a hypergraph is a generalization of a graph in which an edge can join any number of vertices

Based on this, I can see two modes of thought to designing this in:

Swap backend

Because hypergraphs are a generalization of graphs, it seems like the basis of graphs in CXXGraph should be based on hypergraphs - and Graph (max edge vertices = 2) and Hypergraph (max edge vertices = ∞) classes merely provide convenient API access based on the user's preferred flavor of usage.

This also seems like a significant overhaul in-terms of how the backend code would be structured, and could cause performance degradation to normal 2-vertices-per-edge graphs without careful design considerations. However the big win would be code scalability and maintenance wins.

Split architecture

Just create a completely separate class and offer completely new algorithms specialized for hypergraphs (and maybe reimplement some existing ones for compatibility with hypergraphs).

This may be a good route if we discover that there is significant overhead in using hypergraph-based graphs as the library backend, which would otherwise drag down the performance of users only interested in using classic 2-vertex-per-edge graphs.

Introduce Hypergraph #122

Description

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions