Installation

Installation Using pip or conda

DeepGraph can be installed via pip from PyPI:

$ pip install deepgraph

To install with optional dependencies (i.e. “extras”, extending the base functionality of DeepGraph), use:

$ pip install deepgraph[plot,basemap,tables,scipy,networkx]

or simply use:

$ pip install deepgraph[all]

to install all optional dependencies. Extras and their added functionality are described below.

Note

There is one optional dependency that can not (yet) be install via pip: graph-tool. Please refer to the linked project homepage for recommended installation.

Alternatively, if you’re using Conda, install with:

$ conda install -c conda-forge deepgraph

It is not possible to specify optional dependencies (“extras”) directly when installing a Python package using conda. To enable the extra features of deepgraph, you have to manually add the corresponding packages to your conda environment.

For example, if you want to use features that require “scipy”, you would need to install scipy separately using:

$ conda install scipy

To install deepgraph including all optional dependencies, you could also use an environment.yml file, e.g.

name: deepgraph-env
channels:
  - conda-forge
dependencies:
  - python>=3.9
  - numpy>=1.21.6
  - pandas>=1.2
  - matplotlib>=3.1
  - basemap>=2.0
  - pytables>=3.7
  - scipy>=1.5.4
  - networkx>=2.4
  - graph-tool>=2.27
  - deepgraph

Then create the environment with:

$ conda env create -f environment.yml

and activate it via:

$ conda activate deepgraph-env

Installation from Source & Environment Setup

DeepGraph uses uv to manage the package. To get started as a developer, follow the instructions below

1. Install uv

   If you haven’t already, install uv by following the official instructions:

   $ curl -Ls https://astral.sh/uv/install.sh | sh
   

2. Clone the repository

   Use git to clone the repository and cd into it:

   $ git clone https://github.com/deepgraph/deepgraph.git
   $ cd deepgraph
   

3. Set up the development environment

   Use uv to create and manage your virtual environment:

   $ uv sync
   

   This will create a virtual environment, install the required and the developer dependencies, build the DeepGraph package and install it in editable mode.

4. Installing with optional dependencies (“extras”)

   To install extras (i.e. optional dependencies, extending the core functionality of DeepGraph), use the following command:

   $ uv sync --extra <EXTRA>
   

   where <EXTRA> can be one of

   • tables

   • scipy

   • networkx

   • plot

   • basemap

   To install all optional dependencies, simply use:

   $ uv sync --all-extras
   

5. Run tests

   To run tests, use the following command:

   $ uv run pytest .
   

6. Build the documentation

   To build the documentation, run the following commands:

   $ cd doc
   $ make html
   

Dependencies

Package	Minimum supported version
Python	3.9
NumPy	1.21.6
Pandas	1.2

It is highly recommended to install Pandas’ optional dependencies. In particular, it is recommended to install the performance dependencies.

Optional Dependencies / Extras

The following are recommended packages that DeepGraph can use to provide additional functionality.

Dependency	Minimum Version	pip extra	Notes
Matplotlib	3.1	plot	Used by the plotting methods of DeepGraph.
basemap	2.0	basemap	Used by plot_map and plot_map_generator to plot networks on map projections.
PyTables	3.7	tables	Necessary for HDF5-based storage of pandas DataFrames. A DeepGraph may be instantiated with an HDFStore containing a node table in order to iteratively create edges directly from disc (see create_edges and create_edges_ft).
SciPy	1.5.4	scipy	Used by return_cs_graph to convert from DeepGraph’s network representation to sparse adjacency matrices, and by append_cp to append a connected components column to the node table.
NetworkX	2.4	networkx	Used by return_nx_graph and return_nx_multigraph to convert from DeepGraph’s network representation to NetworkX’s network representations.
graph_tool	2.27	N/A	Used by return_gt_graph to convert from DeepGraph’s network representation to Graph-Tool’s network representation. See here for installation instructions.