3 * This file is a part of LEMON, a generic C++ optimization library
5 * Copyright (C) 2003-2008
6 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
7 * (Egervary Research Group on Combinatorial Optimization, EGRES).
9 * Permission to use, modify and distribute this software is granted
10 * provided that this copyright notice appears in all copies. For
11 * precise terms see the accompanying LICENSE file.
13 * This software is provided "AS IS" with no warranty of any kind,
14 * express or implied, and with no claim as to its suitability for any
19 #ifndef LEMON_EDMONDS_KARP_H
20 #define LEMON_EDMONDS_KARP_H
24 /// \brief Implementation of the Edmonds-Karp algorithm.
26 #include <lemon/tolerance.h>
31 /// \brief Default traits class of EdmondsKarp class.
33 /// Default traits class of EdmondsKarp class.
34 /// \param _Graph Graph type.
35 /// \param _CapacityMap Type of capacity map.
36 template <typename _Graph, typename _CapacityMap>
37 struct EdmondsKarpDefaultTraits {
39 /// \brief The graph type the algorithm runs on.
42 /// \brief The type of the map that stores the edge capacities.
44 /// The type of the map that stores the edge capacities.
45 /// It must meet the \ref concepts::ReadMap "ReadMap" concept.
46 typedef _CapacityMap CapacityMap;
48 /// \brief The type of the length of the edges.
49 typedef typename CapacityMap::Value Value;
51 /// \brief The map type that stores the flow values.
53 /// The map type that stores the flow values.
54 /// It must meet the \ref concepts::ReadWriteMap "ReadWriteMap" concept.
55 typedef typename Graph::template EdgeMap<Value> FlowMap;
57 /// \brief Instantiates a FlowMap.
59 /// This function instantiates a \ref FlowMap.
60 /// \param graph The graph, to which we would like to define the flow map.
61 static FlowMap* createFlowMap(const Graph& graph) {
62 return new FlowMap(graph);
65 /// \brief The tolerance used by the algorithm
67 /// The tolerance used by the algorithm to handle inexact computation.
68 typedef Tolerance<Value> Tolerance;
74 /// \brief Edmonds-Karp algorithms class.
76 /// This class provides an implementation of the \e Edmonds-Karp \e
77 /// algorithm producing a flow of maximum value in a directed
78 /// graphs. The Edmonds-Karp algorithm is slower than the Preflow
79 /// algorithm but it has an advantage of the step-by-step execution
80 /// control with feasible flow solutions. The \e source node, the \e
81 /// target node, the \e capacity of the edges and the \e starting \e
82 /// flow value of the edges should be passed to the algorithm
83 /// through the constructor.
85 /// The time complexity of the algorithm is \f$ O(nm^2) \f$ in
86 /// worst case. Always try the preflow algorithm instead of this if
87 /// you just want to compute the optimal flow.
89 /// \param _Graph The directed graph type the algorithm runs on.
90 /// \param _CapacityMap The capacity map type.
91 /// \param _Traits Traits class to set various data types used by
92 /// the algorithm. The default traits class is \ref
93 /// EdmondsKarpDefaultTraits. See \ref EdmondsKarpDefaultTraits for the
94 /// documentation of a Edmonds-Karp traits class.
96 /// \author Balazs Dezso
98 template <typename _Graph, typename _CapacityMap, typename _Traits>
100 template <typename _Graph,
101 typename _CapacityMap = typename _Graph::template EdgeMap<int>,
102 typename _Traits = EdmondsKarpDefaultTraits<_Graph, _CapacityMap> >
107 typedef _Traits Traits;
108 typedef typename Traits::Graph Graph;
109 typedef typename Traits::CapacityMap CapacityMap;
110 typedef typename Traits::Value Value;
112 typedef typename Traits::FlowMap FlowMap;
113 typedef typename Traits::Tolerance Tolerance;
115 /// \brief \ref Exception for the case when the source equals the target.
117 /// \ref Exception for the case when the source equals the target.
