Dan Schult

I think the colored edges should work if you include an edge_colors list with the colors listed in the same order as the edges in G.edges. You should be able...

Something like: ```python import networkx as nx import matplotlib.pyplot as plt G=nx.MultiDiGraph() G.add_edges_from([(0,1),(0,0),(0,2),(2,2)]) G.add_edges_from([(0,0),(2,2)]) ec=['r','g','b','m','c','k'] pos = nx.spring_layout(G) nx.draw_networkx_edges(G, pos, width=3, edge_color=ec) ```

In addition to using a random number `seed` to keep the springlayout consistent, you could save the `pos` coordinates between drawing. For the shell_layout oddness you are running into a...

That one looks like a historical bug to me -- but I can explain what's going on. The historical bug is that for multigraphs we have switched between adding each...

We should fix connectionstyle for multigraphs to use a full list of connectionstyles for all edges. That would match how we handle colors, widths, etc. But maybe we can get...

Thanks for this -- it seems that the cytoscape module has been written so that the result hold valid Python values. The tests then use `json.dumps` on the result to...

If I understand correctly, you are requesting that NetworkX move the calling of `json.dumps` inside the functions named `cytoscape_*` which currently prepare a python object for being input to `json.dumps`....

It looks like none of the function in the `json_graph` subpackage return strings. They return python objects (dicts) that can be serialized (using `json.dumps`) and transferred to other programs like...

@nojhan and @jdreo can you take a look at the doc-string change in #7928 and see if we are leaving out something helpful there? Thanks!

I think you are considering "the other NetworkX readwrite features" to be the same as the `json` features, while we are not considering them the same. The `read`/`write` functions for...