使用java开发一个helloworld级别UDF，打包成udf.jar，存放在/home/hadoop/lib下，代码如下：
package com.luogankun.udf; import org.apache.hadoop.hive.ql.exec.UDF; public class HelloUDF extends UDF { public String evaluate(String str) { try { return "HelloWorld " + str; } catch (Exception e) { return null; } } }
Hive中使用UDF
cd $SPARK_HOME/bin spark-sql --jars /home/hadoop/lib/udf.jar CREATE TEMPORARY FUNCTION hello AS ‘com.luogankun.udf.HelloUDF‘;
select hello(url) from page_views limit 1;
SparkSQL中使用UDF
方式一：在启动spark-sql时通过--jars指定
cd $SPARK_HOME/bin spark-sql --jars /home/hadoop/lib/udf.jar CREATE TEMPORARY FUNCTION hello AS ‘com.luogankun.udf.HelloUDF‘;
select hello(url) from page_views limit 1;
方式二：先启动spark-sql后add jar
cd $SPARK_HOME/bin spark-sql add jar /home/hadoop/lib/udf.jar; CREATE TEMPORARY FUNCTION hello AS ‘com.luogankun.udf.HelloUDF‘;
select hello(url) from page_views limit 1;
在测试过程中发现并不支持该种方式，会报java.lang.ClassNotFoundException: com.luogankun.udf.HelloUDF
如何解决？
1）需要先将udf.jar的路径配置到spark-env.sh的SPARK_CLASSPATH中，形如：
export SPARK_CLASSPATH=$SPARK_CLASSPATH:/home/hadoop/software/mysql-connector-java-5.1.27-bin.jar:/home/hadoop/lib/udf.jar
2）再启动spark-sql，直接CREATE TEMPORARY FUNCTION即可；
cd $SPARK_HOME/bin spark-sql CREATE TEMPORARY FUNCTION hello AS ‘com.luogankun.udf.HelloUDF‘;
select hello(url) from page_views limit 1;
方式三：Thrift JDBC Server中使用UDF
在beeline命令行中执行：
add jar /home/hadoop/lib/udf.jar; CREATE TEMPORARY FUNCTION hello AS ‘com.luogankun.udf.HelloUDF‘;
select hello(url) from page_views limit 1;
原文：http://www.cnblogs.com/luogankun/p/3970153.html