src/lemon/suurballe.h
author alpar
Mon, 21 Feb 2005 14:59:12 +0000
changeset 1164 80bb73097736
parent 1020 f42cb3146ed4
child 1359 1581f961cfaa
permissions -rw-r--r--
A year has passed again.
alpar@906
     1
/* -*- C++ -*-
alpar@921
     2
 * src/lemon/suurballe.h - Part of LEMON, a generic C++ optimization library
alpar@906
     3
 *
alpar@1164
     4
 * Copyright (C) 2005 Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
alpar@906
     5
 * (Egervary Combinatorial Optimization Research Group, EGRES).
alpar@906
     6
 *
alpar@906
     7
 * Permission to use, modify and distribute this software is granted
alpar@906
     8
 * provided that this copyright notice appears in all copies. For
alpar@906
     9
 * precise terms see the accompanying LICENSE file.
alpar@906
    10
 *
alpar@906
    11
 * This software is provided "AS IS" with no warranty of any kind,
alpar@906
    12
 * express or implied, and with no claim as to its suitability for any
alpar@906
    13
 * purpose.
alpar@906
    14
 *
alpar@906
    15
 */
alpar@906
    16
alpar@921
    17
#ifndef LEMON_SUURBALLE_H
alpar@921
    18
#define LEMON_SUURBALLE_H
alpar@899
    19
alpar@899
    20
///\ingroup flowalgs
alpar@899
    21
///\file
alpar@899
    22
///\brief An algorithm for finding k paths of minimal total length.
alpar@899
    23
alpar@899
    24
alpar@921
    25
#include <lemon/maps.h>
alpar@899
    26
#include <vector>
alpar@921
    27
#include <lemon/min_cost_flow.h>
alpar@899
    28
alpar@921
    29
namespace lemon {
alpar@899
    30
alpar@899
    31
/// \addtogroup flowalgs
alpar@899
    32
/// @{
alpar@899
    33
alpar@899
    34
  ///\brief Implementation of an algorithm for finding k edge-disjoint paths between 2 nodes 
alpar@899
    35
  /// of minimal total length 
alpar@899
    36
  ///
alpar@921
    37
  /// The class \ref lemon::Suurballe implements
alpar@899
    38
  /// an algorithm for finding k edge-disjoint paths
alpar@899
    39
  /// from a given source node to a given target node in an
alpar@899
    40
  /// edge-weighted directed graph having minimal total weight (length).
alpar@899
    41
  ///
alpar@899
    42
  ///\warning Length values should be nonnegative.
alpar@899
    43
  /// 
alpar@899
    44
  ///\param Graph The directed graph type the algorithm runs on.
alpar@899
    45
  ///\param LengthMap The type of the length map (values should be nonnegative).
alpar@899
    46
  ///
alpar@968
    47
  ///\note It it questionable whether it is correct to call this method after
alpar@1020
    48
  ///%Suurballe for it is just a special case of Edmonds' and Karp's algorithm
alpar@968
    49
  ///for finding minimum cost flows. In fact, this implementation just
alpar@899
    50
  ///wraps the MinCostFlow algorithms. The paper of both %Suurballe and
alpar@899
    51
  ///Edmonds-Karp published in 1972, therefore it is possibly right to
alpar@899
    52
  ///state that they are
alpar@899
    53
  ///independent results. Most frequently this special case is referred as
alpar@899
    54
  ///%Suurballe method in the literature, especially in communication
alpar@899
    55
  ///network context.
alpar@899
    56
  ///\author Attila Bernath
alpar@899
    57
  template <typename Graph, typename LengthMap>
alpar@899
    58
  class Suurballe{
alpar@899
    59
alpar@899
    60
alpar@987
    61
    typedef typename LengthMap::Value Length;
alpar@899
    62
    
alpar@899
    63
    typedef typename Graph::Node Node;
alpar@899
    64
    typedef typename Graph::NodeIt NodeIt;
alpar@899
    65
    typedef typename Graph::Edge Edge;
alpar@899
    66
    typedef typename Graph::OutEdgeIt OutEdgeIt;
alpar@899
    67
    typedef typename Graph::template EdgeMap<int> EdgeIntMap;
alpar@899
    68
alpar@899
    69
    typedef ConstMap<Edge,int> ConstMap;
alpar@899
    70
alpar@899
    71
    const Graph& G;
alpar@899
    72
marci@941
    73
    Node s;
marci@941
    74
    Node t;
marci@941
    75
alpar@899
    76
    //Auxiliary variables
alpar@899
    77
    //This is the capacity map for the mincostflow problem
alpar@899
    78
    ConstMap const1map;
alpar@899
    79
    //This MinCostFlow instance will actually solve the problem
marci@941
    80
    MinCostFlow<Graph, LengthMap, ConstMap> min_cost_flow;
alpar@899
    81
alpar@899
    82
    //Container to store found paths
alpar@899
    83
    std::vector< std::vector<Edge> > paths;
alpar@899
    84
alpar@899
    85
  public :
alpar@899
    86
alpar@899
    87
marci@941
    88
    /*! \brief The constructor of the class.
alpar@899
    89
    
