I'm happy to announce the availability of LightGraphs, available at https://github.com/sbromberger/LightGraphs.jl <https://github.com/sbromberger/LightGraphs.jl>.
LightGraphs is an optimized simple graphs package designed for fast analysis using standard functions. It seeks to mimic the functionality of established packages like NetworkX, but with better performance. The tradeoff is that extraneous data (that is, attributes not needed for direct graph manipulation) are not supported in the graph structure itself and must be referenced in an external datastore by a vertex or edge. A good number of functions are available for LightGraphs already, and more will be added. Right now, centrality measures, some basic linear algebra functions, shortest-path algorithms, distance metrics, and graph operators are available, as are common small graph structures for experimentation. (Almost-complete) documentation is available via ReadTheDocs. In tests, LightGraphs outperforms NetworkX by as much as 99%. While it is not as flexible as the existing (and excellent) Graphs.jl package, the design tradeoffs have resulted in modest performance gains, especially as graph density increases. If you're looking for a way to represent simple graphs with good performance, LightGraphs might be appropriate. I'd appreciate any feedback and suggestions for improvement.
