gravatar
alpar (Alpar Juttner)
alpar@cs.elte.hu
Merge #340
0 4 0
merge default
4 files changed with 377 insertions and 161 deletions:
↑ Collapse diff ↑
Ignore white space 6 line context
1 1
/* -*- C++ -*-
2 2
 *
3 3
 * This file is a part of LEMON, a generic C++ optimization library
4 4
 *
5 5
 * Copyright (C) 2003-2008
6 6
 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
7 7
 * (Egervary Research Group on Combinatorial Optimization, EGRES).
8 8
 *
9 9
 * Permission to use, modify and distribute this software is granted
10 10
 * provided that this copyright notice appears in all copies. For
11 11
 * precise terms see the accompanying LICENSE file.
12 12
 *
13 13
 * This software is provided "AS IS" with no warranty of any kind,
14 14
 * express or implied, and with no claim as to its suitability for any
15 15
 * purpose.
16 16
 *
17 17
 */
18 18

	
19 19
#ifndef LEMON_CAPACITY_SCALING_H
20 20
#define LEMON_CAPACITY_SCALING_H
21 21

	
22 22
/// \ingroup min_cost_flow_algs
23 23
///
24 24
/// \file
25 25
/// \brief Capacity Scaling algorithm for finding a minimum cost flow.
26 26

	
27 27
#include <vector>
28 28
#include <limits>
29 29
#include <lemon/core.h>
30 30
#include <lemon/bin_heap.h>
31 31

	
32 32
namespace lemon {
33 33

	
34 34
  /// \brief Default traits class of CapacityScaling algorithm.
35 35
  ///
36 36
  /// Default traits class of CapacityScaling algorithm.
37 37
  /// \tparam GR Digraph type.
38 38
  /// \tparam V The number type used for flow amounts, capacity bounds
39 39
  /// and supply values. By default it is \c int.
40 40
  /// \tparam C The number type used for costs and potentials.
41 41
  /// By default it is the same as \c V.
42 42
  template <typename GR, typename V = int, typename C = V>
43 43
  struct CapacityScalingDefaultTraits
44 44
  {
45 45
    /// The type of the digraph
46 46
    typedef GR Digraph;
47 47
    /// The type of the flow amounts, capacity bounds and supply values
48 48
    typedef V Value;
49 49
    /// The type of the arc costs
50 50
    typedef C Cost;
51 51

	
52 52
    /// \brief The type of the heap used for internal Dijkstra computations.
53 53
    ///
54 54
    /// The type of the heap used for internal Dijkstra computations.
55 55
    /// It must conform to the \ref lemon::concepts::Heap "Heap" concept,
56 56
    /// its priority type must be \c Cost and its cross reference type
57 57
    /// must be \ref RangeMap "RangeMap<int>".
58 58
    typedef BinHeap<Cost, RangeMap<int> > Heap;
59 59
  };
60 60

	
61 61
  /// \addtogroup min_cost_flow_algs
62 62
  /// @{
63 63

	
64 64
  /// \brief Implementation of the Capacity Scaling algorithm for
65 65
  /// finding a \ref min_cost_flow "minimum cost flow".
66 66
  ///
67 67
  /// \ref CapacityScaling implements the capacity scaling version
68 68
  /// of the successive shortest path algorithm for finding a
69 69
  /// \ref min_cost_flow "minimum cost flow" \ref amo93networkflows,
70 70
  /// \ref edmondskarp72theoretical. It is an efficient dual
71 71
  /// solution method.
72 72
  ///
73 73
  /// Most of the parameters of the problem (except for the digraph)
74 74
  /// can be given using separate functions, and the algorithm can be
75 75
  /// executed using the \ref run() function. If some parameters are not
76 76
  /// specified, then default values will be used.
77 77
  ///
78 78
  /// \tparam GR The digraph type the algorithm runs on.
79 79
  /// \tparam V The number type used for flow amounts, capacity bounds
80 80
  /// and supply values in the algorithm. By default, it is \c int.
81 81
  /// \tparam C The number type used for costs and potentials in the
82 82
  /// algorithm. By default, it is the same as \c V.
83 83
  /// \tparam TR The traits class that defines various types used by the
84 84
  /// algorithm. By default, it is \ref CapacityScalingDefaultTraits
85 85
  /// "CapacityScalingDefaultTraits<GR, V, C>".
86 86
  /// In most cases, this parameter should not be set directly,
87 87
  /// consider to use the named template parameters instead.
88 88
  ///
89 89
  /// \warning Both number types must be signed and all input data must
90 90
  /// be integer.
91 91
  /// \warning This algorithm does not support negative costs for such
92 92
  /// arcs that have infinite upper bound.
93 93
#ifdef DOXYGEN
94 94
  template <typename GR, typename V, typename C, typename TR>
95 95
#else
96 96
  template < typename GR, typename V = int, typename C = V,
97 97
             typename TR = CapacityScalingDefaultTraits<GR, V, C> >
98 98
#endif
99 99
  class CapacityScaling
100 100
  {
101 101
  public:
102 102

	
103 103
    /// The type of the digraph
104 104
    typedef typename TR::Digraph Digraph;
105 105
    /// The type of the flow amounts, capacity bounds and supply values
106 106
    typedef typename TR::Value Value;
107 107
    /// The type of the arc costs
108 108
    typedef typename TR::Cost Cost;
109 109

	
110 110
    /// The type of the heap used for internal Dijkstra computations
111 111
    typedef typename TR::Heap Heap;
112 112

	
113 113
    /// The \ref CapacityScalingDefaultTraits "traits class" of the algorithm
114 114
    typedef TR Traits;
115 115

	
116 116
  public:
117 117

	
118 118
    /// \brief Problem type constants for the \c run() function.
119 119
    ///
120 120
    /// Enum type containing the problem type constants that can be
121 121
    /// returned by the \ref run() function of the algorithm.
122 122
    enum ProblemType {
123 123
      /// The problem has no feasible solution (flow).
124 124
      INFEASIBLE,
125 125
      /// The problem has optimal solution (i.e. it is feasible and
126 126
      /// bounded), and the algorithm has found optimal flow and node
127 127
      /// potentials (primal and dual solutions).
128 128
      OPTIMAL,
129 129
      /// The digraph contains an arc of negative cost and infinite
130 130
      /// upper bound. It means that the objective function is unbounded
131 131
      /// on that arc, however, note that it could actually be bounded
132 132
      /// over the feasible flows, but this algroithm cannot handle
133 133
      /// these cases.
134 134
      UNBOUNDED
135 135
    };
136 136
  
137 137
  private:
138 138

	
139 139
    TEMPLATE_DIGRAPH_TYPEDEFS(GR);
140 140

	
141 141
    typedef std::vector<int> IntVector;
142
    typedef std::vector<char> BoolVector;
143 142
    typedef std::vector<Value> ValueVector;
144 143
    typedef std::vector<Cost> CostVector;
144
    typedef std::vector<char> BoolVector;
145
    // Note: vector<char> is used instead of vector<bool> for efficiency reasons
145 146

