From 55a246309c0c02811dcdb9aa689549c01303e688 Mon Sep 17 00:00:00 2001
From: litaolemo <593516104@qq.com>
Date: Wed, 23 Sep 2020 16:15:12 +0800
Subject: [PATCH] update

---
 output/out_put_diary_0923.py | 221 +++++++++++++++++++++++++++++++++++
 1 file changed, 221 insertions(+)
 create mode 100644 output/out_put_diary_0923.py

diff --git a/output/out_put_diary_0923.py b/output/out_put_diary_0923.py
new file mode 100644
index 0000000..590879e
--- /dev/null
+++ b/output/out_put_diary_0923.py
@@ -0,0 +1,221 @@
+# -*- coding:UTF-8 -*-
+# @Time  : 2020/9/23 16:10
+# @File  : out_put_diary_0923.py
+# @email : litao@igengmei.com
+# @author : litao
+
+
+
+import hashlib
+import json
+
+import pymysql
+import xlwt, datetime
+import redis
+from meta_base_code.utils.func_from_redis_get_portrait import user_portrait_scan_info,get_user_portrait_tag3_from_redis
+from meta_base_code.utils.func_from_es_get_article import get_es_article_num,get_user_post_from_mysql
+# from pyhive import hive
+from maintenance.func_send_email_with_file import send_file_email
+from typing import Dict, List
+from elasticsearch_7 import Elasticsearch
+from elasticsearch_7.helpers import scan
+import sys
+import time
+from pyspark import SparkConf
+from pyspark.sql import SparkSession, DataFrame
+
+
+# from pyspark.sql.functions import lit
+# import pytispark.pytispark as pti
+startTime = time.time()
+sparkConf = SparkConf()
+sparkConf.set("spark.sql.crossJoin.enabled", True)
+sparkConf.set("spark.debug.maxToStringFields", "100")
+sparkConf.set("spark.tispark.plan.allow_index_double_read", False)
+sparkConf.set("spark.tispark.plan.allow_index_read", True)
+sparkConf.set("spark.hive.mapred.supports.subdirectories", True)
+sparkConf.set("spark.hadoop.mapreduce.input.fileinputformat.input.dir.recursive", True)
+sparkConf.set("spark.serializer", "org.apache.spark.serializer.KryoSerializer")
+sparkConf.set("mapreduce.output.fileoutputformat.compress", False)
+sparkConf.set("mapreduce.map.output.compress", False)
+sparkConf.set("prod.gold.jdbcuri",
+              "jdbc:mysql://172.16.30.136/doris_prod?user=doris&password=o5gbA27hXHHm&rewriteBatchedStatements=true")
+sparkConf.set("prod.mimas.jdbcuri",
+              "jdbc:mysql://172.16.30.138/mimas_prod?user=mimas&password=GJL3UJe1Ck9ggL6aKnZCq4cRvM&rewriteBatchedStatements=true")
+sparkConf.set("prod.gaia.jdbcuri",
+              "jdbc:mysql://172.16.30.143/zhengxing?user=work&password=BJQaT9VzDcuPBqkd&rewriteBatchedStatements=true")
+sparkConf.set("prod.tidb.jdbcuri",
+              "jdbc:mysql://172.16.40.158:4000/eagle?user=st_user&password=aqpuBLYzEV7tML5RPsN1pntUzFy&rewriteBatchedStatements=true")
+sparkConf.set("prod.jerry.jdbcuri",
