Optimizing relational algebra operations using generic equivalence discriminators and lazy products

• Authors: Fritz Henglein
• Venue: Proceedings of the 2010 ACM SIGPLAN workshop on Partial evaluation and program manipulation
• Year: 2010

We show how to efficiently evaluate generic map-filter-product queries, generalizations of select-project-join (SPJ) queries in relational algebra, based on a combination of two novel techniques: generic discrimination-based joins and lazy (formal) products. Discrimination-based joins are based on the notion of (equivalence) discriminator. A discriminator partitions a list of values according to a user-specified equivalence relation on keys the values are associated with. Equivalence relations can be specified in an expressive embedded language for denoting equivalence relations. We show that discriminators can be constructed generically (by structural recursion on equivalence expressions), purely functionally, and efficiently (worst-case linear time). The array-based basic multiset discrimination algorithm of Cai and Paige (1995) provides a base discriminator that is both asymptotically and practically efficient. In contrast to hashing, discrimination is fully abstract (only depends on which equivalences hold on its inputs), and in contrast to comparison-based sorting, it does not require an ordering relation on its inputs. In particular, it is applicable to references (pointers). Furthermore, it has better asymptotic computational complexity than both sorting and hashing. We represent cross-products and unions lazily (symbolically) as formal products of the argument sets (relations). This allows the selection operation to recognize on the fly whenever it is applied to a cross-product and invoke an efficient equijoin implementation. In particular, queries can still be formulated naively, using filter, map and product without an explicit join operation, yet garner the advantages of efficient join-algorithms during evaluation. The techniques subsume many of the optimization techniques based on relational algebra equalities, without need for a query preprocessing phase. They require no indexes and behave purely functionally. They can be considered a form of symbolic execution of set expressions that automate and encapsulate dynamic program transformation of such expressions and lead to asymptotic performance improvements over naive execution in many cases.