How would one save a graph created with the draw method? I presently use the command in ipython with the show='ipython' argument (works great), but it isn't clear to me how to save the image as a .png or .pdf.

Hi chebee7i,

thank you for writing this package.
Could you please make it available on PyPI as well?

Best regards,
Arne

We want to use nxpd with Python 3.4. However, this commit (c962ee5) is not included in ver.0.1.2.

Would you please dump up version in PyPI.

I am trying to run the below code (in Jupyter-notebook) using networkx==2.4 and nxpd==0.2.0

import networkx as nx
import nxpd
from nxpd import draw
from nxpd import nxpdParams
nxpdParams["show"] = "ipynb"

G = nx.DiGraph()
G.add_node("Hello World")
draw(G)

However I get the error

'DiGraph' object has no attribute 'number_of_selfloops'

I am aware this function has been depreciated in networkx==2.4 but are there any plans for a fix to nxpd or any quick fixes I could use in the meantime?

Hi, useful project, thanks!
Is it possible to use latex for node or edge labels?

G = nx.MultiDiGraph(with_labels=False,rankdir='LR')
G.graph['dpi'] = 120
G.add_node(1,label=r"$\sigma_1$",shape = 'circle')
G.add_node(2,label=r"$\sigma_2$",shape = 'circle')
G.add_edge(1,1,headport='nw',tailport='sw',label='sp')
G.add_edge(2,2,headport='se',tailport='ne')
G.add_edge(1,2,label='1-sp')

im = draw(G, layout='circo')
im

which gives this output:

thanks for any help!

The to_pydot method was using a deprecated function: https://github.com/chebee7i/nxpd/blob/master/nxpd/nx_pydot.py#L249 which has been removed in the latest release 2.4.0: https://networkx.github.io/documentation/stable/release/release_2.4.html#deprecations

Wonder if you are interested to migrate the ipyhon intergration to something similar to

https://github.com/rainwoodman/nxsvg/blob/master/src/nxsvg.py#L532

This would get rid of the conditional import in draw_pydot.

Sorry, one more as I scan through the code. I am not sure if this one make much sense.

Apparently all of the graphviz executables nxpd uses can read from stdin. We can likely do away with the temp files with .communicate . I wonder why original pydot chose to write to a temp file though.

However, in a longer term maybe we want to avoid spawn/fork/exec all together; they won't reliably work, for example, in a MPI program. -- I don't quite see why one would be visualizing graphs in a MPI program though.

hi,

trying to install from a kaggle kernel, doesn't seem to find your package,

So, I've tried the standard way to install on kaggle,
https://www.kaggle.com/docs/kernels#modifying-a-kernel-specific-environment

and I get the attached image:

thanks for any help!

One more comment.

A user provided filter_attrs function will allow the user to generate the pydot attributes as we iterate over the graph, without requiring these attributes be saved into the data dict of node nor edges.

This essentially splits the graph data model (G) and the view (filter_attrs). It also reduces chances of having conflicts on the data dict.