Hadoop的MapReduce计算模型的Key，Value值都是采用的自定义Writable类型，我们也可以自定义Writable来实现不同的业务需求。
1.定义：
a.基于DataInput和DataOutput实现简单，高效，序列协议的可序列化对象；
b.在hadoop的MapReduce计算编程模型中，必须实现Writable接口
c.实现通常实现静态read(DataInput)方法，该方法构造一个新的实例，调用readFields(DataInput)并返回实例。
2.接口说明
public interface Writable {
/**
* 序列化对象的field(字段)到DataOutput中
* (序列化过程)
*/
void write(DataOutput out) throws IOException;
/**
* 从DataInput中将对象的field(字段)反序列化
*（反序列化的过程）
*/
void readFields(DataInput in) throws IOException;
}
与java的区别：
java序列户过程是面向流的；hadoop中Writable序列化面向对象的；
3.序列化过程
public static byte[] serialize(Writable writable) throws IOException {
ByteArrayOutputStream out = new ByteArrayOutputStream();
DataOutputStream dataOut = new DataOutputStream(out);
writable.write(dataOut);
dataOut.close();
}
4.反序列化过程：
public static byte[] deserialize(Writable writable, byte[] bytes)throws IOException {
ByteArrayInputStream in = new ByteArrayInputStream(bytes);
DataInputStream dataIn = new DataInputStream(in);
writable.readFields(dataIn);
dataIn.close();
return bytes;
}
分析序列化和反序列化的过程
------------------------------------------------
0.结合MapWritable、IntWritable和Text跟踪底层源码
1.构建MapWritable：
MapWritable _map = new MapWritable();
_map.put(new IntWritable(12), new Text("zhangsan"));
_map.put(new IntWritable(130), new Text("lisi"));
_map.put(new IntWritable(16), new Text("hanmeimei"));
_map.put(new IntWritable(7778878), new Text("lilei"));
2.序列化mapWritable对象，调用serialize(Writable writable)
//序列化过程，调用writable的write()
byte[] tt = serialize(_map);
System.out.println(tt.length);
说明：【MapWritable.class】-->write(DataOutput out)方法
public void write(DataOutput out) throws IOException {
super.write(out); 【说明：调用AbstractMapWritable.class的write()方法：将newclass添加至out中】
// Write out the number of entries in the map
out.writeInt(instance.size()); 【说明：将mapwritable对象的个数写入到out中】
// Then write out each key/value pair 【说明：循环遍历，写入顺序为：KeyClassID+key+ValueClassID+value】
for (Map.Entry<Writable, Writable> e: instance.entrySet()) {
out.writeByte(getId(e.getKey().getClass()));
e.getKey().write(out);
out.writeByte(getId(e.getValue().getClass()));
e.getValue().write(out);
}
}
3.反序列化，定义新的mapWritable对象，调用deserialize(Writable writable, byte[] bytes)
MapWritable map2 = new MapWritable();
deserialize(map2, tt);
System.out.println(map2.size());
说明：【MapWritable.class】-->readFields(DataInput in)方法
public void readFields(DataInput in) throws IOException {
super.readFields(in); 【说明：调用AbstractMapWritable.class的readFields()方法：获取newclass的类型】
// First clear the map. Otherwise we will just accumulate
// entries every time this method is called.
this.instance.clear(); 【说明：清空map对象中的数据】
// Read the number of entries in the map
int entries = in.readInt(); 【说明：通过in读取一个int(整型位)获得序列化中数据的个数】
// Then read each key/value pair 【说明：循环遍历个数，利用反射机制获取keyclass和valueclass，并通过Writable的readFields方法读取key值和value值，并将其写入至instance(当前MapWritable对象中)】
for (int i = 0; i < entries; i++) {
Writable key = (Writable) ReflectionUtils.newInstance(getClass(in.readByte()), getConf());
key.readFields(in);
Writable value = (Writable) ReflectionUtils.newInstance(getClass(in.readByte()), getConf());
value.readFields(in);
instance.put(key, value);
}
}
原文地址：https://www.cnblogs.com/lyr999736/p/9381637.html