Trident学习笔记(二)
aggregator
------------------
聚合动作;聚合操作可以是基于batch、stream、partiton
[聚合方式-分区聚合]
partitionAggregate
分区聚合;基于分区进行聚合运算;作用于分区而不是batch。
mystream.partitionAggregate(new Fields("x"), new Count(), new Fields("count"));
[聚合方式-batch聚合]
aggregate
批次聚合;先将同一batch的所有分区的tuple进行global再分区,将其汇集到一个分区中,再进行聚合运算。
.aggregate(new Fields("a"), new Count(), new Fields("count")); // 批次聚合
聚合函数
[ReducerAggregator]
init();
reduce();
Aggregator
CombinerAggregator
import org.apache.storm.trident.operation.ReducerAggregator;
import org.apache.storm.trident.tuple.TridentTuple; /**
* 自定义sum聚合函数
*/
public class SumReducerAggregator implements ReducerAggregator<Integer> { private static final long serialVersionUID = 1L; @Override
public Integer init() {
return 0;
} @Override
public Integer reduce(Integer curr, TridentTuple tuple) {
return curr + tuple.getInteger(0) + tuple.getInteger(1);
} }
分区聚合
import net.baiqu.shop.report.trident.demo01.PrintFunction;
import org.apache.storm.Config;
import org.apache.storm.LocalCluster;
import org.apache.storm.trident.Stream;
import org.apache.storm.trident.TridentTopology;
import org.apache.storm.trident.testing.FixedBatchSpout;
import org.apache.storm.tuple.Fields;
import org.apache.storm.tuple.Values; public class TridentTopologyApp4 { public static void main(String[] args) {
// 创建topology
TridentTopology topology = new TridentTopology(); // 创建spout
FixedBatchSpout testSpout = new FixedBatchSpout(new Fields("a", "b"), 4,
new Values(1, 2),
new Values(2, 3),
new Values(3, 4),
new Values(4, 5)); // 创建流
Stream stream = topology.newStream("testSpout", testSpout);
stream.partitionAggregate(new Fields("a", "b"), new SumReducerAggregator(), new Fields("sum"))
.shuffle().each(new Fields("sum"), new PrintFunction(), new Fields("result")); // 本地提交
LocalCluster cluster = new LocalCluster();
cluster.submitTopology("TridentDemo4", new Config(), topology.build());
} }
[ReducerAggregator]
init();
reduce();
public interface ReducerAggregator<T> extends Serializable {
T init();
T reduce(T curr, TridentTuple tuple);
}
[Aggregator]
描述同ReducerAggregator.
public interface Aggregator<T> extends Operation {
// 开始聚合之间调用,主要用于保存状态。共下面的两个方法使用
T init(Object batchId, TridentCollector collector);
// 迭代batch的每个tuple, 处理每个tuple后更新state的状态。
void aggreate(T val, TridentTuple tuple, TridentCollector collector);
// 所有tuple处理完成后调用,返回单个tuple给每个batch。
void complete(T val, TridentCollector collector);
}
CombinerAggregator
import org.apache.storm.trident.operation.BaseAggregator;
import org.apache.storm.trident.operation.TridentCollector;
import org.apache.storm.trident.tuple.TridentTuple;
import org.apache.storm.tuple.Values; import java.io.Serializable; /**
* 批次求和函数
*/
public class SumAggregator extends BaseAggregator<SumAggregator.State> { private static final long serialVersionUID = 1L; static class State implements Serializable {
private static final long serialVersionUID = 1L;
int sum = 0;
} @Override
public SumAggregator.State init(Object batchId, TridentCollector collector) {
return new State();
} @Override
public void aggregate(SumAggregator.State state, TridentTuple tuple, TridentCollector collector) {
state.sum = state.sum + tuple.getInteger(0) + tuple.getInteger(1);
} @Override
public void complete(SumAggregator.State state, TridentCollector collector) {
collector.emit(new Values(state.sum));
} }
批次聚合
package net.baiqu.shop.report.trident.demo04; import net.baiqu.shop.report.trident.demo01.PrintFunction;
import org.apache.storm.Config;
import org.apache.storm.LocalCluster;
import org.apache.storm.trident.Stream;
import org.apache.storm.trident.TridentTopology;
import org.apache.storm.trident.testing.FixedBatchSpout;
import org.apache.storm.tuple.Fields;
import org.apache.storm.tuple.Values; public class TridentTopologyApp4 { public static void main(String[] args) {
// 创建topology
TridentTopology topology = new TridentTopology(); // 创建spout
FixedBatchSpout testSpout = new FixedBatchSpout(new Fields("a", "b"), 4,
new Values(1, 2),
new Values(2, 3),
new Values(3, 4),
new Values(4, 5)); // 创建流
Stream stream = topology.newStream("testSpout", testSpout);
stream.aggregate(new Fields("a", "b"), new SumAggregator(), new Fields("sum"))
.shuffle().each(new Fields("sum"), new PrintFunction(), new Fields("result")); // 本地提交
LocalCluster cluster = new LocalCluster();
cluster.submitTopology("TridentDemo4", new Config(), topology.build());
} }
运行结果
PrintFunction:
PrintFunction:
PrintFunction:
......
