Shortest path length exists but shortest path is empty about pandana

[kaidiwang]

I see. Thank you!

[PyMap]

Hi @kaidiwang! Just wondering what are you calling with the call_client? If I try both methods getting the nodes from the net object I'm always getting results for both shortest path methods

[kaidiwang]

The call_client.origin is 5062 and call_client.destination is 6235.

Hi @kaidiwang! Just wondering what are you calling with the call_client? If I try both methods getting the nodes from the net object I'm always getting results for both shortest path methods

[PyMap]

hey @kaidiwang how are you? I found out that what you get with 5062 and 6235 is not happening just for this pair of nodes. There are some other node pairs with the same behavior.

If you look for them in your edges gdf you will see that these pairs are not connected. So, when creating your net object with nodes and edges gdf you are creating a graph where nodes can be connected or not. That's why you get that empty list when calling the shorest path method

[PyMap]

Exploring it further, I would suggest to:

1. Create the net object from the same source (geodataframe of links):

connected_nodes = np.unique(gdf_links[['o_nodeidx','d_nodeidx']])
gdf_nodes_ = gdf_nodes.loc[gdf_nodes.index.isin(connected_nodes)].copy()

net_ = pdna.Network(gdf_nodes_['X'], gdf_nodes_['Y'], 
                                  gdf_links['o_nodeidx'], gdf_links['d_nodeidx'], 
                                  gdf_links[['TIMAU']], twoway = False)

This way, if you try to get the shortest_path_distance from non connected nodes you will get a nan error as feedback. The nodes are not connected, so the error would be the correct output.

The problem with pandana here is that it returns a shortest_path_distance value (that has no sense) instead of 0 or non connected nodes warning.

2. If you want to work with all the nodes (even if some of them are not connected) you can use the low_connectivity method:

number_of_edges = len(np.unique(net.edges_df[['from','to']]))
number_of_nodes = len(net.nodes_df.index.unique())

# directed graph
node_degree = number_of_edges/number_of_nodes
max_imp = gdf_links['TIMAU'].max()

low_connected_nodes = net.low_connectivity_nodes(impedance=max_imp, count=node_degree*2)

The low_connected_nodes would be a list of nodes to filter or exclude

(cc @sablanchard @smmaurer )