僅供參考：

def get_hdfs_dir(input: String, sc: SparkContext): Array[String] = {

//建立[LongWritable, Text, TextInputFormat]的rdd

val fileRDD = sc.newAPIHadoopFile[LongWritable, Text, TextInputFormat](input)

//獲取hadoop的rdd

val hadoopRDD = fileRDD.asInstanceOf[NewHadoopRDD[LongWritable, Text]]

//透過檔案進行分割槽，然後透過不同分割槽來獲取分割槽的path

val fileAdnLine = hadoopRDD.mapPartitionsWithInputSplit((inputSplit: InputSplit, iterator: Iterator[(LongWritable, Text)]) => {

val file = inputSplit.asInstanceOf[FileSplit]

iterator.take(1).map(x => {

file.getPath.toString() //就是當前資料的所在路徑

})

})

val dirOut: Array[String] = fileAdnLine

.distinct() //可能會有很多一樣的path，進去去重操作

.coalesce(1)

// val array: Array[List[Char]] = fileAdnLine //這裡面存放的就是path

.map(lines => {

lines.toString

})

.collect()

dirOut //dirOut 是一個Array型別的資料裡面存放的就是傳入資料夾下的路徑

}