An Optimal Trie Construction Algorithm for Partial — Match Queries

Abstract

This paper presents design algorithms of a trie file to answer partial-match queries. The author assumes in this paper that each field of a record is mapped into several bits which construct a bucket address into which the record is stored.

At each level of a trie, at most 2^d-way branching is possible according to predetermined d bits. The problem is the allocation and the selection of the d bits at each level.

Major results are as follows;

• 1) When data is uniform, an optimal trie is obtained.

• 2) When queries are uniform and data is not uniform, three factors which affect the efficiency are found. A new unbalance measure is defined and a brief description of a trie construction algorithm is presented.

• 3) When neither queries nor data is uniform, a heuristic algorithm which is based on four observations is given.