query about join stage

wuruima · Post by **wuruima** » Thu May 21, 2015 1:20 am

hi friends, I have queries about using join stage.
is it necessary to SORT both links using the keys we want to join?

in parallel job, does the partition wll impact the join result ?

Thanks.

ray.wurlod · Post by **ray.wurlod** » Thu May 21, 2015 4:37 pm

Indirectly yes. If you don't specify sorting, then compiling the job will insert sort operators on each input link to the Join stage.

Data should be partitioned on at least the first join key.