RhodeCode

The usage with class type algorithms is little bit harder. In this

case the function type map adaptors can not be used, because the

function map adaptors give back temporary objects.

\code

  Digraph graph;

  typedef Digraph::ArcMap<double> DoubleArcMap;

  DoubleArcMap length(graph);

  DoubleArcMap speed(graph);

  typedef DivMap<DoubleArcMap, DoubleArcMap> TimeMap;

  TimeMap time(length, speed);

  Dijkstra<Digraph, TimeMap> dijkstra(graph, time);

  dijkstra.run(source, target);

\endcode

We have a length map and a maximum speed map on the arcs of a digraph.

The minimum time to pass the arc can be calculated as the division of

the two maps which can be done implicitly with the \c DivMap template

class. We use the implicit minimum time map as the length map of the

\c Dijkstra algorithm.

*/

/**

@defgroup paths Path Structures

@ingroup datas

\brief %Path structures implemented in LEMON.

This group contains the path structures implemented in LEMON.

LEMON provides flexible data structures to work with paths.

All of them have similar interfaces and they can be copied easily with

assignment operators and copy constructors. This makes it easy and

efficient to have e.g. the Dijkstra algorithm to store its result in

any kind of path structure.

\sa \ref concepts::Path "Path concept"

*/

/**

@defgroup heaps Heap Structures

@ingroup datas

\brief %Heap structures implemented in LEMON.

This group contains the heap structures implemented in LEMON.

LEMON provides several heap classes. They are efficient implementations

of the abstract data type \e priority \e queue. They store items with

specified values called \e priorities in such a way that finding and

removing the item with minimum priority are efficient.

The basic operations are adding and erasing items, changing the priority

of an item, etc.

Heaps are crucial in several algorithms, such as Dijkstra and Prim.

The heap implementations have the same interface, thus any of them can be

used easily in such algorithms.

\sa \ref concepts::Heap "Heap concept"

*/

/**

@defgroup matrices Matrices

@ingroup datas

\brief Two dimensional data storages implemented in LEMON.

This group contains two dimensional data storages implemented in LEMON.

*/

/**

@defgroup auxdat Auxiliary Data Structures

@ingroup datas

\brief Auxiliary data structures implemented in LEMON.

This group contains some data structures implemented in LEMON in

order to make it easier to implement combinatorial algorithms.

*/

/**

@defgroup geomdat Geometric Data Structures

@ingroup auxdat

\brief Geometric data structures implemented in LEMON.

This group contains geometric data structures implemented in LEMON.

 - \ref lemon::dim2::Point "dim2::Point" implements a two dimensional

   vector with the usual operations.

 - \ref lemon::dim2::Box "dim2::Box" can be used to determine the

   rectangular bounding box of a set of \ref lemon::dim2::Point

   "dim2::Point"'s.

*/

/**

@defgroup matrices Matrices

@ingroup auxdat

\brief Two dimensional data storages implemented in LEMON.

This group contains two dimensional data storages implemented in LEMON.

*/

/**

@defgroup algs Algorithms

\brief This group contains the several algorithms

implemented in LEMON.

This group contains the several algorithms

implemented in LEMON.

*/

/**

@defgroup search Graph Search

@ingroup algs

\brief Common graph search algorithms.

This group contains the common graph search algorithms, namely

\e breadth-first \e search (BFS) and \e depth-first \e search (DFS)

\ref clrs01algorithms.

*/

/**

@defgroup shortest_path Shortest Path Algorithms

@ingroup algs

\brief Algorithms for finding shortest paths.

This group contains the algorithms for finding shortest paths in digraphs

\ref clrs01algorithms.

 - \ref Dijkstra algorithm for finding shortest paths from a source node

   when all arc lengths are non-negative.

 - \ref BellmanFord "Bellman-Ford" algorithm for finding shortest paths

   from a source node when arc lenghts can be either positive or negative,

   but the digraph should not contain directed cycles with negative total

   length.

 - \ref FloydWarshall "Floyd-Warshall" and \ref Johnson "Johnson" algorithms

   for solving the \e all-pairs \e shortest \e paths \e problem when arc

   lenghts can be either positive or negative, but the digraph should

   not contain directed cycles with negative total length.

 - \ref Suurballe A successive shortest path algorithm for finding

   arc-disjoint paths between two nodes having minimum total length.

