我要整理一下RDD。排序需要在我的记录的多个字段上,因此我需要一个自定义比较器。
我看到sortBy作为它只接受一个键。我偶然发现了http://codingjunkie.net/spark-secondary-sort/,因此使用repartitionAndSortWithinPartitions实现了同样的目标。
为什么sortBy不接受自定义比较器和排序?为什么我必须重新划分才能使用自定义比较器?
发布于 2017-10-09 04:09:03
Question1:这是方法sortBy的签名
/**
* Return this RDD sorted by the given key function.
*/
def sortBy[K](
f: (T) => K,
ascending: Boolean = true,
numPartitions: Int = this.partitions.length)
(implicit ord: Ordering[K], ctag: ClassTag[K]): RDD[T] = withScope {
this.keyBy[K](f)
.sortByKey(ascending, numPartitions)
.values
}您的RDD数据对象显然是T类型的
请注意,sortBy方法绝对有单键参数字段: f: (T) => K
它接受匿名函数,因此您可以轻松地生成自定义的可比结构,并最大限度地利用具有自己定义良好的比较器的公共数据类型。
例如,如果您的RDDInt,Int,我们称之为data,您可以这样做:
val cmp = (t: (Int, Int)) => (t._1, -t._2)
data.sortBy(cmp)这可以很容易地实现多个领域,对吗?
这将得到一个排序的RDD与第一场上升和第二场下降。
Question2:repartitionAndSortWithinPartitions usage
这是一个特定的rdd操作符,其目标是比调用重新分区并在每个分区中进行排序更有效。
在排序之前,您的程序不需要预先重新分区,它只是在这种特殊的通用模式下进行内部优化,以获得高性能。
详情请参阅文档。
发布于 2017-10-09 05:25:20
正如Spark中所写的那样,repartitionAndSortWithinPartitions比调用重新分区并在每个分区中排序更有效--换句话说,repartitionAndSortWithinPartitions将首先根据提供的分区器重新划分数据,然后按键进行排序:
因此,首先重新分区,然后调用sortBy,给您很好的性能,您可以使用repartitionAndSortWithinPartitions实现。
添加几个排序示例,希望能有所帮助。
Ex 1
val rdd = sc.parallelize(Seq(
| ("math", 55),
| ("math", 56),
| ("english", 57),
| ("english", 58),
| ("science", 59),
| ("science", 54)))
rdd.collect()
//Default Sorting : Ascending order
val sorted1 = rdd.sortByKey()
sorted1.collect()
//Custom Sorting : Descending order (using implicit 'Ordering')
{
| //Let us define an implicit sorting for the method sortByKey()
| //We have used '{' above to limit the scope of the implicit ordering
| implicit val sortIntegersByString = new Ordering[String] {
| override def compare(a: String, b: String) = {
| val result = a.compare(b)
| //We use -ve to sort the key in descending order
| -result
| }
| }
| val sorted2 = rdd.sortByKey()
|
| //Result
| sorted2.collect()
| }
//Default Sorting : Descending order (done using the 'ascending' flag argument)
val sorted3 = rdd.sortByKey(false)
//Result
sorted3.collect()结果:
rdd: org.apache.spark.rdd.RDD[(String, Int)] = ParallelCollectionRDD[101] at parallelize at command-1784487111427703:1
sorted1: org.apache.spark.rdd.RDD[(String, Int)] = ShuffledRDD[104] at sortByKey at command-1784487111427703:12
sorted3: org.apache.spark.rdd.RDD[(String, Int)] = ShuffledRDD[110] at sortByKey at command-1784487111427703:34
res28: Array[(String, Int)] = Array((science,59), (science,54), (math,55), (math,56), (english,57), (english,58))Ex 2
case class Row(var firstName: String, var lastName: String, var city: String)
var rows = List(new Row("Oscar", "Wilde", "London"),
new Row("Otto", "Swift", "Berlin"),
new Row("Carl", "Swift", "Paris"),
new Row("Hans", "Swift", "Dublin"),
new Row("Hugo", "Swift", "Sligo"))
//print ("sort by last name")
//rows.sortBy(_.lastName)
print ("sort by last name and first name")
rows.sortBy(r => (r.lastName, r.firstName))
sort by last name and first namedefined class Row
rows: List[Row] = List(Row(Oscar,Wilde,London), Row(Otto,Swift,Berlin), Row(Carl,Swift,Paris), Row(Hans,Swift,Dublin), Row(Hugo,Swift,Sligo))
res26: List[Row] = List(Row(Carl,Swift,Paris), Row(Hans,Swift,Dublin), Row(Hugo,Swift,Sligo), Row(Otto,Swift,Berlin), Row(Oscar,Wilde,London))RDD与Dataset:
val sqlContext = new org.apache.spark.sql.SQLContext(sc)
case class MyRecord(time: Double, id: String)
val rdd = sc.parallelize(1 to 200, 200).flatMap(x =>Seq.fill(10000)(MyRecord(util.Random.nextDouble, "xxx")))
// sort this RDD by time:
val sorted = rdd.sortBy(x => x.time)
result.count
// convert the original RDD to Dataframe and sort again:
val df = sqlContext.createDataFrame(rdd)
df.registerTempTable("data")
val result = sqlContext.sql("select * from data order by time")
result.counthttps://stackoverflow.com/questions/46638504
复制相似问题