	
146 147
  private:
147 148

	
148 149
    // Data related to the underlying digraph
149 150
    const GR &_graph;
150 151
    int _node_num;
151 152
    int _arc_num;
152 153
    int _res_arc_num;
153 154
    int _root;
154 155

	
155 156
    // Parameters of the problem
156 157
    bool _have_lower;
157 158
    Value _sum_supply;
158 159

	
159 160
    // Data structures for storing the digraph
160 161
    IntNodeMap _node_id;
161 162
    IntArcMap _arc_idf;
162 163
    IntArcMap _arc_idb;
163 164
    IntVector _first_out;
164 165
    BoolVector _forward;
165 166
    IntVector _source;
166 167
    IntVector _target;
167 168
    IntVector _reverse;
168 169

	
169 170
    // Node and arc data
170 171
    ValueVector _lower;
171 172
    ValueVector _upper;
172 173
    CostVector _cost;
173 174
    ValueVector _supply;
174 175

	
175 176
    ValueVector _res_cap;
176 177
    CostVector _pi;
177 178
    ValueVector _excess;
178 179
    IntVector _excess_nodes;
179 180
    IntVector _deficit_nodes;
180 181

	
181 182
    Value _delta;
182 183
    int _factor;
183 184
    IntVector _pred;
184 185

	
185 186
  public:
186 187
  
187 188
    /// \brief Constant for infinite upper bounds (capacities).
188 189
    ///
189 190
    /// Constant for infinite upper bounds (capacities).
190 191
    /// It is \c std::numeric_limits<Value>::infinity() if available,
191 192
    /// \c std::numeric_limits<Value>::max() otherwise.
192 193
    const Value INF;
193 194

	
194 195
  private:
195 196

	
196 197
    // Special implementation of the Dijkstra algorithm for finding
197 198
    // shortest paths in the residual network of the digraph with
198 199
    // respect to the reduced arc costs and modifying the node
199 200
    // potentials according to the found distance labels.
200 201
    class ResidualDijkstra
201 202
    {
202 203
    private:
203 204

	
204 205
      int _node_num;
205 206
      bool _geq;
206 207
      const IntVector &_first_out;
207 208
      const IntVector &_target;
208 209
      const CostVector &_cost;
209 210
      const ValueVector &_res_cap;
210 211
      const ValueVector &_excess;
211 212
      CostVector &_pi;
212 213
      IntVector &_pred;
213 214
      
214 215
      IntVector _proc_nodes;
215 216
      CostVector _dist;
216 217
      
217 218
    public:
218 219

	
219 220
      ResidualDijkstra(CapacityScaling& cs) :
220 221
        _node_num(cs._node_num), _geq(cs._sum_supply < 0),
221 222
        _first_out(cs._first_out), _target(cs._target), _cost(cs._cost),
222 223
        _res_cap(cs._res_cap), _excess(cs._excess), _pi(cs._pi),
223 224
        _pred(cs._pred), _dist(cs._node_num)
224 225
      {}
225 226

	
226 227
      int run(int s, Value delta = 1) {
227 228
        RangeMap<int> heap_cross_ref(_node_num, Heap::PRE_HEAP);
228 229
        Heap heap(heap_cross_ref);
229 230
        heap.push(s, 0);
230 231
        _pred[s] = -1;
231 232
        _proc_nodes.clear();
232 233

	
233 234
        // Process nodes
234 235
        while (!heap.empty() && _excess[heap.top()] > -delta) {
235 236
          int u = heap.top(), v;
236 237
          Cost d = heap.prio() + _pi[u], dn;
237 238
          _dist[u] = heap.prio();
238 239
          _proc_nodes.push_back(u);
239 240
          heap.pop();
240 241

	
241 242
          // Traverse outgoing residual arcs
242 243
          int last_out = _geq ? _first_out[u+1] : _first_out[u+1] - 1;
243 244
          for (int a = _first_out[u]; a != last_out; ++a) {
244 245
            if (_res_cap[a] < delta) continue;
245 246
            v = _target[a];
246 247
            switch (heap.state(v)) {
247 248
              case Heap::PRE_HEAP:
248 249
                heap.push(v, d + _cost[a] - _pi[v]);
249 250
                _pred[v] = a;
250 251
                break;
251 252
              case Heap::IN_HEAP:
252 253
                dn = d + _cost[a] - _pi[v];
253 254
                if (dn < heap[v]) {
254 255
                  heap.decrease(v, dn);
255 256
                  _pred[v] = a;
256 257
                }
257 258
                break;
258 259
              case Heap::POST_HEAP:
259 260
                break;
260 261
            }
261 262
          }
262 263
        }
263 264
        if (heap.empty()) return -1;
264 265

	
265 266
        // Update potentials of processed nodes
266 267
        int t = heap.top();
267 268
        Cost dt = heap.prio();
268 269
        for (int i = 0; i < int(_proc_nodes.size()); ++i) {
269 270
          _pi[_proc_nodes[i]] += _dist[_proc_nodes[i]] - dt;
270 271
        }
271 272

	
272 273
        return t;
273 274
      }
274 275

	
275 276
    }; //class ResidualDijkstra
276 277

	
277 278
  public:
278 279

	
279 280
    /// \name Named Template Parameters
280 281
    /// @{
281 282

	
282 283
    template <typename T>
283 284
    struct SetHeapTraits : public Traits {
284 285
      typedef T Heap;
285 286
    };
286 287

	
287 288
    /// \brief \ref named-templ-param "Named parameter" for setting
288 289
    /// \c Heap type.
289 290
    ///
290 291
    /// \ref named-templ-param "Named parameter" for setting \c Heap
291 292
    /// type, which is used for internal Dijkstra computations.
292 293
    /// It must conform to the \ref lemon::concepts::Heap "Heap" concept,
293 294
    /// its priority type must be \c Cost and its cross reference type
294 295
    /// must be \ref RangeMap "RangeMap<int>".
295 296
    template <typename T>
296 297
    struct SetHeap
297 298
      : public CapacityScaling<GR, V, C, SetHeapTraits<T> > {
298 299
      typedef  CapacityScaling<GR, V, C, SetHeapTraits<T> > Create;
299 300
    };
300 301

	
301 302
    /// @}
302 303

	
303 304
  public:
304 305

	
305 306
    /// \brief Constructor.
306 307
    ///
307 308
    /// The constructor of the class.
308 309
    ///
309 310
    /// \param graph The digraph the algorithm runs on.
310 311
    CapacityScaling(const GR& graph) :
311 312
      _graph(graph), _node_id(graph), _arc_idf(graph), _arc_idb(graph),
312 313
      INF(std::numeric_limits<Value>::has_infinity ?
313 314
          std::numeric_limits<Value>::infinity() :
314 315
          std::numeric_limits<Value>::max())
315 316
    {
316 317
      // Check the number types
317 318
      LEMON_ASSERT(std::numeric_limits<Value>::is_signed,
318 319
        "The flow type of CapacityScaling must be signed");
319 320
      LEMON_ASSERT(std::numeric_limits<Cost>::is_signed,
320 321
        "The cost type of CapacityScaling must be signed");
321 322

