Spark修炼之道（进阶篇）——Spark入门到精通：第五节 Spark编程模型（二)

 /**

 * Return the union of this RDD and another one. Any identical elements will appear multiple

 * times (use `.distinct()` to eliminate them).

 def union(other: RDD[T]): RDD[T] 

RDD与另外一个RDD进行Union操作之后，两个数据集中的存在的重复元素
代码如下：

scala val rdd1=sc.parallelize(1 to 5)

rdd1: org.apache.spark.rdd.RDD[Int] = ParallelCollectionRDD[15] at parallelize at console :21

scala val rdd2=sc.parallelize(4 to 8)

rdd2: org.apache.spark.rdd.RDD[Int] = ParallelCollectionRDD[16] at parallelize at console :21

//存在重复元素

scala rdd1.union(rdd2).collect

res13: Array[Int] = Array(1, 2, 3, 4, 5, 4, 5, 6, 7, 8)

（2）intersection
方法返回两个RDD数据集的交集
函数参数：
/**
* Return the intersection of this RDD and another one. The output will not contain any duplicate
* elements, even if the input RDDs did.
*
* Note that this method performs a shuffle internally.
*/
def intersection(other: RDD[T]): RDD[T]

使用示例：

scala val rdd1=sc.parallelize(1 to 5)

rdd1: org.apache.spark.rdd.RDD[Int] = ParallelCollectionRDD[0] at parallelize at console :21

scala val rdd2=sc.parallelize(4 to 8)

rdd2: org.apache.spark.rdd.RDD[Int] = ParallelCollectionRDD[1] at parallelize at console :21

scala rdd1.union(rdd2).distinct.collect

res0: Array[Int] = Array(6, 1, 7, 8, 2, 3, 4, 5)

（4）groupByKey([numTasks])
输入数据为(K, V) 对, 返回的是 (K, Iterable) ，numTasks指定task数量，该参数是可选的，下面给出的是无参数的groupByKey方法
/**
* Group the values for each key in the RDD into a single sequence. Hash-partitions the
* resulting RDD with the existing partitioner/parallelism level. The ordering of elements
* within each group is not guaranteed, and may even differ each time the resulting RDD is
* evaluated.
*
* Note: This operation may be very expensive. If you are grouping in order to perform an
* aggregation (such as a sum or average) over each key, using [[PairRDDFunctions.aggregateByKey]]
* or [[PairRDDFunctions.reduceByKey]] will provide much better performance.
*/
def groupByKey(): RDD[(K, Iterable[V])]

scala val rdd1=sc.parallelize(1 to 5)

rdd1: org.apache.spark.rdd.RDD[Int] = ParallelCollectionRDD[0] at parallelize at console :21

scala val rdd2=sc.parallelize(4 to 8)

rdd2: org.apache.spark.rdd.RDD[Int] = ParallelCollectionRDD[1] at parallelize at console :21

scala rdd1.union(rdd2).map((_,1)).groupByKey.collect

res2: Array[(Int, Iterable[Int])] = Array((6,CompactBuffer(1)), (1,CompactBuffer(1)), (7,CompactBuffer(1)), (8,CompactBuffer(1)), (2,CompactBuffer(1)), (3,CompactBuffer(1)), (4,CompactBuffer(1, 1)), (5,CompactBuffer(1, 1)))

（5）reduceByKey(func, [numTasks])
reduceByKey函数输入数据为(K, V)对，返回的数据集结果也是（K,V）对，只不过V为经过聚合操作后的值
/**
* Merge the values for each key using an associative reduce function. This will also perform
* the merging locally on each mapper before sending results to a reducer, similarly to a
* “combiner” in MapReduce. Output will be hash-partitioned with numPartitions partitions.
*/
def reduceByKey(func: (V, V) = V, numPartitions: Int): RDD[(K, V)]

使用示例：

scala rdd1.union(rdd2).map((_,1)).reduceByKey(_+_).collect

res4: Array[(Int, Int)] = Array((6,1), (1,1), (7,1), (8,1), (2,1), (3,1), (4,2), (5,2))

（6）sortByKey([ascending], [numTasks])
对输入的数据集按key排序
sortByKey方法定义

/**

 * Sort the RDD by key, so that each partition contains a sorted range of the elements. Calling

 * `collect` or `save` on the resulting RDD will return or output an ordered list of records

 * (in the `save` case, they will be written to multiple `part-X` files in the filesystem, in

 * order of the keys).

