Generalize is_tree and is_forest to directed graphs. #1145
Conversation
Generalize is_tree(), is_forest() to handle directed graphs.
|
Here's a migrated trac ticket #639 for http://en.wikipedia.org/wiki/Edmonds'_algorithm related to arborescences. |
|
Ah great. I'll attach the stuff to that. I didn't implement the method from Gabow, Galil, Spencer, Tarjan...mostly because Edmond's algorithm seemed easier to implement. Still though, NetworkX data structures require a bit of work for algorithm like these. This is, in part, because edges are second class citizens in the primary graph classes. Add to that time-varying graphs that need to maintain edge identity, and it can be a bit painful. |
Is this still happening, or would it be worth separating out the fix and unit test for just the |
|
This is ready for merging, just needs another eye on it to for suggestions/corrections/etc. Take time to read the module docstring. From what I can see, there are two conventions in the literature. I tried to explain what these conventions were and which one NetworkX is using. |
| """ | ||
| for graph in nx.connected_component_subgraphs(G): | ||
| if not nx.is_tree(graph): | ||
| n = G.number_of_nodes() |
|
The documentation is clear to me. |
Generalize is_tree and is_forest to directed graphs. Also fixes #1144.
Also add recognition for branchings and arborescences. I'm preparing another pull request which builds on this code. I was planning to add unit tests in that PR (next week).