| 摘要 |
A computer-implemented method that speeds up relational database qualification processing by emulating the function of a multiple dimension index, including constant expressions and joins involving complex functions. The method comprises the following steps. Establishing a command comprising a plurality of range variables related by join operators. Breaking up the qualification portion of the command into sub-expressions of the general type f(a) (join) f(b), or f(a) join "constant expression", where "a" and "b" are genetic range variables. The sub-expressions are then mapped onto component joins that may be used repeatedly for numerous commands. The component joins may contain boolean operators. Establishing a range variable processing order to resolve processing ambiguities, both in ordering of the range variables and in specialized join types such as outer-joins, and eliminate false roots. Evaluating the component joins to make a partial index for each join, wherein the partial index comprises a pointer table such that it may be used simultaneously with other partial index. Finally, looping over range variables using boolean processing to combine the join vectors, as needed, to complete the qualification. The present method eliminates the need for a user defined index. A processing gain is achieved without the associated increase in storage requirements normally encountered with a multiple dimension index. The measured performance gain for join processing is about two orders of magnitude, while the overall performance gain was measured at about an order of magnitude.
|
