An Efficient Method of Indexing All Topological Orders for a Given DAG

Topological orders of a directed graph are an important concept of graph algorithms. The generation of topological orders is useful for designing graph algorithms and solving scheduling problems. In this paper, we generate and index all topological orders of a given graph. Since topological orders are permutations of vertices, we can use the data structure πDD, which generates and indexes a set of permutations. In this paper, we propose Rot-πDDs, which are a variation of πDDs based on a different interpretation. Compression ratios of Rot-πDDs for representing topological orders are theoretically improved from the original πDDs. We propose an efficient method for constructing a Rot-πDD based on dynamic programming approach. Computational experiments show the amazing efficiencies of a Rot-πDD: a Rot-πDD for 3.7× 10 topological orders has only 2.2× 10 nodes and is constructed in 36 seconds. In addition, the indexed structure of a Rot-πDD allows us to fast post-process such as edge addition and random samplings.