package org.apache.hadoop.mapred;
import java.io.IOException;
import org.apache.hadoop.classification.InterfaceAudience;
import org.apache.hadoop.classification.InterfaceStability;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.*;
import org.apache.hadoop.util.ReflectionUtils;
/** An {@link RecordReader} for {@link SequenceFile}s. */
@InterfaceAudience.Public
@InterfaceStability.Stable
/**
*hadoop中一个读取文件中某个片段(由其内部的start、length决定片段)的类,继承自RecordReader,可以作为InputFomat中 public *RecordReader<K, V> getRecordReader(InputSplit split,JobConf job, Reporter reporter)方法的返回的结果,处理FileSplit 。
* K V对应map中K V,其每次取得K V,要作为map的输入来做的
**/
public class SequenceFileRecordReader<K, V> implements RecordReader<K, V> {
private SequenceFile.Reader in; //读取的Reader
private long start; //文件在Reader的开始位置
private long end; //结束位置,一个hdfs文件在Reader中是连续的?
private boolean more = true;
protected Configuration conf;
public SequenceFileRecordReader(Configuration conf, FileSplit split)
throws IOException {
Path path = split.getPath();
FileSystem fs = path.getFileSystem(conf);
this.in = new SequenceFile.Reader(fs, path, conf);
this.end = split.getStart() + split.getLength();
this.conf = conf;
if (split.getStart() > in.getPosition())
in.sync(split.getStart()); // sync to start
this.start = in.getPosition(); // 一个hdfs的开始位置,但不是在Reader中的开始位置,是以取Reader的开始位置 ??
more = start < end;
}
/** The class of key that must be passed to {@link
* #next(Object, Object)}.. */
public Class getKeyClass() { return in.getKeyClass(); }
/** The class of value that must be passed to {@link
* #next(Object, Object)}.. */
public Class getValueClass() { return in.getValueClass(); }
/**
*每次调用next(K key, V value)方法需要传入相应的K 和 V 对象,是以以下像个方法构造K 、V的实例,
*以便调用next(K key, V value),传给map方法
*
*/
@SuppressWarnings("unchecked")
public K createKey() {
return (K) ReflectionUtils.newInstance(getKeyClass(), conf);
}
@SuppressWarnings("unchecked")
public V createValue() {
return (V) ReflectionUtils.newInstance(getValueClass(), conf);
}
/**
*取值的函数
*/
public synchronized boolean next(K key, V value) throws IOException {
if (!more) return false;
long pos = in.getPosition();
boolean remaining = (in.next(key) != null);
if (remaining) {
getCurrentValue(value);
}
if (pos >= end && in.syncSeen()) {
more = false;
} else {
more = remaining;
}
return more;
}
/**
*用于跳过某个key
*/
protected synchronized boolean next(K key)
throws IOException {
if (!more) return false;
long pos = in.getPosition();
boolean remaining = (in.next(key) != null);
if (pos >= end && in.syncSeen()) {
more = false;
} else {
more = remaining;
}
return more;
}
protected synchronized void getCurrentValue(V value)
throws IOException {
in.getCurrentValue(value);
}
/**
* Return the progress within the input split
* @return 0.0 to 1.0 of the input byte range
*/
public float getProgress() throws IOException {
if (end == start) {
return 0.0f;
} else {
return Math.min(1.0f, (in.getPosition() - start) / (float)(end - start));
}
}
public synchronized long getPos() throws IOException {
return in.getPosition();
}
protected synchronized void seek(long pos) throws IOException {
in.seek(pos);
}
public synchronized void close() throws IOException { in.close(); }
}
hadoop中一个读取文件中某个片段(由其内部的start、length决定片段)的类,继承自RecordReader,可以作为InputFomat中 public RecordReader<K, V> getRecordReader(InputSplit split,JobConf job, Reporter reporter)方法的返回的结果,处理FileSplit 。
可以读取FileSplit中的key、Value对,以便map可以使用 。