Support scalable solution sets on the native heap. Support for this feature should be integrated in two locations:
1. Query engine (input queues for operators).
2. Solution sets resulting from "at-once" evaluation of SPARQL 1.1 subqueries.
The basic mechanism for a scalable "solution set" is a heap, which is just an unordered collection of solutions. If the solution set is "distinct" then the heap can be modeled by a map from the distinct keys to the solution for each such key. Solution sets may be made scalable by buffering them on NIO buffers (on the native process heap).