java - How do you access related entries co-located with data affinity in Hazelcast? -

i trying use hazelcast's map-reduce feature perform aggregate operation, needs access co-located entries. co-location controlled using data-affinity.

imagine classic customer/order model used in hazelcast documentation on data-affinity. in example want return customer summary has customer , sum of orders, example, given data set:

customer_id | name ------------------ 1           | dave 2           | kate   order_id | customer_id | value ------------------------------ 1        | 1           | 5 2        | 1           | 10 3        | 2           | 12 

i want return:

customer_id | name | value -------------------------- 1           | dave | 15 2           | kate | 12 

this simple enough, reason using data-affinity able perform summing logic within respective partition holding data getting orders within partition , therefore avoiding cross jvm communication.

and question, within mapper or similar, how co-located entries in cache?

edit:

after @noctarius' answer , comments, here's code (i've tried make brief possible) highlights point @ want orders current partition.

the order key class looks this:

public class orderkey implements partitionaware<customeridentity> {   ...    @override   public customeridentity getpartitionkey()   {     return this.customeridentity;    }    ... } 

and mapper this:

public class ordersummapper implements mapper<customerkey, customer, customerkey, customerordertotal>, hazelcastinstanceaware {   ...    @override   public void map(customerkey customerkey, customer customer, context<customerkey, customerordertotal> context)   {     predicate ordersforcustomer = new ordersforcustomerpredicate(customerkey);      int totalvalue = 0;      //******************************************************************     //     // given orders co-located customer, how ensure      // call orders runs in current partition?     //     //******************************************************************      (order order : hazelcastinstance.getmap("orders").values(ordersforcustomer))     {       totalvalue += order.getvalue();     }      context.emit(customerkey, new customerordertotal(customer, total));   }    ... } 

the highlighted call hazelcastinstance.getmap("orders").values(ordersforcustomer) ordinarily hit nodes in cluster, because data co-located unncessary overhead.

and me original question, how orders such in current partition returned?

you inject current node's hazelcastinstance mapper , retrieve second data structure read data.

see basic example here: https://github.com/noctarius/hazelcast-mapreduce-presentation/blob/master/src/main/java/com/hazelcast/examples/tutorials/impl/salarymapper.java