specifies that when partitioning data, the distribution of partitions is not determined by a tree, but by a hashing algorithm. As a result, the distribution of the partitions is not as evenly balanced, but it is effective when working with high-cardinality partition keys (in the order of millions of partitions).
data hdfs.transactions(partition=(cust_id year) hash=yes);
set somelib.sometable;
run;