accountArticleRank.py 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356
  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. print(history_title_dict)
  62. for item in tqdm(self.params["publishArticleList"]):
  63. flag = self.pipeline.deal(item, self.accountName, history_title_dict)
  64. if flag:
  65. item["filterReason"] = flag["filterReason"]
  66. self.filter_list.append(item)
  67. else:
  68. self.publishArticleList.append(item)
  69. print("过滤完成")
  70. async def check_params(self):
  71. """
  72. 校验参数
  73. :return:
  74. """
  75. try:
  76. self.accountId = self.params["accountId"]
  77. self.accountName = self.params["accountName"]
  78. self.ghId = self.params["ghId"]
  79. self.strategy = self.params["strategy"]
  80. self.publishNum = self.params["publishNum"]
  81. print("开始校验参数")
  82. self.filter()
  83. print("参数校验成功")
  84. self.logger.log(code="1001", msg="参数校验成功", data=self.params)
  85. return None
  86. except Exception as e:
  87. response = {
  88. "msg": "params error",
  89. "info": "params check failed, params : {} is not correct".format(e),
  90. "code": 0,
  91. }
  92. self.logger.log(
  93. code="1002", msg="参数校验失败--{}".format(e), data=self.params
  94. )
  95. return response
  96. async def basic_rank(self):
  97. """
  98. 基础排序
  99. :return:
  100. """
  101. # 第一步把所有文章标题分为3组
  102. article_list1_ori = [
  103. i for i in self.publishArticleList if "【1】" in i["producePlanName"]
  104. ]
  105. article_list2_ori = [
  106. i for i in self.publishArticleList if "【2】" in i["producePlanName"]
  107. ]
  108. article_list3_ori = [
  109. i
  110. for i in self.publishArticleList
  111. if not i in article_list1_ori and not i in article_list2_ori
  112. ]
  113. # # 全局去重,保留优先级由 L1 --> L2 --> L3
  114. # hash_map = {}
  115. #
  116. # article_list1 = []
  117. # for i in article_list1_ori:
  118. # title = i['title']
  119. # if hash_map.get(title):
  120. # continue
  121. # else:
  122. # article_list1.append(i)
  123. # hash_map[title] = 1
  124. #
  125. # article_list2 = []
  126. # for i in article_list2_ori:
  127. # title = i['title']
  128. # if hash_map.get(title):
  129. # continue
  130. # else:
  131. # article_list2.append(i)
  132. # hash_map[title] = 2
  133. #
  134. # article_list3 = []
  135. # for i in article_list3_ori:
  136. # title = i['title']
  137. # if hash_map.get(title):
  138. # continue
  139. # else:
  140. # article_list3.append(i)
  141. # hash_map[title] = 1
  142. # 第二步对article_list1, article_list3按照得分排序, 对article_list2按照播放量排序
  143. if article_list1_ori:
  144. rank1 = ArticleRank().rank(
  145. account_list=[self.accountName],
  146. text_list=[i["title"] for i in article_list1_ori],
  147. )
  148. score_list1 = rank1[self.accountName]["score_list"]
  149. ranked_1 = []
  150. for index, value in enumerate(score_list1):
  151. obj = article_list1_ori[index]
  152. obj["score"] = value + 1000
  153. ranked_1.append(obj)
  154. ranked_1 = sorted(ranked_1, key=lambda x: x["score"], reverse=True)
  155. else:
  156. ranked_1 = []
  157. # rank2
  158. if article_list2_ori:
  159. for item in article_list2_ori:
  160. item["score"] = 100
  161. ranked_2 = sorted(
  162. article_list2_ori, key=lambda x: x["crawlerViewCount"], reverse=True
  163. )
  164. else:
  165. ranked_2 = []
  166. # rank3
  167. if article_list3_ori:
  168. rank3 = ArticleRank().rank(
  169. account_list=[self.accountName],
  170. text_list=[i["title"] for i in article_list3_ori],
  171. )
  172. score_list3 = rank3[self.accountName]["score_list"]
  173. ranked_3 = []
  174. for index, value in enumerate(score_list3):
  175. obj = article_list3_ori[index]
  176. obj["score"] = value
  177. ranked_3.append(obj)
  178. ranked_3 = sorted(ranked_3, key=lambda x: x["score"], reverse=True)
  179. else:
  180. ranked_3 = []
  181. self.logger.log(
  182. code="1004",
  183. msg="排序完成",
  184. data={"rank1": ranked_1, "rank2": ranked_2, "rank3": ranked_3},
  185. )
  186. return ranked_1, ranked_2, ranked_3
  187. async def rank_v1(self):
  188. """
  189. Rank Version 1
  190. :return:
  191. """
  192. print("开始排序")
  193. try:
  194. ranked_1_d, ranked_2_d, ranked_3_d = await self.basic_rank()
  195. ranked_1, ranked_2, ranked_3 = deduplication(