*/

/**

@defgroup spantree Minimum Spanning Tree Algorithms

@ingroup algs

\brief Algorithms for finding minimum cost spanning trees and arborescences.

This group contains the algorithms for finding minimum cost spanning

trees and arborescences \ref clrs01algorithms.

*/

/**

@defgroup max_flow Maximum Flow Algorithms

@ingroup algs

\brief Algorithms for finding maximum flows.

This group contains the algorithms for finding maximum flows and

feasible circulations \ref clrs01algorithms, \ref amo93networkflows.

The \e maximum \e flow \e problem is to find a flow of maximum value between

a single source and a single target. Formally, there is a \f$G=(V,A)\f$

digraph, a \f$cap: A\rightarrow\mathbf{R}^+_0\f$ capacity function and

\f$s, t \in V\f$ source and target nodes.

A maximum flow is an \f$f: A\rightarrow\mathbf{R}^+_0\f$ solution of the

following optimization problem.

\f[ \max\sum_{sv\in A} f(sv) - \sum_{vs\in A} f(vs) \f]

\f[ \sum_{uv\in A} f(uv) = \sum_{vu\in A} f(vu)

    \quad \forall u\in V\setminus\{s,t\} \f]

\f[ 0 \leq f(uv) \leq cap(uv) \quad \forall uv\in A \f]

LEMON contains several algorithms for solving maximum flow problems:

- \ref EdmondsKarp Edmonds-Karp algorithm

  \ref edmondskarp72theoretical.

- \ref Preflow Goldberg-Tarjan's preflow push-relabel algorithm

  \ref goldberg88newapproach.

- \ref DinitzSleatorTarjan Dinitz's blocking flow algorithm with dynamic trees

  \ref dinic70algorithm, \ref sleator83dynamic.

- \ref GoldbergTarjan !Preflow push-relabel algorithm with dynamic trees

  \ref goldberg88newapproach, \ref sleator83dynamic.

In most cases the \ref Preflow algorithm provides the

fastest method for computing a maximum flow. All implementations

also provide functions to query the minimum cut, which is the dual

problem of maximum flow.

\ref Circulation is a preflow push-relabel algorithm implemented directly 

for finding feasible circulations, which is a somewhat different problem,

but it is strongly related to maximum flow.

For more information, see \ref Circulation.

*/

/**

@defgroup min_cost_flow_algs Minimum Cost Flow Algorithms

@ingroup algs

\brief Algorithms for finding minimum cost flows and circulations.

This group contains the algorithms for finding minimum cost flows and

circulations \ref amo93networkflows. For more information about this

problem and its dual solution, see \ref min_cost_flow

"Minimum Cost Flow Problem".

LEMON contains several algorithms for this problem.

 - \ref NetworkSimplex Primal Network Simplex algorithm with various

   pivot strategies \ref dantzig63linearprog, \ref kellyoneill91netsimplex.

 - \ref CostScaling Push-Relabel and Augment-Relabel algorithms based on

   cost scaling \ref goldberg90approximation, \ref goldberg97efficient,

   \ref bunnagel98efficient.

 - \ref CapacityScaling Successive Shortest %Path algorithm with optional

   capacity scaling \ref edmondskarp72theoretical.

 - \ref CancelAndTighten The Cancel and Tighten algorithm

   \ref goldberg89cyclecanceling.

 - \ref CycleCanceling Cycle-Canceling algorithms

   \ref klein67primal, \ref goldberg89cyclecanceling.

In general NetworkSimplex is the most efficient implementation,

but in special cases other algorithms could be faster.

For example, if the total supply and/or capacities are rather small,

CapacityScaling is usually the fastest algorithm (without effective scaling).

*/

/**

@defgroup min_cut Minimum Cut Algorithms

@ingroup algs

\brief Algorithms for finding minimum cut in graphs.

This group contains the algorithms for finding minimum cut in graphs.

The \e minimum \e cut \e problem is to find a non-empty and non-complete

\f$X\f$ subset of the nodes with minimum overall capacity on

outgoing arcs. Formally, there is a \f$G=(V,A)\f$ digraph, a

\f$cap: A\rightarrow\mathbf{R}^+_0\f$ capacity function. The minimum

cut is the \f$X\f$ solution of the next optimization problem:

\f[ \min_{X \subset V, X\not\in \{\emptyset, V\}}

    \sum_{uv\in A: u\in X, v\not\in X}cap(uv) \f]

LEMON contains several algorithms related to minimum cut problems:

- \ref HaoOrlin "Hao-Orlin algorithm" for calculating minimum cut

  in directed graphs.

