Graph Library Proposal for the C++ Standard

This library is in the alpha stage that may include significant changes to the interface. It is not recommended for general use.

A breaking change in the near future will introduce a new concept of identifier for vertices and edges, similar to the descriptor concept in the Boost Graph Library. It will replace the existing id and references in the API and will allow for more flexibility in the design as well as reducing the number of concepts and functions. Most functions and views will be affected. Phil - September 2024.

Overview

This library designed to provide a useful set of algorithms, views and container(s) for graphs. It also defines a core Graph Container Interface that provide the basis of interacting with an abstract adjacency list graph, and to provide easy integration with external adjacency list graphs.

• Hyper-graphs are outside the scope of this project.
• Comments and questions are welcome and can be directed to GitHub discussions or issues.

Purpose

This prototype library is an implementation of the proposed Graph Library for ISO Standard C++. It has gone through major revisions since it was first introduced in 2019. While we are comfortable of the core design, there is still plenty of activity being done and refinements made in its design and implementation. Experimenting with this library is encouraged, keeping in mind that breaking changes are expected.

Goals

The goals of the library include:

1. Support creation of high-performance, state-of-the-art algorithms.
2. Syntax that is simple, expressive and easy to understand when writing algorithms.
3. Define useful concepts and traits that can be used by algorithms to describe their requirements.
4. Support views for graph traversal commonly used by algorithms.
5. Support optional, user-defined value_types for an edge, vertex and graph.
6. Allow the use of standard containers to define simple graphs.
7. Easy integration of existing graph data structures.
8. Have an open design to allow for extensions in the future:
   1. Support the incoming edges on a vertex (e.g. bidirectional graphs).
   2. Support sparse vertex_ids stored in bi-directional ranges (e.g. map and unordered_map).

Getting Started

This is being actively developed with the latest releases of MSVC (VS2022) on Windows and gcc (13) on Linux/MacOS. Other releases or compilers may or may not work.

Prerequesites

• C++20 compliant compiler that fully supports concepts and ranges. (C++23 is required for building the benchmarks.)
• CMake 26 or later (for CMake Presets)

Quick Start Guide (Linux, WSL, MacOS)

git clone https://github.com/stdgraph/graph-v2.git
cd graph-v2
mkdir build && cd build
cmake ..
make -j8

Editor/IDE Configuration (Windows)

You'll need to assure CMake Presets are enabled in your IDE or other development tool. See https://docs.microsoft.com/en-us/cpp/build/cmake-presets-vs?view=msvc-170 for configuring Microsoft tools.

Description

Graph Algorithms (P3128)

The following algorithms are planned for the Graph Library.

Algorithm	Status
Breadth First Search	revision scheduled
Depth First Search	revision scheduled
Topological Sort	implementation scheduled
Dijkstra Shortest Paths	Completed
Bellman-Ford Shortest Paths	Completed
Triangle Counting	needs implementation
Label propagation (Comm. Detection)	needs implementation
Articulation Points	needs implementation
Bi-Connected Components	needs implementation
Connected Components	needs implementation
Strongly Connected Components	needs implementation
Maximal Independent Set	needs implementation
Jaccard Coefficient	needs implementation
Kruskell Minimum Spanning Tree	needs implementation
Prim Minimum Spanning Tre	needs implementation

Graph Views (P3129 )

View	Description
vertexlist	Iterates over vertices
incidence	Iterates over outgoing edges of a vertex
neighbors	Iterates over outgoing neighbor vertices of a vertex
edgelist	Iterates over edges of a graph
depth_first_search	Iterates over vertices or edges of a seed vertex in depth-first order
breadth_first_search	Iterates over vertices or edges of a seed vertex in breadth-first order
topological_sort	Iterates over vertices or edges of a seed vertex in topological sort order

Graph Containers (P3131)

Container	Description
compressed_graph	Compresed Sparse Row graph. High performance, static structure.
dynamic_graph	Easy to use different containers for vertices and edges.
(std containers)	Use standard containers like vector and list to define a simple graph.

Papers

The following papers make up the current proposal for the Graph Library.

Paper	Description
P3126 Overview	Describes the big picture of what we are proposing.
P3127 Background	Background and Terminology, provides the motivation and theoretical background underlying the proposal.
P3128 Algoritms	Covers the initial algorithms as well as the ones we'd like to see in the future.
P3129 Views	Helpful views for traversing a graph.
P3130 Graph Container Interface	The core interface for uniformly accessing graph data structure and adapting to external graphs.
P3131 Graph Containers	Includes the compressed_graph and how to use standard containers to define simple graphs.
P3337 Graph Comparison	[future] Syntax and performance comparison to the Boost Graph Library.

Acknowledgments

• The NWGraph team for all their collaborations and support, along with providing the algorithm implementations NWGraph Library
• Numerous comments and support from the Machine Learning study group (SG19) in the ISO C++ Standards Committee (WG21).
• Bob Steagal for his gcc-builder & clang-builder scripts
• Jason Turner for his cpp_starter_project
• Vincent La for his cvs-parser
• The ISO C++ Standards Committee (WG21) for C++