理解Spark SQL(三）—— Spark SQL程序举例
上一篇说到，在Spark 2.x当中，实际上SQLContext和HiveContext是过时的，相反是采用SparkSession对象的sql函数来操作SQL语句的。使用这个函数执行SQL语句前需要先调用DataFrame的createOrReplaceTempView注册一个临时表，所以关键是先要将RDD转换成DataFrame。实际上，在Spark中实际声明了
type DataFrame = Dataset[Row]
所以，DataFrame是Dataset[Row]的别名。RDD是提供面向低层次的API，而DataFrame/Dataset提供面向高层次的API（适合于SQL等面向结构化数据的场合）。
下面提供一些Spark SQL程序的例子。
例子一：SparkSQLExam.scala
1 package bruce.bigdata.spark.example 2 3 import org.apache.spark.sql.Row 4 import org.apache.spark.sql.SparkSession 5 import org.apache.spark.sql.types._ 6 7 object SparkSQLExam { 8 9 case class offices(office:Int,city:String,region:String,mgr:Int,target:Double,sales:Double)10 11 def main(args: Array[String]) {12 13 val spark = SparkSession14 .builder15 .appName("SparkSQLExam")16 .getOrCreate()17 18 runSparkSQLExam1(spark)19 runSparkSQLExam2(spark)20 21 spark.stop()22 23 }24 25 26 private def runSparkSQLExam1(spark: SparkSession): Unit = {27 28 import spark.implicits._29 30 val rddOffices=spark.sparkContext.textFile("/user/hive/warehouse/orderdb.db/offices/offices.txt").map(_.split("\t")).map(p=>offices(p(0).trim.toInt,p(1),p(2),p(3).trim.toInt,p(4).trim.toDouble,p(5).trim.toDouble))31 val officesDataFrame = spark.createDataFrame(rddOffices)32 33 officesDataFrame.createOrReplaceTempView("offices")34 spark.sql("select city from offices where region='Eastern'").map(t=>"City: " + t(0)).collect.foreach(println)35 36 37 }38 39 private def runSparkSQLExam2(spark: SparkSession): Unit = {40 41 import spark.implicits._42 import org.apache.spark.sql._43 import org.apache.spark.sql.types._44 45 val schema = new StructType(Array(StructField("office", IntegerType, false), StructField("city", StringType, false), StructField("region", StringType, false), StructField("mgr", IntegerType, true), StructField("target", DoubleType, true), StructField("sales", DoubleType, false)))46 val rowRDD = spark.sparkContext.textFile("/user/hive/warehouse/orderdb.db/offices/offices.txt").map(_.split("\t")).map(p => Row(p(0).trim.toInt,p(1),p(2),p(3).trim.toInt,p(4).trim.toDouble,p(5).trim.toDouble))47 val dataFrame = spark.createDataFrame(rowRDD, schema)48 49 dataFrame.createOrReplaceTempView("offices2") 50 spark.sql("select city from offices2 where region='Western'").map(t=>"City: " + t(0)).collect.foreach(println)51 52 }53 54 }
使用下面的命令进行编译：
[root@BruceCentOS4 scala]# scalac SparkSQLExam.scala
在编译之前，需要在CLASSPATH中增加路径：
export CLASSPATH=$CLASSPATH:$SPARK_HOME/jars/*:$(/opt/hadoop/bin/hadoop classpath)
然后打包成jar文件：
[root@BruceCentOS4 scala]# jar -cvf spark_exam_scala.jar bruce
然后通过spark-submit提交程序到yarn集群执行，为了方便从客户端查看结果，这里采用yarn cient模式运行。
[root@BruceCentOS4 scala]# $SPARK_HOME/bin/spark-submit --class bruce.bigdata.spark.example.SparkSQLExam --master yarn --deploy-mode client spark_exam_scala.jar
运行结果截图：
例子二：SparkSQLExam.scala（需要启动hive metastore）
1 package bruce.bigdata.spark.example 2 3 import org.apache.spark.sql.{SaveMode, SparkSession} 4 5 object SparkHiveExam { 6 7 def main(args: Array[String]) { 8 9 val spark = SparkSession10 .builder()11 .appName("Spark Hive Exam")12 .config("spark.sql.warehouse.dir", "/user/hive/warehouse")13 .enableHiveSupport()14 .getOrCreate()15 16 import spark.implicits._17 18 //使用hql查看hive数据19 spark.sql("show databases").collect.foreach(println)20 spark.sql("use orderdb")21 spark.sql("show tables").collect.foreach(println)22 spark.sql("select city from offices where region='Eastern'").map(t=>"City: " + t(0)).collect.foreach(println)23 24 //将hql查询出的数据保存到另外一张新建的hive表25 //找出订单金额超过1万美元的产品26 spark.sql("""create table products_high_sales(mfr_id string,product_id string,description string) 27 ROW FORMAT DELIMITED FIELDS TERMINATED BY '\t' LINES TERMINATED BY '\n' STORED AS TEXTFILE""")28 spark.sql("""select mfr_id,product_id,description29 from products a inner join orders b30 on a.mfr_id=b.mfr and a.product_id=b.product31 where b.amount>10000""").write.mode(SaveMode.Overwrite).saveAsTable("products_high_sales")32 33 //将HDFS文件数据导入到hive表中 34 spark.sql("""CREATE TABLE IF NOT EXISTS offices2 (office int,city string,region string,mgr int,target double,sales double ) 35 ROW FORMAT DELIMITED FIELDS TERMINATED BY '\t' LINES TERMINATED BY '\n' STORED AS TEXTFILE""")36 spark.sql("LOAD DATA INPATH '/user/hive/warehouse/orderdb.db/offices/offices.txt' INTO TABLE offices2")37 38 spark.stop()39 }40 }
