[英]Boost dominator tree for graph with custom vertex properties
I'm trying to use boost::lengauer_tarjan_dominator_tree with a graph with custom vertex properties , but can't get even a simple example to compile:我正在尝试将boost::lengauer_tarjan_dominator_tree与具有自定义顶点属性的图形一起使用,但甚至无法获得一个简单的编译示例:
#include <vector>
#include <iterator>
#include <boost/graph/adjacency_list.hpp>
#include <boost/graph/dominator_tree.hpp>
#include <boost/property_map/property_map.hpp>
struct GraphNode
{
explicit GraphNode(unsigned i) : index {i} {}
unsigned index;
};
using Graph = boost::adjacency_list<boost::listS, boost::listS,
boost::bidirectionalS, GraphNode, boost::no_property>;
using Vertex = boost::graph_traits<Graph>::vertex_descriptor;
int main()
{
Graph g {};
const auto u = boost::add_vertex(GraphNode {0}, g);
const auto v = boost::add_vertex(GraphNode {1}, g);
const auto x = boost::add_vertex(GraphNode {2}, g);
const auto y = boost::add_vertex(GraphNode {3}, g);
const auto z = boost::add_vertex(GraphNode {4}, g);
boost::add_edge(u, v, g);
boost::add_edge(u, x, g);
boost::add_edge(v, y, g);
boost::add_edge(x, y, g);
boost::add_edge(y, z, g);
const auto index_map = boost::get(&GraphNode::index, g);
std::vector<Vertex> dom_tree_pred_vector(boost::num_vertices(g),
boost::graph_traits<Graph>::null_vertex());
auto dom_tree_pred_map = boost::make_iterator_property_map(std::begin(dom_tree_pred_vector),
index_map);
boost::lengauer_tarjan_dominator_tree(g, u, dom_tree_pred_map);
}
Which I tried to adapt from the example given in the docs.我试图从文档中给出的示例中改编。
Here is part of the error message:这是错误消息的一部分:
/usr/local/include/boost/graph/detail/adjacency_list.hpp:2544:33: error: cannot form a reference to 'void'
typedef const value_type& const_reference;
^
/usr/local/include/boost/graph/dominator_tree.hpp:355:31: error: no matching function for call to 'get'
const IndexMap indexMap = get(vertex_index, g);
I've also tried to pass in the index map explicitly using the second form of the method, without success.我还尝试使用该方法的第二种形式显式传入索引映射,但没有成功。 I've noticed this methods interface seems a little different from other graph methods, such as depth_first_search , where the
vertex_index_map
is a named parameter.我注意到这个方法接口似乎与其他图形方法有点不同,例如depth_first_search ,其中
vertex_index_map
是一个命名参数。
Is it possible to use this method with custom vertex properties?是否可以将此方法与自定义顶点属性一起使用?
The problem - as ever - is with using something other than vecS
for the vertex container.问题 - 一如既往 - 使用
vecS
以外的东西作为顶点容器。 You lose the builtin vertex_index
property, making it mandatory to supply it to the API.您丢失了内置的
vertex_index
属性,因此必须将其提供给 API。
Sadly this algorithm doesn't support custom vertex index maps well.遗憾的是,该算法不能很好地支持自定义顶点索引图。 You tried - correctly - by using
index_map
, but internally the algorithm still looks for vertex_index_t
tagged property.您尝试 - 正确 - 通过使用
index_map
,但在内部算法仍然寻找vertex_index_t
标记的属性。
The only two ways in which I can see this work is,我能看到这项工作的唯一两种方式是,
to do the DFS stage manually (as even the all-argument overload of lengauer_tarjan*
fails to forward the correct index map into the DFS).手动执行 DFS 阶段(因为即使
lengauer_tarjan*
的全参数重载也无法将正确的索引映射转发到 DFS)。 Then you could call the lengauer_tarjan_dominator_tree_without_dfs
implementation and have the result.然后你可以调用
lengauer_tarjan_dominator_tree_without_dfs
实现并获得结果。
Alternatively you could tell the library about your graph's index map.或者,您可以将图形的索引图告诉图书馆。
(Lastly, you could accept fate and use vecS
as the vertex container selector. I suspect this is explicitly not what you wanted.) (最后,你可以接受命运并使用
vecS
作为顶点容器选择器。我怀疑这显然不是你想要的。)
Using the second approach, which may be the most elegant.使用第二种方法,这可能是最优雅的。 Here are the specializations/overloads to add:
以下是要添加的专业化/重载:
namespace boost {
template <>
struct property_map<Graph, vertex_index_t> {
typedef typename property_map<Graph, size_t GraphNode::*>::type type;
typedef typename property_map<Graph, size_t GraphNode::*>::const_type const_type;
};
static auto get(vertex_index_t, Graph& g) { return get(&GraphNode::index, g); }
static auto get(vertex_index_t, Graph const& g) { return get(&GraphNode::index, g); }
}
#include <vector>
#include <iterator>
#include <iostream>
#include <boost/graph/adjacency_list.hpp>
#include <boost/graph/dominator_tree.hpp>
#include <boost/property_map/property_map.hpp>
struct GraphNode
{
explicit GraphNode(size_t i) : index {i} {}
size_t index;
};
using Graph = boost::adjacency_list<boost::listS, boost::listS, boost::bidirectionalS, GraphNode, boost::no_property>;
namespace boost {
template <>
struct property_map<Graph, vertex_index_t> {
typedef typename property_map<Graph, size_t GraphNode::*>::type type;
typedef typename property_map<Graph, size_t GraphNode::*>::const_type const_type;
};
static property_map<Graph, vertex_index_t>::type get(vertex_index_t, Graph& g) { return get(&GraphNode::index, g); }
static property_map<Graph, vertex_index_t>::const_type get(vertex_index_t, Graph const& g) { return get(&GraphNode::index, g); }
}
using Vertex = boost::graph_traits<Graph>::vertex_descriptor;
int main()
{
Graph g {};
const auto u = boost::add_vertex(GraphNode {0}, g);
const auto v = boost::add_vertex(GraphNode {1}, g);
const auto x = boost::add_vertex(GraphNode {2}, g);
const auto y = boost::add_vertex(GraphNode {3}, g);
const auto z = boost::add_vertex(GraphNode {4}, g);
boost::add_edge(u, v, g);
boost::add_edge(u, x, g);
boost::add_edge(v, y, g);
boost::add_edge(x, y, g);
boost::add_edge(y, z, g);
std::vector<Vertex> dom_pred(num_vertices(g), boost::graph_traits<Graph>::null_vertex());
auto index_map = boost::get(&GraphNode::index, g); // equivalent to vertex_index_t now
auto dom_tree_pred_map (boost::make_iterator_property_map(std::begin(dom_pred), index_map));
// Run main algorithm
boost::lengauer_tarjan_dominator_tree(g, u, dom_tree_pred_map);
std::cout << "Result: ";
for (auto v : dom_pred) {
if (v == boost::graph_traits<Graph>::null_vertex())
std::cout << "(root) ";
else
std::cout << g[v].index << " ";
}
}
Prints印刷
Result: (root) 0 0 0 3
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.