119 class InvalidArgument : public lemon::LogicError {
121 virtual const char* what() const throw() {
122 return "lemon::EdmondsKarp::InvalidArgument";
129 GRAPH_TYPEDEFS(typename Graph);
130 typedef typename Graph::template NodeMap<Edge> PredMap;
133 const CapacityMap* _capacity;
135 Node _source, _target;
141 std::vector<Node> _queue;
143 Tolerance _tolerance;
146 void createStructures() {
148 _flow = Traits::createFlowMap(_graph);
152 _pred = new PredMap(_graph);
154 _queue.resize(countNodes(_graph));
157 void destroyStructures() {
168 ///\name Named template parameters
172 template <typename _FlowMap>
173 struct DefFlowMapTraits : public Traits {
174 typedef _FlowMap FlowMap;
175 static FlowMap *createFlowMap(const Graph&) {
176 throw UninitializedParameter();
180 /// \brief \ref named-templ-param "Named parameter" for setting
183 /// \ref named-templ-param "Named parameter" for setting FlowMap
185 template <typename _FlowMap>
187 : public EdmondsKarp<Graph, CapacityMap, DefFlowMapTraits<_FlowMap> > {
188 typedef EdmondsKarp<Graph, CapacityMap, DefFlowMapTraits<_FlowMap> >
201 /// \brief The constructor of the class.
203 /// The constructor of the class.
204 /// \param graph The directed graph the algorithm runs on.
205 /// \param capacity The capacity of the edges.
206 /// \param source The source node.
207 /// \param target The target node.
208 EdmondsKarp(const Graph& graph, const CapacityMap& capacity,
209 Node source, Node target)
210 : _graph(graph), _capacity(&capacity), _source(source), _target(target),
211 _flow(0), _local_flow(false), _pred(0), _tolerance(), _flow_value()
213 if (_source == _target) {
214 throw InvalidArgument();
218 /// \brief Destrcutor.
225 /// \brief Sets the capacity map.
227 /// Sets the capacity map.
228 /// \return \c (*this)
229 EdmondsKarp& capacityMap(const CapacityMap& map) {
234 /// \brief Sets the flow map.
236 /// Sets the flow map.
237 /// \return \c (*this)
238 EdmondsKarp& flowMap(FlowMap& map) {
247 /// \brief Returns the flow map.
249 /// \return The flow map.
250 const FlowMap& flowMap() {
254 /// \brief Sets the source node.
256 /// Sets the source node.
257 /// \return \c (*this)
258 EdmondsKarp& source(const Node& node) {
263 /// \brief Sets the target node.
265 /// Sets the target node.
266 /// \return \c (*this)
267 EdmondsKarp& target(const Node& node) {
272 /// \brief Sets the tolerance used by algorithm.
274 /// Sets the tolerance used by algorithm.
275 EdmondsKarp& tolerance(const Tolerance& tolerance) const {
276 _tolerance = tolerance;
280 /// \brief Returns the tolerance used by algorithm.
282 /// Returns the tolerance used by algorithm.
283 const Tolerance& tolerance() const {
287 /// \name Execution control The simplest way to execute the
288 /// algorithm is to use the \c run() member functions.
290 /// If you need more control on initial solution or
291 /// execution then you have to call one \ref init() function and then
292 /// the start() or multiple times the \c augment() member function.
296 /// \brief Initializes the algorithm
298 /// It sets the flow to empty flow.
301 for (EdgeIt it(_graph); it != INVALID; ++it) {
307 /// \brief Initializes the algorithm
309 /// Initializes the flow to the \c flowMap. The \c flowMap should
310 /// contain a feasible flow, ie. in each node excluding the source
311 /// and the target the incoming flow should be equal to the
313 template <typename FlowMap>
314 void flowInit(const FlowMap& flowMap) {
316 for (EdgeIt e(_graph); e != INVALID; ++e) {
317 _flow->set(e, flowMap[e]);
320 for (OutEdgeIt jt(_graph, _source); jt != INVALID; ++jt) {
321 _flow_value += (*_flow)[jt];
323 for (InEdgeIt jt(_graph, _source); jt != INVALID; ++jt) {
324 _flow_value -= (*_flow)[jt];
328 /// \brief Initializes the algorithm
330 /// Initializes the flow to the \c flowMap. The \c flowMap should
331 /// contain a feasible flow, ie. in each node excluding the source
332 /// and the target the incoming flow should be equal to the
334 /// \return %False when the given flowMap does not contain
336 template <typename FlowMap>
337 bool checkedFlowInit(const FlowMap& flowMap) {
339 for (EdgeIt e(_graph); e != INVALID; ++e) {
340 _flow->set(e, flowMap[e]);
342 for (NodeIt it(_graph); it != INVALID; ++it) {
343 if (it == _source || it == _target) continue;
345 for (OutEdgeIt jt(_graph, it); jt != INVALID; ++jt) {
346 outFlow += (*_flow)[jt];
349 for (InEdgeIt jt(_graph, it); jt != INVALID; ++jt) {
350 inFlow += (*_flow)[jt];
352 if (_tolerance.different(outFlow, inFlow)) {
356 for (EdgeIt it(_graph); it != INVALID; ++it) {
357 if (_tolerance.less((*_flow)[it], 0)) return false;
358 if (_tolerance.less((*_capacity)[it], (*_flow)[it])) return false;
361 for (OutEdgeIt jt(_graph, _source); jt != INVALID; ++jt) {
362 _flow_value += (*_flow)[jt];
364 for (InEdgeIt jt(_graph, _source); jt != INVALID; ++jt) {
365 _flow_value -= (*_flow)[jt];
370 /// \brief Augment the solution on an edge shortest path.
