On Fri, Oct 1, 2010 at 3:08 PM, Jinsong Hu <jinsong...@hotmail.com> wrote: > Hi, There: > I wonder if it is possible to run map-reduce again hive's raw data. > hive supports hql, but sometimes, I want to run map-reduce to do more > sophisticated > processing than those simple hql can handle. In that case, I need to run my > own custom map-reduce job against hive's raw data. > I wonder if that is possible. The key issue is where to find those files > and how to deserialize them. > Can anybody point me into the right location to find the API ? > > Jimmy. >
Jimmy, The files are typically found in /user/hive/warehouse/<tablename> By default they would be TextFiles delimited with ^A. But depending how you defined the table, possibly other delimiters,sequence files they could in a different format.