accountArticleRank.py 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354
  1. """
  2. @author: luojunhui
  3. """
  4. import random
  5. import time
  6. from uuid import uuid4
  7. from tqdm import tqdm
  8. from applications.aliyunLog import AliyunArticleLog
  9. from applications.functions import ArticleRank, title_sim_v2_by_list
  10. from applications.pipeline import LongArticlesPipeline
  11. def deduplication(rank1, rank2, rank3):
  12. """
  13. 标题相似度去重
  14. :return:
  15. """
  16. dup_list = []
  17. final_result = []
  18. for item_list in [rank1, rank2, rank3]:
  19. result = []
  20. if item_list:
  21. for item in item_list:
  22. title = item["title"]
  23. if title_sim_v2_by_list(title, dup_list):
  24. # print("标题重复,已经过滤\t", title)
  25. continue
  26. else:
  27. result.append(item)
  28. dup_list.append(title)
  29. final_result.append(result)
  30. return final_result[0], final_result[1], final_result[2]
  31. class AccountArticleRank(object):
  32. """
  33. 文章排序
  34. """
  35. def __init__(self, params, mysql_client):
  36. """
  37. :param params: 请求参数
  38. :param mysql_client: 数据库链接池
  39. """
  40. self.filter_list = None
  41. self.publishArticleList = None
  42. self.publishNum = None
  43. self.strategy = None
  44. self.ghId = None
  45. self.accountName = None
  46. self.accountId = None
  47. self.params = params
  48. self.mysql_client = mysql_client
  49. self.request_id = "alg-{}-{}".format(uuid4(), int(time.time()))
  50. self.logger = AliyunArticleLog(request_id=self.request_id, alg="ArticleRank")
  51. self.pipeline = LongArticlesPipeline()
  52. def filter(self):
  53. """
  54. 过滤器
  55. """
  56. self.publishArticleList = []
  57. self.filter_list = []
  58. history_title_dict = self.pipeline.history_title(
  59. account_nickname=self.accountName
  60. )
  61. for item in tqdm(self.params["publishArticleList"]):
  62. flag = self.pipeline.deal(item, self.accountName, history_title_dict)
  63. if flag:
  64. item["filterReason"] = flag["filterReason"]
  65. self.filter_list.append(item)
  66. else:
  67. self.publishArticleList.append(item)
  68. print("过滤完成")
  69. async def check_params(self):
  70. """
  71. 校验参数
  72. :return:
  73. """
  74. try:
  75. self.accountId = self.params["accountId"]
  76. self.accountName = self.params["accountName"]
  77. self.ghId = self.params["ghId"]
  78. self.strategy = self.params["strategy"]
  79. self.publishNum = self.params["publishNum"]
  80. print("开始校验参数")
  81. self.filter()
  82. self.logger.log(code="1001", msg="参数校验成功", data=self.params)
  83. return None
  84. except Exception as e:
  85. response = {
  86. "msg": "params error",
  87. "info": "params check failed, params : {} is not correct".format(e),
  88. "code": 0,
  89. }
  90. self.logger.log(
  91. code="1002", msg="参数校验失败--{}".format(e), data=self.params
  92. )
  93. return response
  94. async def basic_rank(self):
  95. """
  96. 基础排序
  97. :return:
  98. """
  99. # 第一步把所有文章标题分为3组
  100. article_list1_ori = [
  101. i for i in self.publishArticleList if "【1】" in i["producePlanName"]
  102. ]
  103. article_list2_ori = [
  104. i for i in self.publishArticleList if "【2】" in i["producePlanName"]
  105. ]
  106. article_list3_ori = [
  107. i
  108. for i in self.publishArticleList
  109. if not i in article_list1_ori and not i in article_list2_ori
  110. ]
  111. # # 全局去重,保留优先级由 L1 --> L2 --> L3