372 /// Augment the solution on an edge shortest path. It search an
373 /// edge shortest path between the source and the target
374 /// in the residual graph with the bfs algoritm.
375 /// Then it increase the flow on this path with the minimal residual
376 /// capacity on the path. If there is not such path it gives back
378 /// \return %False when the augmenting is not success so the
379 /// current flow is a feasible and optimal solution.
381 for (NodeIt n(_graph); n != INVALID; ++n) {
382 _pred->set(n, INVALID);
385 int first = 0, last = 1;
388 _pred->set(_source, OutEdgeIt(_graph, _source));
390 while (first != last && (*_pred)[_target] == INVALID) {
391 Node n = _queue[first++];
393 for (OutEdgeIt e(_graph, n); e != INVALID; ++e) {
394 Value rem = (*_capacity)[e] - (*_flow)[e];
395 Node t = _graph.target(e);
396 if (_tolerance.positive(rem) && (*_pred)[t] == INVALID) {
401 for (InEdgeIt e(_graph, n); e != INVALID; ++e) {
402 Value rem = (*_flow)[e];
403 Node t = _graph.source(e);
404 if (_tolerance.positive(rem) && (*_pred)[t] == INVALID) {
411 if ((*_pred)[_target] != INVALID) {
413 Edge e = (*_pred)[n];
415 Value prem = (*_capacity)[e] - (*_flow)[e];
416 n = _graph.source(e);
417 while (n != _source) {
419 if (_graph.target(e) == n) {
420 Value rem = (*_capacity)[e] - (*_flow)[e];
421 if (rem < prem) prem = rem;
422 n = _graph.source(e);
424 Value rem = (*_flow)[e];
425 if (rem < prem) prem = rem;
426 n = _graph.target(e);
433 _flow->set(e, (*_flow)[e] + prem);
434 n = _graph.source(e);
435 while (n != _source) {
437 if (_graph.target(e) == n) {
438 _flow->set(e, (*_flow)[e] + prem);
439 n = _graph.source(e);
441 _flow->set(e, (*_flow)[e] - prem);
442 n = _graph.target(e);
453 /// \brief Executes the algorithm
455 /// It runs augmenting phases until the optimal solution is reached.
460 /// \brief runs the algorithm.
462 /// It is just a shorthand for:
475 /// \name Query Functions
476 /// The result of the Edmonds-Karp algorithm can be obtained using these
478 /// Before the use of these functions,
479 /// either run() or start() must be called.
483 /// \brief Returns the value of the maximum flow.
485 /// Returns the value of the maximum flow by returning the excess
486 /// of the target node \c t. This value equals to the value of
487 /// the maximum flow already after the first phase.
488 Value flowValue() const {
493 /// \brief Returns the flow on the edge.
495 /// Sets the \c flowMap to the flow on the edges. This method can
496 /// be called after the second phase of algorithm.
497 Value flow(const Edge& edge) const {
498 return (*_flow)[edge];
501 /// \brief Returns true when the node is on the source side of minimum cut.
504 /// Returns true when the node is on the source side of minimum
505 /// cut. This method can be called both after running \ref
506 /// startFirstPhase() and \ref startSecondPhase().
507 bool minCut(const Node& node) const {
508 return (*_pred)[node] != INVALID;
511 /// \brief Returns a minimum value cut.
513 /// Sets \c cut to the characteristic vector of a minimum value cut
514 /// It simply calls the minMinCut member.
515 /// \retval cut Write node bool map.
516 template <typename CutMap>
517 void minCutMap(CutMap& cutMap) const {
518 for (NodeIt n(_graph); n != INVALID; ++n) {
519 cutMap.set(n, (*_pred)[n] != INVALID);
521 cutMap.set(_source, true);