[大数据]sparkSql数据离线处理--整理记录

sparkSql数据离线处理

前言：本文作为本人学习sparkSql离线数据抽取，离线数据处理的学习整理记录，文中参考博客均附上原文链接。

一、Hive环境准备

1、配置文件准备：

/opt/hive/conf/hive-site.xml:(2021/12/31修改，添加了&useSSL=false&useUnicode=true&characterEncoding=utf8支持中文编码)

<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<configuration>
  <property>
    <name>javax.jdo.option.ConnectionURL</name>
    <value>jdbc:mysql://localhost:3306/hive_demo?createDatabaseIfNotExist=true&amp;useSSL=false&amp;useUnicode=true&amp;characterEncoding=utf8</value>
    <description>hive的元数据库 </description>
  </property>
  <property>
    <name>javax.jdo.option.ConnectionDriverName</name>
    <value>com.mysql.jdbc.Driver</value>
    <description>mysql的驱动jar包 </description>
  </property>
  <property>
    <name>javax.jdo.option.ConnectionUserName</name>
    <value>root</value>
    <description>设定数据库的用户名 </description>
  </property>
  <property>
    <name>javax.jdo.option.ConnectionPassword</name>
    <value>xxx</value>
    <description>设定数据库的密码</description>
   </property>
<!--zbt添加-->
   <property>
      <name>hive.exec.max.dynamic.partitions</name>
      <value>100000</value>
      <description>在所有执行MR的节点上，最大一共可以创建多少个动态分区</description>
   </property>
   <property>
      <name>hive.exec.max.dynamic.partitions.pernode</name>
      <value>100000</value>
      <description>在所有执行MR的节点上，最大可以创建多少个动态分区</description>
  </property>
</configuration>

若要在idea环境下运行要把

hdfs-site.xml

core-site.xml

hive-site.xml

放到resources文件夹中

否则hive.exec.max.dynamic.partitions.pernode，hive.exec.max.dynamic.partitions

配置不生效

2、hosts设置

若在不同网络环境下

需设置本地hosts

设置的内容为集群主机名

Ubuntu的hosts文件在 /etc 下

参考资料：(10条消息) java.lang.IllegalArgumentException: java.net.UnknownHostException: xxx_小健的博客-CSDN博客

3、远程连接服务开启

hive --service metastore

参考资料：(13条消息) hive的几种启动方式_lbl的博客-CSDN博客_hive启动

4、其他

mysql服务启动

service mysqld start

防火墙关闭

systemctl stop firewalld

二、IDEA环境准备

1、pom.xml文件

<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0"
         xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
         xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
    <parent>
        <artifactId>sparkDome1</artifactId>
        <groupId>org.example</groupId>
        <version>1.0-SNAPSHOT</version>
    </parent>
    <modelVersion>4.0.0</modelVersion>

    <artifactId>HiveAndMysql</artifactId>

    <properties>
        <maven.compiler.source>8</maven.compiler.source>
        <maven.compiler.target>8</maven.compiler.target>
        <hadoop.version>2.7.7</hadoop.version>
        <spark.version>2.1.1</spark.version>
        <scala.version>2.11</scala.version>
    </properties>

    <dependencies>
        <!--hadoop依赖-->
        <dependency>
            <groupId>org.apache.hadoop</groupId>
            <artifactId>hadoop-client</artifactId>
            <version>${hadoop.version}</version>
        </dependency>
        <dependency>
            <groupId>org.apache.hadoop</groupId>
            <artifactId>hadoop-common</artifactId>
            <version>${hadoop.version}</version>
        </dependency>
        <dependency>
            <groupId>org.apache.hadoop</groupId>
            <artifactId>hadoop-hdfs</artifactId>
            <version>${hadoop.version}</version>
        </dependency>
        <!--scala依赖-->
        <dependency>
            <groupId>org.scala-lang</groupId>
            <artifactId>scala-library</artifactId>
            <version>2.11.0</version>
        </dependency>
        <!--spark依赖-->
        <dependency>
            <groupId>org.apache.spark</groupId>
            <artifactId>spark-core_${scala.version}</artifactId>
            <version>${spark.version}</version>
        </dependency>
        <dependency>
            <groupId>org.apache.spark</groupId>
            <artifactId>spark-sql_${scala.version}</artifactId>
            <version>${spark.version}</version>
        </dependency>
        <dependency>
            <groupId>org.apache.spark</groupId>
            <artifactId>spark-streaming_${scala.version}</artifactId>
            <version>${spark.version}</version>
        </dependency>
        <dependency>
            <groupId>org.apache.spark</groupId>
            <artifactId>spark-mllib_${scala.version}</artifactId>
            <version>${spark.version}</version>
        </dependency>
        <!--hive依赖-->
        <dependency>
            <groupId>org.apache.spark</groupId>
            <artifactId>spark-hive_${scala.version}</artifactId>
            <version>${spark.version}</version>
        </dependency>
        <dependency>
            <groupId>mysql</groupId>
            <artifactId>mysql-connector-java</artifactId>
            <version>5.1.48</version>
        </dependency>
    </dependencies>

    <build>
        <plugins>
            <plugin>
                <groupId>org.scala-tools</groupId>
                <artifactId>maven-scala-plugin</artifactId>
                <version>2.15.2</version>
                <executions>
                    <execution>
                        <id>scala-compile</id>
                        <goals>
                            <goal>compile</goal>
                        </goals>
                        <configuration>
                            <!--includes是一个数组，包含要编译的code-->
                            <includes>
                                <include>**/*.scala</include>
                            </includes>
                        </configuration>
                    </execution>
                    <execution>
                        <id>scala-test-compile</id>
                        <goals>
                            <goal>testCompile</goal>
                        </goals>
                    </execution>
                </executions>
            </plugin>
        </plugins>
    </build>


</project>

2、Hadoop环境

window下运行需要准备Hadoop环境

在代码编写中指定hadoop.home.dir

System.setProperty("hadoop.home.dir","........")

