mzy2240/csrgraph

A tiny library for large graphs

CSRGraphs

This library aims to be a graph analogue to Pandas. It aims to provide fast and memory efficient operations for read-only graphs.

By exploiting CSR Sparse matrices, we can efficiently pack and traverse graphs in memory (as long as we don't add/delete nodes often).

Accessing the underlying matrix

import csrgraph as cg

G = cg.csrgraph(data)
G.mat # This is the underlying scipy.sparse.csr_matrix

For instance, all the procedures in scipy csgraph module here will function directly on the G.mat object.

The rest of the csrgraph class is mainly around managing node names, ingesting, and optimized (numba compiled) methods.

Installation

pip install csrgraph

TODO:

• Add support for directed graphs