Hi Krishna, I get your point. I recommend writing a custom DBWritable class and provide the implementation for the write() method and have this writable instance as the value from you Mapper(if its a Map only job) or Reducer(in case of a Map - Reduce job) . The framework will do the work for you. Regards Ravi On Sun, Mar 1, 2015 at 11:29 PM, Krishna wrote: > Ravi, thanks. > If the target table is salted, do I need to compute the leading byte (as i > understand, its a hash value) in the mapper? > > > On Sunday, March 1, 2015, Ravi Kiran wrote: > >> Hi Krishna, >> >> I assume you have already taken a look at the example here >> http://phoenix.apache.org/phoenix_mr.html >> >> > Is there a need to compute hash byte in the MR job? >> Can you please elaborate a bit more on what hash byte is ? >> >> > Are keys and values stored in BytesWritable before doing a >> "context.write(...)" in the mapper? >> The Key-values from a mapper to reducer are the usual >> Writable/WritableComparable instances and you can definitely write >> BytesWritable . >> >> Regards >> Ravi >> >> On Sun, Mar 1, 2015 at 10:04 PM, Krishna wrote: >> >>> Could someone comment of following questions regarding the usage of >>> PhoenixOutputFormat in a standalone MR job: >>> >>> - Is there a need to compute hash byte in the MR job? >>> - Are keys and values stored in BytesWritable before doing a >>> "context.write(...)" in the mapper? >>> >>> >>> >>