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)<i style="color:transparent">本文来源gaodai$ma#com搞$$代**码网$</i>; 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。
搞代码网（gaodaima.com）提供的所有资源部分来自互联网，如果有侵犯您的版权或其他权益，请说明详细缘由并提供版权或权益证明然后发送到邮箱[email protected]，我们会在看到邮件的第一时间内为您处理，或直接联系QQ：872152909。本网站采用BY-NC-SA协议进行授权
转载请注明原文链接：hive 复杂 UDAF 使用方法
转载请注明原文链接：hive 复杂 UDAF 使用方法