- \ref NagamochiIbaraki "Nagamochi-Ibaraki algorithm" for

  calculating minimum cut in undirected graphs.

- \ref GomoryHu "Gomory-Hu tree computation" for calculating

  all-pairs minimum cut in undirected graphs.

If you want to find minimum cut just between two distinict nodes,

see the \ref max_flow "maximum flow problem".

*/

/**

@defgroup min_mean_cycle Minimum Mean Cycle Algorithms

@ingroup algs

\brief Algorithms for finding minimum mean cycles.

The \e minimum \e mean \e cycle \e problem is to find a directed cycle

of minimum mean length (cost) in a digraph.

The mean length of a cycle is the average length of its arcs, i.e. the

ratio between the total length of the cycle and the number of arcs on it.

This problem has an important connection to \e conservative \e length

\e functions, too. A length function on the arcs of a digraph is called

conservative if and only if there is no directed cycle of negative total

length. For an arbitrary length function, the negative of the minimum

cycle mean is the smallest \f$\epsilon\f$ value so that increasing the

arc lengths uniformly by \f$\epsilon\f$ results in a conservative length

function.

LEMON contains three algorithms for solving the minimum mean cycle problem:

- \ref HartmannOrlin "Hartmann-Orlin"'s algorithm, which is an improved

In practice, the Howard algorithm proved to be by far the most efficient

one, though the best known theoretical bound on its running time is

exponential.

Both Karp and HartmannOrlin algorithms run in time O(ne) and use space

O(n<sup>2</sup>+e), but the latter one is typically faster due to the

applied early termination scheme.

*/

/**

@defgroup matching Matching Algorithms

@ingroup algs

\brief Algorithms for finding matchings in graphs and bipartite graphs.

This group contains the algorithms for calculating

matchings in graphs and bipartite graphs. The general matching problem is

finding a subset of the edges for which each node has at most one incident

edge.

There are several different algorithms for calculate matchings in

graphs.  The matching problems in bipartite graphs are generally

easier than in general graphs. The goal of the matching optimization

can be finding maximum cardinality, maximum weight or minimum cost

matching. The search can be constrained to find perfect or

maximum cardinality matching.

The matching algorithms implemented in LEMON:

- \ref MaxBipartiteMatching Hopcroft-Karp augmenting path algorithm

  for calculating maximum cardinality matching in bipartite graphs.

- \ref PrBipartiteMatching Push-relabel algorithm

  for calculating maximum cardinality matching in bipartite graphs.

- \ref MaxWeightedBipartiteMatching

  Successive shortest path algorithm for calculating maximum weighted

  matching and maximum weighted bipartite matching in bipartite graphs.

- \ref MinCostMaxBipartiteMatching

  Successive shortest path algorithm for calculating minimum cost maximum

  matching in bipartite graphs.

- \ref MaxMatching Edmond's blossom shrinking algorithm for calculating

  maximum cardinality matching in general graphs.

- \ref MaxWeightedMatching Edmond's blossom shrinking algorithm for calculating

  maximum weighted matching in general graphs.

- \ref MaxWeightedPerfectMatching

  Edmond's blossom shrinking algorithm for calculating maximum weighted

  perfect matching in general graphs.

\image html bipartite_matching.png

\image latex bipartite_matching.eps "Bipartite Matching" width=\textwidth

*/

/**

@defgroup graph_properties Connectivity and Other Graph Properties

@ingroup algs

\brief Algorithms for discovering the graph properties

This group contains the algorithms for discovering the graph properties

like connectivity, bipartiteness, euler property, simplicity etc.

\image html connected_components.png

\image latex connected_components.eps "Connected components" width=\textwidth

*/

/**

@defgroup planar Planarity Embedding and Drawing

@ingroup algs

\brief Algorithms for planarity checking, embedding and drawing

This group contains the algorithms for planarity checking,

embedding and drawing.

\image html planar.png

\image latex planar.eps "Plane graph" width=\textwidth

*/

/**

@defgroup approx Approximation Algorithms

@ingroup algs

\brief Approximation algorithms.

This group contains the approximation and heuristic algorithms

implemented in LEMON.

*/

/**

@defgroup auxalg Auxiliary Algorithms

@ingroup algs

\brief Auxiliary algorithms implemented in LEMON.