 // TODO: this currently doesnt work on P other than Tuple2!

 def sortByKey(ascending: Boolean = true, numPartitions: Int = self.partitions.length)

 : RDD[(K, V)]

使用示例：

scala var data = sc.parallelize(List((1,3),(1,2),(1, 4),(2,3),(7,9),(2,4)))

data: org.apache.spark.rdd.RDD[(Int, Int)] = ParallelCollectionRDD[20] at parallelize at console :21

scala data.sortByKey(true).collect

res10: Array[(Int, Int)] = Array((1,3), (1,2), (1,4), (2,3), (2,4), (7,9))

（7）join(otherDataset, [numTasks])
对于数据集类型为 (K, V) 及 (K, W)的RDD，join操作后返回类型为 (K, (V, W))，join函数有三种：
def join[W](other: RDD[(K, W)], partitioner: Partitioner): RDD[(K, (V, W))]
def leftOuterJoin[W](
other: RDD[(K, W)],
partitioner: Partitioner): RDD[(K, (V, Option[W]))]
def rightOuterJoin[W](other: RDD[(K, W)], partitioner: Partitioner)
: RDD[(K, (Option[V], W))]

使用示例：

scala val rdd1=sc.parallelize(Array((1,2),(1,3))

rdd1: org.apache.spark.rdd.RDD[(Int, Int)] = ParallelCollectionRDD[24] at parallelize at console :21

scala val rdd2=sc.parallelize(Array((1,3)))

rdd2: org.apache.spark.rdd.RDD[(Int, Int)] = ParallelCollectionRDD[32] at parallelize at console :21

scala rdd1.join(rdd2).collect

res13: Array[(Int, (Int, Int))] = Array((1,(3,3)), (1,(2,3)))

scala rdd1.leftOuterJoin(rdd2).collect

res15: Array[(Int, (Int, Option[Int]))] = Array((1,(3,Some(3))), (1,(2,Some(3))))

scala rdd1.rightOuterJoin(rdd2).collect

res16: Array[(Int, (Option[Int], Int))] = Array((1,(Some(3),3)), (1,(Some(2),3)))

（8）cogroup(otherDataset, [numTasks])
如果输入的RDD类型为(K, V) 和(K, W)，则返回的RDD类型为 (K, (Iterable, Iterable)) . 该操作与 groupWith等同

方法定义：
/**
* For each key k in this or other, return a resulting RDD that contains a tuple with the
* list of values for that key in this as well as other.
*/
def cogroup[W](other: RDD[(K, W)], partitioner: Partitioner)
: RDD[(K, (Iterable[V], Iterable[W]))]

scala val rdd1=sc.parallelize(Array((1,2),(1,3))

rdd1: org.apache.spark.rdd.RDD[(Int, Int)] = ParallelCollectionRDD[24] at parallelize at console :21

scala val rdd2=sc.parallelize(Array((1,3)))

rdd2: org.apache.spark.rdd.RDD[(Int, Int)] = ParallelCollectionRDD[32] at parallelize at console :21

scala rdd1.cogroup(rdd2).collect

res17: Array[(Int, (Iterable[Int], Iterable[Int]))] = Array((1,(CompactBuffer(3, 2),CompactBuffer(3))))

scala rdd1.groupWith(rdd2).collect

res18: Array[(Int, (Iterable[Int], Iterable[Int]))] = Array((1,(CompactBuffer(2, 3),CompactBuffer(3))))

（9）cartesian(otherDataset)
求两个RDD数据集间的笛卡尔积
函数定义：
/**
* Return the Cartesian product of this RDD and another one, that is, the RDD of all pairs of
* elements (a, b) where a is in this and b is in other.
*/
def cartesian[U: ClassTag](other: RDD[U]): RDD[(T, U)]

scala val rdd1=sc.parallelize(Array(1,2,3,4))

rdd1: org.apache.spark.rdd.RDD[Int] = ParallelCollectionRDD[52] at parallelize at console :21

scala val rdd2=sc.parallelize(Array(5,6))

rdd2: org.apache.spark.rdd.RDD[Int] = ParallelCollectionRDD[53] at parallelize at console :21

scala rdd1.cartesian(rdd2).collect

res21: Array[(Int, Int)] = Array((1,5), (1,6), (2,5), (2,6), (3,5), (4,5), (3,6), (4,6))