	
322 323
      // Reset data structures
323 324
      reset();
324 325
    }
325 326

	
326 327
    /// \name Parameters
327 328
    /// The parameters of the algorithm can be specified using these
328 329
    /// functions.
329 330

	
330 331
    /// @{
331 332

	
332 333
    /// \brief Set the lower bounds on the arcs.
333 334
    ///
334 335
    /// This function sets the lower bounds on the arcs.
335 336
    /// If it is not used before calling \ref run(), the lower bounds
336 337
    /// will be set to zero on all arcs.
337 338
    ///
338 339
    /// \param map An arc map storing the lower bounds.
339 340
    /// Its \c Value type must be convertible to the \c Value type
340 341
    /// of the algorithm.
341 342
    ///
342 343
    /// \return <tt>(*this)</tt>
343 344
    template <typename LowerMap>
344 345
    CapacityScaling& lowerMap(const LowerMap& map) {
345 346
      _have_lower = true;
346 347
      for (ArcIt a(_graph); a != INVALID; ++a) {
347 348
        _lower[_arc_idf[a]] = map[a];
348 349
        _lower[_arc_idb[a]] = map[a];
349 350
      }
350 351
      return *this;
351 352
    }
352 353

	
353 354
    /// \brief Set the upper bounds (capacities) on the arcs.
354 355
    ///
355 356
    /// This function sets the upper bounds (capacities) on the arcs.
356 357
    /// If it is not used before calling \ref run(), the upper bounds
357 358
    /// will be set to \ref INF on all arcs (i.e. the flow value will be
358 359
    /// unbounded from above).
359 360
    ///
360 361
    /// \param map An arc map storing the upper bounds.
361 362
    /// Its \c Value type must be convertible to the \c Value type
362 363
    /// of the algorithm.
363 364
    ///
364 365
    /// \return <tt>(*this)</tt>
365 366
    template<typename UpperMap>
366 367
    CapacityScaling& upperMap(const UpperMap& map) {
367 368
      for (ArcIt a(_graph); a != INVALID; ++a) {
368 369
        _upper[_arc_idf[a]] = map[a];
369 370
      }
370 371
      return *this;
371 372
    }
372 373

	
373 374
    /// \brief Set the costs of the arcs.
374 375
    ///
375 376
    /// This function sets the costs of the arcs.
376 377
    /// If it is not used before calling \ref run(), the costs
377 378
    /// will be set to \c 1 on all arcs.
378 379
    ///
379 380
    /// \param map An arc map storing the costs.
380 381
    /// Its \c Value type must be convertible to the \c Cost type
381 382
    /// of the algorithm.
382 383
    ///
383 384
    /// \return <tt>(*this)</tt>
384 385
    template<typename CostMap>
385 386
    CapacityScaling& costMap(const CostMap& map) {
386 387
      for (ArcIt a(_graph); a != INVALID; ++a) {
387 388
        _cost[_arc_idf[a]] =  map[a];
388 389
        _cost[_arc_idb[a]] = -map[a];
389 390
      }
390 391
      return *this;
391 392
    }
392 393

	
393 394
    /// \brief Set the supply values of the nodes.
394 395
    ///
395 396
    /// This function sets the supply values of the nodes.
396 397
    /// If neither this function nor \ref stSupply() is used before
397 398
    /// calling \ref run(), the supply of each node will be set to zero.
398 399
    ///
399 400
    /// \param map A node map storing the supply values.
400 401
    /// Its \c Value type must be convertible to the \c Value type
401 402
    /// of the algorithm.
402 403
    ///
403 404
    /// \return <tt>(*this)</tt>
404 405
    template<typename SupplyMap>
405 406
    CapacityScaling& supplyMap(const SupplyMap& map) {
406 407
      for (NodeIt n(_graph); n != INVALID; ++n) {
407 408
        _supply[_node_id[n]] = map[n];
408 409
      }
409 410
      return *this;
410 411
    }
411 412

	
412 413
    /// \brief Set single source and target nodes and a supply value.
413 414
    ///
414 415
    /// This function sets a single source node and a single target node
415 416
    /// and the required flow value.
416 417
    /// If neither this function nor \ref supplyMap() is used before
417 418
    /// calling \ref run(), the supply of each node will be set to zero.
418 419
    ///
419 420
    /// Using this function has the same effect as using \ref supplyMap()
420 421
    /// with such a map in which \c k is assigned to \c s, \c -k is
421 422
    /// assigned to \c t and all other nodes have zero supply value.
422 423
    ///
423 424
    /// \param s The source node.
424 425
    /// \param t The target node.
425 426
    /// \param k The required amount of flow from node \c s to node \c t
426 427
    /// (i.e. the supply of \c s and the demand of \c t).
427 428
    ///
428 429
    /// \return <tt>(*this)</tt>
429 430
    CapacityScaling& stSupply(const Node& s, const Node& t, Value k) {
430 431
      for (int i = 0; i != _node_num; ++i) {
431 432
        _supply[i] = 0;
432 433
      }
433 434
      _supply[_node_id[s]] =  k;
434 435
      _supply[_node_id[t]] = -k;
435 436
      return *this;
436 437
    }
437 438
    
438 439
    /// @}
439 440

	
440 441
    /// \name Execution control
441 442
    /// The algorithm can be executed using \ref run().
442 443

	
443 444
    /// @{
444 445

	
445 446
    /// \brief Run the algorithm.
446 447
    ///
447 448
    /// This function runs the algorithm.
448 449
    /// The paramters can be specified using functions \ref lowerMap(),
449 450
    /// \ref upperMap(), \ref costMap(), \ref supplyMap(), \ref stSupply().
450 451
    /// For example,
451 452
    /// \code
452 453
    ///   CapacityScaling<ListDigraph> cs(graph);
453 454
    ///   cs.lowerMap(lower).upperMap(upper).costMap(cost)
454 455
    ///     .supplyMap(sup).run();
455 456
    /// \endcode
456 457
    ///
457 458
    /// This function can be called more than once. All the given parameters
458 459
    /// are kept for the next call, unless \ref resetParams() or \ref reset()
459 460
    /// is used, thus only the modified parameters have to be set again.
460 461
    /// If the underlying digraph was also modified after the construction
461 462
    /// of the class (or the last \ref reset() call), then the \ref reset()
462 463
    /// function must be called.
463 464
    ///
464 465
    /// \param factor The capacity scaling factor. It must be larger than
465 466
    /// one to use scaling. If it is less or equal to one, then scaling
466 467
    /// will be disabled.
467 468
    ///
468 469
    /// \return \c INFEASIBLE if no feasible flow exists,
469 470
    /// \n \c OPTIMAL if the problem has optimal solution
470 471
    /// (i.e. it is feasible and bounded), and the algorithm has found
471 472
    /// optimal flow and node potentials (primal and dual solutions),
472 473
    /// \n \c UNBOUNDED if the digraph contains an arc of negative cost
473 474
    /// and infinite upper bound. It means that the objective function
474 475
    /// is unbounded on that arc, however, note that it could actually be
475 476
    /// bounded over the feasible flows, but this algroithm cannot handle
476 477
    /// these cases.
477 478
    ///
478 479
    /// \see ProblemType
479 480
    /// \see resetParams(), reset()
480 481
    ProblemType run(int factor = 4) {
481 482
      _factor = factor;
482 483
      ProblemType pt = init();
483 484
      if (pt != OPTIMAL) return pt;
484 485
      return start();
485 486
    }
486 487

