Thanks Zheng. It helps. -Gang
----- 原始邮件 ---- 发件人： Zheng Shao <[email protected]> 收件人： [email protected]; [email protected] 发送日期： 2009/11/12 (周四) 7:10:33 下午 主 题： Re: How to use comstom mapreduce script Yes you can compile it into a jar, and insert the command line "java /xxx/my.jar" into Hive queries. http://www.slideshare.net/ragho/hive-user-meeting-august-2009-facebook Page 72 has an example IF your map/reduce function is simple, you can probably write a Hive UDF instead. In the near future, Hive will support UDTF and you can write a UDTF for any map/reduce function. This will make it more efficient than the current way (running "java /xxx/my.jar"). Zheng 2009/11/13 Gang Luo <[email protected]>: > Hi all, > It is said that we can plug our own map/reduce into hive. I got two > questions here. > 1) Can I write this 'script' in Java? If so, should I compile it and make it > into a jar file first? > 2) Should I load this script file into HDFS first? > > Thanks. > > --Gang > > ________________________________ > 好玩贺卡等你发，邮箱贺卡全新上线！ -- Yours, Zheng ___________________________________________________________ 好玩贺卡等你发，邮箱贺卡全新上线！ http://card.mail.cn.yahoo.com/
