如何使用字节数组作为RDD中的键? [英] How to use byte array as key in RDD?

问题描述

我想使用Array [Byte]作为RDD中的键.例如:

I want to use Array[Byte] as Key from RDD. For example:

val rdd1:RDD[((Array[Byte]), (String, Int)] = from src rdd
val rdd2:RDD[((Array[Byte]), (String, Int)] = from dest rdd

val resultRdd = rdd1.join(rdd2)

我想使用Array [Byte]作为键在rdd1和rdd2上执行联接操作 但总是得到resultRdd.count()=0.

I want to perform join operation on rdd1 and rdd2 using Array[Byte] as Key but always getting resultRdd.count() = 0.

因此，我尝试序列化Array [Byte]，并且我希望看到它工作正常，仅适用于小型数据集.

So I tried to serialize the Array[Byte] and It is working fine as I want to see, only for small Dataset.

val serRdd1= rdd1.map { case (k,v) =>  (new SerByteArr(k), v) }
val serRdd2= rdd2.map { case (k,v) =>  (new SerByteArr(k), v) }

class SerByteArr(val bytes: Array[Byte]) extends Serializable {
   override val hashCode = bytes.deep.hashCode
   override def equals(obj:Any) = obj.isInstanceOf[SerByteArr] && obj.asInstanceOf[SerByteArr].bytes.deep == this.bytes.deep
 }

对于大型数据集，获取java.lang.OutOfMemoryError:超出了GC开销限制， 创建对象(新的SerByteArr(k))时出现问题.

For Large dataset, getting java.lang.OutOfMemoryError: GC overhead limit exceeded, Problem is occuring in creating the object(new SerByteArr(k)).

如何避免GC限制超出错误.有人帮我吗?

How to avoid the GC limit exceed error. Does anyone help me?

推荐答案

您可以对数组WrappedArray[Byte]使用内置的scala包装器.可以使用toSeq方法将数组转换为WrappedArray. WrappedArray已正确实现equals和hashCode，因此具有相同元素的两个不同数组被视为相等.

You can use a built-in scala wrapper for arrays, WrappedArray[Byte]. An array can be converted to a WrappedArray by using toSeq method. WrappedArray has properly implemented equals and hashCode, so two different arrays with the same elements are considered as equal.

scala> val a = Array(1,2,3,4,5)
a: Array[Int] = Array(1, 2, 3, 4, 5)

scala> val b = Array(1,2,3,4,5)
b: Array[Int] = Array(1, 2, 3, 4, 5)

scala> a == b
res0: Boolean = false

scala> a.toSeq
res1: Seq[Int] = WrappedArray(1, 2, 3, 4, 5)

scala> a.toSeq == b.toSeq
res2: Boolean = true

这篇关于如何使用字节数组作为RDD中的键?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！