python-igraph

python-igraph is the Python interface to igraph, a high-performance C library for creating, manipulating, and analyzing graphs. It is built for scale — the heavy lifting happens in optimized C, so you can work with networks of millions of vertices and edges from ergonomic Python code.

The library covers the full breadth of network science: graph construction and I/O, centrality and connectivity measures, community detection, shortest paths, subgraph isomorphism, and layout algorithms for visualization. Widely used in academic research and data science, python-igraph pairs a mature algorithmic core with plotting support via Cairo and Matplotlib.

What You Get

• A rich Graph class for directed and undirected, weighted and attributed graphs
• Hundreds of algorithms: centrality, shortest paths, connectivity, flows, and isomorphism
• Community detection methods including Louvain, Leiden, walktrap, and label propagation
• Graph layout and visualization via Cairo and Matplotlib backends
• Import/export for many formats (GraphML, GML, edge lists, Pajek, and more)

Common Use Cases

• Analyzing social, biological, or infrastructure networks at scale
• Detecting communities and clusters in large graphs
• Computing centrality and connectivity metrics for research
• Visualizing network structure with publication-quality layouts

Under The Hood

Architecture — The package is a thin, Pythonic layer over the C igraph core. A compiled extension module, src/_igraph, exposes the low-level graph engine, while src/igraph/ builds the high-level API on top of it: datatypes.py and __init__.py define the Graph class, with functionality split across focused modules — community.py, clustering.py, structural.py, cut.py, matching.py, automorphisms.py, plus io/ for format readers/writers and drawing/ for layout and plotting. This keeps performance-critical work in C while presenting an idiomatic object-oriented interface.

Tech Stack — Python bindings compiled against the C igraph library via a CPython extension, with build configuration in pyproject.toml/setup.py and vendored sources under vendor/. Visualization is powered by Cairo (via pycairo) and optionally Matplotlib and plotly; testing and multi-version builds are driven by tox.

Code Quality — The repository carries an extensive tests/ suite run across many Python versions through tox and CI, ships type stubs, and hosts thorough documentation on ReadTheDocs. As a long-lived scientific package it maintains careful backward compatibility and a detailed CHANGELOG.

API Design — The developer experience is centered on a single expressive Graph object: constructors accept edge lists, adjacency matrices, and named formats, and analysis methods read naturally (g.degree(), g.community_multilevel(), g.shortest_paths()). Plotting integrates directly through plot(g), so going from graph to figure is a short path, though the sheer breadth of algorithms means there is a lot of surface area to learn.