+              "jdbc:mysql://172.16.40.158:4000/jerry_prod?user=st_user&password=aqpuBLYzEV7tML5RPsN1pntUzFy&rewriteBatchedStatements=true")
+sparkConf.set("prod.tispark.pd.addresses", "172.16.40.158:2379")
+sparkConf.set("prod.tispark.pd.addresses", "172.16.40.170:4000")
+sparkConf.set("prod.tidb.database", "jerry_prod")
+sparkConf.setAppName("test")
+
+spark = (SparkSession.builder.config(conf=sparkConf).config("spark.sql.extensions", "org.apache.spark.sql.TiExtensions")
+         .config("spark.tispark.pd.addresses", "172.16.40.170:2379").enableHiveSupport().getOrCreate())
+
+spark.sql("ADD JAR hdfs:///user/hive/share/lib/udf/brickhouse-0.7.1-SNAPSHOT.jar")
+spark.sql("ADD JAR hdfs:///user/hive/share/lib/udf/hive-udf-1.0-SNAPSHOT.jar")
+spark.sql("CREATE TEMPORARY FUNCTION json_map AS 'brickhouse.udf.json.JsonMapUDF'")
+spark.sql("CREATE TEMPORARY FUNCTION is_json AS 'com.gmei.hive.common.udf.UDFJsonFormatCheck'")
+spark.sql("CREATE TEMPORARY FUNCTION arrayMerge AS 'com.gmei.hive.common.udf.UDFArryMerge'")
+
+task_list = []
+task_days = 2
+
+
+def con_sql(sql):
+    # 从数据库的表里获取数据
+
+    db = pymysql.connect(host='172.16.40.158', port=4000, user='st_user', passwd='aqpuBLYzEV7tML5RPsN1pntUzFy',
+                         db='jerry_prod')
+    cursor = db.cursor()
+    cursor.execute(sql)
+    result = cursor.fetchall()
+    db.close()
+    return result
+
+
+second_demands_zero_dict = {
+    # "answer":{},
+    # "tractate":{},
+    "diary":{},
+}
+project_zero_dict = {
+    # "answer":{},
+    # "tractate":{},
+    "diary":{},
+}
+t = 1
+day_num = 0 - t
+now = (datetime.datetime.now() + datetime.timedelta(days=day_num))
+last_30_day_str = (now + datetime.timedelta(days=-31)).strftime("%Y%m%d")
+today_str = now.strftime("%Y%m%d")
+today_str_format = now.strftime("%Y-%m-%d")
+yesterday_str = (now + datetime.timedelta(days=-1)).strftime("%Y%m%d")
+yesterday_str_format = (now + datetime.timedelta(days=-1)).strftime("%Y-%m-%d")
+one_week_age_str = (now + datetime.timedelta(days=-7)).strftime("%Y%m%d")
+sql = """select first_device from online.ml_user_history_detail where partition_date = {today_str} and last_active_date >= {last_30_day_str}
+""".format(today_str=today_str,last_30_day_str=last_30_day_str)
+print(sql)
+new_urser_device_id_df = spark.sql(sql)
+new_urser_device_id_df.createOrReplaceTempView("device_id_view")
+new_urser_device_id_df.show(1)
+sql_res = new_urser_device_id_df.collect()
+
+bulk_dict = {
+    0: [0, 0, 0],
+    10: [0, 0, 0],
+    50: [0, 0, 0],
+    100: [0, 0, 0],
+    200: [0, 0, 0],
+    500: [0, 0, 0],
+    1000: [0, 0, 0],
+}
+
+
+task_list = []
+sql = """
+select card_id from strategy_content_exposure_index where card_content_type="diary" and preciseexposure_num>=50 and ctr>=0.05 and avg_page_stay>=20 and create_day="2020-09-17";   
