Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Pygraphviz / networkx set node level or layer

I have a dataset that represents a kind of genealogy tree. Each node has 2 parents (except first generation, they have no parents). For a given node, its parents can be from any previous generation. For example a node in generation n, can have a parent in n-1, and another parent in n-5. A node can be parent of several other nodes.

So basically, for every node I know its generation and its parents.

I am trying to represent this graph keeping the nodes from the same generation in the same line. Each generation has 10 nodes, except the first one.

So far I am trying the "dot" layout. When I only input two generations it does a good job, but when I input 3 generations, it is "too smart", and rearranges the nodes in some way.

For example, the following data, which represents 3 generations:

[(ObjectId('530b2ad783a1a15e695c9711'), ObjectId('530b804883a1a15e695c972b')), (ObjectId('530b2ad783a1a15e695c971c'), ObjectId('530b804883a1a15e695c972b')), (ObjectId('530b2ad783a1a15e695c9709'), ObjectId('530b804883a1a15e695c9727')), (ObjectId('530b2ad783a1a15e695c970a'), ObjectId('530b804883a1a15e695c9727')), (ObjectId('530b2ad783a1a15e695c970b'), ObjectId('530b804883a1a15e695c9724')), (ObjectId('530b2ad783a1a15e695c970f'), ObjectId('530b804883a1a15e695c9724')), (ObjectId('530b2ad783a1a15e695c9714'), ObjectId('530b804883a1a15e695c9729')), (ObjectId('530b2ad783a1a15e695c970e'), ObjectId('530b804883a1a15e695c9729')), (ObjectId('530b2ad783a1a15e695c9711'), ObjectId('530b804883a1a15e695c9723')), (ObjectId('530b2ad783a1a15e695c970e'), ObjectId('530b804883a1a15e695c9723')), (ObjectId('530b2ad783a1a15e695c971c'), ObjectId('530b804883a1a15e695c9728')), (ObjectId('530b2ad783a1a15e695c9719'), ObjectId('530b804883a1a15e695c9728')), (ObjectId('530b2ad783a1a15e695c9714'), ObjectId('530b804883a1a15e695c9726')), (ObjectId('530b2ad783a1a15e695c9713'), ObjectId('530b804883a1a15e695c9726')), (ObjectId('530b2ad783a1a15e695c9716'), ObjectId('530b804883a1a15e695c9722')), (ObjectId('530b2ad783a1a15e695c9719'), ObjectId('530b804883a1a15e695c9722')), (ObjectId('530b2ad783a1a15e695c970d'), ObjectId('530b804883a1a15e695c9725')), (ObjectId('530b2ad783a1a15e695c9715'), ObjectId('530b804883a1a15e695c9725')), (ObjectId('530b804883a1a15e695c9724'), ObjectId('530ba27c83a1a15e695c972d')), (ObjectId('530b2ad783a1a15e695c9713'), ObjectId('530ba27c83a1a15e695c972d')), (ObjectId('530b804883a1a15e695c9724'), ObjectId('530ba27c83a1a15e695c972e')), (ObjectId('530b2ad783a1a15e695c9709'), ObjectId('530ba27c83a1a15e695c972e')), (ObjectId('530b804883a1a15e695c9727'), ObjectId('530ba27c83a1a15e695c972f')), (ObjectId('530b2ad783a1a15e695c9709'), ObjectId('530ba27c83a1a15e695c972f')), (ObjectId('530b2ad783a1a15e695c9713'), ObjectId('530ba27c83a1a15e695c9730')), (ObjectId('530b2ad783a1a15e695c9709'), ObjectId('530ba27c83a1a15e695c9730')), (ObjectId('530b2ad783a1a15e695c9713'), ObjectId('530ba27c83a1a15e695c9731')), (ObjectId('530b804883a1a15e695c9725'), ObjectId('530ba27c83a1a15e695c9731')), (ObjectId('530b804883a1a15e695c9722'), ObjectId('530ba27c83a1a15e695c9732')), (ObjectId('530b2ad783a1a15e695c970b'), ObjectId('530ba27c83a1a15e695c9732')), (ObjectId('530b2ad783a1a15e695c9711'), ObjectId('530ba27c83a1a15e695c9733')), (ObjectId('530b2ad783a1a15e695c971f'), ObjectId('530ba27c83a1a15e695c9733')), (ObjectId('530b804883a1a15e695c972b'), ObjectId('530ba27c83a1a15e695c9734')), (ObjectId('530b2ad783a1a15e695c9713'), ObjectId('530ba27c83a1a15e695c9734')), (ObjectId('530b2ad783a1a15e695c9709'), ObjectId('530ba27c83a1a15e695c9735')), (ObjectId('530b804883a1a15e695c9724'), ObjectId('530ba27c83a1a15e695c9735')), (ObjectId('530b2ad783a1a15e695c9713'), ObjectId('530ba27c83a1a15e695c9736')), (ObjectId('530b804883a1a15e695c9723'), ObjectId('530ba27c83a1a15e695c9736')), (ObjectId('530b2ad783a1a15e695c9716'), ObjectId('530b804883a1a15e695c9722')), (ObjectId('530b2ad783a1a15e695c9719'), ObjectId('530b804883a1a15e695c9722')), (ObjectId('530b2ad783a1a15e695c9711'), ObjectId('530b804883a1a15e695c9723')), (ObjectId('530b2ad783a1a15e695c970e'), ObjectId('530b804883a1a15e695c9723')), (ObjectId('530b2ad783a1a15e695c970b'), ObjectId('530b804883a1a15e695c9724')), (ObjectId('530b2ad783a1a15e695c970f'), ObjectId('530b804883a1a15e695c9724')), (ObjectId('530b2ad783a1a15e695c970d'), ObjectId('530b804883a1a15e695c9725')), (ObjectId('530b2ad783a1a15e695c9715'), ObjectId('530b804883a1a15e695c9725')), (ObjectId('530b2ad783a1a15e695c9714'), ObjectId('530b804883a1a15e695c9726')), (ObjectId('530b2ad783a1a15e695c9713'), ObjectId('530b804883a1a15e695c9726')), (ObjectId('530b2ad783a1a15e695c9709'), ObjectId('530b804883a1a15e695c9727')), (ObjectId('530b2ad783a1a15e695c970a'), ObjectId('530b804883a1a15e695c9727')), (ObjectId('530b2ad783a1a15e695c971c'), ObjectId('530b804883a1a15e695c9728')), (ObjectId('530b2ad783a1a15e695c9719'), ObjectId('530b804883a1a15e695c9728')), (ObjectId('530b2ad783a1a15e695c9714'), ObjectId('530b804883a1a15e695c9729')), (ObjectId('530b2ad783a1a15e695c970e'), ObjectId('530b804883a1a15e695c9729')), (ObjectId('530b2ad783a1a15e695c9715'), ObjectId('530b804883a1a15e695c972a')), (ObjectId('530b2ad783a1a15e695c970b'), ObjectId('530b804883a1a15e695c972a')), (ObjectId('530b2ad783a1a15e695c9711'), ObjectId('530b804883a1a15e695c972b')), (ObjectId('530b2ad783a1a15e695c971c'), ObjectId('530b804883a1a15e695c972b'))]