This group contains some algorithms implemented in LEMON

in order to make it easier to implement complex algorithms.

*/

/**

@defgroup gen_opt_group General Optimization Tools

\brief This group contains some general optimization frameworks

implemented in LEMON.

This group contains some general optimization frameworks

implemented in LEMON.

*/

/**

@defgroup lp_group LP and MIP Solvers

@ingroup gen_opt_group

\brief LP and MIP solver interfaces for LEMON.

This group contains LP and MIP solver interfaces for LEMON.

Various LP solvers could be used in the same manner with this

high-level interface.

The currently supported solvers are \ref glpk, \ref clp, \ref cbc,

\ref cplex, \ref soplex.

*/

/**

@defgroup lp_utils Tools for Lp and Mip Solvers

@ingroup lp_group

\brief Helper tools to the Lp and Mip solvers.

This group adds some helper tools to general optimization framework

implemented in LEMON.

*/

/**

@defgroup metah Metaheuristics

@ingroup gen_opt_group

\brief Metaheuristics for LEMON library.

This group contains some metaheuristic optimization tools.

*/

/**

@defgroup utils Tools and Utilities

\brief Tools and utilities for programming in LEMON

Tools and utilities for programming in LEMON.

*/

/**

@defgroup gutils Basic Graph Utilities

@ingroup utils

\brief Simple basic graph utilities.

This group contains some simple basic graph utilities.

*/

/**

@defgroup misc Miscellaneous Tools

@ingroup utils

\brief Tools for development, debugging and testing.

This group contains several useful tools for development,

debugging and testing.

*/

/**

@defgroup timecount Time Measuring and Counting

@ingroup misc

\brief Simple tools for measuring the performance of algorithms.

This group contains simple tools for measuring the performance

of algorithms.

*/

/**

@defgroup exceptions Exceptions

@ingroup utils

\brief Exceptions defined in LEMON.

This group contains the exceptions defined in LEMON.

*/

/**

@defgroup io_group Input-Output

\brief Graph Input-Output methods

This group contains the tools for importing and exporting graphs

and graph related data. Now it supports the \ref lgf-format

"LEMON Graph Format", the \c DIMACS format and the encapsulated

postscript (EPS) format.

*/

/**

@defgroup lemon_io LEMON Graph Format

@ingroup io_group

\brief Reading and writing LEMON Graph Format.

This group contains methods for reading and writing

\ref lgf-format "LEMON Graph Format".

*/

/**

@defgroup eps_io Postscript Exporting

@ingroup io_group

\brief General \c EPS drawer and graph exporter

This group contains general \c EPS drawing methods and special

graph exporting tools.

*/

/**

@defgroup dimacs_group DIMACS Format

@ingroup io_group

\brief Read and write files in DIMACS format

Tools to read a digraph from or write it to a file in DIMACS format data.

*/

/**

@defgroup nauty_group NAUTY Format

@ingroup io_group

\brief Read \e Nauty format

Tool to read graphs from \e Nauty format data.

*/

/**

@defgroup concept Concepts

\brief Skeleton classes and concept checking classes

This group contains the data/algorithm skeletons and concept checking

classes implemented in LEMON.

The purpose of the classes in this group is fourfold.

- These classes contain the documentations of the %concepts. In order

  to avoid document multiplications, an implementation of a concept

  simply refers to the corresponding concept class.

- These classes declare every functions, <tt>typedef</tt>s etc. an

  implementation of the %concepts should provide, however completely

  without implementations and real data structures behind the

  interface. On the other hand they should provide nothing else. All

  the algorithms working on a data structure meeting a certain concept

  should compile with these classes. (Though it will not run properly,

  of course.) In this way it is easily to check if an algorithm

  doesn't use any extra feature of a certain implementation.

- The concept descriptor classes also provide a <em>checker class</em>

  that makes it possible to check whether a certain implementation of a

  concept indeed provides all the required features.

- Finally, They can serve as a skeleton of a new implementation of a concept.

*/

/**

@defgroup graph_concepts Graph Structure Concepts

@ingroup concept

\brief Skeleton and concept checking classes for graph structures

This group contains the skeletons and concept checking classes of

graph structures.

*/

/**

@defgroup map_concepts Map Concepts

@ingroup concept

/* -*- C++ -*-

 *

 * This file is a part of LEMON, a generic C++ optimization library

 *

 * Copyright (C) 2003-2008

 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport

 * (Egervary Research Group on Combinatorial Optimization, EGRES).

