从 dataproc 中的 google 存储读取文件

Question

我正在尝试将 scala spark 作业从 hadoop 集群迁移到 GCP，我有这段读取文件并创建 ArrayBuffer[String]

的代码片段

    import java.io._
    import org.apache.hadoop.conf.Configuration
    import org.apache.hadoop.fs.Path
    import org.apache.hadoop.fs.FSDataInputStream

    val filename="it.txt.1604607878987"
    val fs = FileSystem.get(new Configuration())
    val dataInputStream: FSDataInputStream = fs.open(new Path(filename))
    val sourceEDR=new BufferedReader(new InputStreamReader(dataInputStream, "UTF-8")); }
    val outputEDRFile = ArrayBuffer[String]()
    buffer = new Array[Char](300)
    var num_of_chars = 0
    while (sourceEDR.read(buffer) > -1) {
        val str = new String(buffer)
        num_of_chars += str.length
        outputEDRFile += (str + "\n");}
        println(num_of_chars)

集群中的代码运行s 并给了我 3025000 个字符，我尝试运行 dataproc 中的代码：

       val path_gs = new Path("gs://my-bucket")
       val filename="it.txt.1604607878987"

       val fs = path_gs.getFileSystem(new Configuration())        
       val dataInputStream: FSDataInputStream = fs.open(new Path(filename))
       val sourceEDR =new BufferedReader(new InputStreamReader(dataInputStream, "UTF-8")); }

       val outputEDRFile = ArrayBuffer[String]()
       buffer = new Array[Char](300)
       var num_of_chars = 0
       while (sourceEDR.read(buffer) > -1) {
       val str = new String(buffer)
       num_of_chars += str.length
       outputEDRFile += (str + "\n");}
       println(num_of_chars)

它给出了 3175025 个字符，我认为文件内容中添加了空格，或者我必须使用另一个接口从 dataproc 中的 google 存储读取文件？我也尝试了其他编码选项，但结果相同。有帮助吗？

Answer 1

我没有找到使用缓冲区的解决方案，所以我尝试逐个读取一个字符，它对我有用：

var i = 0
var r=0
val response = new StringBuilder
while ( ({r=sourceEDR.read(); r} != -1)) {
  val ch= r.asInstanceOf[Char]
  if(response.length < 300) { response.append(ch)}
  else {  val str = response.toString().replaceAll("[\r\n]", " ")
    i += str.length
    outputEDRFile += (str + "\n");
    response.setLength(0)
    response.append(ch)
  }
}
val str = response.toString().replaceAll("[\r\n]", " ")
i += str.length
outputEDRFile += (str + "\n");

从 dataproc 中的 google 存储读取文件

Read a file from google storage in dataproc

java

hadoop

google-cloud-platform

google-cloud-storage

google-cloud-dataproc