How to read text files with Graphviz? - graphviz

Question

For example, I have put the weights of the edges in the txt file below:
0 1 2
1 0 3
2 3 0
The desired result is:
{
node0--node1[weight=1];
node0--node2[weight=2];
node1--node2[weight=3];
}

Answer 1

Graphviz understands text files with DOT language, not DSV-like formats as in your example.
So you need to convert your file to DOT notation and that's a topic for a another question, but as a proof of concept and illustration of the fact (as mentioned in the comments to your question) that Graphviz may not be the best solution for visualization 50 000 000 edges in one image (you can ask app for clustering on the Software Recommendations website), I will show an example of generating a text file with random weights for ~20 000 edges between 200 nodes, then reading that file and converting it to DOT language (by graphviz 0.19.1 package) and producing output .dot/.gv, .canon or .png file formats. The program was executed within 5 minutes.
Python 3.9.1 script:
import graphviz as gv
import random
# Count of nodes:
size = 200
# Range of edges weights (upper limit):
weights_range = 10
# Generate txt file with random weights:
arr = [['0' for x in range(size)] for y in range(size)]
col = 1
for row in range (0, size):
i = col
while i < size:
arr[row][i] = str(random.randrange(weights_range))
arr[i][row] = arr[row][i]
i+=1
col+=1
with open('data.txt', 'w+') as file:
for row in range(0, len(arr)):
file.write(' '.join(arr[row]) +'\n')
# Create empty graph object:
Gr = gv.Graph(format='canon', # For PNG output change to 'png'.
filename='graph.gv',
engine='neato', # Layout engine.
graph_attr=dict(splines='true',
sep='5',
overlap='scale'),
node_attr=dict(shape='circle',
margin='0',
fontsize='10',
width='0.2',
height='0.2',
fixedsize='true'),
edge_attr=dict(arrowsize='0.4'))
# Read data from file and add nodes/edges to graph object:
file = open('data.txt', 'r')
row = 0
col_shift = 1 # To shift the beginning of each line to the right,
# so as not to go over the diagonal line of zeros.
for line in file:
line = line.strip()
if len(line) == 0:
continue
weights = line.split(' ')
col=col_shift
while col < len(weights):
# Variant 1. Long names for nodes:
# Gr.edge('node'+str(row),'node'+str(col),weight=weights[col])
# Variant 2. Short names for nodes:
Gr.edge(str(row),str(col),weight=weights[col])
col+=1
row+=1
col_shift+=1
file.close()
Gr.render() # Rendering graph to text file or image.
Result .canon (19 916 lines, ~500 KB) (I replaced the node-naming node42 output with 42 to reduce the file size):
graph {
graph [overlap=scale,
sep=5,
splines=true
];
node [fixedsize=true,
fontsize=10,
height=0.2,
label="\N",
margin=0,
shape=circle,
width=0.2
];
edge [arrowsize=0.4];
0 -- 1 [weight=8];
0 -- 2 [weight=1];
0 -- 3 [weight=5];
0 -- 4 [weight=2];
0 -- 5 [weight=7];
...
196 -- 198 [weight=0];
196 -- 199 [weight=3];
197 -- 198 [weight=0];
197 -- 199 [weight=5];
198 -- 199 [weight=9];
}
Result PNG (~20 000 edges between 200 nodes, ~5 MB):
P.S. However, if you want to create such an image, the following questions may help you properly adjust graph settings for faster rendering and clear picture (these are optional and not exactly related to current question, so I don't want SO to auto-link them to current question, so I give their numbers instead of links):
GraphViz Dot very long duration of generation
/questions/10766100
laying out a large graph with graphviz
/questions/13417411
Graphviz is too slow
/questions/64130276
Is it possible to generate a small GraphViz chart?
/questions/68321537
Compacting a digraph in Graphviz using Dot Language
/questions/8610710
Reducing the size (as in area) of the graph generated by graphviz
/questions/3428448