 *

 * Permission to use, modify and distribute this software is granted

 * provided that this copyright notice appears in all copies. For

 * precise terms see the accompanying LICENSE file.

 *

 * This software is provided "AS IS" with no warranty of any kind,

 * express or implied, and with no claim as to its suitability for any

 * purpose.

 *

 */

#ifndef LEMON_HARTMANN_ORLIN_H

#define LEMON_HARTMANN_ORLIN_H

/// \ingroup min_mean_cycle

///

/// \file

/// \brief Hartmann-Orlin's algorithm for finding a minimum mean cycle.

#include <vector>

#include <limits>

#include <lemon/core.h>

#include <lemon/path.h>

#include <lemon/tolerance.h>

#include <lemon/connectivity.h>

namespace lemon {

  /// \brief Default traits class of HartmannOrlin algorithm.

  ///

  /// Default traits class of HartmannOrlin algorithm.

  /// \tparam GR The type of the digraph.

  /// \tparam LEN The type of the length map.

  /// It must conform to the \ref concepts::Rea_data "Rea_data" concept.

#ifdef DOXYGEN

  template <typename GR, typename LEN>

#else

  template <typename GR, typename LEN,

    bool integer = std::numeric_limits<typename LEN::Value>::is_integer>

#endif

  struct HartmannOrlinDefaultTraits

  {

    /// The type of the digraph

    typedef GR Digraph;

    /// The type of the length map

    typedef LEN LengthMap;

    /// The type of the arc lengths

    typedef typename LengthMap::Value Value;

    /// \brief The large value type used for internal computations

    ///

    /// The large value type used for internal computations.

    /// It is \c long \c long if the \c Value type is integer,

    /// otherwise it is \c double.

    /// \c Value must be convertible to \c LargeValue.

    typedef double LargeValue;

    /// The tolerance type used for internal computations

    typedef lemon::Tolerance<LargeValue> Tolerance;

    /// \brief The path type of the found cycles

    ///

    /// The path type of the found cycles.

    /// It must conform to the \ref lemon::concepts::Path "Path" concept

    /// and it must have an \c addBack() function.

    typedef lemon::Path<Digraph> Path;

  };

  // Default traits class for integer value types

  template <typename GR, typename LEN>

  struct HartmannOrlinDefaultTraits<GR, LEN, true>

  {

    typedef GR Digraph;

    typedef LEN LengthMap;

    typedef typename LengthMap::Value Value;

#ifdef LEMON_HAVE_LONG_LONG

    typedef long long LargeValue;

#else

    typedef long LargeValue;

#endif

    typedef lemon::Tolerance<LargeValue> Tolerance;

    typedef lemon::Path<Digraph> Path;

  };

  /// \addtogroup min_mean_cycle

  /// @{

  /// \brief Implementation of the Hartmann-Orlin algorithm for finding

  /// a minimum mean cycle.

  ///

  /// This class implements the Hartmann-Orlin algorithm for finding

  /// It is an improved version of \ref Karp "Karp"'s original algorithm,

  /// it applies an efficient early termination scheme.

  /// It runs in time O(ne) and uses space O(n<sup>2</sup>+e).

  ///

  /// \tparam GR The type of the digraph the algorithm runs on.

  /// \tparam LEN The type of the length map. The default

  /// map type is \ref concepts::Digraph::ArcMap "GR::ArcMap<int>".

#ifdef DOXYGEN

  template <typename GR, typename LEN, typename TR>

#else

  template < typename GR,

             typename LEN = typename GR::template ArcMap<int>,

             typename TR = HartmannOrlinDefaultTraits<GR, LEN> >

#endif

  class HartmannOrlin

  {

  public:

    /// The type of the digraph

    typedef typename TR::Digraph Digraph;

    /// The type of the length map

    typedef typename TR::LengthMap LengthMap;

    /// The type of the arc lengths

    typedef typename TR::Value Value;

    /// \brief The large value type

    ///

    /// The large value type used for internal computations.

    /// Using the \ref HartmannOrlinDefaultTraits "default traits class",

    /// it is \c long \c long if the \c Value type is integer,

    /// otherwise it is \c double.

    typedef typename TR::LargeValue LargeValue;

    /// The tolerance type

    typedef typename TR::Tolerance Tolerance;

    /// \brief The path type of the found cycles

    ///

    /// The path type of the found cycles.

