Hi, Piggybacking on Gang’s reply, to add files / dirs recursively you can use the filestatus, liststatus to determine if its a file or dir and add as needed ( check FileStatus API for this ) There is a patch which does this for FileInputFormat
http://issues.apache.org/jira/browse/MAPREDUCE-1501 Amogh On 3/23/10 6:25 PM, "Gang Luo" <[email protected]> wrote: Hi Oleg, you can use FileInputFormat.addInputPath(JobConf, Path) multiple times in your program to add arbitrary paths. Instead, if you use FileInputFormat.setInputPath, there could be only one input path. If you are talking about output, the path you give is an output directory, all the output files (part-00000, part-00001...) will be generated in that directory. -Gang ----- 原始邮件 ---- 发件人： Oleg Ruchovets <[email protected]> 收件人： [email protected] 发送日期： 2010/3/23 (周二) 6:18:34 上午 主 题： execute mapreduce job on multiple hdfs files Hi , All examples that I found executes mapreduce job on a single file but in my situation I have more than one. Suppose I have such folder on HDFS which contains some files: /my_hadoop_hdfs/my_folder: /my_hadoop_hdfs/my_folder/file1.txt /my_hadoop_hdfs/my_folder/file2.txt /my_hadoop_hdfs/my_folder/file3.txt How can I execute hadoop mapreduce on file1.txt , file2.txt and file3.txt? Is it possible to provide to hadoop job folder as parameter and all files will be produced by mapreduce job? Thanks In Advance Oleg
