Description

szhorvat

opened on Jun 25, 2023

Issue body actions

When plotting a graph/matrix, reciprocal directed edges overlap. The result is that only the weight of one of them is visible, which is confusing. I believe NetworkX does support curved edges, which could be a solution. However, generally, visualization does not seem to be a priority for NetworkX, so perhaps it's not the best idea to rely on it in the long term.

M = gb.Matrix.from_edgelist([(0,1,10),(1,0,20)], nrows=2, ncols=2)
gb.viz.draw(M)

Both weights, 10 and 20, should be visible, and it should be clear which belongs to which edge.

I think this is important since draw() visualizes matrices specifically as directed graphs, with no undirected option being present.

Part of: pyOpenSci/software-submission#81

Reacted by Sultan Orazbayev

Activity

1. eriknw commented on Jun 26, 2023

   

   eriknw

   on Jun 26, 2023

   Member

   More actions

   Yeah, this isn't good. This version of draw was part of the initial PoC that @jim22k created years ago. We do plan to give vizualization a revamp and to show examples by including notebooks in the docs (#407), b/c viz is great for user applications and make great examples.

   It might be nifty to be able to use netgraph or graph-tool (or ...) for viz.

   • https://netgraph.readthedocs.io/en/latest/sphinx_gallery_output/index.html
   • https://graph-tool.skewed.de/

   I think @paul-tqh-nguyen may be interested in looking into visualization
2. szhorvat commented on Jun 26, 2023

   

   szhorvat

   on Jun 26, 2023

   Author

   More actions

   In case you're considering multiple viz backends, igraph does have good supports for both self-loops and multi-edges, and how has a matplotlib back-end.
3. eriknw commented on Jun 29, 2023

   

   eriknw

   on Jun 29, 2023

   Member

   More actions

   Thanks for letting us know about viz in igraph!

   After a quick scan of examples, it's not clear to me how viz from igraph would be able to handle showing the edge attributes of reciprocal edges any better: https://python.igraph.org/en/latest/tutorials/index.html
4. szhorvat commented on Jun 29, 2023

   

   szhorvat

   on Jun 29, 2023

   Author

   More actions

   I'm sorry about that, it seems that contrary to my memory, multi-edge handling is not yet implemented with the matplotlib backend. It does work with the Cairo backend.
5. szhorvat commented on Jun 30, 2023

   

   szhorvat

   on Jun 30, 2023

   Author

   More actions

   @eriknw I need to correct myself again. It seems that multigraph plotting in igraph is broken in the development version, but it's working fine in the released version. Python is not my main language, so I was a bit confused.

   To be clear: Yes, multigraph plotting works fine in igraph with either the Cairo or matplotlib backends.

   Example:

   import igraph as ig
   import matplotlib.pyplot as plt
   
   g = ig.Graph([(0,1), (1,0), (1,2), (2,0)], directed=True)
   
   import matplotlib.pyplot as plt
   fig, ax= plt.subplots()
   ig.plot(g, edge_label=['a','b','c','d'], target=ax)
   

   

   Reacted by Sultan Orazbayev and Erik Welch
6. eriknw commented on Jul 11, 2023

   

   eriknw

   on Jul 11, 2023

   Member

   More actions

   Nice, thanks!
7. eriknw

   added 2 commits that reference this issue on Aug 6, 2026

   Fix viz.datashade cell positioning and separate reciprocal edges in v…

   80a2f44

   Fix viz.datashade cell positioning and separate reciprocal edges in v…

   1f1e16d

Issue actions