Performing group-by before join query processing

摘要

Assume that we have an SQL query containing joins and a group-by.nThe standard way of evaluating this type of query is to first performnall the joins and then the group-by operation. However, it may benpossible to perform the group-by early, that is, to push the group-bynoperation past one or more joins. Early grouping may reduce the querynprocessing cost by reducing the amount of data participating in joins.nWe formally define the problem, adhering strictly to the semantics ofnNULL and duplicate elimination in SQL2, and prove necessary andnsufficient conditions for deciding when this transformation is valid. Innpractice, it may be expensive or even impossible to test whether thenconditions are satisfied. Therefore, we also present a more practicalnalgorithm that tests a simpler, sufficient condition. This algorithm isnfast and detects a large subclass of transformable queries