Spark得到两个RDD值集合有包含关系的映射

val data = sc.textFile("/user/wuzhongqiang/clean_data/baidubaike_source.20180801/").cache()
//1.以左边的为key，进行分组，统计每一个key所对应的值集合
val groupData = data.map(item => {
    val key = item._1 
    val value = item._1
    (key, value)
}).groupByKey

//2. 读取链接映射文件至map
//(AKey, BKey) 
val projectionMap = sc.textFile("hdfs://projection").cache()
// (AKey, BKey) 
val aData = projectionMap.map(item => (item._1, item._2))
// (BKey, AKey) 
val bData = projectionMap.map(item => (item._2, item._1))
// (AKey, (BKey, AValueSet)) 
val aKeyJoinData = aData.join(groupData)
// (BKey, (AKey, BValueSet)) 
var bKeyJoinData = bData.join(groupData)
// 交换新老键的位置，为后面的join做准备 (AKey, (BKey, BValueSet) )
bKeyJoinData = nbKeyJoinData.map(item => {(item._2._1, (item._1, item._2._2))})
//结果形式为(AKey, ((BKey, AValueSet), (BKey, BValueSet)))
val aBData = aKeyJoinData.join(bKeyJoinData)
aKeyJoinData.take(1).foreach(println)
// oldUrl -> newUrl
val resultMap = aBData.map(item => {
    val aValueSet = item._2._1._2.to[Set]
    val bValueAttrSet = item._2._2._2.to[Set]
    val subtractSet = aValueSet -- bValueAttrSet
    //若新的URL属性可以完全覆盖旧的url属性， 即 oldAttrSet与newAttrSet的差集为空
    if(subtractSet.isEmpty)
    	(item._1, item._2._1._1)
    else ("", "")
}).filter(_._1 != "")
resultMap.take(1).foreach(println)