marci@941
    90
    \param _G The directed graph the algorithm runs on. 
marci@941
    91
    \param _length The length (weight or cost) of the edges. 
marci@941
    92
    \param _s Source node.
marci@941
    93
    \param _t Target node.
marci@941
    94
    */
marci@941
    95
    Suurballe(Graph& _G, LengthMap& _length, Node _s, Node _t) : 
marci@941
    96
      G(_G), s(_s), t(_t), const1map(1), 
marci@941
    97
      min_cost_flow(_G, _length, const1map, _s, _t) { }
alpar@899
    98
alpar@899
    99
    ///Runs the algorithm.
alpar@899
   100
alpar@899
   101
    ///Runs the algorithm.
alpar@899
   102
    ///Returns k if there are at least k edge-disjoint paths from s to t.
marci@941
   103
    ///Otherwise it returns the number of edge-disjoint paths found 
marci@941
   104
    ///from s to t.
alpar@899
   105
    ///
alpar@899
   106
    ///\param k How many paths are we looking for?
alpar@899
   107
    ///
marci@941
   108
    int run(int k) {
marci@941
   109
      int i = min_cost_flow.run(k);
alpar@899
   110
alpar@899
   111
      //Let's find the paths
alpar@899
   112
      //We put the paths into stl vectors (as an inner representation). 
alpar@899
   113
      //In the meantime we lose the information stored in 'reversed'.
alpar@899
   114
      //We suppose the lengths to be positive now.
alpar@899
   115
marci@941
   116
      //We don't want to change the flow of min_cost_flow, so we make a copy
alpar@899
   117
      //The name here suggests that the flow has only 0/1 values.
alpar@899
   118
      EdgeIntMap reversed(G); 
alpar@899
   119
alpar@899
   120
      for(typename Graph::EdgeIt e(G); e!=INVALID; ++e) 
marci@941
   121
	reversed[e] = min_cost_flow.getFlow()[e];
alpar@899
   122
      
alpar@899
   123
      paths.clear();
alpar@899
   124
      //total_length=0;
alpar@899
   125
      paths.resize(k);
alpar@899
   126
      for (int j=0; j<i; ++j){
alpar@899
   127
	Node n=s;
alpar@899
   128
alpar@899
   129
	while (n!=t){
alpar@899
   130
klao@946
   131
	  OutEdgeIt e(G, n);
alpar@899
   132
	  
alpar@899
   133
	  while (!reversed[e]){
alpar@899
   134
	    ++e;
alpar@899
   135
	  }
alpar@986
   136
	  n = G.target(e);
alpar@899
   137
	  paths[j].push_back(e);
alpar@899
   138
	  //total_length += length[e];
alpar@899
   139
	  reversed[e] = 1-reversed[e];
alpar@899
   140
	}
alpar@899
   141
	
alpar@899
   142
      }
alpar@899
   143
      return i;
alpar@899
   144
    }
alpar@899
   145
alpar@899
   146
    
marci@941
   147
    ///Returns the total length of the paths.
alpar@899
   148
    
alpar@899
   149
    ///This function gives back the total length of the found paths.
alpar@899
   150
    Length totalLength(){
marci@941
   151
      return min_cost_flow.totalLength();
alpar@899
   152
    }
alpar@899
   153
alpar@899
   154
    ///Returns the found flow.
alpar@899
   155
alpar@899
   156
    ///This function returns a const reference to the EdgeMap \c flow.
marci@941
   157
    const EdgeIntMap &getFlow() const { return min_cost_flow.flow;}
alpar@899
   158
alpar@899
   159
    /// Returns the optimal dual solution
alpar@899
   160
    
alpar@899
   161
    ///This function returns a const reference to the NodeMap
alpar@899
   162
    ///\c potential (the dual solution).
marci@941
   163
    const EdgeIntMap &getPotential() const { return min_cost_flow.potential;}
alpar@899
   164
alpar@899
   165
    ///Checks whether the complementary slackness holds.
alpar@899
   166
alpar@899
   167
    ///This function checks, whether the given solution is optimal.
alpar@899
   168
    ///Currently this function only checks optimality,
alpar@899
   169
    ///doesn't bother with feasibility
alpar@899
   170
    ///It is meant for testing purposes.
alpar@899
   171
    bool checkComplementarySlackness(){
marci@941
   172
      return min_cost_flow.checkComplementarySlackness();
alpar@899
   173
    }
alpar@899
   174
alpar@899
   175
    ///Read the found paths.
alpar@899
   176
    
alpar@899
   177
    ///This function gives back the \c j-th path in argument p.
alpar@899
   178
    ///Assumes that \c run() has been run and nothing changed since then.
alpar@899
   179
    /// \warning It is assumed that \c p is constructed to
alpar@899
   180
    ///be a path of graph \c G.
alpar@899
   181
    ///If \c j is not less than the result of previous \c run,
alpar@899
   182
    ///then the result here will be an empty path (\c j can be 0 as well).
alpar@899
   183
    ///
alpar@921
   184
    ///\param Path The type of the path structure to put the result to (must meet lemon path concept).
alpar@899
   185
    ///\param p The path to put the result to 
alpar@899
   186
    ///\param j Which path you want to get from the found paths (in a real application you would get the found paths iteratively)
alpar@899
   187
    template<typename Path>
alpar@899
   188
    void getPath(Path& p, size_t j){
alpar@899
   189
alpar@899
   190
      p.clear();
alpar@899
   191
      if (j>paths.size()-1){
alpar@899
   192
	return;
alpar@899
   193
      }
alpar@899
   194
      typename Path::Builder B(p);
alpar@899
   195
      for(typename std::vector<Edge>::iterator i=paths[j].begin(); 
alpar@899
   196
	  i!=paths[j].end(); ++i ){
alpar@899
   197
	B.pushBack(*i);
alpar@899
   198
      }
alpar@899
   199
alpar@899
   200
      B.commit();
alpar@899
   201
    }
alpar@899
   202
alpar@899
   203
  }; //class Suurballe
alpar@899
   204
alpar@899
   205
  ///@}
alpar@899
   206
alpar@921
   207
} //namespace lemon
alpar@899
   208
alpar@921
   209
#endif //LEMON_SUURBALLE_H