3、其他

Scala插件依赖需先下载好

注意环境与集群对应，本文档的环境为Scala-11

三、代码编写

1、全量抽取

import org.apache.spark.sql.SparkSession

/**
   * mysql->hive 全量抽取
   */
   object ShopTest {

  def main(args: Array[String]): Unit = {
	//设置用户名，防止因为权限不足无法创建文件
    System.setProperty("HADOOP_USER_NAME", "root")
    //获取实例对象
    val spark = SparkSession.builder()
      .appName("ShopTest")
      .master("local[*]")
      .config("spark.sql.warehouse.dir", "hdfs://xx.xxx.x.x:8020/user/hive/warehouse")
      .config("hive.metastore.uris", "thrift://xx.xxx.x.x:9083")
      .enableHiveSupport()
      .getOrCreate()
    //jdbc连接配置
    val mysqlMap = Map(
      "url" -> "jdbc:mysql://xx.xxx.x.x:3306/clown_db?useSSL=false",
      "user" -> "root",
      "password" -> "xxx",
      "driver" -> "com.mysql.jdbc.Driver"
    )
    //使用jdbc抽取mysql表数据
    val inputTable = spark.read.format("jdbc")
      .options(mysqlMap)
      .option("dbtable", "EcData_tb_1")
      .load()
    
    //    inputTable.show()
      
    //将mysql表数据创建为临时表
    inputTable.createOrReplaceTempView("inputTable")
    //hive动态分区开启
    spark.sqlContext.sql("set hive.exec.dynamic.partition = true")
    //hive分区模式设置，默认为strict严格模式，若设置分区必须要有一个静态分区
    //需要设置为nonstrict模式，可以都是动态分区
    spark.sqlContext.sql("set hive.exec.dynamic.partition.mode = nonstrict")
    //hive分区数设置，目前版本已无法在程序中设置，参考上文Hive环境准备-配置文件准备
    spark.sqlContext.sql("set hive.exec.max.dynamic.partitions.pernode = 10000")
    spark.sqlContext.sql("set hive.exec.max.dynamic.partitions = 10000")

    // mysql表结构，通过desc table tb_name;命令可获取

    /*    +-------------+---------+------+-----+---------+-------+
        | Field       | Type    | Null | Key | Default | Extra |
        +-------------+---------+------+-----+---------+-------+
        | InvoiceNo   | text    | YES  |     | NULL    |       |
        | StockCode   | text    | YES  |     | NULL    |       |
        | Description | text    | YES  |     | NULL    |       |
        | Quantity    | int(11) | YES  |     | NULL    |       |
        | InvoiceDate | text    | YES  |     | NULL    |       |
        | UnitPrice   | double  | YES  |     | NULL    |       |
        | CustomerID  | int(11) | YES  |     | NULL    |       |
        | Country     | text    | YES  |     | NULL    |       |
        +-------------+---------+------+-----+---------+-------+*/
    
    //于hive数据库，ods层中创建表
    spark.sqlContext.sql(
      """
        |create table if not exists clown_test_db.ShopTest_ods_tb_1
        |(
        | InvoiceNo string ,
        | StockCode string ,
        | Description string ,
        | Quantity int ,
        | InvoiceDate string ,
        | UnitPrice double ,
        | CustomerID int ,
        | Country string
        |)
        |partitioned by (country_pid string,customer_pid int)
        |row format delimited
        |fields terminated by '\t'  //本数据中字段值存在','不能用','作为分隔符
        |lines terminated by '\n'
        |stored as textfile
        |""".stripMargin)
      
    //使用sql-insert into 语句将mysql数据全部导入hive表中
    spark.sqlContext.sql(
      """
        |insert into table clown_test_db.ShopTest_ods_tb_1 partition(country_pid,customer_pid)
        |select *,Country,CustomerID from inputTable
        |""".stripMargin)

  }

}

2、增量抽取

import java.text.SimpleDateFormat

import org.apache.spark.sql.{SaveMode, SparkSession}

/**

 * hive_ods -> hive_dwd 增量抽取
   */
   object ShopTest2 {

  def main(args: Array[String]): Unit = {
    System.setProperty("HADOOP_USER_NAME","root")

    val spark = SparkSession.builder()
      .appName("ShopTest2")
      .master("local[*]")
      .config("spark.sql.warehouse.dir","hdfs://xx.xxx.x.x:8020/user/hive/warehouse")
      .config("hive.metastore.uris","thrift://xx.xxx.x.x:9083")
      .enableHiveSupport()
      .getOrCreate()
    
    /*    +-------------+---------+------+-----+---------+-------+
    | Field       | Type    | Null | Key | Default | Extra |
    +-------------+---------+------+-----+---------+-------+
    | InvoiceNo   | text    | YES  |     | NULL    |       |
    | StockCode   | text    | YES  |     | NULL    |       |
    | Description | text    | YES  |     | NULL    |       |
    | Quantity    | int(11) | YES  |     | NULL    |       |
    | InvoiceDate | text    | YES  |     | NULL    |       |
    | UnitPrice   | double  | YES  |     | NULL    |       |
    | CustomerID  | int(11) | YES  |     | NULL    |       |
    | Country     | text    | YES  |     | NULL    |       |
    +-------------+---------+------+-----+---------+-------+*/

