hive之UDF编程

一、UDF函数可以直接应用于select语句，对查询结构做格式化处理后，再输出内容。

二、编写UDF函数的时候需要注意一下几点：

a）自定义UDF需要继承org.apache.hadoop.hive.ql.UDF。

b）需要实现evaluate函数，evaluate函数支持重载。

三、编写UDF函数代码

0.要继承org.apache.hadoop.hive.ql.exec.UDF类实现evaluate 方法。

public class NationUDF extends UDF {public static Map<String,String> nationMap = new HashMap<String,String>();static{nationMap.put("China", "中国");nationMap.put("Japan", "小日本");nationMap.put("USA", "美帝");}Text t = new Text();//1000 sum(income)//返回值：中国  getNation(nation)public Text evaluate(Text nation){String nation_e = nation.toString();String name = nationMap.get(nation_e);if(name == null){name = "火星人";}t.set(name);return t;}}

0.1打jar包

右键-->export-->java/jar file-->next-->勾选-->finish;

注：打jar包时，注意jdk版本的问题，centOS的hadoop框架下jdk可以向下兼容，也就说，hadoop框架的jdk版本>=UDFjar的版本。

四、自定义函数调用过程：
1.添加jar包（在hive命令行里面执行）
hive> add jar /root/NUDF.jar;

2.创建临时函数（所谓的temporary，就是函数在本会话（此次hive客户端）有效）
hive> create temporary function getNation as 'com.heres.hive.NationUDF';

3.调用
hive> select id, name, getNation(nation) from beauties;

4.利用自定义函数查询结果

hive>  select id,name,size,getNation(nation) from beauties order by size desc;

Total jobs = 1
Launching Job 1 out of 1
Number of reduce tasks determined at compile time: 1
In order to change the average load for a reducer (in bytes):
  set hive.exec.reducers.bytes.per.reducer=<number>
In order to limit the maximum number of reducers:
  set hive.exec.reducers.max=<number>
In order to set a constant number of reducers:
  set mapreduce.job.reduces=<number>
Starting Job = job_1494132664539_0002, Tracking URL = http://heres04:8088/proxy/application_1494132664539_0002/
Kill Command = /heres/hadoop-2.2.0/bin/hadoop job  -kill job_1494132664539_0002
Hadoop job information for Stage-1: number of mappers: 1; number of reducers: 1
2017-05-07 13:03:37,424 Stage-1 map = 0%,  reduce = 0%
2017-05-07 13:03:51,342 Stage-1 map = 100%,  reduce = 0%, Cumulative CPU 2.57 sec
2017-05-07 13:04:00,996 Stage-1 map = 100%,  reduce = 100%, Cumulative CPU 4.41 sec
MapReduce Total cumulative CPU time: 4 seconds 410 msec
Ended Job = job_1494132664539_0002
MapReduce Jobs Launched: 
Job 0: Map: 1  Reduce: 1   Cumulative CPU: 4.41 sec   HDFS Read: 328 HDFS Write: 160 SUCCESS
Total MapReduce CPU Time Spent: 4 seconds 410 msec
OK
1       bgyjy   56.6565 小日本
4       bing    56.56   火星人
3       liu     45.0    火星人
3       ewrwe   43.9    小日本
1       glm     34.0    火星人
2       lina    30.9    火星人
2       jzmb    23.232  小日本
Time taken: 37.869 seconds, Fetched: 7 row(s)

5、将查询结果保存到HDFS中

create table result row format delimited fields terminated by '\t' as select id, getNation(nation) from beauties;

6、销毁临时函数：

hive>DROP TEMPORARY FUNCTION getNation;