	
487 488
    /// \brief Reset all the parameters that have been given before.
488 489
    ///
489 490
    /// This function resets all the paramaters that have been given
490 491
    /// before using functions \ref lowerMap(), \ref upperMap(),
491 492
    /// \ref costMap(), \ref supplyMap(), \ref stSupply().
492 493
    ///
493 494
    /// It is useful for multiple \ref run() calls. Basically, all the given
494 495
    /// parameters are kept for the next \ref run() call, unless
495 496
    /// \ref resetParams() or \ref reset() is used.
496 497
    /// If the underlying digraph was also modified after the construction
497 498
    /// of the class or the last \ref reset() call, then the \ref reset()
498 499
    /// function must be used, otherwise \ref resetParams() is sufficient.
499 500
    ///
500 501
    /// For example,
501 502
    /// \code
502 503
    ///   CapacityScaling<ListDigraph> cs(graph);
503 504
    ///
504 505
    ///   // First run
505 506
    ///   cs.lowerMap(lower).upperMap(upper).costMap(cost)
506 507
    ///     .supplyMap(sup).run();
507 508
    ///
508 509
    ///   // Run again with modified cost map (resetParams() is not called,
509 510
    ///   // so only the cost map have to be set again)
510 511
    ///   cost[e] += 100;
511 512
    ///   cs.costMap(cost).run();
512 513
    ///
513 514
    ///   // Run again from scratch using resetParams()
514 515
    ///   // (the lower bounds will be set to zero on all arcs)
515 516
    ///   cs.resetParams();
516 517
    ///   cs.upperMap(capacity).costMap(cost)
517 518
    ///     .supplyMap(sup).run();
518 519
    /// \endcode
519 520
    ///
520 521
    /// \return <tt>(*this)</tt>
521 522
    ///
522 523
    /// \see reset(), run()
523 524
    CapacityScaling& resetParams() {
524 525
      for (int i = 0; i != _node_num; ++i) {
525 526
        _supply[i] = 0;
526 527
      }
527 528
      for (int j = 0; j != _res_arc_num; ++j) {
528 529
        _lower[j] = 0;
529 530
        _upper[j] = INF;
530 531
        _cost[j] = _forward[j] ? 1 : -1;
531 532
      }
532 533
      _have_lower = false;
533 534
      return *this;
534 535
    }
535 536

	
536 537
    /// \brief Reset the internal data structures and all the parameters
537 538
    /// that have been given before.
538 539
    ///
539 540
    /// This function resets the internal data structures and all the
540 541
    /// paramaters that have been given before using functions \ref lowerMap(),
541 542
    /// \ref upperMap(), \ref costMap(), \ref supplyMap(), \ref stSupply().
542 543
    ///
543 544
    /// It is useful for multiple \ref run() calls. Basically, all the given
544 545
    /// parameters are kept for the next \ref run() call, unless
545 546
    /// \ref resetParams() or \ref reset() is used.
546 547
    /// If the underlying digraph was also modified after the construction
547 548
    /// of the class or the last \ref reset() call, then the \ref reset()
548 549
    /// function must be used, otherwise \ref resetParams() is sufficient.
549 550
    ///
550 551
    /// See \ref resetParams() for examples.
551 552
    ///
552 553
    /// \return <tt>(*this)</tt>
553 554
    ///
554 555
    /// \see resetParams(), run()
555 556
    CapacityScaling& reset() {
556 557
      // Resize vectors
557 558
      _node_num = countNodes(_graph);
558 559
      _arc_num = countArcs(_graph);
559 560
      _res_arc_num = 2 * (_arc_num + _node_num);
560 561
      _root = _node_num;
561 562
      ++_node_num;
562 563

	
563 564
      _first_out.resize(_node_num + 1);
564 565
      _forward.resize(_res_arc_num);
565 566
      _source.resize(_res_arc_num);
566 567
      _target.resize(_res_arc_num);
567 568
      _reverse.resize(_res_arc_num);
568 569

	
569 570
      _lower.resize(_res_arc_num);
570 571
      _upper.resize(_res_arc_num);
571 572
      _cost.resize(_res_arc_num);
572 573
      _supply.resize(_node_num);
573 574
      
574 575
      _res_cap.resize(_res_arc_num);
575 576
      _pi.resize(_node_num);
576 577
      _excess.resize(_node_num);
577 578
      _pred.resize(_node_num);
578 579

	
579 580
      // Copy the graph
580 581
      int i = 0, j = 0, k = 2 * _arc_num + _node_num - 1;
581 582
      for (NodeIt n(_graph); n != INVALID; ++n, ++i) {
582 583
        _node_id[n] = i;
583 584
      }
584 585
      i = 0;
585 586
      for (NodeIt n(_graph); n != INVALID; ++n, ++i) {
586 587
        _first_out[i] = j;
587 588
        for (OutArcIt a(_graph, n); a != INVALID; ++a, ++j) {
588 589
          _arc_idf[a] = j;
589 590
          _forward[j] = true;
590 591
          _source[j] = i;
591 592
          _target[j] = _node_id[_graph.runningNode(a)];
592 593
        }
593 594
        for (InArcIt a(_graph, n); a != INVALID; ++a, ++j) {
594 595
          _arc_idb[a] = j;
595 596
          _forward[j] = false;
596 597
          _source[j] = i;
597 598
          _target[j] = _node_id[_graph.runningNode(a)];
598 599
        }
599 600
        _forward[j] = false;
600 601
        _source[j] = i;
601 602
        _target[j] = _root;
602 603
        _reverse[j] = k;
603 604
        _forward[k] = true;
604 605
        _source[k] = _root;
605 606
        _target[k] = i;
606 607
        _reverse[k] = j;
607 608
        ++j; ++k;
608 609
      }
609 610
      _first_out[i] = j;
610 611
      _first_out[_node_num] = k;
611 612
      for (ArcIt a(_graph); a != INVALID; ++a) {
612 613
        int fi = _arc_idf[a];
613 614
        int bi = _arc_idb[a];
614 615
        _reverse[fi] = bi;
615 616
        _reverse[bi] = fi;
616 617
      }
617 618
      
618 619
      // Reset parameters
619 620
      resetParams();
620 621
      return *this;
621 622
    }
622 623

	
623 624
    /// @}
624 625

	
625 626
    /// \name Query Functions
626 627
    /// The results of the algorithm can be obtained using these
627 628
    /// functions.\n
628 629
    /// The \ref run() function must be called before using them.
629 630