  196. ranked_1_d, ranked_2_d, ranked_3_d
  197. )
  198. print("去重成功")
  199. try:
  200. L = []
  201. if ranked_1:
  202. target = random.choice(ranked_1[:5])
  203. L.append(target)
  204. if ranked_2:
  205. L.append(ranked_2[0])
  206. else:
  207. if ranked_2:
  208. if len(ranked_2) > 1:
  209. for i in ranked_2[:2]:
  210. L.append(i)
  211. else:
  212. L.append(ranked_2[0])
  213. # L only 1
  214. for item in ranked_3:
  215. L.append(item)
  216. # L 1 and 3
  217. result = {
  218. "accountId": self.accountId,
  219. "accountName": self.accountName,
  220. "ghId": self.ghId,
  221. "strategy": self.strategy,
  222. "publishNum": self.publishNum,
  223. "rank_list": L[: self.publishNum],
  224. "filter_list": self.filter_list,
  225. }
  226. self.logger.log(code=1006, msg="rank successfully", data=result)
  227. response = {"status": "Rank Success", "data": result, "code": 1}
  228. except Exception as e:
  229. result = {
  230. "accountId": self.accountId,
  231. "accountName": self.accountName,
  232. "ghId": self.ghId,
  233. "strategy": self.strategy,
  234. "publishNum": self.publishNum,
  235. "rank_list": self.publishArticleList[: self.publishNum],
  236. "filter_list": self.filter_list,
  237. }
  238. self.logger.log(
  239. code=1007, msg="rank failed because of {}".format(e), data=result
  240. )
  241. print("排序成功")
  242. response = {"status": "Rank Fail", "data": result, "code": 1}
  243. return response
  244. except:
  245. result = {"code": 2, "info": "account is not exist"}
  246. return result
  247. async def rank_v2(self):
  248. """
  249. Rank Version 2
  250. :return:
  251. """
  252. try:
  253. ranks = ArticleRank().rank(
  254. account_list=[self.accountName],
  255. text_list=[i["title"] for i in self.publishArticleList],
  256. )
  257. score_list1 = ranks[self.accountName]["score_list"]
  258. ranked_v2 = []
  259. for index, value in enumerate(score_list1):
  260. obj = self.publishArticleList[index]
  261. obj["score"] = value
  262. ranked_v2.append(obj)
  263. ranked_v2 = sorted(ranked_v2, key=lambda x: (-x["score"], -x['crawlerViewCount']))
  264. result = {
  265. "accountId": self.accountId,
  266. "accountName": self.accountName,
  267. "ghId": self.ghId,
  268. "strategy": self.strategy,
  269. "publishNum": self.publishNum,
  270. "rank_list": ranked_v2[: self.publishNum],
  271. "filter_list": self.filter_list,
  272. }
  273. response = {"status": "Rank Success", "data": result, "code": 1}
  274. except Exception as e:
  275. result = {
  276. "accountId": self.accountId,
  277. "accountName": self.accountName,
  278. "ghId": self.ghId,
  279. "strategy": self.strategy,
  280. "publishNum": self.publishNum,
  281. "rank_list": self.publishArticleList[: self.publishNum],
  282. "filter_list": self.filter_list,
  283. }
  284. response = {"status": "Rank Fail Because Of {}".format(e), "data": result, "code": 1}
  285. return response
  286. async def rank_v3(self):
  287. """
  288. Rank Version 3
  289. :return:
  290. """
  291. return await self.rank_v1()
  292. async def rank_v4(self):
  293. """
  294. Rank Version 4
  295. :return:
  296. """
  297. return await self.rank_v1()
  298. async def rank_v5(self):
  299. """
  300. Rank Version 5
  301. :return:
  302. """
  303. return await self.rank_v1()
  304. async def choose_strategy(self):
  305. """
  306. 选择排序策略
  307. :return:
  308. """
  309. match self.strategy:
  310. case "ArticleRankV1":
  311. self.logger.log(code="1003", msg="命中排序策略1")
  312. return await self.rank_v1()
  313. case "ArticleRankV2":
  314. self.logger.log(code="1003", msg="命中排序策略2")
  315. return await self.rank_v2()
  316. case "ArticleRankV3":
  317. self.logger.log(code="1003", msg="命中排序策略3")
  318. return await self.rank_v3()
  319. case "ArticleRankV4":
  320. self.logger.log(code="1003", msg="命中排序策略4")
  321. return await self.rank_v4()
  322. case "ArticleRankV5":
  323. self.logger.log(code="1003", msg="命中排序策略5")
  324. return await self.rank_v5()
  325. async def deal(self):
  326. """
  327. Deal Function
  328. :return:
  329. """
  330. error_params = await self.check_params()
  331. if error_params:
  332. return error_params
  333. else:
  334. print("参数校验成功")
  335. return await self.choose_strategy()