我正在运行用Scala 2.9.3编写的Spark应用程序。以下是创建节点树的功能network
。每个节点都有一组唯一的邻居,它们是该特定节点的子节点。我面临的问题是在for
循环内对象current
是不同的对象(通过其不同的地址显而易见)。我怎样才能防止这种情况,并且在外部宣布的同一个对象上执行for
循环?Scala:对象范围
def network (root: Node) : Tree =
{
var tree = new Tree(root)
var queue = ListBuffer[Node](root)
while (!queue.isEmpty && queue(0).level<maxlen)
{
var current: Node = queue.remove(0)
println(">>>>>>>>>>>>>>>>>> Current1: "+current)
var neigh = findNeighbor(current.userID)
for (n <- neigh)
{
if(tree.search(n._1) == null)
{
var c = new Node(n._1, current.level+1, n._2, n._3)
current.addChild(c)
println(">>>>>>>>>>>>>>>>>> Current2: "+current)
}
}
println(">>>>>>>>>>>>>>>>>> Current3: "+current)
queue ++= current.neighbors
}
return tree
}
下面是代码的输出。在由Current1,Current2和Current3表示的3个位置检查值为current
。我们观察到,当前1 == Current3
[[email protected] network]$ sbt run
Loading /usr/local/sbt/bin/sbt-launch-lib.bash
[info] Set current project to Network (in build file:/usr/local/spark/test/vijay/network/)
[info] Compiling 1 Scala source to /usr/local/spark/test/vijay/network/target/scala-2.9.3/classes...
[info] Running Network
SLF4J: Failed to load class "org.slf4j.impl.StaticLoggerBinder".
SLF4J: Defaulting to no-operation (NOP) logger implementation
SLF4J: See http://www.slf4j.org/codes.html#StaticLoggerBinder for further details.
>>>>>>>>>>>>>>>>>> Current1: [email protected]
13/10/13 14:23:50 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
13/10/13 14:23:50 WARN snappy.LoadSnappy: Snappy native library not loaded
13/10/13 14:23:50 INFO mapred.FileInputFormat: Total input paths to process : 1
>>>>>>>>>>>>>>>>>> Current2: [email protected]
>>>>>>>>>>>>>>>>>> Current2: [email protected]
>>>>>>>>>>>>>>>>>> Current2: [email protected]
>>>>>>>>>>>>>>>>>> Current2: [email protected]
>>>>>>>>>>>>>>>>>> Current3: [email protected]
[success] Total time: 11 s, completed Oct 13, 2013 2:23:51 PM
附加信息:findNeighbor
返回一个包含邻居的元组节点userID
一个RDD。 tree.search
函数检查树,以查找n
是否已存在于树中。只有在树中不存在节点时才将其作为子节点添加。 所有功能都按预期工作。
我认为这是各种优化的一个功能,Scala内部支持懒惰评估等等。此外,这可能是一个很好的学习练习,但我认为创建此树会打破Spark在RDD抽象中进行优化的目的。 – Vidya