storm trident实战 分组聚合
程序员文章站
2022-07-02 09:55:19
...
一、前言
groupBy分组操作,根据指定属性进行分组,如果后面是aggregate()的话,先根据partitionBy分区,在每个partition上分组,分完组后,在每个分组上进行聚合。
二、实战
main:
public static void main(String[] args) throws Exception {
/**
* 多设置几个并行度,分组后如果分组不够,那么将有并行度空闲跑者
*/
@SuppressWarnings("unchecked")
FixedBatchSpout spout = new FixedBatchSpout(new Fields("sentence"), 3,
new Values("a"), new Values("b"), new Values("a"),new Values("c"),
new Values("c"),new Values("c"),new Values("d"));
spout.setCycle(false);
TridentTopology tridentTopology = new TridentTopology();
tridentTopology
.newStream("spout", spout)
.parallelismHint(3)
.shuffle()
.groupBy(new Fields("sentence"))
.aggregate(new Fields("sentence"), new MyAgg(),
new Fields("Map"))
.parallelismHint(5)
.each(new Fields("sentence","Map"), new MyBolt());
Config config = new Config();
config.setDebug(false);
StormSubmitter.submitTopology("trident_groupby_aggregate_many", config,
tridentTopology.build());
}
MyAgg:
package com.storm.trident.groupby.先分组后聚合; import java.util.HashMap; import java.util.Map; import org.apache.storm.trident.operation.BaseAggregator; import org.apache.storm.trident.operation.TridentCollector; import org.apache.storm.trident.operation.TridentOperationContext; import org.apache.storm.trident.tuple.TridentTuple; import org.apache.storm.tuple.Values; public class MyAgg extends BaseAggregator<Map<String, Integer>> { private static final long serialVersionUID = 1L; /** * 属于哪个分区 */ private int partitionId; /** * 分区数量 */ private int numPartitions; @SuppressWarnings("rawtypes") @Override public void prepare(Map conf, TridentOperationContext context) { partitionId = context.getPartitionIndex(); numPartitions = context.numPartitions(); } @Override public void aggregate(Map<String, Integer> val, TridentTuple tuple, TridentCollector collector) { String word = tuple.getString(0); Integer value = val.get(word); if (value == null) { value = 0; } value++; // 把数据保存到一个map对象中 val.put(word, value); val.put(word + "属于哪个分区", partitionId); System.out.println("I am partition [" + partitionId + "] and I have kept a tweet by: " + numPartitions); } @Override public void complete(Map<String, Integer> val, TridentCollector collector) { collector.emit(new Values(val)); } @Override public Map<String, Integer> init(Object arg0, TridentCollector arg1) { return new HashMap<String, Integer>(); } }
MyBolt:
package com.storm.trident.groupby.先分组后聚合; import java.util.Map; import java.util.Map.Entry; import org.apache.storm.trident.operation.BaseFilter; import org.apache.storm.trident.tuple.TridentTuple; public class MyBolt extends BaseFilter { /** * */ private static final long serialVersionUID = 1L; @SuppressWarnings("unchecked") @Override public boolean isKeep(TridentTuple tuple) { System.out.println("打印出来的tuple:" + tuple); Map<String, Integer> value = ((Map<String,Integer>) tuple.getValue(1)); for (Entry<String, Integer> entry : value.entrySet()) { System.out.println("key:"+ entry.getKey()+",value:" + entry.getValue()); } return false; } }
三、测试
打包在storm集群里跑
查看log日志,主要日志如下
2016-12-22 18:36:11.293 STDIO [INFO] I am partition [3] and I have kept a tweet by: 5 2016-12-22 18:36:11.302 STDIO [INFO] I am partition [4] and I have kept a tweet by: 5 2016-12-22 18:36:11.304 STDIO [INFO] 打印出来的tuple:[b, {b属于哪个分区=4, b=1}] 2016-12-22 18:36:11.306 STDIO [INFO] key:b属于哪个分区,value:4 2016-12-22 18:36:11.317 STDIO [INFO] I am partition [3] and I have kept a tweet by: 5 2016-12-22 18:36:11.321 STDIO [INFO] key:b,value:1 2016-12-22 18:36:11.335 STDIO [INFO] 打印出来的tuple:[a, {a属于哪个分区=3, a=2}] 2016-12-22 18:36:11.341 STDIO [INFO] key:a属于哪个分区,value:3 2016-12-22 18:36:11.344 STDIO [INFO] key:a,value:2 2016-12-22 18:36:11.423 STDIO [INFO] I am partition [0] and I have kept a tweet by: 5 2016-12-22 18:36:11.424 STDIO [INFO] I am partition [0] and I have kept a tweet by: 5 2016-12-22 18:36:11.425 STDIO [INFO] I am partition [0] and I have kept a tweet by: 5 2016-12-22 18:36:11.433 STDIO [INFO] 打印出来的tuple:[c, {c=3, c属于哪个分区=0}] 2016-12-22 18:36:11.433 STDIO [INFO] key:c,value:3 2016-12-22 18:36:11.439 STDIO [INFO] key:c属于哪个分区,value:0 2016-12-22 18:36:11.923 STDIO [INFO] I am partition [1] and I have kept a tweet by: 5 2016-12-22 18:36:11.933 STDIO [INFO] 打印出来的tuple:[d, {d=1, d属于哪个分区=1}] 2016-12-22 18:36:11.939 STDIO [INFO] key:d,value:1 2016-12-22 18:36:11.942 STDIO [INFO] key:d属于哪个分区,value:1
下一篇: 超强防护套装