produces:enter image description here

All the nodes that are "root" and therefore don't receive an edge should be placed on the first line, for example, but it is placing some of them on the second level.

Once I try to make bigger graphs, with 10 generations, all the hierarchy makes no sense at all.

is there any parameter or way to specify the level or layer of a node? In the data I provided that info is not there, but I can easily produce it, the problem is that I don't know how to send that info to pygraphviz or networkx.

I also want to use this data to produce graphs with twopi like this one http://networkx.github.io/documentation/latest/examples/drawing/lanl_routes.html

where the layers are represented on a circular way.

like image 742
Dr Sokoban Avatar asked Mar 05 '14 19:03

Dr Sokoban


People also ask

How does pygraphviz handle data storage?

All of the data storage is handled by internal graphviz data structures. The python code in pygraphviz is just a wrapper for the C language API of libcgraph. That is, you could write a C language code that does exactly the same thing as the pygraphviz code about (and produces the same output).

What is a graph created with NetworkX progstring?

A graph created with NetworkX progstring Name of Graphviz layout program rootstring, optional Root node for twopi layout argsstring, optional Extra arguments to Graphviz layout program Returns node_posdict

What are the options for layout in GraphViz?

The name of the GraphViz program to use for layout. Options depend on GraphViz version but may include: ‘dot’, ‘twopi’, ‘fdp’, ‘sfdp’, ‘circo’ The node of G from which to start some layout algorithms. Dictionary of (x, y) positions keyed by node.

Is it possible to write pygraphviz code in C?

That is, you could write a C language code that does exactly the same thing as the pygraphviz code about (and produces the same output). The dot output is produced by a call to libcgraph's agwrite function.


2 Answers

Use a Graphviz subgraph with attribute rank=same. e.g.

import networkx as nx
import pygraphviz as pgv # pygraphviz should be available

G = nx.DiGraph()
G.add_edge('a','aa')
G.add_edge('a','ab')
G.add_edge('a','bbc')
G.add_edge('b','ab')
G.add_edge('b','bb')
G.add_edge('c','bbc')
G.add_edge('bb','bba')
G.add_edge('bb','bbc')
A = nx.to_agraph(G)
one = A.add_subgraph(['a','b','c'],rank='same')
two = A.add_subgraph(['aa','ab','bb'],rank='same')
three = A.add_subgraph(['bba','bbc'],rank='same')
A.draw('example.png', prog='dot')

enter image description here

like image 140
Aric Avatar answered Oct 27 '22 11:10

Aric


As of 2017, the function to_agraph is no longer exposed at the nx.level. Now you must call nx.nx_agraph.to_agraph()

like image 35
Dominik Fischer Avatar answered Oct 27 '22 11:10

Dominik Fischer