Spark SQL大数据处理并写入Elasticsearch

 1 # coding: utf-8
 2 import sys
 3 import os
 4 
 5 pre_current_dir = os.path.dirname(os.getcwd())
 6 sys.path.append(pre_current_dir)
 7 from pyspark.sql import SparkSession
 8 from pyspark.sql.types import *
 9 from pyspark.sql.functions import udf
10 from settings import ES_CONF
11 
12 current_dir = os.path.dirname(os.path.realpath(__file__))
13 
14 spark = SparkSession.builder.appName("weather_result").getOrCreate()
15 
16 
17 def get_health_level(value):
18     """
19     PM2.5对应健康级别
20     :param value:
21     :return:
22     """
23     if 0 <= value <= 50:
24         return "Very Good"
25     elif 50 < value <= 100:
26         return "Good"
27     elif 100 < value <= 150:
28         return "Unhealthy for Sensi"
29     elif value <= 200:
30         return "Unhealthy"
31     elif 200 < value <= 300:
32         return "Very Unhealthy"
33     elif 300 < value <= 500:
34         return "Hazardous"
35     elif value > 500:
36         return "Extreme danger"
37     else:
38         return None
39 
40 
41 def get_weather_result():
42     """
43     获取Spark SQL分析后的数据
44     :return:
45     """
46     # load所需字段的数据到DF
47     df_2017 = spark.read.format("csv") 48         .option("header", "true") 49         .option("inferSchema", "true") 50         .load("file://{}/data/Beijing2017_PM25.csv".format(current_dir)) 51         .select("Year", "Month", "Day", "Hour", "Value", "QC Name")
52 
53     # 查看Schema
54     df_2017.printSchema()
55 
56     # 通过udf将字符型health_level转换为column
57     level_function_udf = udf(get_health_level, StringType())
58 
59     # 新建列healthy_level 并healthy_level分组
60     group_2017 = df_2017.withColumn(
61         "healthy_level", level_function_udf(df_2017[‘Value‘])
62     ).groupBy("healthy_level").count()
63 
64     # 新建列days和percentage 并计算它们对应的值
65     result_2017 = group_2017.select("healthy_level", "count") 66         .withColumn("days", group_2017[‘count‘] / 24) 67         .withColumn("percentage", group_2017[‘count‘] / df_2017.count())
68     result_2017.show()
69 
70     return result_2017
71 
72 
73 def write_result_es():
74     """
75     将SparkSQL计算结果写入到ES
76     :return:
77     """
78     result_2017 = get_weather_result()
79     # ES_CONF配置 ES的node和index
80     result_2017.write.format("org.elasticsearch.spark.sql") 81         .option("es.nodes", "{}".format(ES_CONF[‘ELASTIC_HOST‘])) 82         .mode("overwrite") 83         .save("{}/pm_value".format(ES_CONF[‘WEATHER_INDEX_NAME‘]))
84 
85 
86 write_result_es()
87 spark.stop()