有要处理的请求"
  345. )