Hello,
Thank you for this great and maintained project.
While trying to use the algorithm on networkx directed graph (by running nx_subdue),
I've noticed that the internal presentation of the graph is undirected (i.e. the 'directed' property of the edges is 'False').
In the previous version it caused wrong pattern results. In the last Subdue version (with overlap parameter) it doesn't influence the results anymore, as far as I saw, but I'm afraid it can influence other things.
I'm adding a link to the problematic line, which I changed in my fork:
https://github.com/bar-ta/Subdue/blob/master/src/Graph.py#L93
It seems like we use a different version of networkx maybe. I'm using the last version: networkx==2.5. Which one do you use?
Thank you,
Bar.
Hello,
Thank you for this great and maintained project.
While trying to use the algorithm on networkx directed graph (by running nx_subdue),
I've noticed that the internal presentation of the graph is undirected (i.e. the 'directed' property of the edges is 'False').
In the previous version it caused wrong pattern results. In the last Subdue version (with overlap parameter) it doesn't influence the results anymore, as far as I saw, but I'm afraid it can influence other things.
I'm adding a link to the problematic line, which I changed in my fork:
https://github.com/bar-ta/Subdue/blob/master/src/Graph.py#L93
It seems like we use a different version of networkx maybe. I'm using the last version: networkx==2.5. Which one do you use?
Thank you,
Bar.