从spark WordCount demo中学习算子：mapflatMapreduceByKey

Posted 2023-02-22 nefu-ljw

文章目录

• • • spark map和flatMap
    • 应用：Word Count
    • reduceByKey的用法

spark map和flatMap

val rdd = sc.parallelize(List("coffee panda","happy panda","happiest panda party"))

（1）map

rdd.map(_.split(" ")).collect
// 等价于 rdd.map(x => x.split(" ")).collect

map结果

Array[Array[String]] = Array(Array(coffee, panda), Array(happy, panda), Array(happiest, panda, party))

（2）flatMap

rdd.flatMap(_.split(" ")).collect
// 等价于 rdd.flatMap(x => x.split(" ")).collect

flatMap结果

Array[String] = Array(coffee, panda, happy, panda, happiest, panda, party)

应用：Word Count

val rdd = sc.parallelize(List("coffee panda","happy panda","happiest panda party"))

val wordWithCount1 = rdd.flatMap(_.split(" ")).map(x=>(x,1))
wordWithCount1.collect
// Array[(String, Int)] = Array((coffee,1), (panda,1), (happy,1), (panda,1), (happiest,1), (panda,1), (party,1))
// map(x=>(x,1)) 等价于 map((_, 1))

val wordCounts = wordWithCount1.reduceByKey((x, y) => (x + y)).collect
// Array[(String, Int)] = Array((coffee,1), (happiest,1), (panda,3), (party,1), (happy,1))
// reduceByKey((x, y) => (x + y)) 等价于 reduceByKey(_+_)
// x,y代表同一个key的两个不同value

Scala允许使用”占位符”下划线”_”来替代一个或多个参数，只要这个参数值函数定义中只出现一次，Scala编译器可以推断出参数。

reduceByKey的用法

Spark的RDD的reduceByKey是使用一个相关的函数来合并每个相同key的value值的一个算子。要求前一步传入的数据必须是(key,value)类型。

Merge the values for each key using an associative and commutative reduce function. This will also perform the merging locally on each mapper before sending results to a reducer, similarly to a “combiner” in MapReduce.

https://blog.csdn.net/dong_lxkm/article/details/103125693

我理解的reduceByKey((x, y) => (x + y))，就是对于相同key的两个value，暂且将其命名为x和y，定义这两个value之间的计算关系为求和。（注意：聚合都是两两聚合的，超过两个会报错）

reduceByKey的shuffle: 相同的key移到同一个分区

MapPartitionsRDD是预聚合（groupByKey没有这种预聚合操作，所以速度慢）

同样的，比如改成reduceByKey((x, y) => if (x > y) x else y)，那么就是两个value之间取max，最终得到的结果是对于相同的多个key，只取得了它们的values中最大的那个。

// 定义映射表的数组mp
val mp:Array[(String, Int)] = Array(
    "coffee"->4, 
    "panda"->3, 
    "panda"->2, 
    "happiest"->2, 
    "panda"->1, 
    "happiest"->1)
val wordCounts = sc.parallelize(mp)
wordCounts.reduceByKey((x, y) => if (x > y) x else y).collect
// Array[(String, Int)] = Array((coffee,4), (happiest,2), (panda,3))