Node colors does not work for large graphs on Jupyter/IJulia

[Jacob-Stevens-Haas]

I'm looking at a version of the connectome("neural wiring") of the C. elegans worm, a graph of 297 vertices. I'm trying to display the graph with each vertex colored according to it's eccentricity, with more central vertices in a hotter color.

However, gplot() colors each node black, even with the nodefillc parameter assigned. It seems that this is an issue with plotting large graphs. Test code, adapted from the documentation:

using LightGraphs
using GraphPlot

n=195
g=erdos_renyi(n,.05)
membership=rand([1,2], nv(g));
nodecolor = [colorant"lightseagreen", colorant"orange"]
nodefillc = nodecolor[membership]
gplot(g, nodefillc=nodefillc)

When n is larger than 195, the nodes are all filled in black, as opposed to green and orange.

[sbromberger]

I'm unable to reproduce (your code with n=300):

[Jacob-Stevens-Haas]

Thanks for the quick reply. How interesting. Are you using something other than the Colors package? I added it to the previous post. Here's the images I'm getting, FYI: n=150 n=300

Interestingly, when I use the other example from the documentation: nodefillc = distinguishable_colors((g), colorant"blue") it seems to work: In both cases, typeof(nodefillc) returns Array{ColorTypes.RGB{FixedPointNumbers.Normed{UInt8,8}},1}. They're both length 300.

Edit: I'm using Jupyter notebooks. Judging by the background, you're using Juno? Let me try that.

[sbromberger]

Are you using something other than the Colors package?

Not that I'm aware.

I was doing this within vscode, so maybe that makes a difference?

[Jacob-Stevens-Haas]

Struggling to get plots working in Juno. Can GraphPlot work as a backend to Plots.jl? I see your examples on other issues reference Compose.jl, rather than Plots.jl

EDIT: I was able to get draw(SVG("testplot2.svg", 8cm, 8cm), gplot(g, nodefillc=nodefillc)) to work.
Opening the file, I see that the vertices are colored correctly. (still could use help displaying in the plots window in Juno or as a gui in the REPL)

That being said, this issue clearly applies to the Jupyter/IJulia+GraphPlot combination. I'll update the title accordingly. Not sure exactly how gplot() works, so I don't know if it's a problem with large Compose objects, specifically with gplot, or with IJulia.

I'm relatively new to GraphPlot, so I'll try other node and edge parameters to see if they also break with large graphs

[sbromberger]

Since this isn't specific to gplot, I think we can close this out. It's more likely some bug in jupyter/ijulia or some strange edge case. Let me know if you'd like to continue troubleshooting and I'm happy to reopen.

[dmos62]

Can still reproduce. Also unsure where the cause is. I've a new clue though. Currently I'm rendering to file as a workaround, but I've noticed that if I render the same plot object (the return value of gplot(...) call) more than once, after the first time the nodes will be colored black. Using JupyterLab.

My snippet:

p = gplot(sg, nodefillc=nodefillc);
draw(SVG("plot.svg", 50cm, 50cm), p);