  112. # hash_map = {}
  113. #
  114. # article_list1 = []
  115. # for i in article_list1_ori:
  116. # title = i['title']
  117. # if hash_map.get(title):
  118. # continue
  119. # else:
  120. # article_list1.append(i)
  121. # hash_map[title] = 1
  122. #
  123. # article_list2 = []
  124. # for i in article_list2_ori:
  125. # title = i['title']
  126. # if hash_map.get(title):
  127. # continue
  128. # else:
  129. # article_list2.append(i)
  130. # hash_map[title] = 2
  131. #
  132. # article_list3 = []
  133. # for i in article_list3_ori:
  134. # title = i['title']
  135. # if hash_map.get(title):
  136. # continue
  137. # else:
  138. # article_list3.append(i)
  139. # hash_map[title] = 1
  140. # 第二步对article_list1, article_list3按照得分排序, 对article_list2按照播放量排序
  141. if article_list1_ori:
  142. rank1 = ArticleRank().rank(
  143. account_list=[self.accountName],
  144. text_list=[i["title"] for i in article_list1_ori],
  145. )
  146. score_list1 = rank1[self.accountName]["score_list"]
  147. ranked_1 = []
  148. for index, value in enumerate(score_list1):
  149. obj = article_list1_ori[index]
  150. obj["score"] = value + 1000
  151. ranked_1.append(obj)
  152. ranked_1 = sorted(ranked_1, key=lambda x: x["score"], reverse=True)
  153. else:
  154. ranked_1 = []
  155. # rank2
  156. if article_list2_ori:
  157. for item in article_list2_ori:
  158. item["score"] = 100
  159. ranked_2 = sorted(
  160. article_list2_ori, key=lambda x: x["crawlerViewCount"], reverse=True
  161. )
  162. else:
  163. ranked_2 = []
  164. # rank3
  165. if article_list3_ori:
  166. rank3 = ArticleRank().rank(
  167. account_list=[self.accountName],
  168. text_list=[i["title"] for i in article_list3_ori],
  169. )
  170. score_list3 = rank3[self.accountName]["score_list"]
  171. ranked_3 = []
  172. for index, value in enumerate(score_list3):
  173. obj = article_list3_ori[index]
  174. obj["score"] = value
  175. ranked_3.append(obj)
  176. ranked_3 = sorted(ranked_3, key=lambda x: x["score"], reverse=True)
  177. else:
  178. ranked_3 = []
  179. self.logger.log(
  180. code="1004",
  181. msg="排序完成",
  182. data={"rank1": ranked_1, "rank2": ranked_2, "rank3": ranked_3},
  183. )
  184. return ranked_1, ranked_2, ranked_3
  185. async def rank_v1(self):
  186. """
  187. Rank Version 1
  188. :return:
  189. """
  190. print("开始排序")
  191. try:
  192. ranked_1_d, ranked_2_d, ranked_3_d = await self.basic_rank()
  193. ranked_1, ranked_2, ranked_3 = deduplication(
  194. ranked_1_d, ranked_2_d, ranked_3_d
  195. )
  196. print("去重成功")
  197. try:
  198. L = []
  199. if ranked_1:
  200. target = random.choice(ranked_1[:5])
  201. L.append(target)
  202. if ranked_2:
  203. L.append(ranked_2[0])
  204. else:
  205. if ranked_2:
  206. if len(ranked_2) > 1:
  207. for i in ranked_2[:2]:
  208. L.append(i)
  209. else:
  210. L.append(ranked_2[0])
  211. # L only 1
  212. for item in ranked_3:
  213. L.append(item)
  214. # L 1 and 3
  215. result = {
  216. "accountId": self.accountId,
  217. "accountName": self.accountName,
  218. "ghId": self.ghId,
  219. "strategy": self.strategy,
  220. "publishNum": self.publishNum,
  221. "rank_list": L[: self.publishNum],
  222. "filter_list": self.filter_list,
  223. }
  224. self.logger.log(code=1006, msg="rank successfully", data=result)
  225. response = {"status": "Rank Success", "data": result, "code": 1}