	//隐式转换，sql方法导入
    import spark.implicits._
    import org.apache.spark.sql.functions._
    
    spark.sqlContext.sql("set hive.exec.dynamic.partition = true")
    spark.sqlContext.sql("set hive.exec.dynamic.partition.mode = nonstrict")
    
    //直接通过sql语句获取到hive ods层中的表数据
    val inputData = spark.sqlContext.sql("select * from clown_test_db.ShopTest_ods_tb_1")
    
    //设置时间条件
    val timeStr = "2011/01/01 00:00"
    val timeTemp = new SimpleDateFormat("yyyy/MM/dd HH:mm").parse(timeStr).getTime//单位为ms
    println(timeTemp)
    
    //未转换前的数据格式为：12/8/2010 9:53
    val timeFormat = inputData
      .withColumn("InvoiceDate",unix_timestamp($"InvoiceDate","MM/dd/yyyy HH:mm"))//时间戳获取，单位为s
      .where(s"InvoiceDate>$timeTemp/1000")//增量条件判断
      .withColumn("InvoiceDate",from_unixtime($"InvoiceDate","yyyy/MM/dd HH:mm"))//时间格式转换
      .where("Country='United Kingdom' or Country = 'Finland'")//筛选出国家名为United Kingdom 或 Finland的数据
    
    //由于该ods层表与目标dwd层表结构相同，直接用like语句创建结构相同的dwd表
    spark.sqlContext.sql(
      """
        |create table if not exists clown_dwd_db.shoptest_dwd_tb_1
        |like clown_test_db.ShopTest_ods_tb_1
        |""".stripMargin)


    //使用sparkSql算子将数据由ods表数据增量抽取到dwd表中
    timeFormat.write.format("hive")
      .mode(SaveMode.Append)
      .insertInto("clown_dwd_db.shoptest_dwd_tb_1")
  }

}

3、数据清洗

import org.apache.spark.sql.{SaveMode, SparkSession}

/**
 * hive_dwd->hive_dwd 缺失值剔除与填充
 */
object ShopTest3 {
 /*+-------------+---------+------+-----+---------+-------+
   | Field       | Type    | Null | Key | Default | Extra |
   +-------------+---------+------+-----+---------+-------+
   | InvoiceNo   | text    | YES  |     | NULL    |       |
   | StockCode   | text    | YES  |     | NULL    |       |
   | Description | text    | YES  |     | NULL    |       |
   | Quantity    | int(11) | YES  |     | NULL    |       |
   | InvoiceDate | text    | YES  |     | NULL    |       |
   | UnitPrice   | double  | YES  |     | NULL    |       |
   | CustomerID  | int(11) | YES  |     | NULL    |       |
   | Country     | text    | YES  |     | NULL    |       |
   +-------------+---------+------+-----+---------+-------+*/
def main(args: Array[String]): Unit = {
   System.setProperty("HADOOP_USER_NAME","root")

   val spark = SparkSession.builder()
     .appName("ShopTest3")
     .master("local[*]")
     .config("spark.sql.warehouse.dir","hdfs://xx.xxx.x.x:8020/user/hive/warehouse")
     .config("hive.metastore.uris","thrift://xx.xxx.x.x:9083")
     .enableHiveSupport()
     .getOrCreate()

   import spark.implicits._
   import org.apache.spark.sql.functions._

   spark.sqlContext.sql("set hive.exec.dynamic.partition = true")
   spark.sqlContext.sql("set hive.exec.dynamic.partition.mode = nonstrict")

   val data = spark.sqlContext.sql("select * from clown_dwd_db.shoptest_dwd_tb_1")

   spark.sqlContext.sql(
     """
       |create table if not exists clown_dwd_db.shopTest_dwd_tb_3
       |(
       | InvoiceNo string ,
       | StockCode string ,
       | Description string ,
       | Quantity int ,
       | InvoiceDate string ,
       | UnitPrice double ,
       | CustomerID int ,
       | Country string
       |)
       |partitioned by (country_pid string)
       |row format delimited
       |fields terminated by '\t'
       |lines terminated by '\n'
       |stored as textfile
       |""".stripMargin)

  //使用na.fill对缺失值进行填充
  //使用na.drop对缺失值进行剔除
   data.na.fill(
     Map(
       "Country"->"Country_Null",
       "CustomerID"->0
     )
   )
     .na.drop(
     Seq("UnitPrice","Quantity")
   )      .selectExpr("InvoiceNo","StockCode","Description","Quantity","InvoiceDate","UnitPrice","CustomerID","Country","Country")//由于数据中存在分区表字段，且该字段关联数据已改变，需要重新进行赋值
     .limit(10000)
     .write
     .format("hive")
     .mode(SaveMode.Append)
     .insertInto("clown_dwd_db.shopTest_dwd_tb_3")

  }
}

4、指标计算

import org.apache.spark.sql.SparkSession

/**
  * sparkSql算子实现指标计算
  */
object ShopTest4 {
  /*    +-------------+---------+------+-----+---------+-------+
| Field       | Type    | Null | Key | Default | Extra |
+-------------+---------+------+-----+---------+-------+
| InvoiceNo   | text    | YES  |     | NULL    |       |
| StockCode   | text    | YES  |     | NULL    |       |
| Description | text    | YES  |     | NULL    |       |
| Quantity    | int(11) | YES  |     | NULL    |       |
| InvoiceDate | text    | YES  |     | NULL    |       |
| UnitPrice   | double  | YES  |     | NULL    |       |
| CustomerID  | int(11) | YES  |     | NULL    |       |
| Country     | text    | YES  |     | NULL    |       |
+-------------+---------+------+-----+---------+-------+*/
  def main(args: Array[String]): Unit = {

