Database system with improved methods for filtering duplicates from a tuple stream

摘要

A Client/Server Database system is described which includes a Database Server providing methods eliminating duplicates from an ordered tuple stream (e.g., resulting from a query involving a database "join"), without the need for performing an expensive sort operation. Specifically, the system provides a "filter" which eliminates duplicates without having to perform a sort. The filter, which is implemented as an optimization at the level of the query processor, comprises two basic pieces. The first piece, INIT.sub.-- FILTER, simply serves to initialize the filter--that is, the piece sets a flag that forces the filter to pass the first tuple encountered and to construct a first key from it. The second piece, FILTER, serves as the actual filter, when the system scans the tuple stream. If the current tuple has the same key as the preceding tuple, then the current tuple is thrown away. Otherwise, the current tuple is passed and a new key is constructed from it. The positions of both INIT.sub.-- FILTER and FILTER in a given join order are important. INIT.sub.-- FILTER immediately preceeds the scan which initializes the filter; FILTER immediately follows the scan which actually performs the filtering.