Optimal Incremental Algorithms for Top-k Joins with User-Defined Join Constraints

We investigate the problem of incremental joins of multiple ranked data streams when the join condition is a list of arbitrary user-defined predicates on the input tuples. We propose an algorithm J∗ for ranked input joins over user-defined join predicates. The basic version of the algorithm uses only sequential access into the database and is easily pipelinable—that is, the output of one join query can be fed as the input of another. We also propose a J∗ PA algorithm that can exploit available database indexes for efficient random access based on the join predicates, as well as give 2-approximation versions for both of the above algorithms. Finally, we prove strong optimality results for J∗ and its approximated version, and we study their performance empirically.