（10）coalesce(numPartitions)
将RDD的分区数减至指定的numPartitions分区数

函数定义：
/**
* Return a new RDD that is reduced into numPartitions partitions.
*
* This results in a narrow dependency, e.g. if you go from 1000 partitions
* to 100 partitions, there will not be a shuffle, instead each of the 100
* new partitions will claim 10 of the current partitions.
*
* However, if you’re doing a drastic coalesce, e.g. to numPartitions = 1,
* this may result in your computation taking place on fewer nodes than
* you like (e.g. one node in the case of numPartitions = 1). To avoid this,
* you can pass shuffle = true. This will add a shuffle step, but means the
* current upstream partitions will be executed in parallel (per whatever
* the current partitioning is).
*
* Note: With shuffle = true, you can actually coalesce to a larger number
* of partitions. This is useful if you have a small number of partitions,
* say 100, potentially with a few partitions being abnormally large. Calling
* coalesce(1000, shuffle = true) will result in 1000 partitions with the
* data distributed using a hash partitioner.
*/
def coalesce(numPartitions: Int, shuffle: Boolean = false)(implicit ord: Ordering[T] = null)
: RDD[T]

示例代码：

scala val rdd1=sc.parallelize(1 to 100,3)

rdd1: org.apache.spark.rdd.RDD[Int] = ParallelCollectionRDD[55] at parallelize at console :21

scala val rdd2=rdd1.coalesce(2)

rdd2: org.apache.spark.rdd.RDD[Int] = CoalescedRDD[56] at coalesce at console :23

repartition(numPartitions)，功能与coalesce函数相同，实质上它调用的就是coalesce函数，只不是shuffle = true，意味着可能会导致大量的网络开销。
方法定义：
/**
* Return a new RDD that has exactly numPartitions partitions.
*
* Can increase or decrease the level of parallelism in this RDD. Internally, this uses
* a shuffle to redistribute data.
*
* If you are decreasing the number of partitions in this RDD, consider using coalesce,
* which can avoid performing a shuffle.
*/
def repartition(numPartitions: Int)(implicit ord: Ordering[T] = null): RDD[T] = withScope {
coalesce(numPartitions, shuffle = true)
}

【Spark Mllib】分类模型——各分类模型使用 一. 数据集 这个数据集源自 Kaggle 比赛,由 StumbleUpon 提供。比赛的问题涉及网页中推荐的页面是短暂(短暂存在,很快就不流行了)还是长久(长时间流行)。
客户流失？来看看大厂如何基于spark+机器学习构建千万数据规模上的用户留存模型 ⛵ 如何在海量用户中精准预测哪些客户即将流失？本文结合音乐流媒体平台 Sparkify 数据，详细讲解一个客户流失建模预测案例的全流程：探索性数据分析 EDA、数据处理、进一步数据探索、建模优化、结果评估。【代码与数据集亲测可运行】
Spark和MapReduce任务计算模型 【前言：本文主要从任务处理的运行模式为角度，分析Spark计算模型，希望帮助大家对Spark有一个更深入的了解。同时拿MapReduce和Spark计算模型做对比，强化对Spark和MapReduce理解】
6月23日 Spark 社区技术直播【半小时，将你的Spark SQL模型变为在线服务】 SparkSQL在机器学习场景中应用模型从批量到实时面临的问题 - SparkSQL 转换成实时执行成本高 - 离线特征和在线特征保持一致困难 - 离线效果与在线效果差距大 我们是如何解决这些问题 相对传统实现方式我们优势 SparkSQL实时上线demo
EMR Spark Relational Cache如何支持雪花模型中的关联匹配 我们需要找到一种方式可以通过单个Relational Cache支持优化多个关联查询的方式，从而在加速用户查询的同时，减少创建和更新relational cache的代价。Record Preserve Join是支持这种优化的非常有效的方式。
EMR Spark Relational Cache如何支持雪花模型中的关联匹配 在Spark中，Join通常是代价比较大，尤其是shuffle join。Relational Cache将反范式化表（即关联后的大表）保存为relational cache，便可以使用cache重写执行计划，提高查询效率。