    /// Using the \ref HartmannOrlinDefaultTraits "default traits class",

    /// it is \ref lemon::Path "Path<Digraph>".

    typedef typename TR::Path Path;

    /// The \ref HartmannOrlinDefaultTraits "traits class" of the algorithm

    typedef TR Traits;

  private:

    TEMPLATE_DIGRAPH_TYPEDEFS(Digraph);

    // Data sturcture for path data

    struct PathData

    {

      LargeValue dist;

      Arc pred;

      PathData(LargeValue d, Arc p = INVALID) :

        dist(d), pred(p) {}

    };

    typedef typename Digraph::template NodeMap<std::vector<PathData> >

      PathDataNodeMap;

  private:

    // The digraph the algorithm runs on

    const Digraph &_gr;

    // The length of the arcs

    const LengthMap &_length;

    // Data for storing the strongly connected components

    int _comp_num;

    typename Digraph::template NodeMap<int> _comp;

    std::vector<std::vector<Node> > _comp_nodes;

    std::vector<Node>* _nodes;

    typename Digraph::template NodeMap<std::vector<Arc> > _out_arcs;

    // Data for the found cycles

    bool _curr_found, _best_found;

    LargeValue _curr_length, _best_length;

    int _curr_size, _best_size;

    Node _curr_node, _best_node;

    int _curr_level, _best_level;

    Path *_cycle_path;

    bool _local_path;

    // Node map for storing path data

    PathDataNodeMap _data;

    // The processed nodes in the last round

    std::vector<Node> _process;

    Tolerance _tolerance;

    // Infinite constant

    const LargeValue INF;

  public:

    /// \name Named Template Parameters

    /// @{

    template <typename T>

    struct SetLargeValueTraits : public Traits {

      typedef T LargeValue;

      typedef lemon::Tolerance<T> Tolerance;

    };

    /// \brief \ref named-templ-param "Named parameter" for setting

    /// \c LargeValue type.

    ///

    /// \ref named-templ-param "Named parameter" for setting \c LargeValue

    /// type. It is used for internal computations in the algorithm.

    template <typename T>

    struct SetLargeValue

      : public HartmannOrlin<GR, LEN, SetLargeValueTraits<T> > {

      typedef HartmannOrlin<GR, LEN, SetLargeValueTraits<T> > Create;

    };

    template <typename T>

    struct SetPathTraits : public Traits {

      typedef T Path;

    };

    /// \brief \ref named-templ-param "Named parameter" for setting

    /// \c %Path type.

    ///

    /// \ref named-templ-param "Named parameter" for setting the \c %Path

    /// type of the found cycles.

    /// It must conform to the \ref lemon::concepts::Path "Path" concept

    /// and it must have an \c addFront() function.

    template <typename T>

    struct SetPath

      : public HartmannOrlin<GR, LEN, SetPathTraits<T> > {

      typedef HartmannOrlin<GR, LEN, SetPathTraits<T> > Create;

    };

    /// @}

  public:

    /// \brief Constructor.

    ///

    /// The constructor of the class.

    ///

    /// \param digraph The digraph the algorithm runs on.

    /// \param length The lengths (costs) of the arcs.

    HartmannOrlin( const Digraph &digraph,

                   const LengthMap &length ) :

      _gr(digraph), _length(length), _comp(digraph), _out_arcs(digraph),

      _best_found(false), _best_length(0), _best_size(1),

      _cycle_path(NULL), _local_path(false), _data(digraph),

      INF(std::numeric_limits<LargeValue>::has_infinity ?

          std::numeric_limits<LargeValue>::infinity() :

          std::numeric_limits<LargeValue>::max())

    {}

    /// Destructor.

    ~HartmannOrlin() {

      if (_local_path) delete _cycle_path;

    }

    /// \brief Set the path structure for storing the found cycle.

    ///

    /// This function sets an external path structure for storing the

    /// found cycle.

    ///

    /// If you don't call this function before calling \ref run() or

    /// \ref findMinMean(), it will allocate a local \ref Path "path"

    /// structure. The destuctor deallocates this automatically

    /// allocated object, of course.

    ///

    /// \note The algorithm calls only the \ref lemon::Path::addFront()

    /// "addFront()" function of the given path structure.

