On 8/19/17 3:28 AM, Remil Mohanan wrote:
I am trying to pass multiple non key values from mapper to reducer.
The only way to pass data from the mapper to the reducer is through passing key-values. One common trick is to designate a special key as the out-of-band information key and then use a custom sorting comparator to make sure that key comes first in the sort order. I'm sure you can find examples online.
Similarly for reading and writing a file inside the hdfs system other than normal read and write.
I don't understand. Reading and writing a file in HDFS from an MR task works exactly the same as doing it from a stand-alone program. You probably want to do it in the setup() method, though.
Daniel --------------------------------------------------------------------- To unsubscribe, e-mail: mapreduce-dev-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-dev-h...@hadoop.apache.org
