【云星数据---Apache Flink实战系列(精品版)】：Apache Flink批处理API详解与编程实战009--DateSet实用API详解009

DateSet的API详解九

join

def join[O](other: DataSet[O], strategy: JoinHint): UnfinishedJoinOperation[T, O]
def join[O](other: DataSet[O]): UnfinishedJoinOperation[T, O]

Creates a new DataSet by joining this DataSet with the other DataSet.

将两个DataSet进行join操作

join示例一：

执行程序：

//1.创建一个 DataSet其元素为[(Int,String)]类型
val input1: DataSet[(Int, String)] =  benv.fromElements(
(2,"zhagnsan"),(3,"lisi"),(4,"wangwu"),(5,"zhaoliu"))

//2.创建一个 DataSet其元素为[(Double, Int)]类型
val input2: DataSet[(Double, Int)] =  benv.fromElements(
(1850.98,4),(1950.98,5),(2350.98,6),(3850.98,3))

//3.两个DataSet进行join操作，条件是input1(0)==input2(1)
val result = input1.join(input2).where(0).equalTo(1)

//4.显示结果
result.collect

执行结果：

res56: Seq[((Int, String), (Double, Int))] = Buffer(
((4,wangwu),(1850.98,4)), 
((5,zhaoliu),(1950.98,5)), 
((3,lisi),(3850.98,3)))

web ui中的执行效果：

join示例二：

A Join transformation can also call a user-defined join function to process joining tuples. 
A join function receives one element of the first input DataSet and one element of the second 
input DataSet and returns exactly one element.

The following code performs a join of DataSet with custom java objects and a Tuple DataSet using 
key-selector functions and shows how to use a user-defined join function:

执行程序：

//1.定义case class
case class Rating(name: String, category: String, points: Int)

//2.定义DataSet[Rating]
val ratings: DataSet[Rating] = benv.fromElements(
Rating("moon","youny1",3),Rating("sun","youny2",4),
Rating("cat","youny3",1),Rating("dog","youny4",5))

//3.创建DataSet[(String, Double)] 
val weights: DataSet[(String, Double)] = benv.fromElements(
("youny1",4.3),("youny2",7.2),
("youny3",9.0),("youny4",1.5))

//4.使用方法进行join
val weightedRatings = ratings.join(weights).where("category").equalTo(0) {
  (rating, weight) => (rating.name, rating.points + weight._2)
}

//5.显示结果
weightedRatings.collect

程序解析：

res57: Seq[(String, Double)] = Buffer((moon,7.3), (sun,11.2), (cat,10.0), (dog,6.5))

web ui中的执行效果：

join示例三：？？？

A Join transformation can also call a user-defined join function to process joining tuples. 
A join function receives one element of the first input DataSet and one element of the second 
input DataSet and returns exactly one element.

The following code performs a join of DataSet with custom java objects and a Tuple DataSet using 
key-selector functions and shows how to use a user-defined join function:

执行程序：

case class Rating(name: String, category: String, points: Int)
val ratings: DataSet[Rating] = benv.fromElements(
Rating("moon","youny1",3),Rating("sun","youny2",4),
Rating("cat","youny3",1),Rating("dog","youny4",5))

val weights: DataSet[(String, Double)] = benv.fromElements(
("youny1",4.3),("youny2",7.2),
("youny3",9.0),("youny4",1.5))

val weightedRatings = ratings.join(weights).where("category").equalTo(0) {
  (rating, weight, out: Collector[(String, Double)]) =>
    if (weight._2 > 0.1) out.collect(rating.name, rating.points * weight._2)
}

weightedRatings.collect

join示例四：执行join操作时暗示数据大小

在执行join操作时暗示数据大小，可以帮助flink优化它的执行策略，提高执行效率。

执行程序：

//1.定义DataSet[(Int, String)]
val input1: DataSet[(Int, String)] = 
benv.fromElements((3,"zhangsan"),(2,"lisi"),(4,"wangwu"),(6,"zhaoliu"))

//2.定义 DataSet[(Int, String)]
val input2: DataSet[(Int, String)] = 
benv.fromElements((4000,"zhangsan"),(70000,"lisi"),(4600,"wangwu"),(53000,"zhaoliu"))

// 3.暗示第二个输入很小
val result1 = input1.joinWithTiny(input2).where(1).equalTo(1)
result1.collect

// 4.暗示第二个输入很大
val result2 = input1.joinWithHuge(input2).where(1).equalTo(1)
result2.collect

执行结果：

Scala-Flink> result1.collect
res12: Seq[((Int, String), (Int, String))] = Buffer(
((3,zhangsan),(4000,zhangsan)), ((2,lisi),(70000,lisi)), 
((4,wangwu),(4600,wangwu)), ((6,zhaoliu),(53000,zhaoliu)))


Scala-Flink> result2.collect
res13: Seq[((Int, String), (Int, String))] = Buffer(
((3,zhangsan),(4000,zhangsan)), ((2,lisi),(70000,lisi)), 
((4,wangwu),(4600,wangwu)), ((6,zhaoliu),(53000,zhaoliu)))

web ui中的执行效果：

join示例五：执行join操作时暗示数据大小

flink有很多种执行join的策略，你可以指定一个执行策略，以便提高执行效率。

执行程序：

//1.定义两个 DataSet
val input1: DataSet[(Int, String)] = 
benv.fromElements((3,"zhangsan"),(2,"lisi"),(4,"wangwu"),(6,"zhaoliu"))
val input2: DataSet[(Int, String)] = 
benv.fromElements((4000,"zhangsan"),(70000,"lisi"),(4600,"wangwu"),(53000,"zhaoliu"))

//2.暗示input2很小
val result1 = input1.join(input2, JoinHint.BROADCAST_HASH_FIRST).where(1).equalTo(1)

//3.显示结果
result1.collect

执行结果：

res15: Seq[((Int, String), (Int, String))] = Buffer(
((3,zhangsan),(4000,zhangsan)),
((2,lisi),(70000,lisi)), 
((4,wangwu),(4600,wangwu)),
((6,zhaoliu),(53000,zhaoliu)))

暗示项说明：

暗示有如下选项：
1.JoinHint.OPTIMIZER_CHOOSES:
    没有明确暗示，让系统自行选择。
2.JoinHint.BROADCAST_HASH_FIRST
    把第一个输入转化成一个哈希表，并广播出去。适用于第一个输入数据较小的情况。
3.JoinHint.BROADCAST_HASH_SECOND:
    把第二个输入转化成一个哈希表，并广播出去。适用于第二个输入数据较小的情况。
4.JoinHint.REPARTITION_HASH_FIRST:（defalut）
    1.如果输入没有分区，系统将把输入重分区。
    2.系统将把第一个输入转化成一个哈希表广播出去。
    3.两个输入依然比较大。
    4.适用于第一个输入小于第二个输入的情况。
5.JoinHint.REPARTITION_HASH_SECOND:
    1.如果输入没有分区，系统将把输入重分区。
    2.系统将把第二个输入转化成一个哈希表广播出去。
    3.两个输入依然比较大。
    4.适用于第二个输入小于第一个输入的情况。
6.JoinHint.REPARTITION_SORT_MERGE:
    1.如果输入没有分区，系统将把输入重分区。
    2.如果输入没有排序，系统将吧输入重排序。
    3.系统将合并两个排序好的输入。
    4.适用于一个或两个分区已经排序好的情况。