Hadoop problem with Custom RecordReader

[marcin biczak]

Hi all

I am using Hadoop version 0.20.2003. I wanted to create my own InputFormat and Reader. I wanted to do it with the use of new org.apache.hadoop.mapreduce.* classes. Unfortunately I have encountered some problems.

public class PatentLineRecordReder extends RecordReader<Text, Patent> {
  ...
  @Override
  public void initialize(InputSplit genericSplit, TaskAttemptContext context)
    ...
    inputByteCounter = context.getCounter(FileInputFormat.COUNTER_GROUP, FileInputFormat.BYTES_READ);
    ...
  }
}

Inside the visible line I have 3 strange problems:
1. cannot find symbol : variable COUNTER_GROUP
2 cannot find symbol : variable BYTES_READ
3. cannot find symbol : method getCounter(java.lang.String,java.lang.String)

I've checked the sources and all 3 "symbols" are there. Does anyone know what is the problem? Also I have came across interesting thing. Mainly according to hadoop javadoc org.apache.hadoop.mapreduce.TaskAttemptContext is:

public class TaskAttemptContext extends JobContext implements Progressable {...}

While the src file defines it as:

public interface TaskAttemptContext extends JobContext, Progressable {...}

For the record my "amazing" PatentLineRecordReder class is just a copy/paste from the org.apache.hadoop.mapreduce.lib.input.LineRecordReader, with minor changes in the initialize() method.

Does anyone came across similar problems, or rather should I ask "What am I doing wrong?".

For any help I will be grateful.