1

I have a very large DAG of strings (~200k). I would like to find the longest path that exists in this graph. The below code is how I've set up the graph (from the list of strings new_list).

#create new empty graph
g = nx.DiGraph()

#add all words to graph
for word in new_list:
    g.add_node(word)

#fill graph with valid word chains
for word in g.nodes():
    for c in string.ascii_lowercase:
        new_word = alphagramatize(word+c) #add char c to word in alphagram order
        if(binary_search(new_list, new_word) != -1):
            g.add_edge(word, new_word)

I have attempted the naive approach of checking the path distance from every node to every other node... this is clearly impractical and will not terminate.

I have also attempted to rework the longest_path() code from this thread, to no avail.

I have read much on what I could understand about performing a topological sort and ordering of the graph, but am having trouble implementing this. Networkx provides a function topological_sort(g), so that work is done for me. However, how do I proceed from here now that I have a topo_sorted graph?

Community
  • 1
  • 1
Forerunner117
  • 11
  • 4
  • What doesn't work with the code at http://stackoverflow.com/questions/17985202/networkx-efficiently-find-absolute-longest-path-in-digraph ? – Aric Oct 22 '13 at 18:17
  • @Aric It seems only to output a single node in my graph with a distance. I am searching for the entire path (all nodes in the longest path). I notice yours adds paths (`G.add_path()`) rather than adding edges like mine (`G.add_edge()`). Should I rework my code to generate paths rather than edges? – Forerunner117 Oct 22 '13 at 19:37
  • No adding edges is fine. I updated that answer with (hopefully) a corrected version :-). That one was profoundly wrong except for the special case I tested it on... – Aric Oct 22 '13 at 21:49
  • @Aric Your updated code works wonders! Thank you sir! – Forerunner117 Oct 22 '13 at 23:21

0 Answers0