平均值批次聚合函数
import org.apache.storm.trident.operation.BaseAggregator;
import org.apache.storm.trident.operation.TridentCollector;
import org.apache.storm.trident.tuple.TridentTuple;
import org.apache.storm.tuple.Values; import java.io.Serializable; /**
* 批次求平均值函数
*/
public class AvgAggregator extends BaseAggregator<AvgAggregator.State> { private static final long serialVersionUID = 1L; static class State implements Serializable {
private static final long serialVersionUID = 1L;
// 元组值的总和
float sum = 0;
// 元组个数
int count = 0;
} /**
* 初始化状态
*/
@Override
public AvgAggregator.State init(Object batchId, TridentCollector collector) {
return new State();
} /**
* 在state变量中维护状态
*/
@Override
public void aggregate(AvgAggregator.State state, TridentTuple tuple, TridentCollector collector) {
state.count = state.count + 2;
state.sum = state.sum + tuple.getInteger(0) + tuple.getInteger(1);
} /**
* 处理完成所有元组之后,返回一个具有单个值的tuple
*/
@Override
public void complete(AvgAggregator.State state, TridentCollector collector) {
collector.emit(new Values(state.sum / state.count));
} }
批次聚合求平均值
import net.baiqu.shop.report.trident.demo01.PrintFunction;
import org.apache.storm.Config;
import org.apache.storm.LocalCluster;
import org.apache.storm.trident.Stream;
import org.apache.storm.trident.TridentTopology;
import org.apache.storm.trident.testing.FixedBatchSpout;
import org.apache.storm.tuple.Fields;
import org.apache.storm.tuple.Values; public class TridentTopologyApp4 { public static void main(String[] args) {
// 创建topology
TridentTopology topology = new TridentTopology(); // 创建spout
FixedBatchSpout testSpout = new FixedBatchSpout(new Fields("a", "b"), 4,
new Values(1, 2),
new Values(2, 3),
new Values(3, 4),
new Values(4, 5)); // 创建流
Stream stream = topology.newStream("testSpout", testSpout);
stream.aggregate(new Fields("a", "b"), new AvgAggregator(), new Fields("avg"))
.shuffle().each(new Fields("avg"), new PrintFunction(), new Fields("result")); // 本地提交
LocalCluster cluster = new LocalCluster();
cluster.submitTopology("TridentDemo4", new Config(), topology.build());
} }
[CombinerAggregator]
在每个partition运行分区聚合,然后再进行global再分区将同一对batch的所有tuple分到一个partition中,最后再这一个partition中进行聚合运算,并产生结果进行输出。
该种方式的网络流量占用少于前两种方式。
public interface CombinerAggregator<T> extents Serializable {
// 在每个tuple上运行,并接受字段值
T init(TridentTuple tuple);
// 合成tuple的值,并输出一个值的tuple
T combine(T val1, T vak2);
// 如果分区不含有tuple,调用该方法.
T zero();
}
合成聚合函数
import clojure.lang.Numbers;
import org.apache.storm.trident.operation.CombinerAggregator;
import org.apache.storm.trident.tuple.TridentTuple; /**
* 合成聚合函数
*/
public class SumCombinerAggregator implements CombinerAggregator<Number> { private static final long serialVersionUID = 1L; @Override
public Number init(TridentTuple tuple) {
return (Number) tuple.getValue(0);
} @Override
public Number combine(Number val1, Number val2) {
return Numbers.add(val1, val2);
} @Override
public Number zero() {
return 0;
} }
topology
import net.baiqu.shop.report.trident.demo01.PrintFunction;
import org.apache.storm.Config;
import org.apache.storm.LocalCluster;
import org.apache.storm.trident.Stream;
import org.apache.storm.trident.TridentTopology;
import org.apache.storm.trident.testing.FixedBatchSpout;
import org.apache.storm.tuple.Fields;
import org.apache.storm.tuple.Values; public class TridentTopologyApp4 { public static void main(String[] args) {
// 创建topology
TridentTopology topology = new TridentTopology(); // 创建spout
FixedBatchSpout testSpout = new FixedBatchSpout(new Fields("a", "b"), 4,
new Values(1, 2),
new Values(2, 3),
new Values(3, 4),
new Values(4, 5)); // 创建流
Stream stream = topology.newStream("testSpout", testSpout);
stream.aggregate(new Fields("a", "b"), new SumCombinerAggregator(), new Fields("sum"))
.shuffle().each(new Fields("sum"), new PrintFunction(), new Fields("result")); // 本地提交
LocalCluster cluster = new LocalCluster();
cluster.submitTopology("TridentDemo4", new Config(), topology.build());
} }
输出结果
PrintFunction:
PrintFunction:
PrintFunction:
......
最新文章
- Another app is currently holding the yum lock
- 基于Python的TestAgent实现
- Mongodb 的基本使用
- 分享50款 Android 移动应用程序图标【下篇】
- sd_cms置顶新闻,背景颜色突击显示
- 如何解决paramiko执行与否的问题
- 【转载】Kafka High Availability
- [BZOJ 1042] [HAOI2008] 硬币购物 【DP + 容斥】
- Discuz X2.5 用户名包含被系统屏蔽的字符[解决方法]
- (转) C# Activator.CreateInstance()方法使用
- 如何中途停止RMAN备份任务
- 如何使用SQLite数据库 匹配一个字符串的子串?
- Netty(二) 从线程模型的角度看 Netty 为什么是高性能的?
- 一个简单的TensorFlow可视化MNIST数据集识别程序
- ubuntu 下搭建redis和php的redis的拓展
- mac安装navicat mysql破解版
- wps去除首字母自动大写
- background 和渐变 总结
- ubuntu安装redis 和可视化工具
- 【六】注入框架RoboGuice使用:(Singletons And ContextSingletons)