storm中declareOutputFields方法的使用规则及与fieldsGrouping方法的联系

    declareOutputFields方法中声明了该bolt/spout输出的字段个数，供下游使用，在该bolt中的execute方法中，emit发射的字段个数必须和声明的相同，否则报错：Tuple created with wrong number of fields. Expected 2 fields but got 1 fields。

        fieldsGrouping方法设定从上一个bolt中获取哪些fields。fieldsGrouping("2", new Fields("ahah","bb"))，Fields中的参数名字必须和上一个bolt中声明的字段名字相同，否则报错：“Topology submission exception. (topology name='HelloStorm') #<InvalidTopologyException InvalidTopologyException(msg:Component: [3] subscribes from stream: [default] of component [2] with non-existent fields: #{"ahah1"})>”

测试Spout代码如下：

public class Produce extends BaseRichSpout {
	private SpoutOutputCollector collector;
	@Override
	public void nextTuple() {
		String[] str1=new String[]{"aaa","bbb","ccc","ddd"};
		String[] str2=new String[]{"ee","ff","gg","kk"};
		Random random=new Random();
		String word1=str1[random.nextInt(str1.length)];
		String word2=str2[random.nextInt(str2.length)];
		collector.emit(new Values(word1,word2));
		try {
			Thread.sleep(100);
		} catch (InterruptedException e) {
			e.printStackTrace();
		}
	}

	@Override
	public void open(Map arg0, TopologyContext arg1, SpoutOutputCollector arg2) {
		this.collector=arg2;
	}
	@Override
	public void declareOutputFields(OutputFieldsDeclarer arg0) {
		arg0.declare(new Fields("ppee","ttt"));
	}
}

测试Output代码如下：

public class Output extends BaseBasicBolt{
	@Override
	public void execute(Tuple tuple, BasicOutputCollector arg1) {
		String str=tuple.getString(0);
		String str1=tuple.getString(1);
		arg1.emit(new Values(str,str1));
		/*此处发射的字段个数必须要和declareOutputFields中声明的字段个数相同，否则报错：
		 *  Tuple created with wrong number of fields. Expected 2 fields but got 1 fields
		 */
	}

	@Override
	public void declareOutputFields(OutputFieldsDeclarer arg0) {
		arg0.declare(new Fields("ahah","bb"));
	}
}

测试writer代码如下：

public class Writer extends BaseBasicBolt{

	@Override
	public void execute(Tuple arg0, BasicOutputCollector arg1) {
		// TODO Auto-generated method stub
		String str=arg0.getString(0);
		String str1=arg0.getString(1);
		System.out.println(str+"   "+str1+"  "+this);
	}

	@Override
	public void prepare(Map stormConf, TopologyContext context) {
	}

	@Override
	public void declareOutputFields(OutputFieldsDeclarer arg0) {
		arg0.declare(new Fields("hehe"));
	}
}

测试Main类如下：

public class main {

	public static void main(String[] args) throws Exception {
		TopologyBuilder builder=new TopologyBuilder();
		builder.setSpout("1", new Produce(),1);
		builder.setBolt("2", new Output(), 4).shuffleGrouping("1");
		builder.setBolt("3", new Writer(), 8).fieldsGrouping("2", new Fields("ahah","bb"));
		/*设定从上一个bolt中获取"ahah"字段，这个“ahah”必须要和上一个bolt中的declare声明的字段相同，
		 * 否则报错，“Topology submission exception. (topology name='HelloStorm') 
		 * #<InvalidTopologyException InvalidTopologyException(msg:Component: [3] subscribes from 
		 * stream: [default] of component [2] with non-existent fields: #{"ahah1"})>”
		 * 
		 */
		
		Config con=new Config();
		con.setDebug(false);
		con.setNumWorkers(5);
//		StormSubmitter.submitTopology("first", con, builder.createTopology());
		LocalCluster cluster = new LocalCluster();
		cluster.submitTopology("HelloStorm", con, builder.createTopology());
	}
}

    在运行测试的过程中发现一个有趣的现象，在builder.setBolt("3", new Writer(), 8).fieldsGrouping("2", new Fields("ahah","bb"))中的Fields中如果以第二个字段"bb"分组，则只打印出一个对象的哈希值，如果以第一个字段''ahah''或者两个字段则正常打印出多个对象的哈希值。《后续：经leader指导，是因为发射数据量的问题，发射数据改成"word+System.currentTimeMillis()",则按正常设置生成相应的Task对象》

    fieldsGrouping分组的含义是同id的tuple同时只能在一个Task上。