| 摘要 |
A server performing an indexing method of data management to create and maintain indexes more efficiently than existing indexing approaches is described. The server is disposed between an application program and a DBMS and is coupled to a data base located within the DBMS. The data base has an ordered set of data values stored in memory. Each data value has a bit pattern and an identifier associated therewith. The server creates a plurality of bit vectors such that the number of bit vectors created equals the longest length bit pattern for the values. The server accesses one of the values stored in the data base. Each bit of the bit pattern for the value is then assigned by the server to a unique position in successive bit vectors. The bits are assigned to identical unique positions in each of the successive bit vectors. The server repeats the above-described accessing and assigning steps for each remaining value of the set to form an index of bit vectors for the values. Methods are provided for improving the performance of database queries when using bit-vector or HighNonGroup (HNG) indexes. Such queries include, for instance, aggregate operations specified in an SQL statement, such as SUM, MAX, MIN, and AVG operations. Specific methods described include optimizing "range" comparisons by reducing bit operations, optimization of MAX and MIN operations, optimization of SUM and AVG operations, implementation of a "Datepart" index, and execution of SUBSTRING predicates in an HNG index.
|
