Clear and automatic DAG visualization

Question

I am trying to draw a DAG (directed acyclic graph) starting from a dataframe with 212 attributes. I am able to draw it using graphviz

import graphviz

G = graphviz.Digraph()

for col in graph_df.columns:
    for row in graph_df.index:
        if graph_df.loc[row, col] == 1:
            G.edge(col, row)

G.render('graph', format='pdf', cleanup=True)

But the final pdf looks like this

If I zoom I can read everything, but I would like to stretch vertically the final DAG to have it tidier. Can you help me please?

Answer 1

You can use ratio and size, if you need vertically long graph display:

G = graphviz.Digraph(graph_attr={'rankdir': 'TB', 'ratio': 'fill', 'size': '50,200'})

Code:

import pandas as pd
import graphviz


def get_plot(graph_df):
    G = graphviz.Digraph(graph_attr={'rankdir': 'TB', 'ratio': 'fill', 'size': '10,30'})
    for i, col in enumerate(graph_df.columns):
        for j, row in enumerate(graph_df.index):
            if i != j and graph_df.loc[row, col] == 1:  # Ensure not connecting a node to itself
                G.edge(col, row)

    G.render('graph', format='pdf', cleanup=True)


V = 26
nodes = [chr(ord('A') + i) for i in range(V)]
edges = {name: [1 if i != j else 0 for j in range(V)] for i, name in enumerate(nodes)}
df = pd.DataFrame(edges, index=nodes)
get_plot(df)

Actual Size

• You can change the size to get what's expected.

---

• node_attr() and graph_attr() has everything that you need to customize your view:

import pandas as pd
import graphviz


def get_plot(graph_df):
    G = graphviz.Digraph(
        graph_attr={
            'rankdir': 'TB',
            'ratio': 'fill',
            'size': '100,200',
            'nodesep': '1.0',
            'ranksep': '1.5'
        },
        node_attr={
            'fixedsize': 'true',
            'width': '10.0',
            'height': '10.0',
            'fontsize': '300'
        }
    )
    for i, col in enumerate(graph_df.columns):
        for j, row in enumerate(graph_df.index):
            if i != j and graph_df.loc[row, col] == 1:
                G.edge(col, row)

    G.render('graph', format='pdf', cleanup=True)


V = 26
nodes = [chr(ord('A') + i) for i in range(V)]
edges = {name: [1 if i != j else 0 for j in range(V)] for i, name in enumerate(nodes)}
df = pd.DataFrame(edges, index=nodes)
get_plot(df)

10% view (of the actual size)