BGL Dijkstra Shortest Paths with Bundled Properties - c++

Question

I'm trying to use the dijkstra shortest path algorithm in BGL to compute a simple ST path on an unweighted undirected graph. I may care about edge weights in the future, but for now I just want to consider edge traversals to be a uniform cost.
I am also tracking multiple edge and vertex properties so I've based what I've done so far on the bundled properties example that seemed to be the closest to what I'm attempting to do.
Now I'm trying to figure out how to get dijkstra working so I can do my ST search but I am getting stuck on getting the right parameters set up for it.
Here's a simplified example of the code I have so far:
#include <iostream>
#include <vector>
#include <boost/config.hpp>
#include <boost/graph/graph_traits.hpp>
#include <boost/graph/adjacency_list.hpp>
#include <boost/graph/dijkstra_shortest_paths.hpp>
#include <boost/property_map/property_map.hpp>
// Create a struct to hold properties for each vertex
typedef struct VertexProperties
{
int p1;
} VertexProperties;
// Create a struct to hold properties for each edge
typedef struct EdgeProperties
{
int p1;
} EdgeProperties;
// Define the type of the graph
typedef boost::adjacency_list<boost::vecS, boost::vecS, boost::undirectedS, VertexProperties, EdgeProperties> Graph;
int main(int,char*[])
{
// Create a graph object
Graph g;
// Add vertices
Graph::vertex_descriptor v0 = boost::add_vertex(g);
Graph::vertex_descriptor v1 = boost::add_vertex(g);
Graph::vertex_descriptor v2 = boost::add_vertex(g);
// Set vertex properties
g[v0].p1 = 1;
g[v1].p1 = 2;
g[v2].p1 = 3;
// Add edges
std::pair<Graph::edge_descriptor, bool> e01 = boost::add_edge(v0, v1, g);
std::pair<Graph::edge_descriptor, bool> e02 = boost::add_edge(v1, v2, g);
// Set edge properties
g[e01.first].p1 = 1;
g[e02.first].p1 = 2;
std::cout << "num_verts: " << boost::num_vertices(g) << std::endl;
std::cout << "num_edges: " << boost::num_edges(g) << std::endl;
// compute ST shortest paths here...
return 0;
}
I'm getting tripped up on the right parameters for the call to dijkstra's algorithm. They take the graph, a starting vertex, and then a predecessor map and distance map. The examples I've seen so far, like this one set up their graph with just an edge weight without the bundled edge properties, which simplifies things.
Ultimately, I'm after the ST shortest path so I'd need to recover the path from S to T. From the looks of things, we need to set up a predecessor map and then we can use that to extract the path from a particular T back to S?
I should also note that the environment I'm in does not allow C++11 language features. :(
Any help here would be greatly appreciated!

Answer 1

So the question was "how to use a bundled property as weight map with Boost Graph Library?".
Good. You use property maps. The bundled property can be accessed with a little bit of funky syntax documented right on the "Bundled Properties" page: http://www.boost.org/doc/libs/1_58_0/libs/graph/doc/bundles.html, see heading "Property maps from bundled properties".
Now for a quick demo:
// set up a weight map:
auto weights = boost::get(&EdgeProperties::p1, g);
Passing the minimum amount of arguments to dijkstra:
// you can pass it to dijkstra using direct or named params. Let's do the simplest
boost::dijkstra_shortest_paths(g, v0, boost::no_named_parameters() .weight_map(weights));
You will want to add more parameters, but hey, this is your start :)
Live On Coliru
#include <iostream>
#include <vector>
#include <boost/config.hpp>
#include <boost/graph/graph_traits.hpp>
#include <boost/graph/adjacency_list.hpp>
#include <boost/graph/dijkstra_shortest_paths.hpp>
#include <boost/property_map/property_map.hpp>
#include <boost/graph/graph_utility.hpp>
// Create a struct to hold properties for each vertex
struct VertexProperties { int p1; };
// Create a struct to hold properties for each edge
struct EdgeProperties { int p1; };
// Define the type of the graph
typedef boost::adjacency_list<boost::vecS, boost::vecS, boost::undirectedS, VertexProperties, EdgeProperties> Graph;
int main() {
// Create a graph object
Graph g;
// Add vertices
auto v0 = boost::add_vertex({1}, g),
v1 = boost::add_vertex({2}, g),
v2 = boost::add_vertex({3}, g);
// Add edges
boost::add_edge(v0, v1, EdgeProperties{1}, g);
boost::add_edge(v1, v2, EdgeProperties{2}, g);
boost::print_graph(g, boost::get(&VertexProperties::p1, g));
// set up a weight map:
auto weights = boost::get(&EdgeProperties::p1, g);
// you can pass itprint_graph`enter code here` to dijkstra using direct or named params. Let's do the simplest
boost::dijkstra_shortest_paths(g, v0, boost::no_named_parameters() .weight_map(weights));
}
You'll note that I simplified the initialization of the vertex/edge properties as well. The print_graph utility is neat if you want to have an idea of what the graph "looks" like (short of using Graphviz).
The output on Coliru is:
1 <--> 2
2 <--> 1 3
3 <--> 2

