def maximum_acyclic_subgraph_helper_recurse(self, graph):
'''
Graph is an adjacency matrix, and is continuously changed.
Calls make_cut to determine the cut, the computes the edges over each
cut (from A to B and B to A, in the manner described above), removing
edges in the graph and calling the helper function recursively on the
new graphs.
'''
linearizing_agent = DAGSolver(graph)
topo_sort = linearizing_agent.topological_sort()
while topo_sort is None:
cut_one, cut_two = self.make_cut(graph)
edges_one_to_two = self.compute_edges_over_cut(graph, cut_one)
edges_two_to_one = self.compute_edges_over_cut(graph, cut_two)
if len(edges_one_to_two) >= len(edges_two_to_one):
for vertex_one, vertex_two in edges_two_to_one:
graph[vertex_one][vertex_two] = 0
else:
for vertex_one, vertex_two in edges_one_to_two:
graph[vertex_one][vertex_two] = 0
linearizing_agent = DAGSolver(graph)
topo_sort = linearizing_agent.topological_sort()
return topo_sort
def obtain_library_solution(self, scc_adj_matrix, force_ip=False, force_eades=False):
num_vertices = len(scc_adj_matrix)
library_graph = Graph().Adjacency(scc_adj_matrix)
if not force_eades and (num_vertices < 10 or force_ip):
removed_edges = library_graph.feedback_arc_set(method='ip')
else:
removed_edges = library_graph.feedback_arc_set(method='eades')
library_graph.delete_edges(removed_edges)
library_graph_adj_matrix = library_graph.get_adjacency()._get_data()
library_graph_dag_solver = DAGSolver(library_graph_adj_matrix)
solution = library_graph_dag_solver.topological_sort()
return solution
