hive 复杂 UDAF 使用方法

周海汉 /文 2013.3.27 前文《hive mapreduce script用法示例》 示例了mapreduce脚本。本文采用较复杂的方式自定义hive聚合函数。 package com.abloz.hive;/** * @author zhouhh * @date 2013-3-27 * note: for count value =1 */import org.apache.hadoop.hi

周海汉 /文
2013.3.27

前文《hive mapreduce script用法示例》
示例了mapreduce脚本。本文采用较复杂的方式自定义hive聚合函数。

package com.abloz.hive;
/**
 * @author zhouhh
 * @date 2013-3-27
 * note: for count value >=1
 */
import org.apache.hadoop.hive.ql.exec.UDAF;
import org.apache.hadoop.hive.ql.exec.UDAFEvaluator;
import org.apache.hadoop.io.IntWritable;
import org.apache.hadoop.io.Text;
import org.apache.log4j.Logger;
public class JJPokerStat extends UDAF {
	public static class JJCountUDAFEvaluator implements UDAFEvaluator {
		public static final Logger log = Logger.getLogger(JJPokerStat.class);
		public static class PartialResult {
			int total=0;
			int win=0;
			int fold=0;
			int allin=0;
		}
		private PartialResult result;
		public void init() {
			result = new PartialResult();
		}
		public boolean calwin(int rbet,String chipwon)
		{
			if(chipwon.equalsIgnoreCase("NULL"))
			{
				return false;
			}
			String[] cw = chipwon.split("\\|");
			int chipwons=0;
			//log.info("calwin:"+chipwon);
			for(String v:cw)
			{
				String[] c = v.split(":");
				//log.info("calwin:v "+v+",c.length:"+c.length);
				if(c.length>1)
				{
					chipwons += Integer.parseInt(c[1]);
				}
			}
			//log.info("calwin:chipwons:"+chipwons+",rbet:"+rbet);
			if(chipwons>rbet)
			{
				return true;
			}
			return false;
		}
		public boolean iterate(IntWritable rbet,Text chipwon,IntWritable f,IntWritable a) {
			if ( rbet == null || chipwon == null || f == null || a == null) {
				return true;
			}
			boolean win = calwin(rbet.get(),chipwon.toString());
			if(result == null)
			{
				result = new PartialResult();
			}
			result.total++;
			if(win)
			{
				result.win++;
			}
			int v = f.get();
			if (v>=1)
			{
				result.fold++;
			}
			v = a.get();
			if (v>=1)
			{
				result.allin++;
			}
			return true;
		}
		public PartialResult terminatePartial() {
			return result;
		}
		public boolean merge(PartialResult other) {
			if(other == null)
			{
				return true;
			}
			result.total+=other.total;
			result.win += other.win;
			result.fold += other.fold;
			result.allin += other.allin;
			return true;
		}
		public Text terminate() {
			if(result == null) {
				return new Text("0\t0\t0\t0");
			}
			String s=""+result.total+"\t"+result.win+"\t"+result.fold+"\t"+result.allin;
			return new Text(s);
		}
	}
}

相关博文：

1. hive mapreduce script用法示例
2. 手工移除.META.表的错误信息
3. 编程方式将hbase数据复制到mysql

原文地址：hive 复杂 UDAF 使用方法, 感谢原作者分享。