Output-Sensitive Construction of CDAWGs from BWT-Runs

Abstract

The compact directed acyclic word graph (CDAWG) of a string can be viewed in two equivalent ways: as the edge-compacted DAWG of the string, and as the DAG obtained from the suffix tree by merging the nodes whose subtrees are isomorphic. By exploiting these two views in opposite directions, we show how to build, for the (reversed) input string of length $n$, the CDAWG with $e_L$ edges in $O(e_L\log n\log(n/r))$ time with $O(r\log(n/r)+e_L)$ words of working space, provided that the fully functional compressed suffix tree of Gagie, Navarro, and Prezza of size $O(r\log(n/r))$ is available. Here, $r$ denotes the number of BWT-runs of the input string.