[size=large]前提Spark集群已经搭建完毕,如果不知道怎么搭建 注意提交作业,需要使用sbt打包成一个jar,然后在主任务里面添加jar包的路径远程提交即可,无须到远程集群上执行测试,本次测试使用的是Spark的Standalone方式 sbt依赖如下: [/size]
name := "spark-hello"
version := "1.0"
scalaVersion := "2.11.7"
//使用公司的私服
resolvers += "Local Maven Repository" at "http://dev.bizbook-inc.com:8083/nexus/content/groups/public/"
//使用内部仓储
externalResolvers := Resolver.withDefaultResolvers(resolvers.value, mavenCentral = false)
//Hadoop的依赖
libraryDependencies += "org.apache.hadoop" % "hadoop-client" % "2.7.1"
//Spark的依赖
libraryDependencies += "org.apache.spark" % "spark-core_2.11" % "1.4.1"
//Spark SQL 依赖
libraryDependencies += "org.apache.spark" % "spark-sql_2.11" % "1.4.1"
//java servlet 依赖
libraryDependencies += "javax.servlet" % "javax.servlet-api" % "3.0.1"
[size=large]demo1:使用Scala读取HDFS的数据:
[/size]
/** *
* Spark读取来自HDFS的数据
*/
def readDataFromHDFS(): Unit ={
//以standalone方式运行,提交到远程的spark集群上面
val conf = new SparkConf().setMaster("spark://h1:7077").setAppName("load hdfs data")
conf.setJars(Seq(jarPaths));
//得到一个Sprak上下文
val sc = new SparkContext(conf)
val textFile=sc.textFile("hdfs://h1:8020/user/webmaster/crawldb/etl_monitor/part-m-00000")
//获取第一条数据
//val data=textFile.first()
// println(data)
//遍历打印
/**
* collect() 方法 游标方式迭代收集每行数据
* take(5) 取前topN条数据
* foreach() 迭代打印
* stop() 关闭链接
*/
textFile.collect().take(5).foreach( line => println(line) )
//关闭资源
sc.stop()
}
[size=large]demo2:使用Scala 在客户端造数据,测试Spark Sql:[/size]
def mappingLocalSQL1() {
val conf = new SparkConf().setMaster("spark://h1:7077").setAppName("hdfs data count")
conf.setJars(Seq(jarPaths));
val sc = new SparkContext(conf)
val sqlContext=new SQLContext(sc);
//导入隐式sql的schema转换
import sqlContext.implicits._
val df = sc.parallelize((1 to 100).map(i => Record(i, s"val_$i"))).toDF()
df.registerTempTable("records")
println("Result of SELECT *:")
sqlContext.sql("SELECT * FROM records").collect().foreach(println)
//聚合查询
val count = sqlContext.sql("SELECT COUNT(*) FROM records").collect().head.getLong(0)
println(s"COUNT(*): $count")
sc.stop()
}
[size=large]Spark SQL 映射实体类的方式读取HDFS方式和字段,注意在Scala的Objcet最上面有个case 类定义,一定要放在
这里,不然会出问题:[/size]
[img]http://dl2.iteye.com/upload/attachment/0111/2931/7566e94d-2d93-3b5d-8e18-b5744b45587a.png[/img]
[size=large]demo2:使用Scala 远程读取HDFS文件,并映射成Spark表,以Spark Sql方式,读取top10:[/size]
val jarPaths="target/scala-2.11/spark-hello_2.11-1.0.jar"
/**Spark SQL映射的到实体类的方式**/
def mapSQL2(): Unit ={
//使用一个类,参数都是可选类型,如果没有值,就默认为NULL
//SparkConf指定master和任务名
val conf = new SparkConf().setMaster("spark://h1:7077").setAppName("spark sql query hdfs file")
//设置上传需要jar包
conf.setJars(Seq(jarPaths));
//获取Spark上下文
val sc = new SparkContext(conf)
//得到SQL上下文
val sqlContext=new SQLContext(sc);
//必须导入此行代码,才能隐式转换成表格
import sqlContext.implicits._
//读取一个hdfs上的文件,并根据某个分隔符split成数组
//然后根据长度映射成对应字段值,并处理数组越界问题
val model=sc.textFile("hdfs://h1:8020/user/webmaster/crawldb/etl_monitor/part-m-00000").map(_.split("\1"))
.map( p => ( if (p.length==4) Model(Some(p(0)), Some(p(1)), Some(p(2)), Some(p(3).toLong))
else if (p.length==3) Model(Some(p(0)), Some(p(1)), Some(p(2)),None)
else if (p.length==2) Model(Some(p(0)), Some(p(1)),None,None)
else Model( Some(p(0)),None,None,None )
)).toDF()//转换成DF
//注册临时表
model.registerTempTable("monitor")
//执行sql查询
val it = sqlContext.sql("SELECT rowkey,title,dtime FROM monitor limit 10 ")
// val it = sqlContext.sql("SELECT rowkey,title,dtime FROM monitor WHERE title IS NULL AND dtime IS NOT NULL ")
println("开始")
it.collect().take(8).foreach(line => println(line))
println("结束")
sc.stop();
}
[size=large]在IDEA的控制台,可以输出如下结果:[/size]
[img]http://dl2.iteye.com/upload/attachment/0111/2933/345f0fce-0ba8-371e-99b3-dd7273826d91.png[/img]
[b][color=green][size=large]