Order-preserving key transformations

Abstract

File organizations based on conventional hash functions provide faster access to the stored records in comparison with tree-like file structures. Tree structures such as B + -trees and ISAM do provide for sequential processing, but require considerable storage for the indices. When sequential processing is needed a table that performs an order-preserving transformation on keys can be used. H is an order-preserving key transform if H(K 1 ) ⩾ H(K 2 ), for all keys K 1 > K 2 . We present methodologies for constructing such key transforms, and illustrate them for some real-life key sets. Storage requirements for the table needed to carry out the transformation are less than those needed for the indices.