    ///

    /// \return <tt>(*this)</tt>

    HartmannOrlin& cycle(Path &path) {

      if (_local_path) {

        delete _cycle_path;

        _local_path = false;

      }

      _cycle_path = &path;

      return *this;

    }

    /// \brief Set the tolerance used by the algorithm.

    ///

    /// This function sets the tolerance object used by the algorithm.

    ///

    /// \return <tt>(*this)</tt>

    HartmannOrlin& tolerance(const Tolerance& tolerance) {

      _tolerance = tolerance;

      return *this;

    }

    /// \brief Return a const reference to the tolerance.

    ///

    /// This function returns a const reference to the tolerance object

    /// used by the algorithm.

    const Tolerance& tolerance() const {

      return _tolerance;

    }

    /// \name Execution control

    /// The simplest way to execute the algorithm is to call the \ref run()

    /// function.\n

    /// If you only need the minimum mean length, you may call

    /// \ref findMinMean().

    /// @{

    /// \brief Run the algorithm.

    ///

    /// This function runs the algorithm.

    /// It can be called more than once (e.g. if the underlying digraph

    /// and/or the arc lengths have been modified).

    ///

    /// \return \c true if a directed cycle exists in the digraph.

    ///

    /// \note <tt>mmc.run()</tt> is just a shortcut of the following code.

    /// \code

    ///   return mmc.findMinMean() && mmc.findCycle();

    /// \endcode

    bool run() {

      return findMinMean() && findCycle();

    }

    /// \brief Find the minimum cycle mean.

    ///

    /// This function finds the minimum mean length of the directed

    /// cycles in the digraph.

    ///

    /// \return \c true if a directed cycle exists in the digraph.

    bool findMinMean() {

      // Initialization and find strongly connected components

      init();

      findComponents();

      // Find the minimum cycle mean in the components

      for (int comp = 0; comp < _comp_num; ++comp) {

        if (!initComponent(comp)) continue;

        processRounds();

        // Update the best cycle (global minimum mean cycle)

        if ( _curr_found && (!_best_found || 

             _curr_length * _best_size < _best_length * _curr_size) ) {

          _best_found = true;

          _best_length = _curr_length;

          _best_size = _curr_size;

          _best_node = _curr_node;

          _best_level = _curr_level;

        }

      }

      return _best_found;

    }

    /// \brief Find a minimum mean directed cycle.

/* -*- C++ -*-

 *

 * This file is a part of LEMON, a generic C++ optimization library

 *

 * Copyright (C) 2003-2008

 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport

 * (Egervary Research Group on Combinatorial Optimization, EGRES).

 *

 * Permission to use, modify and distribute this software is granted

 * provided that this copyright notice appears in all copies. For

 * precise terms see the accompanying LICENSE file.

 *

 * This software is provided "AS IS" with no warranty of any kind,

 * express or implied, and with no claim as to its suitability for any

 * purpose.

 *

 */

#ifndef LEMON_HOWARD_H

#define LEMON_HOWARD_H

/// \ingroup min_mean_cycle

///

/// \file

/// \brief Howard's algorithm for finding a minimum mean cycle.

#include <vector>

#include <limits>

#include <lemon/core.h>

#include <lemon/path.h>

#include <lemon/tolerance.h>

#include <lemon/connectivity.h>

namespace lemon {

  /// \brief Default traits class of Howard class.

  ///

  /// Default traits class of Howard class.

  /// \tparam GR The type of the digraph.

  /// \tparam LEN The type of the length map.

  /// It must conform to the \ref concepts::ReadMap "ReadMap" concept.

#ifdef DOXYGEN

  template <typename GR, typename LEN>

#else

  template <typename GR, typename LEN,

    bool integer = std::numeric_limits<typename LEN::Value>::is_integer>

#endif

  struct HowardDefaultTraits

  {

    /// The type of the digraph

    typedef GR Digraph;

    /// The type of the length map

    typedef LEN LengthMap;

    /// The type of the arc lengths

    typedef typename LengthMap::Value Value;

    /// \brief The large value type used for internal computations

    ///

    /// The large value type used for internal computations.

    /// It is \c long \c long if the \c Value type is integer,

    /// otherwise it is \c double.

    /// \c Value must be convertible to \c LargeValue.

    typedef double LargeValue;

    /// The tolerance type used for internal computations

    typedef lemon::Tolerance<LargeValue> Tolerance;

    /// \brief The path type of the found cycles

    ///

    /// The path type of the found cycles.

