On Apr 29, 2008, at 7:39 AM, Amar Kamat wrote:
You can override the configure() method in the map class to get the output filename. Use FileOutputFormat.getOutputPath(conf) to get the output filename. This will work for 0.17 and later. For earlier versions you can use conf.getOutputPath().
In Hadoop 0.15 and 0.16, if a task asks for the output path, they do not get the original path, but a working directory. Unfortunately, the number of levels of working directories changed from 1 to 2, so it depends which version you are using. In 0.17, this has been straightened out and you can get either the working directory or the original path. (FileInputFormat.getOutputPath and getWorkOutputPath)
-- Owen
