Groupwise Processing of Relational Queries

摘要

In this paper, we define and examine a particular class of queries called group queries. Group queries are natural queries in many decision-support applications. The main characteristic of a group query is that it can be executed in a group-by-group fashion. In other words, the underlying relation(s) can be partitioned (based on some set of attributes) into disjoint groups, and each group can be processed separately. We give a syntactic criterion to identify these queries and prove its sufficiency. We also prove the strong result that every group query has an equivalent formulation that satisfies our syntactic criterion. We describe a general evaluation technique for group queries, and demonstrate how an optimizer can determine this plan. We then consider more complex queries whose components are group queries with potentially different partitioning attributes. We give two methods to identify group query components within such a query. We also give some performance results for group queries expressed in standard SQL, comparing a commercial database system with our optimized plan on top of the same commercial system. These results indicate that there are significant potential performance improvements.