ON SEhI.W,TIC REEFS M-l-l EFFICIENT PROCESINC; OF ~RRELATION QUERIES WITH AGGREGATES

Recent transformation algorithms for speeding up processing of nested SQL-like queries with aggregates are reviewed with respect to the correctness of aggregates over empty sets. It turns out that for a particular subset of such queries these algorithms fail to compute consistent answers. Unfortunately there seems to be no uniform way to do these transformations efficiently and correctly under all circumstances. Also the algorithms for QUEL are reexamined regarding their correctness. It is shown that for a specific subset of QUEL-queries with aggregates a clearer semantics can be associated. Finally, benchmark results for lngres show that considerable performance advantages may be gained for such query types by using dynamic filters. The consequence of all these observations is that more research is required to integrate correlation queries with aggregates into a unified operator tree model.