    System.setProperty("HADOOP_USER_NAME", "root")
    
    val spark = SparkSession.builder()
      .appName("ShopTest4")
      .master("local[*]")
      .config("spark.sql.warehouse.dir", "hdfs://xx.xxx.x.x:8020/user/hive/warehouse")
      .config("hive.metastore.uris", "thrift://xx.xxx.x.x:9083")
      .enableHiveSupport()
      .getOrCreate()
    
    import spark.implicits._
    import org.apache.spark.sql.functions._
    
    spark.sqlContext.sql("set hive.exec.dynamic.partition = true")
    spark.sqlContext.sql("set hive.exec.dynamic.partition.mode = nonstrict")
    
    val data = spark.sqlContext.sql("select * from clown_dwd_db.shopTest_dwd_tb_3")
    
    /**
     * 统计每个国家的客户数，输出结果。
     * 排序后输出客户最多的10个国家
     */
    
        data.dropDuplicates("CustomerID","Country")//去重
          .withColumn("x",lit(1))//添加一列数据都为1
          .groupBy("Country")//聚合国家字段
          .sum("x")//对1数据字段进行累加
          .show(20)
    
    /**
     * 统计各个国家的总销售额分布情况
     */
    
    data.withColumn("x", $"Quantity" * $"UnitPrice")//添加销售额字段，值为数量*单价
      .groupBy("Country")//聚合国家字段
      .sum("x")//计算总销售额
      .withColumn("sum(x)", round($"sum(x)", 2))//对结果字段进行四舍五入到两位，但round会对最后一位0省略，最好使用其他函数
    /*若题目要求输出格式可进行rdd转换
      .rdd
      .map(x=>x.mkString(","))
      .foreach(println(_))
    */


    /**
     * 统计每种商品的销量，输出结果
     * 排序后输出销量最高的10种商品
     */
    data.groupBy("StockCode")//聚合商品编码字段
      .sum("Quantity")//计算销量
      .coalesce(1)//将spark分区设置为1，防止后面排序混乱
      .orderBy(desc("sum(Quantity)"))//由大到小排序
      .show(10)
    
    /**
     * 统计月销售额随时间的变化趋势
     * [月份,销售额]
     */
    data.withColumn("InvoiceDate",substring_index($"InvoiceDate","/",2))//由于数据在增量抽取阶段已进行时间格式转换，可直接进行切割得出 年份/月份 的格式，substring_index与split不同
      .withColumn("x",$"Quantity"*$"UnitPrice")//计算销售额
      .groupBy("InvoiceDate")//对月份进行聚合
      .sum("x")//计算总销售额
      .coalesce(1)//设置spark分区为1
      .orderBy(desc("InvoiceDate"))//由大到小排序
      .withColumn("sum(x)",round($"sum(x)",2))//四舍五入到2位
      .show(100)
    
    /**
     * 统计商品描述中，排名前300（Top300）的热门关键词
     */
    data.select(col("Description"))//商品将描述字段单独查询
      .flatMap(x=>x.toString().split("\\W"))//进行flatMap 切割后展平，切割\\W为正则匹配模式，匹配所有符号
      .withColumn("x",lit(1))//增加1的数据列
      .groupBy("value")//展平后字段名为value，进行聚合
      .sum("x")//累加1数据
      .where("value != '' ")//筛除空白数据
      .coalesce(1)//设置spark分区为1
      .orderBy(desc("sum(x)"))//由大到小排序
      .show(300)//展示300条

  }

}

import org.apache.spark.sql.SparkSession

/**
  *  sql语句实现指标计算
  */

object ShopTest5 {
  /*    +-------------+---------+------+-----+---------+-------+
| Field       | Type    | Null | Key | Default | Extra |
+-------------+---------+------+-----+---------+-------+
| InvoiceNo   | text    | YES  |     | NULL    |       |
| StockCode   | text    | YES  |     | NULL    |       |
| Description | text    | YES  |     | NULL    |       |
| Quantity    | int(11) | YES  |     | NULL    |       |
| InvoiceDate | text    | YES  |     | NULL    |       |
| UnitPrice   | double  | YES  |     | NULL    |       |
| CustomerID  | int(11) | YES  |     | NULL    |       |
| Country     | text    | YES  |     | NULL    |       |
+-------------+---------+------+-----+---------+-------+*/
  def main(args: Array[String]): Unit = {

    System.setProperty("HADOOP_USER_NAME", "root")
    
    val spark = SparkSession.builder()
      .appName("ShopTest5")
      .master("local[*]")
      .config("spark.sql.warehouse.dir", "hdfs://xx.xxx.x.x:8020/user/hive/warehouse")
      .config("hive.metastore.uris", "thrift://xx.xxx.x.x:9083")
      .enableHiveSupport()
      .getOrCreate()
    
    import spark.implicits._
    import org.apache.spark.sql.functions._
    
    spark.sqlContext.sql("set hive.exec.dynamic.partition = true")
    spark.sqlContext.sql("set hive.exec.dynamic.partition.mode = nonstrict")
    
    val data = spark.sqlContext.sql("select * from clown_dwd_db.shopTest_dwd_tb_3")
    
    data.createOrReplaceTempView("dataTable")
    
