On the implementation of compact DAWG's

There are several data structures that allow searching for a pattern <i>P</i> in a preprocessed text <i>T</i> in time dependent just on the length of <i>P</i>. In this paper we present an implementation of CDAWG's-- Compact Direct Acyclic Word Graphs. While the previous implementations of CDAWG's required from 7<i>n</i> to 23<i>n</i> bytes of memory space, ours achieves 1.7<i>n</i> to 5<i>n</i> for a text <i>T</i> of length <i>n</i>. The implementation is suitable for large data files, since it minimizes the number of disk accesses. If disk accesses are not to be optimized, space requirements can be further decreased.