    /// It must conform to the \ref lemon::concepts::Path "Path" concept

    /// and it must have an \c addBack() function.

    typedef lemon::Path<Digraph> Path;

  };

  // Default traits class for integer value types

  template <typename GR, typename LEN>

  struct HowardDefaultTraits<GR, LEN, true>

  {

    typedef GR Digraph;

    typedef LEN LengthMap;

    typedef typename LengthMap::Value Value;

#ifdef LEMON_HAVE_LONG_LONG

    typedef long long LargeValue;

#else

    typedef long LargeValue;

#endif

    typedef lemon::Tolerance<LargeValue> Tolerance;

    typedef lemon::Path<Digraph> Path;

  };

  /// \addtogroup min_mean_cycle

  /// @{

  /// \brief Implementation of Howard's algorithm for finding a minimum

  /// mean cycle.

  ///

  /// This class implements Howard's policy iteration algorithm for finding

  /// This class provides the most efficient algorithm for the

  /// minimum mean cycle problem, though the best known theoretical

  /// bound on its running time is exponential.

  ///

  /// \tparam GR The type of the digraph the algorithm runs on.

  /// \tparam LEN The type of the length map. The default

  /// map type is \ref concepts::Digraph::ArcMap "GR::ArcMap<int>".

#ifdef DOXYGEN

  template <typename GR, typename LEN, typename TR>

#else

  template < typename GR,

             typename LEN = typename GR::template ArcMap<int>,

             typename TR = HowardDefaultTraits<GR, LEN> >

#endif

  class Howard

  {

  public:

    /// The type of the digraph

    typedef typename TR::Digraph Digraph;

    /// The type of the length map

    typedef typename TR::LengthMap LengthMap;

    /// The type of the arc lengths

    typedef typename TR::Value Value;

    /// \brief The large value type

    ///

    /// The large value type used for internal computations.

    /// Using the \ref HowardDefaultTraits "default traits class",

    /// it is \c long \c long if the \c Value type is integer,

    /// otherwise it is \c double.

    typedef typename TR::LargeValue LargeValue;

    /// The tolerance type

    typedef typename TR::Tolerance Tolerance;

    /// \brief The path type of the found cycles

    ///

    /// The path type of the found cycles.

    /// Using the \ref HowardDefaultTraits "default traits class",

    /// it is \ref lemon::Path "Path<Digraph>".

    typedef typename TR::Path Path;

    /// The \ref HowardDefaultTraits "traits class" of the algorithm

    typedef TR Traits;

  private:

    TEMPLATE_DIGRAPH_TYPEDEFS(Digraph);

    // The digraph the algorithm runs on

    const Digraph &_gr;

    // The length of the arcs

    const LengthMap &_length;

    // Data for the found cycles

    bool _curr_found, _best_found;

    LargeValue _curr_length, _best_length;

    int _curr_size, _best_size;

    Node _curr_node, _best_node;

    Path *_cycle_path;

    bool _local_path;

    // Internal data used by the algorithm

    typename Digraph::template NodeMap<Arc> _policy;

    typename Digraph::template NodeMap<bool> _reached;

    typename Digraph::template NodeMap<int> _level;

    typename Digraph::template NodeMap<LargeValue> _dist;

    // Data for storing the strongly connected components

    int _comp_num;

    typename Digraph::template NodeMap<int> _comp;

    std::vector<std::vector<Node> > _comp_nodes;

    std::vector<Node>* _nodes;

    typename Digraph::template NodeMap<std::vector<Arc> > _in_arcs;

    // Queue used for BFS search

    std::vector<Node> _queue;

    int _qfront, _qback;

    Tolerance _tolerance;

    // Infinite constant

    const LargeValue INF;

  public:

    /// \name Named Template Parameters

    /// @{

    template <typename T>

    struct SetLargeValueTraits : public Traits {

      typedef T LargeValue;

      typedef lemon::Tolerance<T> Tolerance;

    };

    /// \brief \ref named-templ-param "Named parameter" for setting

    /// \c LargeValue type.

    ///

    /// \ref named-templ-param "Named parameter" for setting \c LargeValue

    /// type. It is used for internal computations in the algorithm.

    template <typename T>

    struct SetLargeValue

      : public Howard<GR, LEN, SetLargeValueTraits<T> > {

      typedef Howard<GR, LEN, SetLargeValueTraits<T> > Create;

    };