    /**
     * 统计每个国家的客户数，输出结果。
     * 排序后输出客户最多的10个国家
     */

   //对去重后的Country，CustomerID进行聚合统计即可得出各个国家的客户数
   spark.sqlContext.sql(
     """
       |select Country,count(distinct Country,CustomerID) from dataTable group by Country
       |""".stripMargin)
     .show()

    /**
     * 统计各个国家的总销售额分布情况
     */
    
    spark.sqlContext.sql(
      """
        |select Country ,round(sum(Quantity*UnitPrice),2)
        |from dataTable
        |group by Country
        |""".stripMargin)
      .show()
    
    /**
     * 统计每种商品的销量，输出结果
     * 排序后输出销量最高的10种商品
     */
    
    spark.sqlContext.sql(
      """
        |select StockCode,round(sum(Quantity*UnitPrice),2) as xl
        |from dataTable
        |group by StockCode
        |order by xl desc
        |""".stripMargin)
      .show(10)
    
    /**
     * 统计月销售额随时间的变化趋势
     * [月份,销售额]
     */
    
    //group by执行优先度可能高于 as 重命名，因此as后的名字无法用于group by 聚合
    spark.sqlContext.sql(
      """
        |select substring_index(InvoiceDate,"/",2) as time,round(sum(Quantity*UnitPrice),2) as sum
        |from dataTable
        |group by substring_index(InvoiceDate,"/",2)
        |order by substring_index(InvoiceDate,"/",2)
        |""".stripMargin)
      .show()
    
    /**
     * 统计商品描述中，排名前300（Top300）的热门关键词
     */
    
    //目前认为该题用sql解法没有必要
      //- - 
  }

}

四、其他

1、hive分区的增删改查

参考资料：(15条消息) HIve学习：Hive分区修改_u011047968的专栏-CSDN博客_hive修改分区

hive表新增分区：[]内的不必要

alter table tb_name add partition (pid1 = ‘’,pid2 = ) [location ‘xxx’] 

多个分区

alter table tb_name add partition (pid1 = ‘’,pid2 = ) partition (pid1 = ‘’,pid2 = ) [location ‘xxx’] 

hive表修改分区：

alter table tb_name partition(pid1='') rename to partition(pid1='');/*修改分区名*/
alter table tb_name partition(pid1='') set location 'hdfs://master:8020/....';/*修改分区路径，注意使用绝对路径*/  
alter table tb_name partition column (pid1 string);/*修改分区字段数据类型*/

hive表删除分区：

alter table tb_name drop partition (pid1 = ‘’,pid2 = )[ partition (pid1 = ‘’,pid2 = )…] 

hive分区值查询：

show partitions tb_name;

2、spark打包运行

命令：

spark-submit --class ShopTest4 --master spark://xx.xxx.x.x:7077 --executor-memory 512m --total-executor-cores 1 untitled-1.0-SNAPSHOT.jar

若使用了jdbc连接，需要指明驱动jar包 mysql-connector-java-5.1.48.jar

spark-submit --jars mysql-connector-java-5.1.48.jar --class ShopTest --master spark://xx.xxx.x.x:7077 --executor-memory 512m --total-executor-cores 1 untitled-1.0-SNAPSHOT.jar

或者将mysql驱动放至 $‘spark_home’/jars 目录下

3、时间格式

时间模式字符串用来指定时间格式。在此模式中，所有的 ASCII 字母被保留为模式字母，定义如下：

4、Scala正则表达式

Scala 的正则表达式继承了 Java 的语法规则，Java 则大部分使用了 Perl 语言的规则。

下表我们给出了常用的一些正则表达式规则：（注意：\需要转义，算子中写为\，sql语句中写为\\\)

正则表达式实例

常用可以应用正则的函数：

.split("")切割字符串

.regexp_extract(string subject, string pattern, int index) 将字符串subject按照pattern正则表达式的规则拆分，返回index指定的字符

.regexp_replace(string A, string B, string C) 将字符串A中的符合Java正则表达式B的部分替换为C

.equals("")匹配

5、SQL like与rlike

like为通配符匹配，不是正则

%：匹配零个及多个任意字符

_：与任意单字符匹配

[]：匹配一个范围

[^]：排除一个范围

rlike为正则匹配

regexp与rlike功能相似

参考资料：(15条消息) sparksql 正则匹配总结_Andrew LD-CSDN博客_spark 正则表达式

6、中文数据

关于csv文件若包含中文，可在读取时设置option参数

/**
 * 注意option的设置
 * 读取本地文件需要加上file:///否则默认读hdfs文件
 */
val inputData = spark.sqlContext.read.format("csv")
  .option("sep","\t")
  .option("encoding","GBK")
  .option("header","true")
 
.load("file:///C:\\Users\\61907\\Desktop\\BigData\\Spark\\sparkDome1\\HiveAndMysql\\src\\main\\resources\\cov19.csv")

jdbc读取数据库数据时，若有中文需设置jdbc连接参数

&useUnicode=true&characterEncoding=utf8

//    jdbc中文编码设置
    val mysqlMap = Map(
      "url"->"jdbc:mysql://xx.xxx.x.x:3306/clown_db?useSSL=false&useUnicode=true&characterEncoding=utf8",
      "user"->"root",
      "password"->"xxx",
      "driver"->"com.mysql.jdbc.Driver"
    )

关于hive中存储中文数据,中文注释,中文分区(索引)

