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：
hive mapreduce script用法示例手工移除.meta.表的错误信息编程方式将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。