+    """
+second_demands_count_dict, tags_v3_count_dict, second_demands_card_id_list,tags_v3_card_id_list,second_demands_tractate_dict,tags_v3_tractate_dict=get_user_post_from_mysql(sql)
+# second_demands_count_dict= {'地包天': 4, '近视': 20, '洗眉': 16, '阴道缩紧': 338, '避孕': 1, '缩窄下巴': 1, '翘睫': 2, '腿形矫正': 7, '脱手脚毛': 80, '注射物取出': 1, '脱敏': 2, '脱腋毛': 6, '祛纹身': 8, '填充面部': 7, '生私密毛发': 339, '祛红血丝': 2, '自体脂肪修复': 3, '除螨': 1, '生眉毛': 10, '网红脸': 1, '缩胸': 11, '阴茎美化': 11, '耳洞': 3, '生头发': 85, 'AI测试': 3, '凸嘴': 3, '缩短下巴': 6, '纹身': 11, '生睫毛': 2, '假体取出': 1, '鼻孔矫正': 1, '下颌缘提升': 2, '奥美定': 1, '皮肤病': 6, '性快感': 3, '眼线': 10, '颏肌放松': 3, '洗眼线': 2, '产后恢复': 1, '祛腋臭': 20, '脱背毛': 79, '脱毛发': 2, '鼻尖延长': 12, '腿部加长': 2, '洁面': 2, '鼻中隔延长': 3, '唇腭裂': 2, '脱唇毛': 3, '填充卧蚕': 3, '丰眼窝': 3, '脱发际线': 5, '脸型': 241, '脱私密毛发': 1, '缩鼻背': 5, '生发际线': 4, '脱腿毛': 4, '生胡须': 1, '鼻部缩短': 2, '健康调理': 4}
+
+# tags_v3_count_dict = {'牙齿': 3, '缩下巴': 3, '洗眉': 12, '玻尿酸': 1, '超声溶脂': 2, '注射物取出': 1, '生长因子': 3, 'G点注射': 3, '正骨术': 1, '真皮填充卧蚕': 2, '激光祛皱': 4, '胶原蛋白填充面部': 3, '脱私密毛发': 1, '瘦身': 3, '玻尿酸丰唇珠': 1, '半飞秒激光术': 1, '乳牙': 1, '产后修复': 1, '视力检查': 2, '干细胞疗法': 2, '童颜针': 4, '准分子激光术': 1, '中医按摩': 1, '祛黑头': 1, '祛黑眼圈': 4, '自体软骨垫鼻尖': 14, '洗眼线': 2, '耳部矫正': 2, '新手精选': 1, '植私密毛发': 338, '玻尿酸填充卧蚕': 1, '阴茎增大': 6, '乳房缩小': 1, '植胡须': 1, '基因检测': 1, '鼻孔矫正': 1, '下巴前推': 10, '激光脱毛': 2, 'PRP生发': 80, '玻尿酸祛皱': 1, '注射祛腋臭': 14, '面膜': 4, '半永久纹眉1': 2, '毛囊检测': 4, '激光脱腋毛': 2, '翘睫': 2, '下巴截骨术': 2, '激光脱唇毛': 1, '激光祛腋臭': 1, '全飞秒激光术': 1, '拔罐': 1, '美臀': 1, '激光祛纹身': 7, '喷砂洗牙': 1, '半永久纹发际线': 7, '自体脂肪面部填充': 2, '祛斑': 1, '私密紧致': 1, '激光脱手脚毛': 80, '腿形矫正': 7, '激光脱背毛': 80, '埋线缩鼻翼': 2, '半永久': 4, '填充面部': 2, '生私密毛发': 1, '洗唇线': 2, '自体脂肪': 1, '半永久纹眼线': 10, '胶原蛋白注射': 2, '肉毒素治疗多汗': 1, '黑脸娃娃': 6, '玻尿酸丰眼窝': 2, '微笑唇': 1, '打耳洞': 3, '睫毛增长': 2, '双眼皮': 8, '全飞秒': 3, '自体脂肪私密紧致': 337, '吸脂': 2, '皮肤病': 6, '口腔溃疡': 1, '激光洗眉': 5, '药物脱毛': 5, '断骨增高': 2, '额头缩小': 4, '肤质检测': 1, '激光近视矫正': 4, '自体脂肪填充修复': 3, '肉毒素颏肌放松': 1, '祛眼袋': 1, '晶体植入': 3, '阴茎延长': 5, '激光脱发际线': 1, '唇珠唇弓': 2, '包皮手术': 5, '唇腭裂': 4, '乳头缩小': 338, '臀部整形': 1, '植眉': 10, '阴茎增粗': 9, '抗衰紧致': 1, '缩鼻背': 2, '手术祛腋臭': 7, '射频祛眼袋': 4, '上眼睑祛脂': 3, '鼻部硅胶假体取出': 1, '激光脱腿毛': 3, '发质护理': 6, '抗衰': 1}
+
+print(second_demands_count_dict,tags_v3_count_dict)
+time.sleep(10)
+
+
+second_demands_tag_count = {}
+projects_demands_tag_count = {}
+total_tag_count = {}
+total_tag_count_pro = {}
+temp_null_count = 0
+for redis_count,spark_res in enumerate(sql_res):
+    # if redis_count >= 50:break
+    second_demands = []
+    projects = []
+    total_answer_content_num = 0
+    total_tractate_content_num = 0
+    total_diary_content_num = 0
+    # print(sql_res)
+    try:
+        res = get_user_portrait_tag3_from_redis(spark_res.first_device)
+    except:
+        continue
+    if res.get("second_demands"):