Answer 2

I'm adding a 'finished' version of the dijkstra shortest paths search that computes the shortest path from S to T for archival purposes.
I'm sure there are better "boost" ways to do this, but it works on my end.
http://www.boost.org/doc/libs/1_58_0/libs/graph/doc/bundles.html was a really helpful link.
///
/// #file bgl_st_example.cpp
///
/// #brief bundled property example
///
/// #ref http://programmingexamples.net/wiki/CPP/Boost/BGL/BundledProperties
///
#include <iostream>
#include <vector>
#include <boost/config.hpp>
#include <boost/graph/graph_traits.hpp>
#include <boost/graph/adjacency_list.hpp>
#include <boost/graph/dijkstra_shortest_paths.hpp>
#include <boost/property_map/property_map.hpp>
#include <boost/graph/graph_utility.hpp>
// Create a struct to hold properties for each vertex
typedef struct vertex_properties
{
std::string label;
int p1;
} vertex_properties_t;
// Create a struct to hold properties for each edge
typedef struct edge_properties
{
std::string label;
int p1;
int weight;
} edge_properties_t;
// Define the type of the graph
typedef boost::adjacency_list<boost::vecS, boost::vecS, boost::undirectedS, vertex_properties_t, edge_properties_t> graph_t;
typedef graph_t::vertex_descriptor vertex_descriptor_t;
typedef graph_t::edge_descriptor edge_descriptor_t;
typedef boost::property_map<graph_t, boost::vertex_index_t>::type index_map_t;
typedef boost::iterator_property_map<vertex_descriptor_t*, index_map_t*, vertex_descriptor_t, vertex_descriptor_t&> predecessor_map_t;
// The graph, with edge weights labeled.
//
// v1 --(1)-- v2
// | \_ |
// | \ |
// (1) (3) (2)
// | \_ |
// | \ |
// v4 --(1)-- v3
//
//
int main(int,char*[])
{
// Create a graph object
graph_t g;
// Add vertices
vertex_descriptor_t v1 = boost::add_vertex(g);
vertex_descriptor_t v2 = boost::add_vertex(g);
vertex_descriptor_t v3 = boost::add_vertex(g);
vertex_descriptor_t v4 = boost::add_vertex(g);
// Set vertex properties
g[v1].p1 = 1; g[v1].label = "v1";
g[v2].p1 = 2; g[v2].label = "v2";
g[v3].p1 = 3; g[v3].label = "v3";
g[v4].p1 = 4; g[v4].label = "v4";
// Add edges
std::pair<edge_descriptor_t, bool> e01 = boost::add_edge(v1, v2, g);
std::pair<edge_descriptor_t, bool> e02 = boost::add_edge(v2, v3, g);
std::pair<edge_descriptor_t, bool> e03 = boost::add_edge(v3, v4, g);
std::pair<edge_descriptor_t, bool> e04 = boost::add_edge(v4, v1, g);
std::pair<edge_descriptor_t, bool> e05 = boost::add_edge(v1, v3, g);
// Set edge properties
g[e01.first].p1 = 1; g[e01.first].weight = 1; g[e01.first].label = "v1-v2";
g[e02.first].p1 = 2; g[e02.first].weight = 2; g[e02.first].label = "v2-v3";
g[e03.first].p1 = 3; g[e03.first].weight = 1; g[e03.first].label = "v3-v4";
g[e04.first].p1 = 4; g[e04.first].weight = 1; g[e04.first].label = "v4-v1";
g[e05.first].p1 = 5; g[e05.first].weight = 3; g[e05.first].label = "v1-v3";
// Print out some useful information
std::cout << "Graph:" << std::endl;
boost::print_graph(g, boost::get(&vertex_properties_t::label,g));
std::cout << "num_verts: " << boost::num_vertices(g) << std::endl;
std::cout << "num_edges: " << boost::num_edges(g) << std::endl;
// BGL Dijkstra's Shortest Paths here...
std::vector<int> distances( boost::num_vertices(g));
std::vector<vertex_descriptor_t> predecessors(boost::num_vertices(g));
boost::dijkstra_shortest_paths(g, v1,
boost::weight_map(boost::get(&edge_properties_t::weight,g))
.distance_map(boost::make_iterator_property_map(distances.begin(), boost::get(boost::vertex_index,g)))
.predecessor_map(boost::make_iterator_property_map(predecessors.begin(), boost::get(boost::vertex_index,g)))
);
// Extract the shortest path from v1 to v3.
typedef std::vector<edge_descriptor_t> path_t;
path_t path;
vertex_descriptor_t v = v3;
for(vertex_descriptor_t u = predecessors[v]; u != v; v=u, u=predecessors[v])
{
std::pair<edge_descriptor_t,bool> edge_pair = boost::edge(u,v,g);
path.push_back( edge_pair.first );
}
std::cout << std::endl;
std::cout << "Shortest Path from v1 to v3:" << std::endl;
for(path_t::reverse_iterator riter = path.rbegin(); riter != path.rend(); ++riter)
{
vertex_descriptor_t u_tmp = boost::source(*riter, g);
vertex_descriptor_t v_tmp = boost::target(*riter, g);
edge_descriptor_t e_tmp = boost::edge(u_tmp, v_tmp, g).first;
std::cout << " " << g[u_tmp].label << " -> " << g[v_tmp].label << " (weight: " << g[e_tmp].weight << ")" << std::endl;
}
return 0;
}
Here's a CMakeLists.txt file that works for me:
cmake_minimum_required(VERSION 2.8)
project ( bgl_example )
find_package( Boost REQUIRED COMPONENTS )
include_directories( ${Boost_INCLUDE_DIR} )
add_executable( bgl_st_example bgl_st_example.cpp)
target_link_libraries( bgl_st_example ${Boost_LIBRARIES} )
The resulting output that I see:
Graph:
v1 <--> v2 v4 v3
v2 <--> v1 v3
v3 <--> v2 v4 v1
v4 <--> v3 v1
num_verts: 4
num_edges: 5
Shortest Path from v1 to v3:
v1 -> v4 (weight: 1)
v4 -> v3 (weight: 1)