	
630 631
    /// @{
631 632

	
632 633
    /// \brief Return the total cost of the found flow.
633 634
    ///
634 635
    /// This function returns the total cost of the found flow.
635 636
    /// Its complexity is O(e).
636 637
    ///
637 638
    /// \note The return type of the function can be specified as a
638 639
    /// template parameter. For example,
639 640
    /// \code
640 641
    ///   cs.totalCost<double>();
641 642
    /// \endcode
642 643
    /// It is useful if the total cost cannot be stored in the \c Cost
643 644
    /// type of the algorithm, which is the default return type of the
644 645
    /// function.
645 646
    ///
646 647
    /// \pre \ref run() must be called before using this function.
647 648
    template <typename Number>
648 649
    Number totalCost() const {
649 650
      Number c = 0;
650 651
      for (ArcIt a(_graph); a != INVALID; ++a) {
651 652
        int i = _arc_idb[a];
652 653
        c += static_cast<Number>(_res_cap[i]) *
653 654
             (-static_cast<Number>(_cost[i]));
654 655
      }
655 656
      return c;
656 657
    }
657 658

	
658 659
#ifndef DOXYGEN
659 660
    Cost totalCost() const {
660 661
      return totalCost<Cost>();
661 662
    }
662 663
#endif
663 664

	
664 665
    /// \brief Return the flow on the given arc.
665 666
    ///
666 667
    /// This function returns the flow on the given arc.
667 668
    ///
668 669
    /// \pre \ref run() must be called before using this function.
669 670
    Value flow(const Arc& a) const {
670 671
      return _res_cap[_arc_idb[a]];
671 672
    }
672 673

	
673 674
    /// \brief Return the flow map (the primal solution).
674 675
    ///
675 676
    /// This function copies the flow value on each arc into the given
676 677
    /// map. The \c Value type of the algorithm must be convertible to
677 678
    /// the \c Value type of the map.
678 679
    ///
679 680
    /// \pre \ref run() must be called before using this function.
680 681
    template <typename FlowMap>
681 682
    void flowMap(FlowMap &map) const {
682 683
      for (ArcIt a(_graph); a != INVALID; ++a) {
683 684
        map.set(a, _res_cap[_arc_idb[a]]);
684 685
      }
685 686
    }
686 687

	
687 688
    /// \brief Return the potential (dual value) of the given node.
688 689
    ///
689 690
    /// This function returns the potential (dual value) of the
690 691
    /// given node.
691 692
    ///
692 693
    /// \pre \ref run() must be called before using this function.
693 694
    Cost potential(const Node& n) const {
694 695
      return _pi[_node_id[n]];
695 696
    }
696 697

	
697 698
    /// \brief Return the potential map (the dual solution).
698 699
    ///
699 700
    /// This function copies the potential (dual value) of each node
700 701
    /// into the given map.
701 702
    /// The \c Cost type of the algorithm must be convertible to the
702 703
    /// \c Value type of the map.
703 704
    ///
704 705
    /// \pre \ref run() must be called before using this function.
705 706
    template <typename PotentialMap>
706 707
    void potentialMap(PotentialMap &map) const {
707 708
      for (NodeIt n(_graph); n != INVALID; ++n) {
708 709
        map.set(n, _pi[_node_id[n]]);
709 710
      }
710 711
    }
711 712

	
712 713
    /// @}
713 714

	
714 715
  private:
715 716

	
716 717
    // Initialize the algorithm
717 718
    ProblemType init() {
718 719
      if (_node_num <= 1) return INFEASIBLE;
719 720

	
720 721
      // Check the sum of supply values
721 722
      _sum_supply = 0;
722 723
      for (int i = 0; i != _root; ++i) {
723 724
        _sum_supply += _supply[i];
724 725
      }
725 726
      if (_sum_supply > 0) return INFEASIBLE;
726 727
      
727 728
      // Initialize vectors
728 729
      for (int i = 0; i != _root; ++i) {
729 730
        _pi[i] = 0;
730 731
        _excess[i] = _supply[i];
731 732
      }
732 733

	
733 734
      // Remove non-zero lower bounds
734 735
      const Value MAX = std::numeric_limits<Value>::max();
735 736
      int last_out;
736 737
      if (_have_lower) {
737 738
        for (int i = 0; i != _root; ++i) {
738 739
          last_out = _first_out[i+1];
739 740
          for (int j = _first_out[i]; j != last_out; ++j) {
740 741
            if (_forward[j]) {
741 742
              Value c = _lower[j];
742 743
              if (c >= 0) {
743 744
                _res_cap[j] = _upper[j] < MAX ? _upper[j] - c : INF;
744 745
              } else {
745 746
                _res_cap[j] = _upper[j] < MAX + c ? _upper[j] - c : INF;
746 747
              }
747 748
              _excess[i] -= c;
748 749
              _excess[_target[j]] += c;
749 750
            } else {
750 751
              _res_cap[j] = 0;
751 752
            }
752 753
          }
753 754
        }
754 755
      } else {
755 756
        for (int j = 0; j != _res_arc_num; ++j) {
756 757
          _res_cap[j] = _forward[j] ? _upper[j] : 0;
757 758
        }
758 759
      }
759 760

	
760 761
      // Handle negative costs
761 762
      for (int i = 0; i != _root; ++i) {
762 763
        last_out = _first_out[i+1] - 1;
763 764
        for (int j = _first_out[i]; j != last_out; ++j) {
764 765
          Value rc = _res_cap[j];
765 766
          if (_cost[j] < 0 && rc > 0) {
766 767
            if (rc >= MAX) return UNBOUNDED;
767 768
            _excess[i] -= rc;
768 769
            _excess[_target[j]] += rc;
769 770
            _res_cap[j] = 0;
770 771
            _res_cap[_reverse[j]] += rc;
771 772
          }
772 773
        }
773 774
      }
774 775
      
775 776
      // Handle GEQ supply type
776 777
      if (_sum_supply < 0) {
777 778
        _pi[_root] = 0;
778 779
        _excess[_root] = -_sum_supply;
779 780
        for (int a = _first_out[_root]; a != _res_arc_num; ++a) {
780 781
          int ra = _reverse[a];
781 782
          _res_cap[a] = -_sum_supply + 1;
782 783
          _res_cap[ra] = 0;
783 784
          _cost[a] = 0;
784 785
          _cost[ra] = 0;
785 786
        }
786 787
      } else {
787 788
        _pi[_root] = 0;
788 789
        _excess[_root] = 0;
789 790
        for (int a = _first_out[_root]; a != _res_arc_num; ++a) {
790 791
          int ra = _reverse[a];
791 792
          _res_cap[a] = 1;
792 793
          _res_cap[ra] = 0;
793 794
          _cost[a] = 0;
794 795
          _cost[ra] = 0;
795 796
        }
796 797
      }
797 798