+        second_demands = res.get("second_demands")
+        # print(count_res)
+        for tag in second_demands:
+            if tag in second_demands_tag_count:
+                second_demands_tag_count[tag] += 1
+            else:
+                second_demands_tag_count[tag] = 1
+            if tag in second_demands_count_dict:
+                total_tractate_content_num += second_demands_count_dict[tag]
+    if res.get("projects"):
+        projects = res.get("projects")
+        # print(count_res)
+        for tag in projects:
+            if tag in projects_demands_tag_count:
+                projects_demands_tag_count[tag] += 1
+            else:
+                projects_demands_tag_count[tag] = 1
+
+            if tag in tags_v3_count_dict:
+                total_tractate_content_num += tags_v3_count_dict[tag]
+    # print(total_answer_content_num, total_tractate_content_num, total_diary_content_num)
+    tmp_count_num = 0
+
+
+    if 0 <= total_tractate_content_num < 10:
+        bulk_dict[0][1] += 1
+        if not second_demands and not projects:
+            temp_null_count += 1
+        if second_demands:
+            for tag in second_demands:
+                if tag in total_tag_count:
+                    total_tag_count[tag] += 1
+                else:
+                    total_tag_count[tag] = 1
+        if projects:
+            for tag in projects:
+                if tag in total_tag_count_pro:
+                    total_tag_count_pro[tag] += 1
+                else:
+                    total_tag_count_pro[tag] = 1
+    elif 10 <= total_tractate_content_num < 50:
+        bulk_dict[10][1] += 1
+    elif 50 <= total_tractate_content_num < 100:
+        bulk_dict[50][1] += 1
+    elif 100 <= total_tractate_content_num < 200:
+        bulk_dict[100][1] += 1
+    elif 200 <= total_tractate_content_num < 500:
+        bulk_dict[200][1] += 1
+    elif 500 <= total_tractate_content_num < 1000:
+        bulk_dict[500][1] += 1
+    else:
+        bulk_dict[1000][1] += 1
+
+    # if redis_count % 5000 == 0:
+    #     print(redis_count,bulk_dict)
+    #     print(temp_null_count)
+    #     print(second_demands_tag_count,projects_demands_tag_count)
+print("total_tag_count" , total_tag_count)
+print("total_tag_count_pro" ,total_tag_count_pro)
+print("bulk_dict", bulk_dict)
+print("temp_null_count", temp_null_count)
+#
+print("second_demands_tag_count",second_demands_tag_count)
+print("projects_demands_tag_count",projects_demands_tag_count)
+with open("log.log","w",encoding='utf-8') as f:
+    f.write(str(total_tag_count))
+    f.write(str(total_tag_count_pro))
+    f.write(str(second_demands_tag_count))
+    f.write(str(projects_demands_tag_count))
+
+
+
-- 
2.18.0