Ⅰ~Ⅲ参考资料：

(16条消息) hive设置中文编码格式utf-8_2020xyz的博客-CSDN博客_hive建表指定编码格式

(16条消息) hive修改使用utf8编码支持中文字符集_那又怎样？的博客-CSDN博客_hive默认字符集编码

Ⅰ.元数据库设置

元数据库需设置为utf-8编码

##创建hive元数据库hive，并指定utf-8编码格式
mysql>create database hive DEFAULT CHARSET utf8 COLLATE utf8_general_ci;


##修改已存在的hive元数据库，字符编码格式为utf-8
mysql>alter database hive character set utf8;     


##进入hive元数据库
mysql>use hive;

##查看元数据库字符编码格式
mysql>show variables like 'character_set_database';  

Ⅱ.相关表设置

1).修改字段注释字符集

mysql>alter table COLUMNS_V2 modify column COMMENT varchar(256) character set utf8;

2).修改表注释字符集

mysql>alter table TABLE_PARAMS modify column PARAM_VALUE varchar(4000) character set utf8;

类似的，PARAM_KEY若需要中文也可设置为utf8

3).修改分区表参数，以支持分区能够用中文表示

mysql>alter table PARTITION_PARAMS modify column PARAM_VALUE varchar(4000) character set utf8;


mysql>alter table PARTITION_KEYS modify column PKEY_COMMENT varchar(4000) character set utf8;

另外,PARTITIONS表中存放分区名的字段也需要修改为utf8

mysql>alter table PARTITIONS modify column PART_name varchar(4000) character set utf8;

4).修改索引注解

mysql>alter table INDEX_PARAMS modify column PARAM_VALUE varchar(250) character set utf8;

Ⅲ.hive-site.xml配置文件设置

需要在jdbc连接中设置支持中文编码

&useSSL=false&useUnicode=true&characterEncoding=utf8

其中&需要使用&转义

参考资料：(16条消息) 【已解决】The reference to entity “useSSL” must end with the ‘;’ delimiter_清宵尚温的博客-CSDN博客

/opt/hive/conf/hive-site.xml:

<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<?xml-stylesheet type="text/xsl" href="configuration.xsl"?>
<configuration>
  <property>
    <name>javax.jdo.option.ConnectionURL</name>
    <value>jdbc:mysql://localhost:3306/hive_demo?createDatabaseIfNotExist=true&amp;useSSL=false&amp;useUnicode=true&amp;characterEncoding=utf8</value>
    <description>hive的元数据库 </description>
  </property>
  <property>
    <name>javax.jdo.option.ConnectionDriverName</name>
    <value>com.mysql.jdbc.Driver</value>
    <description>mysql的驱动jar包 </description>
  </property>
  <property>
    <name>javax.jdo.option.ConnectionUserName</name>
    <value>root</value>
    <description>设定数据库的用户名 </description>
  </property>
  <property>
    <name>javax.jdo.option.ConnectionPassword</name>
    <value>xxx</value>
    <description>设定数据库的密码</description>
   </property>
<!--zbt添加-->
   <property>
      <name>hive.exec.max.dynamic.partitions</name>
      <value>100000</value>
      <description>在所有执行MR的节点上，最大一共可以创建多少个动态分区</description>
   </property>
   <property>
      <name>hive.exec.max.dynamic.partitions.pernode</name>
      <value>100000</value>
      <description>在所有执行MR的节点上，最大可以创建多少个动态分区</description>
  </property>
</configuration>

Ⅳ.未解决问题

hdfs文件系统中显示

虽然正常显示中文但在文件夹中会出现

Path does not exist on HDFS or WebHDFS is disabled. Please check your path or enable WebHDFS

