java - 如何让 hadoop 忽略\n 输入文件中的字符？

Question

我正在使用 Hadoop 的 map reduce 函数编写倒排索引创建器。我的输入文件中的某些行已将字符\n 作为实际字符写入其中(不是 ASCII 10，而是两个实际字符“\”和“n”)。出于某种我不明白的原因，这似乎导致 map 函数将我的行分成两行。

这是我的一些文件中的一些示例行。

32155: Wyldwood Radio: On the Move WILL begin on Friday May 1st, as originally planned!\n\nWe had some complications with... http://t.co/g8STpuHn5Q

5: RT @immoumita: #SaveJalSatyagrahi\nJal Satyagraha 'holding on to the truth by water' https://t.co/x3XgRvCE5H via @4nks

15161: RT @immoumita: #SaveJalSatyagrahi\nJal Satyagraha 'holding on to the truth by water' https://t.co/x3XgRvCE5H via @4nks

这是输出:

co :78516: tweets0001:30679;2, ... , tweets0001:We had some complications with... http;1, ...

x3XgRvCE5H :2: tweets0000:Jal Satyagraha 'holding on to the truth by water' https;2

下面是我的 map reduce:

map

public static class Map extends MapReduceBase implements Mapper<LongWritable, Text, Text, Text> {
     private final static Text word = new Text();
   private final static Text location = new Text();

     public void map(LongWritable key, Text value, OutputCollector<Text, Text> output, Reporter reporter) throws IOException {

     String line = value.toString();

     int colon_index = line.indexOf(":");
     if(colon_index > 0)
     {
       String tweet_num = line.substring(0,colon_index);
       line = line.substring(colon_index + 1);

       StringTokenizer tokenizer = new StringTokenizer(line," !@$%^&*()-+=\"\\:;/?><.,{}[]|`~");
       FileSplit fileSplit = (FileSplit)reporter.getInputSplit();
       String filename = fileSplit.getPath().getName();
       location.set(filename + ":" + tweet_num);
       while (tokenizer.hasMoreTokens()) {
         word.set(tokenizer.nextToken());
         output.collect(word, location);
       }
     }
}

减少

public static class Reduce extends MapReduceBase implements Reducer<Text, Text, Text, Text> {
     public void reduce(Text key, Iterator<Text> values, OutputCollector<Text, Text> output, Reporter reporter) throws IOException {
       boolean first = true;
     int count = 0;
     StringBuilder locations = new StringBuilder();
     HashMap<String,Integer> frequencies = new HashMap<String, Integer>();


       while (values.hasNext()) {
        String location = values.next().toString();
        if(frequencies.containsKey(location)){
          int frequency = frequencies.get(location).intValue() + 1;
          frequencies.put(location,new Integer(frequency));
        }
        else{
          frequencies.put(location,new Integer(1));
        }
        count++;
       }
     for(String location : frequencies.keySet()){
       int frequency = frequencies.get(location).intValue();
       if(!first)
        locations.append(", ");
       locations.append(location);
       locations.append(";"+frequency);
       first = false;
     }
     StringBuilder finalString = new StringBuilder();
     finalString.append(":"+String.valueOf(count)+": ");
     finalString.append(locations.toString());
       output.collect(key, new Text(finalString.toString()));
     }
   }

一般的数据流是将每一行映射到一个 {Word,filename:line_number} 对，然后通过计算它出现的频率来减少这些对。输出应该是:

Word-->:occurences:filename1:line_number:occurences_on_this_line, filename2....

map reduce 部分工作得很好，你甚至可以从我的示例中看到第 5 行和第 15161 行的推文都包含字符串 x3XgRvCE5H，而且，因为我的 Mapper 在之前查找冒号附加一个行号并且这两条推文包含相同的文本，它们都映射到相同的索引位置，给出“频率”值 2。

所以，我的问题是:如何让 Hadoop 的输入格式不将字符“\n”读取为换行符？毕竟，它们不是 ASCII 10，即实际的换行符、换行符，而是两个单独的字符。

Answer 1

您必须扩展 FileInputFormat 并编写一个新类来覆盖该行为。例如:

public class ClientTrafficInputFormat extends FileInputFormat {

    @Override
    public RecordReader createRecordReader(InputSplit split, TaskAttemptContext context)
            throws IOException, InterruptedException {

        return new ClientTrafficRecordReader();
    }

}

RecordReader 也应该被覆盖

public class ClientTrafficRecordReader extends
        RecordReader<ClientTrafficKeyWritable, ClientTrafficValueWritable> {

    ...

    private LineRecordReader reader = new LineRecordReader(); // create your own RecordReader this is where you have to mention not to use '\n' but it should be read as "\"and "n"

    @Override
    public void initialize(InputSplit is, TaskAttemptContext tac) throws IOException,
            InterruptedException {

        reader.initialize(is, tac);

    }
     ...
    @Override
public boolean nextKeyValue() throws IOException, InterruptedException {
      //customize your input 
 }