Graph and Search

C++ class templates for constructing graphs and search. This library is distributed under MIT license.

1. Design

Assuming a graph G = (V, E) contains n vertices and m edges:

Data Structure

A “Graph” class contains a collection of “Vertex” objects (stored in vertexmap of type VertexMapType) and each “Vertex” contains a list of edges (stored in edgesto of type EdgeListType) connecting to its adjacent vertices.

The overall space complexity of this graph implementation is O(m+n).

Time Complexity

• here O(1), O(n) only accounts for the operation to remove the vertex from the vertex map. Additionally it may take up to O(m^2) to remove all edges connected to the vertex both from upstream and downstream. Possible improvement can be made to reduce the O(m^2) complexity by using a different data structure for vertics_from and edges_to lists.
• for a sparse graph, the number of edges that a vertex contains should be much less than m

Graph Search

The dynamic priority queue is implemented as a binary heap, thus the time complexity of a graph search is O((m+n)*log(n)).

2. Dependencies

• A compiler that supports C++11

This is a header-only library. There are multiple ways you can integrate this library to your project:

1. Simply copy content of the “src” folder to your project and include “graph/graph.hpp”.
2. If you’re using CMake, you could integrate the library to your project by adding this repository as a git submodule. Then use “add_subdirectory()” in your CMakeLists.txt to add the library to your build tree.
3. You can build and install this library to your system path and use CMake “find_package(graph REQUIRED)” to find the library and add dependency by using “target_link_libraries(your_app PRIVATE xmotion::graph)”.

3. Build the demo & pack the library

A “.deb” installation package can be generated if you want to install the library to your system. Relevant CMake configuration files will also be installed so that you can easily use “find_package()” command to find and use the library in your project. Note that the library is exported as “rdu::graph” to avoid naming conflicts with other libraries.

$ git clone --recursive https://github.com/rxdu/libgraph.git
$ mkdir build && cd build
$ cmake --build .
$ cpack

Demo programs will be built and put into “build/bin” folder. A “.deb” package would be generated inside “build” folder. Install the library with the “.deb” package using command “dpkg”, for example

$ sudo dpkg -i graph_1.1_amd64.deb

4. Build document

You need to have doxygen to build the document.

$ sudo apt-get install doxygen
$ cd docs
$ doxygen doxygen/Doxyfile

Outlines of core data structures for the purpose of API reference are given at https://rdu.im/libgraph/ .

5. Construct a graph

You can associate any object to a vertex, but you need to provide an index function for the State struct/class. The index function is used to generate a unique index for a state and it’s necessary for checking whether two given states are the same so that only one vertex is created for one unique state inside the graph.

A default indexer is provided and could be used if you have a member function “GetId()” or a member variable “id_” or “id” that provide you with a unique value of the object.

You can also define your own index function if the default one is not suitable for your application.

// struct YourStateIndexFunction is a functor that defines operator "()". 
struct YourStateIndexFunction
{
    // you should have this one if "State" type 
    // of your Graph is a raw pointer type
    int64_t operator()(YourStateType* state)
    {
        // Generate <unique-value> with state
        return <unique-value>;
    }
    // you should have this one if "State" type 
    // of your Graph is a value type
    int64_t operator()(const YourStateType& state)
    {
        // Generate <unique-value> with state
        return <unique-value>;
    }
};

See “simple_graph_demo.cpp” in “demo” folder for a working example.

6. Known limitations

• TODO List