可能是中文路径导致的错误，但该错误目前未影响到分区表的正常操作，具体影响仍需实验。

Ⅴ.暴力脚本- -

参考资料：(16条消息) hive分区字段含中文导致的报错_一定要努力努力再努力的博客-CSDN博客_hive分区字段是中文

alter database hive_meta default character set utf8;
alter table BUCKETING_COLS default character set utf8;
alter table CDS default character set utf8;
alter table COLUMNS_V2 default character set utf8;
alter table DATABASE_PARAMS default character set utf8;
alter table DBS default character set utf8;
alter table FUNCS default character set utf8;
alter table FUNC_RU default character set utf8;
alter table GLOBAL_PRIVS default character set utf8;
alter table PARTITIONS default character set utf8;
alter table PARTITION_KEYS default character set utf8;
alter table PARTITION_KEY_VALS default character set utf8;
alter table PARTITION_PARAMS default character set utf8;
-- alter table PART_COL_STATS default character set utf8;
alter table ROLES default character set utf8;
alter table SDS default character set utf8;
alter table SD_PARAMS default character set utf8;
alter table SEQUENCE_TABLE default character set utf8;
alter table SERDES default character set utf8;
alter table SERDE_PARAMS default character set utf8;
alter table SKEWED_COL_NAMES default character set utf8;
alter table SKEWED_COL_VALUE_LOC_MAP default character set utf8;
alter table SKEWED_STRING_LIST default character set utf8;
alter table SKEWED_STRING_LIST_VALUES default character set utf8;
alter table SKEWED_VALUES default character set utf8;
alter table SORT_COLS default character set utf8;
alter table TABLE_PARAMS default character set utf8;
alter table TAB_COL_STATS default character set utf8;
alter table TBLS default character set utf8;
alter table VERSION default character set utf8;
alter table BUCKETING_COLS convert to character set utf8;
alter table CDS convert to character set utf8;
alter table COLUMNS_V2 convert to character set utf8;
alter table DATABASE_PARAMS convert to character set utf8;
alter table DBS convert to character set utf8;
alter table FUNCS convert to character set utf8;
alter table FUNC_RU convert to character set utf8;
alter table GLOBAL_PRIVS convert to character set utf8;
alter table PARTITIONS convert to character set utf8;
alter table PARTITION_KEYS convert to character set utf8;
alter table PARTITION_KEY_VALS convert to character set utf8;
alter table PARTITION_PARAMS convert to character set utf8;
-- alter table PART_COL_STATS convert to character set utf8;
alter table ROLES convert to character set utf8;
alter table SDS convert to character set utf8;
alter table SD_PARAMS convert to character set utf8;
alter table SEQUENCE_TABLE convert to character set utf8;
alter table SERDES convert to character set utf8;
alter table SERDE_PARAMS convert to character set utf8;
alter table SKEWED_COL_NAMES convert to character set utf8;
alter table SKEWED_COL_VALUE_LOC_MAP convert to character set utf8;
alter table SKEWED_STRING_LIST convert to character set utf8;
alter table SKEWED_STRING_LIST_VALUES convert to character set utf8;
alter table SKEWED_VALUES convert to character set utf8;
alter table SORT_COLS convert to character set utf8;
alter table TABLE_PARAMS convert to character set utf8;
alter table TAB_COL_STATS convert to character set utf8;
alter table TBLS convert to character set utf8;
alter table VERSION convert to character set utf8;
-- alter table PART_COL_STATS convert to character set utf8;
SET character_set_client = utf8 ;
-- SET character_set_connection = utf8 ;

-- alter table PART_COL_STATS convert to character set utf8;
SET character_set_database = utf8 ;
SET character_set_results = utf8 ;
SET character_set_server = utf8 ;
-- SET collation_connection = utf8 ;
-- SET collation_database = utf8 ;
-- SET collation_server = utf8 ;
SET NAMES 'utf8';

只复制了博客中修改表字段的部分

看看就好，最好还是根据需求修改。

Ⅵ.实例

import org.apache.spark.sql.{SaveMode, SparkSession}

object CNHivePartitionTest {

  def main(args: Array[String]): Unit = {

    System.setProperty("hadoop.home.dir", "D:\\BaiduNetdiskDownload\\hadoop-2.7.3")
    System.setProperty("HADOOP_USER_NAME", "root")

    val spark = SparkSession.builder()
      .appName("Cov19DataDome4")
      .master("local[*]")
      .config("spark.sql.warehouse.dir", "hdfs://xx.xxx.x.x:8020/user/hive/warehouse")
      .config("hive.metastore.uris", "thrift://xx.xxx.x.x:9083")
      .enableHiveSupport()
      .getOrCreate()

    import spark.implicits._
    import org.apache.spark.sql.functions._

    spark.sqlContext.sql("set hive.exec.dynamic.partition = true")
    spark.sqlContext.sql("set hive.exec.dynamic.partition.mode = nonstrict")

    val mysqlMap = Map(
      "url" -> "jdbc:mysql://xx.xxx.x.x:3306/clown_db?useSSL=false&useUnicode=true&characterEncoding=utf8",
      "user" -> "root",
      "password" -> "xxx",
      "driver" -> "com.mysql.jdbc.Driver"
    )

    val mysqlData = spark.read.format("jdbc")
      .options(mysqlMap)
      .option("dbtable","tc_hotel2")
      .load()

    spark.sqlContext.sql(
      """
        |create table if not exists clown_test_db.CNTest
        |(
        |  `hname` string,
        |  `hbrand` string,
        |  `province` string,
        |  `city` string,
        |  `starlevel` string,
        |  `rating` string,
        |  `comment_count` string,
        |  `price` string
        |)
        |partitioned by (pid string)
        |row format delimited
        |fields terminated by '\t'
        |lines terminated by '\n'
        |stored as textfile
        |""".stripMargin)

    mysqlData
      .select(col("*"),col("province"))
      .write
      .format("hive")
      .mode(SaveMode.Append)
      .insertInto("clown_test_db.CNTest")
  }

}

7、表连接join/union

参考资料：https://blog.csdn.net/m0_37809146/article/details/91282446

val tb1 = spark.read.format("jdbc")
  .options(mysqlMap)
  .option("dbtable", "cov19_test_tb")
  .load()

val tb2 = spark.read.format("jdbc")
  .options(mysqlMap)
  .option("dbtable", "cov19_test_tb_2")
  .load()
  .withColumnRenamed("", "")

/**
 * inner 交集，只会联合给出字段都存在的数据
 */
tb1.join(tb2, Seq("provinceName", "cityName"), "inner")
//      .show(100)
/**
 * right 右链接，展示右边表所有数据
 */
tb1.join(tb2, Seq("provinceName", "cityName"), "right")
//      .show(100)
/**
 * left 左链接，展示左边表所有数据
 */
tb1.join(tb2, Seq("provinceName", "cityName"), "left")
//      .show(100)

val testTb1 = tb1.withColumnRenamed("cityName", "tb1CN")
val testTb2 = tb2.withColumnRenamed("cityName", "tb1CN")

//默认 inner连接，进行连接的条件字段必须两边表都存在
testTb1.join(testTb2, "tb1CN")
//      .show()


/**
 * right_outer 右外连接，相当于左连接
 */
tb1.join(tb2, Seq("provinceName", "cityName"), "right_outer")
//      .show(100)

