On the Longest Common Cartesian Substring Problem

Abstract

A Cartesian tree is associated with a string of numbers and is structured as a heap from which the original string can be recovered. Although Cartesian trees have been introduced 40 years ago, the Cartesian tree matching problem appeared very recently. It consists in finding all substrings of given text, which have the same Cartesian tree as that of a given pattern. In this paper, we address the problem of computing the longest common Cartesian substrings of two strings and present three methods for such problem. Our first method is based on a classical suffix tree construction and solves the problem in randomized linear time and linear space, although the space overhead is quite prohibitive in the case of large strings. Our second solution is based on classical dynamic programming, and our third solution is based on a constructive approach. Both of them run in quadratic worst case time but are more space economical in practice. From our experimental results, it turns out that our second solution runs faster than the standard suffix tree solution for short strings, whereas our third solution is more suitable for large strings, when storing a full suffix tree becomes prohibitive.