	
798 799
      // Initialize delta value
799 800
      if (_factor > 1) {
800 801
        // With scaling
801
        Value max_sup = 0, max_dem = 0;
802
        for (int i = 0; i != _node_num; ++i) {
802
        Value max_sup = 0, max_dem = 0, max_cap = 0;
803
        for (int i = 0; i != _root; ++i) {
803 804
          Value ex = _excess[i];
804 805
          if ( ex > max_sup) max_sup =  ex;
805 806
          if (-ex > max_dem) max_dem = -ex;
806
        }
807
        Value max_cap = 0;
808
        for (int j = 0; j != _res_arc_num; ++j) {
809
          if (_res_cap[j] > max_cap) max_cap = _res_cap[j];
807
          int last_out = _first_out[i+1] - 1;
808
          for (int j = _first_out[i]; j != last_out; ++j) {
809
            if (_res_cap[j] > max_cap) max_cap = _res_cap[j];
810
          }
810 811
        }
811 812
        max_sup = std::min(std::min(max_sup, max_dem), max_cap);
812 813
        for (_delta = 1; 2 * _delta <= max_sup; _delta *= 2) ;
813 814
      } else {
814 815
        // Without scaling
815 816
        _delta = 1;
816 817
      }
817 818

	
818 819
      return OPTIMAL;
819 820
    }
820 821

	
821 822
    ProblemType start() {
822 823
      // Execute the algorithm
823 824
      ProblemType pt;
824 825
      if (_delta > 1)
825 826
        pt = startWithScaling();
826 827
      else
827 828
        pt = startWithoutScaling();
828 829

	
829 830
      // Handle non-zero lower bounds
830 831
      if (_have_lower) {
831 832
        int limit = _first_out[_root];
832 833
        for (int j = 0; j != limit; ++j) {
833 834
          if (!_forward[j]) _res_cap[j] += _lower[j];
834 835
        }
835 836
      }
836 837

	
837 838
      // Shift potentials if necessary
838 839
      Cost pr = _pi[_root];
839 840
      if (_sum_supply < 0 || pr > 0) {
840 841
        for (int i = 0; i != _node_num; ++i) {
841 842
          _pi[i] -= pr;
842 843
        }        
843 844
      }
844 845
      
845 846
      return pt;
846 847
    }
847 848

	
848 849
    // Execute the capacity scaling algorithm
849 850
    ProblemType startWithScaling() {
850 851
      // Perform capacity scaling phases
851 852
      int s, t;
852 853
      ResidualDijkstra _dijkstra(*this);
853 854
      while (true) {
854 855
        // Saturate all arcs not satisfying the optimality condition
855 856
        int last_out;
856 857
        for (int u = 0; u != _node_num; ++u) {
857 858
          last_out = _sum_supply < 0 ?
858 859
            _first_out[u+1] : _first_out[u+1] - 1;
859 860
          for (int a = _first_out[u]; a != last_out; ++a) {
860 861
            int v = _target[a];
861 862
            Cost c = _cost[a] + _pi[u] - _pi[v];
862 863
            Value rc = _res_cap[a];
863 864
            if (c < 0 && rc >= _delta) {
864 865
              _excess[u] -= rc;
865 866
              _excess[v] += rc;
866 867
              _res_cap[a] = 0;
867 868
              _res_cap[_reverse[a]] += rc;
868 869
            }
869 870
          }
870 871
        }
871 872

	
872 873
        // Find excess nodes and deficit nodes
873 874
        _excess_nodes.clear();
874 875
        _deficit_nodes.clear();
875 876
        for (int u = 0; u != _node_num; ++u) {
876 877
          Value ex = _excess[u];
877 878
          if (ex >=  _delta) _excess_nodes.push_back(u);
878 879
          if (ex <= -_delta) _deficit_nodes.push_back(u);
879 880
        }
880 881
        int next_node = 0, next_def_node = 0;
881 882

	
882 883
        // Find augmenting shortest paths
883 884
        while (next_node < int(_excess_nodes.size())) {
884 885
          // Check deficit nodes
885 886
          if (_delta > 1) {
886 887
            bool delta_deficit = false;
887 888
            for ( ; next_def_node < int(_deficit_nodes.size());
888 889
                    ++next_def_node ) {
889 890
              if (_excess[_deficit_nodes[next_def_node]] <= -_delta) {
890 891
                delta_deficit = true;
891 892
                break;
892 893
              }
893 894
            }
894 895
            if (!delta_deficit) break;
895 896
          }
896 897

	
897 898
          // Run Dijkstra in the residual network
898 899
          s = _excess_nodes[next_node];
899 900
          if ((t = _dijkstra.run(s, _delta)) == -1) {
900 901
            if (_delta > 1) {
901 902
              ++next_node;
902 903
              continue;
903 904
            }
904 905
            return INFEASIBLE;
905 906
          }
906 907

	
907 908
          // Augment along a shortest path from s to t
908 909
          Value d = std::min(_excess[s], -_excess[t]);
909 910
          int u = t;
910 911
          int a;
911 912
          if (d > _delta) {
912 913
            while ((a = _pred[u]) != -1) {
913 914
              if (_res_cap[a] < d) d = _res_cap[a];
914 915
              u = _source[a];
915 916
            }
916 917
          }
917 918
          u = t;
918 919
          while ((a = _pred[u]) != -1) {
919 920
            _res_cap[a] -= d;
920 921
            _res_cap[_reverse[a]] += d;
921 922
            u = _source[a];
922 923
          }
923 924
          _excess[s] -= d;
924 925
          _excess[t] += d;
925 926

	
926 927
          if (_excess[s] < _delta) ++next_node;
927 928
        }
928 929

	
929 930
        if (_delta == 1) break;
930 931
        _delta = _delta <= _factor ? 1 : _delta / _factor;
931 932
      }
932 933

	
933 934
      return OPTIMAL;
934 935
    }
935 936

	
936 937
    // Execute the successive shortest path algorithm
937 938
    ProblemType startWithoutScaling() {
938 939
      // Find excess nodes
939 940
      _excess_nodes.clear();
940 941
      for (int i = 0; i != _node_num; ++i) {
941 942
        if (_excess[i] > 0) _excess_nodes.push_back(i);
942 943
      }
943 944
      if (_excess_nodes.size() == 0) return OPTIMAL;
944 945
      int next_node = 0;
945 946

	
946 947
      // Find shortest paths
947 948
      int s, t;
948 949
      ResidualDijkstra _dijkstra(*this);
949 950
      while ( _excess[_excess_nodes[next_node]] > 0 ||
950 951
              ++next_node < int(_excess_nodes.size()) )
951 952
      {
952 953
        // Run Dijkstra in the residual network
953 954
        s = _excess_nodes[next_node];
954 955
        if ((t = _dijkstra.run(s)) == -1) return INFEASIBLE;
955 956

	
956 957
        // Augment along a shortest path from s to t
957 958
        Value d = std::min(_excess[s], -_excess[t]);
958 959
        int u = t;
959 960
        int a;
960 961
        if (d > 1) {
961 962
          while ((a = _pred[u]) != -1) {
962 963
            if (_res_cap[a] < d) d = _res_cap[a];
963 964
            u = _source[a];
964 965
          }
965 966
        }
966 967
        u = t;
967 968
        while ((a = _pred[u]) != -1) {
968 969
          _res_cap[a] -= d;
969 970
          _res_cap[_reverse[a]] += d;
970 971
          u = _source[a];
971 972
        }
972 973
        _excess[s] -= d;
973 974
        _excess[t] += d;
974 975
      }
975 976

