New Contributor FAQ#
A collection of frequently-asked questions by newcomers to open-source development and first-time contributors to NetworkX.
Q: I’m new to open source and would like to contribute to NetworkX. How do I get started?#
To contribute to NetworkX, you will need three things:
The source code
A development environment
An idea of what you’d like to contribute
Steps 1 & 2 are covered extensively in Development Workflow. There is no generic answer for step 3. There are many ways that NetworkX can be improved, from adding new algorithms, improving existing algorithms, improving the test suite (e.g. increasing test coverage), and improving the documentation. The “best” way to find a place to start is to follow your own personal interests! That said, a few places to check for ideas on where to get started:
The Algorithms discussion includes a listing of algorithms that users would like to have but that are not yet included in NetworkX.
Q: I’ve found an issue I’m interested in, can I have it assigned to me?#
NetworkX doesn’t typically assign issues to contributors. If you find an issue or feature request on the issue tracker that you’d like to work on, you should first check the issue thread to see if there are any linked pull requests. If not, then feel free to open a new PR to address the issue - no need to ask for permission - and don’t forget to reference the issue number in the PR comments so that others know you are now working on it!
Q: How do I contribute an example to the Gallery?#
The example gallery is great place to contribute, particularly if you have an
interesting application or visualization that uses NetworkX.
The gallery is generated using sphinx-gallery
from Python scripts stored in the
For instance, let’s say I’d like to contribute an example of visualizing a
complete graph using a
Assuming you have already followed the procedure for
setting up a development environment, start by
creating a new branch:
git checkout -b complete-graph-circular-layout-example
It’s generally a good idea to give your branch a descriptive name so that it’s easy to remember what you are working on.
Now you can begin work on your example. Sticking with the circular layout idea,
you might create a file in
with the following contents:
import networkx as nx
import matplotlib.pyplot as plt
G = nx.complete_graph(10) # A complete graph with 10 nodes
It may not be clear where exactly an example belongs. Our circular
layout example is very simple, so perhaps it belongs in
It would also make sense for it to be in
examples/drawing since it deals
with visualization. Don’t worry if you’re not sure: questions like this will
be resolved during the review process.
At this point, your contribution is ready to be reviewed. You can make the
changes on your
complete-graph-circular-layout-example branch visible to
other NetworkX developers by
creating a pull request.
The developer guide has more details on creating pull requests.
Q: I want to work on a specific function. How do I find it in the source code?#
Assuming you have followed the instructions for setting up the development workflow, there are several ways of determining where the in the source code a particular function or class is defined.
For example, let’s say you are interested in making a change to the
kamada_kawai_layout function, so you need to know
where it is defined. In an IPython terminal, you can use
? — the source file is
listed in the
In : import networkx as nx
In : nx.kamada_kawai_layout?
Signature: <clipped for brevity>
Docstring: <clipped for brevity>
Command line utilities like
git grep are also very useful.
For example, from the NetworkX source directory:
$ grep -r "def kamada_kawai_layout" .
Q: What is the policy for deciding whether to include a new algorithm?#
There is no official policy setting explicit inclusion criteria for new algorithms in NetworkX. New algorithms are more likely to be included if they have been published and are cited by others. More important than number of citations is how well proposed additions fit the project Mission and Values.
Testing is also an important factor in determining whether algorithms should be included. Proposals that include thorough tests which illustrate expected behavior are much easier to review, and therefore likely to progress more rapidly.
Thorough does not mean exhaustive. The quality of unit tests is much more important than quantity. Thorough tests should address questions like:
Does the algorithm support different graph types (undirected, directed, multigraphs)?
How does the algorithm behave with disconnected inputs and graphs which contain self-loops?
Are there explicit test cases outlined in the literature which can be incorporated in the test suite?