有没有办法保证 NetworkX 的分层输出？

You can use pygraphviz to get close:

>>> import pygraphviz
>>> import networkx
>>> import networkx as nx
>>> G = nx.Graph()
>>> G.add_node("ROOT")
>>> for i in xrange(5):
...     G.add_node("Child_%i" % i)
...     G.add_node("Grandchild_%i" % i)
...     G.add_node("Greatgrandchild_%i" % i)
...     G.add_edge("ROOT", "Child_%i" % i)
...     G.add_edge("Child_%i" % i, "Grandchild_%i" % i)
...     G.add_edge("Grandchild_%i" % i, "Greatgrandchild_%i" % i)


>>> A = nx.to_agraph(G)
>>> A.layout('dot', args='-Nfontsize=10 -Nwidth=".2" -Nheight=".2" -Nmargin=0 -Gfontsize=8')
>>> A.draw('test.png')

Result:

Note I copied the graphviz options from the link you posted above. I'm not sure why the 4th child is drawn on top instead of in strictly vertical format. Maybe someone who knows more about the Graphviz options can help with that.

If you use a directed graph then the Graphviz dot layout will do something like you want with the tree. Here is some code similar to the above solutions that shows how to do that

import networkx as nx
from networkx.drawing.nx_agraph import graphviz_layout
import matplotlib.pyplot as plt
G = nx.DiGraph()


G.add_node("ROOT")


for i in range(5):
G.add_node("Child_%i" % i)
G.add_node("Grandchild_%i" % i)
G.add_node("Greatgrandchild_%i" % i)


G.add_edge("ROOT", "Child_%i" % i)
G.add_edge("Child_%i" % i, "Grandchild_%i" % i)
G.add_edge("Grandchild_%i" % i, "Greatgrandchild_%i" % i)


# write dot file to use with graphviz
# run "dot -Tpng test.dot >test.png"
nx.nx_agraph.write_dot(G,'test.dot')


# same layout using matplotlib with no labels
plt.title('draw_networkx')
pos=graphviz_layout(G, prog='dot')
nx.draw(G, pos, with_labels=False, arrows=False)
plt.savefig('nx_test.png')

UPDATED

Here is a version updated for networkx-2.0 (and with upcoming networkx-2.1 draws arrows too).

import networkx as nx
from networkx.drawing.nx_agraph import write_dot, graphviz_layout
import matplotlib.pyplot as plt
G = nx.DiGraph()


G.add_node("ROOT")


for i in range(5):
G.add_node("Child_%i" % i)
G.add_node("Grandchild_%i" % i)
G.add_node("Greatgrandchild_%i" % i)


G.add_edge("ROOT", "Child_%i" % i)
G.add_edge("Child_%i" % i, "Grandchild_%i" % i)
G.add_edge("Grandchild_%i" % i, "Greatgrandchild_%i" % i)


# write dot file to use with graphviz
# run "dot -Tpng test.dot >test.png"
write_dot(G,'test.dot')


# same layout using matplotlib with no labels
plt.title('draw_networkx')
pos =graphviz_layout(G, prog='dot')
nx.draw(G, pos, with_labels=False, arrows=True)
plt.savefig('nx_test.png')

You can use grandalf for a python-only solution, if you don't want to install graphviz.

Also, this type of visualization is called a layered graph drawing or Sugiyama-style graph drawing, which can display many kinds of graphs, including non-trees.

See my answer to a different question for details and implementation.

See also the hierarchical structures created by @Abdallah Sobehy, w/o the need to use any extra module/library (like 'pygraphviz'), at preserving the left and right child while printing python graphs using networkx

Maybe the graph is not so nice as the above, but it does the requested job!

You could do it manually just with only networkx. Traverse tree and for each children coordinate x change as if right parent.x+1/np.log(y_coord + 1) if left parent.x-1... and y = parent.y - 1 Or any other algorithm to get coord and eg. See image here.

import networkx as nx
import matplotlib.pyplot as plt
coord = {0: [8, 0], 2: [9, -1], 1: [8, -2], 4: [10, -2], 3: [9, -3], 5: [11, -3]}
eg = [[0, 2], [2, 1], [2, 4], [4, 3], [4, 5]]
G=nx.Graph()
G.add_edges_from(eg)
nx.draw(G, with_labels=True, node_size=1500, node_color="skyblue", pos=coord)
plt.show()