	
976 977
      return OPTIMAL;
977 978
    }
978 979

	
979 980
  }; //class CapacityScaling
980 981

	
981 982
  ///@}
982 983

	
983 984
} //namespace lemon
984 985

	
985 986
#endif //LEMON_CAPACITY_SCALING_H
Ignore white space 6 line context
1 1
/* -*- C++ -*-
2 2
 *
3 3
 * This file is a part of LEMON, a generic C++ optimization library
4 4
 *
5 5
 * Copyright (C) 2003-2008
6 6
 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
7 7
 * (Egervary Research Group on Combinatorial Optimization, EGRES).
8 8
 *
9 9
 * Permission to use, modify and distribute this software is granted
10 10
 * provided that this copyright notice appears in all copies. For
11 11
 * precise terms see the accompanying LICENSE file.
12 12
 *
13 13
 * This software is provided "AS IS" with no warranty of any kind,
14 14
 * express or implied, and with no claim as to its suitability for any
15 15
 * purpose.
16 16
 *
17 17
 */
18 18

	
19 19
#ifndef LEMON_COST_SCALING_H
20 20
#define LEMON_COST_SCALING_H
21 21

	
22 22
/// \ingroup min_cost_flow_algs
23 23
/// \file
24 24
/// \brief Cost scaling algorithm for finding a minimum cost flow.
25 25

	
26 26
#include <vector>
27 27
#include <deque>
28 28
#include <limits>
29 29

	
30 30
#include <lemon/core.h>
31 31
#include <lemon/maps.h>
32 32
#include <lemon/math.h>
33 33
#include <lemon/static_graph.h>
34 34
#include <lemon/circulation.h>
35 35
#include <lemon/bellman_ford.h>
36 36

	
37 37
namespace lemon {
38 38

	
39 39
  /// \brief Default traits class of CostScaling algorithm.
40 40
  ///
41 41
  /// Default traits class of CostScaling algorithm.
42 42
  /// \tparam GR Digraph type.
43 43
  /// \tparam V The number type used for flow amounts, capacity bounds
44 44
  /// and supply values. By default it is \c int.
45 45
  /// \tparam C The number type used for costs and potentials.
46 46
  /// By default it is the same as \c V.
47 47
#ifdef DOXYGEN
48 48
  template <typename GR, typename V = int, typename C = V>
49 49
#else
50 50
  template < typename GR, typename V = int, typename C = V,
51 51
             bool integer = std::numeric_limits<C>::is_integer >
52 52
#endif
53 53
  struct CostScalingDefaultTraits
54 54
  {
55 55
    /// The type of the digraph
56 56
    typedef GR Digraph;
57 57
    /// The type of the flow amounts, capacity bounds and supply values
58 58
    typedef V Value;
59 59
    /// The type of the arc costs
60 60
    typedef C Cost;
61 61

	
62 62
    /// \brief The large cost type used for internal computations
63 63
    ///
64 64
    /// The large cost type used for internal computations.
65 65
    /// It is \c long \c long if the \c Cost type is integer,
66 66
    /// otherwise it is \c double.
67 67
    /// \c Cost must be convertible to \c LargeCost.
68 68
    typedef double LargeCost;
69 69
  };
70 70

	
71 71
  // Default traits class for integer cost types
72 72
  template <typename GR, typename V, typename C>
73 73
  struct CostScalingDefaultTraits<GR, V, C, true>
74 74
  {
75 75
    typedef GR Digraph;
76 76
    typedef V Value;
77 77
    typedef C Cost;
78 78
#ifdef LEMON_HAVE_LONG_LONG
79 79
    typedef long long LargeCost;
80 80
#else
81 81
    typedef long LargeCost;
82 82
#endif
83 83
  };
84 84

	
85 85

	
86 86
  /// \addtogroup min_cost_flow_algs
87 87
  /// @{
88 88

	
89 89
  /// \brief Implementation of the Cost Scaling algorithm for
90 90
  /// finding a \ref min_cost_flow "minimum cost flow".
91 91
  ///
92 92
  /// \ref CostScaling implements a cost scaling algorithm that performs
93 93
  /// push/augment and relabel operations for finding a \ref min_cost_flow
94 94
  /// "minimum cost flow" \ref amo93networkflows, \ref goldberg90approximation,
95 95
  /// \ref goldberg97efficient, \ref bunnagel98efficient. 
96 96
  /// It is a highly efficient primal-dual solution method, which
97 97
  /// can be viewed as the generalization of the \ref Preflow
98 98
  /// "preflow push-relabel" algorithm for the maximum flow problem.
99 99
  ///
100 100
  /// Most of the parameters of the problem (except for the digraph)
101 101
  /// can be given using separate functions, and the algorithm can be
102 102
  /// executed using the \ref run() function. If some parameters are not
103 103
  /// specified, then default values will be used.
104 104
  ///
105 105
  /// \tparam GR The digraph type the algorithm runs on.
106 106
  /// \tparam V The number type used for flow amounts, capacity bounds
107 107
  /// and supply values in the algorithm. By default, it is \c int.
108 108
  /// \tparam C The number type used for costs and potentials in the
109 109
  /// algorithm. By default, it is the same as \c V.
110 110
  /// \tparam TR The traits class that defines various types used by the
111 111
  /// algorithm. By default, it is \ref CostScalingDefaultTraits
112 112
  /// "CostScalingDefaultTraits<GR, V, C>".
113 113
  /// In most cases, this parameter should not be set directly,
114 114
  /// consider to use the named template parameters instead.
115 115
  ///
116 116
  /// \warning Both number types must be signed and all input data must
117 117
  /// be integer.
118 118
  /// \warning This algorithm does not support negative costs for such
119 119
  /// arcs that have infinite upper bound.
120 120
  ///
121 121
  /// \note %CostScaling provides three different internal methods,
122 122
  /// from which the most efficient one is used by default.
123 123
  /// For more information, see \ref Method.
124 124
#ifdef DOXYGEN
125 125
  template <typename GR, typename V, typename C, typename TR>
126 126
#else
127 127
  template < typename GR, typename V = int, typename C = V,
128 128
             typename TR = CostScalingDefaultTraits<GR, V, C> >
129 129
#endif
130 130
  class CostScaling
131 131
  {
132 132
  public:
133 133

	
134 134
    /// The type of the digraph
135 135
    typedef typename TR::Digraph Digraph;
136 136
    /// The type of the flow amounts, capacity bounds and supply values
137 137
    typedef typename TR::Value Value;
138 138
    /// The type of the arc costs
139 139
    typedef typename TR::Cost Cost;
140 140

	
141 141
    /// \brief The large cost type
142 142
    ///
143 143
    /// The large cost type used for internal computations.
144 144
    /// By default, it is \c long \c long if the \c Cost type is integer,
145 145
    /// otherwise it is \c double.
146 146
    typedef typename TR::LargeCost LargeCost;
147 147