使用下面的命令进行编译：
[root@BruceCentOS4 scala]# scalac SparkHiveExam.scala
使用下面的命令打包：
[root@BruceCentOS4 scala]# jar -cvf spark_exam_scala.jar bruce
使用下面的命令运行：
[root@BruceCentOS4 scala]# $SPARK_HOME/bin/spark-submit --class bruce.bigdata.spark.example.SparkHiveExam --master yarn --deploy-mode client spark_exam_scala.jar
程序运行结果：
另外上述程序运行后，hive中多了2张表：
例子三：spark_sql_exam.py
1 from __future__ import print_function 2 3 from pyspark.sql import SparkSession 4 from pyspark.sql.types import * 5 6 7 if __name__ == "__main__": 8 spark = SparkSession 9 .builder 10 .appName("Python Spark SQL exam") 11 .config("spark.some.config.option", "some-value") 12 .getOrCreate()13 14 schema = StructType([StructField("office", IntegerType(), False), StructField("city", StringType(), False), 15 StructField("region", StringType(), False), StructField("mgr", IntegerType(), True), 16 StructField("Target", DoubleType(), True), StructField("sales", DoubleType(), False)])17 18 rowRDD = spark.sparkContext.textFile("/user/hive/warehouse/orderdb.db/offices/offices.txt").map(lambda p: p.split("\t")) 19 .map(lambda p: (int(p[0].strip()), p[1], p[2], int(p[3].strip()), float(p[4].strip()), float(p[5].strip())))20 21 dataFrame = spark.createDataFrame(rowRDD, schema)22 dataFrame.createOrReplaceTempView("offices")23 spark.sql("select city from offices where region='Eastern'").show()24 25 spark.stop()
执行命令运行程序：
[root@BruceCentOS4 spark]# $SPARK_HOME/bin/spark-submit --master yarn --deploy-mode client spark_sql_exam.py
程序运行结果：
例子四：JavaSparkSQLExam.java
1 package bruce.bigdata.spark.example; 2 3 import java.util.ArrayList; 4 import java.util.List; 5 6 import org.apache.spark.api.java.JavaRDD; 7 import org.apache.spark.api.java.function.Function; 8 import org.apache.spark.api.java.function.MapFunction; 9 import org.apache.spark.sql.Dataset;10 import org.apache.spark.sql.Row;11 import org.apache.spark.sql.RowFactory;12 import org.apache.spark.sql.SparkSession;13 import org.apache.spark.sql.types.DataTypes;14 import org.apache.spark.sql.types.StructField;15 import org.apache.spark.sql.types.StructType;16 import org.apache.spark.sql.AnalysisException;17 18 19 public class JavaSparkSQLExam {20 public static void main(String[] args) throws AnalysisException {21 SparkSession spark = SparkSession22 .builder()23 .appName("Java Spark SQL exam")24 .config("spark.some.config.option", "some-value")25 .getOrCreate(); 26 27 List<StructField> fields = new ArrayList<>();28 fields.add(DataTypes.createStructField("office", DataTypes.IntegerType, false));29 fields.add(DataTypes.createStructField("city", DataTypes.StringType, false));30 fields.add(DataTypes.createStructField("region", DataTypes.StringType, false));31 fields.add(DataTypes.createStructField("mgr", DataTypes.IntegerType, true));32 fields.add(DataTypes.createStructField("target", DataTypes.DoubleType, true));33 fields.add(DataTypes.createStructField("sales", DataTypes.DoubleType, false));34 35 StructType schema = DataTypes.createStructType(fields);36 37 38 JavaRDD<String> officesRDD = spark.sparkContext()39 .textFile("/user/hive/warehouse/orderdb.db/offices/offices.txt", 1)40 .toJavaRDD();41 42 JavaRDD<Row> rowRDD = officesRDD.map((Function<String, Row>) record -> {43 String[] attributes = record.split("\t");44 return RowFactory.create(Integer.valueOf(attributes[0].trim()), attributes[1], attributes[2], Integer.valueOf(attributes[3].trim()), Double.valueOf(attributes[4].trim()), Double.valueOf(attributes[5].trim()));45 });46 47 Dataset<Row> dataFrame = spark.createDataFrame(rowRDD, schema);48 49 dataFrame.createOrReplaceTempView("offices");50 Dataset<Row> results = spark.sql("select city from offices where region='Eastern'");51 results.collectAsList().forEach(r -> System.out.println(r));52 53 spark.stop();54 }55 }
编译打包后通过如下命令执行：
[root@BruceCentOS4 spark]# $SPARK_HOME/bin/spark-submit --class bruce.bigdata.spark.example.JavaSparkSQLExam --master yarn --deploy-mode client spark_exam_java.jar
运行结果：
上面是一些关于Spark SQL程序的一些例子，分别采用了Scala/Python/Java来编写的。另外除了这三种语言，Spark还支持R语言编写程序，因为我自己也不熟悉，就不举例了。不管用什么语言，其实API都是基本一致的，主要是采用DataFrame和Dataset的高层次API来调用和执行SQL。使用这些API，可以轻松的将结构化数据转化成SQL来操作，同时也能够方便的操作Hive中的数据。