  226. except Exception as e:
  227. result = {
  228. "accountId": self.accountId,
  229. "accountName": self.accountName,
  230. "ghId": self.ghId,
  231. "strategy": self.strategy,
  232. "publishNum": self.publishNum,
  233. "rank_list": self.publishArticleList[: self.publishNum],
  234. "filter_list": self.filter_list,
  235. }
  236. self.logger.log(
  237. code=1007, msg="rank failed because of {}".format(e), data=result
  238. )
  239. print("排序成功")
  240. response = {"status": "Rank Fail", "data": result, "code": 1}
  241. return response
  242. except:
  243. result = {"code": 2, "info": "account is not exist"}
  244. return result
  245. async def rank_v2(self):
  246. """
  247. Rank Version 2
  248. :return:
  249. """
  250. try:
  251. ranks = ArticleRank().rank(
  252. account_list=[self.accountName],
  253. text_list=[i["title"] for i in self.publishArticleList],
  254. )
  255. score_list1 = ranks[self.accountName]["score_list"]
  256. ranked_v2 = []
  257. for index, value in enumerate(score_list1):
  258. obj = self.publishArticleList[index]
  259. obj["score"] = value
  260. ranked_v2.append(obj)
  261. ranked_v2 = sorted(ranked_v2, key=lambda x: (-x["score"], -x['crawlerViewCount']))
  262. result = {
  263. "accountId": self.accountId,
  264. "accountName": self.accountName,
  265. "ghId": self.ghId,
  266. "strategy": self.strategy,
  267. "publishNum": self.publishNum,
  268. "rank_list": ranked_v2[: self.publishNum],
  269. "filter_list": self.filter_list,
  270. }
  271. response = {"status": "Rank Success", "data": result, "code": 1}
  272. except Exception as e:
  273. result = {
  274. "accountId": self.accountId,
  275. "accountName": self.accountName,
  276. "ghId": self.ghId,
  277. "strategy": self.strategy,
  278. "publishNum": self.publishNum,
  279. "rank_list": self.publishArticleList[: self.publishNum],
  280. "filter_list": self.filter_list,
  281. }
  282. response = {"status": "Rank Fail Because Of {}".format(e), "data": result, "code": 1}
  283. return response
  284. async def rank_v3(self):
  285. """
  286. Rank Version 3
  287. :return:
  288. """
  289. return await self.rank_v1()
  290. async def rank_v4(self):
  291. """
  292. Rank Version 4
  293. :return:
  294. """
  295. return await self.rank_v1()
  296. async def rank_v5(self):
  297. """
  298. Rank Version 5
  299. :return:
  300. """
  301. return await self.rank_v1()
  302. async def choose_strategy(self):
  303. """
  304. 选择排序策略
  305. :return:
  306. """
  307. match self.strategy:
  308. case "ArticleRankV1":
  309. self.logger.log(code="1003", msg="命中排序策略1")
  310. return await self.rank_v1()
  311. case "ArticleRankV2":
  312. self.logger.log(code="1003", msg="命中排序策略2")
  313. return await self.rank_v2()
  314. case "ArticleRankV3":
  315. self.logger.log(code="1003", msg="命中排序策略3")
  316. return await self.rank_v3()
  317. case "ArticleRankV4":
  318. self.logger.log(code="1003", msg="命中排序策略4")
  319. return await self.rank_v4()
  320. case "ArticleRankV5":
  321. self.logger.log(code="1003", msg="命中排序策略5")
  322. return await self.rank_v5()
  323. async def deal(self):
  324. """
  325. Deal Function
  326. :return:
  327. """
  328. error_params = await self.check_params()
  329. if error_params:
  330. return error_params
  331. else:
  332. print("参数校验成功")
  333. return await self.choose_strategy()