Constructing Su x Arrays of Large Texts

Recently, Sadakane [12] proposes a new fast and memory e cient algorithm for sorting su xes of a text in lexicographic order. It is important to sort su xes because an array of indexes of su xes is called sufx array and it is a memory e cient alternative of the su x tree. Sorting su xes is also used for the Burrows-Wheeler transformation in the Block Sorting text compression, therefore fast sorting algorithms are desired. In full-text databases, of course the length of texts are quite large, and this algorithm makes it possible to use the su x array data structure and the compression scheme for such larger texts. In this paper, we compare algorithms for making su x arrays of Bentley-Sedgewick, Andersson-Nilsson and Karp-Miller-Rosenberg and making su x trees of Larsson on speed and required memory and compare them with our new algorithm which is fast and memory e cient by combining them.