Compressed index for dynamic text

Abstract

This paper investigates how to index a text which is subject to updates. The best solution in the literature is based on suffix tree using O(n log n) bits of storage, where n is the length of the text. It supports finding all occurrences of a pattern P in O(|P| + occ) time, where occ is the number of occurrences. Each text update consists of inserting or deleting a substring of length y and can be supported in O(y + √n) time. In this paper, we initiate the study of compressed index using only O(n log |Σ|) bits of space, where Σ denotes the alphabet. Our solution supports finding all occurrences of a pattern P in O(|P|log2 n(logε n + log |ΣE|) + occ log1+ε n) time, while insertion or deletion of a substring of length y can be done in O((y + √n) log 2+εn) amortized time, where 0 < ε ≤ 1. The core part of our data structure is based on the recent work on Compressed Suffix Trees (CST) and Compressed Suffix Arrays (CSA).