5

I have a weighted circular layout plot. I wanted to make edges start from the outside of the node, but cannot find a way to do so. I tried setting alpha=1, but that didn't give me the desired outcome. The image below displays what I'm getting right now

enter image description here

This is the following code I have for nodes right now:

for n in G.nodes():
    if n in set1:
        G.nodes[n]['color'] = '#7a8eff'
    elif n in set2:
        G.nodes[n]['color'] = '#eb2c30'
    elif n in set3:
        G.nodes[n]['color'] = '#7300ff'
    else:
        G.nodes[n]['color'] = '#730a15'

colors = [node[1]['color'] for node in G.nodes(data=True)]
nx.draw_networkx_nodes(G, pos, node_size=1000, node_color=colors)


# edges
for edge in G.edges():
    source, target = edge
    rad = 0.25
    node_color_dict = dict(G.nodes(data='color'))
    if node_color_dict[source] == node_color_dict[target]:
        arrowprops=dict(lw=G.edges[(source,target)]['weight'],
                        arrowstyle="-",
                        color='blue',
                        connectionstyle=f"arc3,rad={rad}",
                        linestyle= '-',
                        alpha=0.45)
        ax.annotate("",
                    xy=pos[source],
                    xytext=pos[target],
                    arrowprops=arrowprops
                   )
    else:
        arrowprops=dict(lw=G.edges[(source,target)]['weight'],
                        arrowstyle="-",
                        color='purple',
                        connectionstyle=f"arc3,rad={rad}",
                        linestyle= '-',
                        alpha=0.45)
        ax.annotate("",
                    xy=pos[source],
                    xytext=pos[target],
                    arrowprops=arrowprops
                   )
# labels
nx.draw_networkx_labels(G, pos, font_size=11, font_family="monospace", font_color='white', font_weight='bold', alpha=1.0)

nandz123
  • 159
  • 12
  • We need the full code to help you, if I only use `nx.draw`, I could not reproduce the issue. I guess you first draw nodes and then the edges. Changing this order will probably resolve the issue. If you add a minimal reproducible example, it's way easier to help you. – Sparky05 Nov 12 '20 at 16:12
  • @Sparky05 i added more code, hope it's helpful – nandz123 Nov 12 '20 at 16:50
  • In that case adding `zorder` to the arrowprops should solve your issue. I've extended the code example below – Sparky05 Nov 13 '20 at 09:41
  • @Sparky05 it works, but the only issue i'm facing is that now the node labels don't show up. i'm simply using: `nx.draw_networkx_labels(G, pos, font_size=11, font_family="monospace", font_color='white', font_weight='bold')` to draw labels – nandz123 Nov 15 '20 at 08:51
  • I've now build a complete example with edges, nodes, and node labels. – Sparky05 Nov 16 '20 at 08:16

1 Answers1

5

A full example, which produces the desired result:

import networkx as nx
import matplotlib.pylab as pl


G = nx.karate_club_graph()
pos = nx.kamada_kawai_layout(G)

set1 = set(node for node in G if G.nodes[node]["club"] == "Mr. Hi")
set2 = set(node for node in G if G.nodes[node]["club"] != "Mr. Hi")

for n in G.nodes():
    if n in set1:
        G.nodes[n]['color'] = '#7a8eff'
    elif n in set2:
        G.nodes[n]['color'] = '#eb2c30'

for u, v in G.edges():
    G.edges[(u,v)]["weight"] = 1

colors = [node[1]['color'] for node in G.nodes(data=True)]
nodes_draw = nx.draw_networkx_nodes(G, pos, node_size=1000, node_color=colors)

ax = pl.gca()

# draw in the order, edges, nodes, node labels
zorder_edges = 3
zorder_nodes = 4
zorder_node_labels = 5

# edges
for edge in G.edges():
    source, target = edge
    rad = 0.25
    node_color_dict = dict(G.nodes(data='color'))
    if node_color_dict[source] == node_color_dict[target]:
        arrowprops=dict(lw=G.edges[(source,target)]['weight'],
                        arrowstyle="-",
                        color='blue',
                        connectionstyle=f"arc3,rad={rad}",
                        linestyle= '-',
                        #alpha=0.45,
                        zorder=zorder_edges,
                        )
        ax.annotate("",
                    xy=pos[source],
                    xytext=pos[target],
                    arrowprops=arrowprops
                   )
    else:
        arrowprops=dict(lw=G.edges[(source,target)]['weight'],
                        arrowstyle="-",
                        color='purple',
                        connectionstyle=f"arc3,rad={rad}",
                        linestyle= '-',
                        #alpha=0.45,
                        zorder=zorder_edges,
                        )
        ax.annotate("",
                    xy=pos[source],
                    xytext=pos[target],
                    arrowprops=arrowprops
                   )
# labels
node_labels_dict = nx.draw_networkx_labels(G, pos, font_size=11, font_family="monospace", font_color='white', font_weight='bold',
                                           #alpha=1.0
                                           )

nodes_draw.set_zorder(zorder_nodes)
for node_labels_draw in node_labels_dict.values():
    node_labels_draw.set_zorder(zorder_node_labels)

pl.axis("off")
# do don't cut off nodes
ax.set_xlim([1.1*x for x in ax.get_xlim()])
ax.set_ylim([1.1*y for y in ax.get_ylim()])
pl.show()

The result:

Visualization of the resulting network

Background

You can change the zorder of the created matplotlib objects:

import networkx as nx
import matplotlib.pylab as pl
# an example graph with string (names) as nodes
g = nx.karate_club_graph()
pos = nx.kamada_kawai_layout(g)

e = nx.draw_networkx_edges(g, pos=pos, )
n = nx.draw_networkx_nodes(g, pos=pos, )

e.set_zorder(5) 
n.set_zorder(10)
pl.show()

In case you use some advanced edge drawing, add the zorder parameter to the arrowprops parameter (all possible parameters) of annotate, e.g.

arrowprops=dict(lw=G.edges[(source,target)]['weight'],
                        arrowstyle="-",
                        color='blue',
                        connectionstyle=f"arc3,rad={rad}",
                        linestyle= '-',
                        alpha=0.45,
                        zorder=0)

I've included this answer to avoid cutting nodes at the border.

Sparky05
  • 4,692
  • 1
  • 10
  • 27