==作者：YB-Chi==

1. idea新建个maven项目，引入依赖

<dependencies>
       <dependency>
           <groupId>org.apache.hive</groupId>
           <artifactId>hive-exec</artifactId>
           <version>2.1.1</version>
           <exclusions>
               <exclusion>
                   <groupId>*</groupId>
                   <artifactId>*</artifactId>
               </exclusion>
           </exclusions>
       </dependency>
   </dependencies>

2. 编写函数类，继承UDF，并实现evaluate方法

import org.apache.hadoop.hive.ql.exec.UDF;
public class split_distinct extends UDF {

    public String evaluate(String line, String split) {
        if (null == line) return null;
        StringBuffer sb = new StringBuffer();
        String[] sp = line.split(split);
        for (String s : sp) {
            if (null != s && sb.indexOf(s) < 0) sb.append(s).append(",");
        }
        return sb.deleteCharAt(sb.length() - 1).toString();
    }
}

3. maven-package打包，上传不带依赖的包并测试

hive>add jar /home/module/testdata/HiveUDF-1.0-SNAPSHOT.jar;
CREATE TEMPORARY FUNCTION split_distinct AS 'split_distinct';
#测试
select split_distinct('[["C8:13:8B:81:32:C1"],["C8:13:8B:81:32:C1"],["C8:13:8B:81:32:C1"],["C8:13:8B:81:32:C2"]]',',');

4. 测试成功后，注册永久函数

#建立hdfs文件夹
hdfs dfs -mkdir /out/hive_udf_jars
#上传
hdfs dfs -put HiveUDF-1.0-SNAPSHOT.jar /out/hive_udf_jars/
#注册
create function split_distinct as 'split_distinct' using jar 'hdfs://xbsafe102/out/hive_udf_jars/HiveUDF-1.0-SNAPSHOT.jar';
#删除函数
drop function split_distinct;
#注意更新jar包需要重启hive或者spark-sql客户端！！！！！