/**
 * left_outer 左外连接，相当于右连接
 */
tb1.join(tb2, Seq("provinceName", "cityName"), "left_outer")
//      .show(100)

/**
 * 外连接 类似把左右连接出的集合加起来- -
 */
tb1.join(tb2, Seq("provinceName", "cityName"), "outer")
//      .show(100)

/**
 * 全连接
 */
tb1.join(tb2, Seq("provinceName", "cityName"), "full")
//      .show(100)

/**
 * 全外连接
 */
tb1.join(tb2, Seq("provinceName", "cityName"), "full_outer")
//      .show(100)

/**
 * 交集
 */
tb1.join(tb2, Seq("provinceName", "cityName"), "left_semi")
  .show(100)

/**
 * 差集
 */
tb1.join(tb2, Seq("provinceName", "cityName"), "left_anti")
  .show(100)


/**
 * https://blog.csdn.net/wcc27857285/article/details/86439313
 * 其他知识点：
 * HAVING 子句
 * 在 SQL 中增加 HAVING 子句原因是，WHERE 关键字无法与聚合函数一起使用。
 *
 * SQL HAVING 语法
 * SELECT column_name, aggregate_function(column_name)
 * FROM table_name
 * WHERE column_name operator value
 * GROUP BY column_name
 * HAVING aggregate_function(column_name) operator value
 *
 *
 * --- JOIN ON
 * JOIN写连接字段
 * ON写匹配条件
 *
 */

8、自定义UDF，UDAF函数

Spark 2.4.0编程指南–Spark SQL UDF和UDAF-阿里云开发者社区 (aliyun.com)

(17条消息) Spark 2.3.0 用户自定义聚合函数UserDefinedAggregateFunction和Aggregator_L-CSDN博客

(17条消息) UDF和UDAF，UDTF的区别_山海-CSDN博客_udf和udtf区别

[(17条消息) Spark] 自定义函数 udf & pandas_udf_風の唄を聴け的博客-CSDN博客_pandas spark udf

9、数据集获取

UCI机器学习知识库：包括近300个不同大小和类型的数据集，可用于分类、回归、聚类和推荐系统任务。数据集列表位于：http://archive.ics.uci.edu/ml/

Amazon AWS公开数据集：包含的通常是大型数据集，可通过Amazon S3访问。这些数据集包括人类基因组项目、Common Craw网页语料库、维基百科数据和Google Books Ngrams。相关信息可参见：http://aws.amazon.com/publicdatasets/

Kaggle：这里集合了Kaggle举行的各种机器学习竞赛所用的数据集。它们覆盖分类、回归、排名、推荐系统以及图像分析领域，可从

Competitions区域下载：http://www.kaggle.com/competitions

KDnuggets：这里包含一个详细的公开数据集列表，其中一些上面提到过的。该列表位于：http://www.kdnuggets.com/datasets/index.html

10、数仓分层概念

参考资料：(10条消息) 数据仓库–数据分层（ETL、ODS、DW、APP、DIM）_hello_java_lcl的博客-CSDN博客_dim层

五、实战复盘

1、2022/1/3

题目：

数据源：

csv文件（未修改）

mysql表格（增加脏数据）

环境准备：

1.mysql数据表格 2.hive目标表 3.pom文件

完成速度：

3h+

遇到问题：

1.data->mysql，数据保存

SaveMode.Overwrite 保存至mysql数据库，不仅会覆盖数据格式，字段名也会被覆盖

在做题途中遇到了保存SaveMode.Append失败的错误，修改为Overwrite 不报错，原因不明

是否解决： ?

出现错误

Unknown column 'sum' in 'field list'

原因是字段名与mysql数据库目标表中的字段名不同

修改字段名相同即可

.withColumnRenamed("sum","total_price")

在hive中是否有相同特性？

2.Join等表连接的使用

Join，union仍不熟悉 select子查询也比较生疏

是否解决： ?？

join理解下图足够

union联合要求字段相同 否则报错

3.Date计算

参考资料：https://blog.csdn.net/wybshyy/article/details/52064337

使用datediff不需要转换时间格式

是否解决： ?

参考资料：

(18条消息) Spark-SQL常用内置日期时间函数_绿萝蔓蔓绕枝生-CSDN博客_sparksql 时间函数

(18条消息) sparksql 时间函数_OH LEI``-CSDN博客_sparksql时间函数

datediff 计算两个时间差天数 结果返回一个整数

对时间格式可能有要求例如‘2021/1/4‘这样的时间格式无法被计算（sql中，算子貌似没有这个问题）

sql写法：

spark.sql(
  """
    |select datediff('2021-1-4','2020-12-30')
    |""".stripMargin).show()

算子写法：

.withColumn("o",datediff(col("delivery_date"),col("order_date")))

months_between计算两个时间差月数 结果返回一个浮点数

sql写法：

spark.sql(
  """
    |select months_between('2021-1-4','2020-12-30')
    |""".stripMargin).show()

返回：0.16129032

若想返回整数月份可以将天数删除：

spark.sql(
  """
    |select months_between('2021-1','2020-12')
    |""".stripMargin).show()

返回：1.0

算子写法：

.withColumn("o",months_between(col("delivery_date"),col("order_date")))

直接用时间戳相减通过计算也可以

spark.sql(
  """
    |select (unix_timestamp('2022/1/1','yyyy/MM/dd') - unix_timestamp('2021/12/31','yyyy/MM/dd'))/60/60/24
    |""".stripMargin).show()