	
148 148
    /// The \ref CostScalingDefaultTraits "traits class" of the algorithm
149 149
    typedef TR Traits;
150 150

	
151 151
  public:
152 152

	
153 153
    /// \brief Problem type constants for the \c run() function.
154 154
    ///
155 155
    /// Enum type containing the problem type constants that can be
156 156
    /// returned by the \ref run() function of the algorithm.
157 157
    enum ProblemType {
158 158
      /// The problem has no feasible solution (flow).
159 159
      INFEASIBLE,
160 160
      /// The problem has optimal solution (i.e. it is feasible and
161 161
      /// bounded), and the algorithm has found optimal flow and node
162 162
      /// potentials (primal and dual solutions).
163 163
      OPTIMAL,
164 164
      /// The digraph contains an arc of negative cost and infinite
165 165
      /// upper bound. It means that the objective function is unbounded
166 166
      /// on that arc, however, note that it could actually be bounded
167 167
      /// over the feasible flows, but this algroithm cannot handle
168 168
      /// these cases.
169 169
      UNBOUNDED
170 170
    };
171 171

	
172 172
    /// \brief Constants for selecting the internal method.
173 173
    ///
174 174
    /// Enum type containing constants for selecting the internal method
175 175
    /// for the \ref run() function.
176 176
    ///
177 177
    /// \ref CostScaling provides three internal methods that differ mainly
178 178
    /// in their base operations, which are used in conjunction with the
179 179
    /// relabel operation.
180 180
    /// By default, the so called \ref PARTIAL_AUGMENT
181 181
    /// "Partial Augment-Relabel" method is used, which proved to be
182 182
    /// the most efficient and the most robust on various test inputs.
183 183
    /// However, the other methods can be selected using the \ref run()
184 184
    /// function with the proper parameter.
185 185
    enum Method {
186 186
      /// Local push operations are used, i.e. flow is moved only on one
187 187
      /// admissible arc at once.
188 188
      PUSH,
189 189
      /// Augment operations are used, i.e. flow is moved on admissible
190 190
      /// paths from a node with excess to a node with deficit.
191 191
      AUGMENT,
192 192
      /// Partial augment operations are used, i.e. flow is moved on 
193 193
      /// admissible paths started from a node with excess, but the
194 194
      /// lengths of these paths are limited. This method can be viewed
195 195
      /// as a combined version of the previous two operations.
196 196
      PARTIAL_AUGMENT
197 197
    };
198 198

	
199 199
  private:
200 200

	
201 201
    TEMPLATE_DIGRAPH_TYPEDEFS(GR);
202 202

	
203 203
    typedef std::vector<int> IntVector;
204
    typedef std::vector<char> BoolVector;
205 204
    typedef std::vector<Value> ValueVector;
206 205
    typedef std::vector<Cost> CostVector;
207 206
    typedef std::vector<LargeCost> LargeCostVector;
207
    typedef std::vector<char> BoolVector;
208
    // Note: vector<char> is used instead of vector<bool> for efficiency reasons
208 209

	
209 210
  private:
210 211
  
211 212
    template <typename KT, typename VT>
212 213
    class StaticVectorMap {
213 214
    public:
214 215
      typedef KT Key;
215 216
      typedef VT Value;
216 217
      
217 218
      StaticVectorMap(std::vector<Value>& v) : _v(v) {}
218 219
      
219 220
      const Value& operator[](const Key& key) const {
220 221
        return _v[StaticDigraph::id(key)];
221 222
      }
222 223

	
223 224
      Value& operator[](const Key& key) {
224 225
        return _v[StaticDigraph::id(key)];
225 226
      }
226 227
      
227 228
      void set(const Key& key, const Value& val) {
228 229
        _v[StaticDigraph::id(key)] = val;
229 230
      }
230 231

	
231 232
    private:
232 233
      std::vector<Value>& _v;
233 234
    };
234 235

	
235 236
    typedef StaticVectorMap<StaticDigraph::Node, LargeCost> LargeCostNodeMap;
236 237
    typedef StaticVectorMap<StaticDigraph::Arc, LargeCost> LargeCostArcMap;
237 238

	
238 239
  private:
239 240

	
240 241
    // Data related to the underlying digraph
241 242
    const GR &_graph;
242 243
    int _node_num;
243 244
    int _arc_num;
244 245
    int _res_node_num;
245 246
    int _res_arc_num;
246 247
    int _root;
247 248

	
248 249
    // Parameters of the problem
249 250
    bool _have_lower;
250 251
    Value _sum_supply;
252
    int _sup_node_num;
251 253

	
252 254
    // Data structures for storing the digraph
253 255
    IntNodeMap _node_id;
254 256
    IntArcMap _arc_idf;
255 257
    IntArcMap _arc_idb;
256 258
    IntVector _first_out;
257 259
    BoolVector _forward;
258 260
    IntVector _source;
259 261
    IntVector _target;
260 262
    IntVector _reverse;
261 263

	
262 264
    // Node and arc data
263 265
    ValueVector _lower;
264 266
    ValueVector _upper;
265 267
    CostVector _scost;
266 268
    ValueVector _supply;
267 269

	
268 270
    ValueVector _res_cap;
269 271
    LargeCostVector _cost;
270 272
    LargeCostVector _pi;
271 273
    ValueVector _excess;
272 274
    IntVector _next_out;
273 275
    std::deque<int> _active_nodes;
274 276

	
275 277
    // Data for scaling
276 278
    LargeCost _epsilon;
277 279
    int _alpha;
278 280

	
281
    IntVector _buckets;
282
    IntVector _bucket_next;
283
    IntVector _bucket_prev;
284
    IntVector _rank;
285
    int _max_rank;
286
  
279 287
    // Data for a StaticDigraph structure
280 288
    typedef std::pair<int, int> IntPair;
281 289
    StaticDigraph _sgr;
282 290
    std::vector<IntPair> _arc_vec;
283 291
    std::vector<LargeCost> _cost_vec;
284 292
    LargeCostArcMap _cost_map;
285 293
    LargeCostNodeMap _pi_map;
286 294
  
287 295
  public:
288 296
  
289 297
    /// \brief Constant for infinite upper bounds (capacities).
290 298
    ///
291 299
    /// Constant for infinite upper bounds (capacities).
292 300
    /// It is \c std::numeric_limits<Value>::infinity() if available,
293 301
    /// \c std::numeric_limits<Value>::max() otherwise.
294 302
    const Value INF;
295 303

	
296 304
  public:
297 305

	
298 306
    /// \name Named Template Parameters
299 307
    /// @{
300 308

	
301 309
    template <typename T>
302 310
    struct SetLargeCostTraits : public Traits {
303 311
      typedef T LargeCost;
304 312
    };
305 313

	
306 314
    /// \brief \ref named-templ-param "Named parameter" for setting
307 315
    /// \c LargeCost type.
308 316
    ///
309 317
    /// \ref named-templ-param "Named parameter" for setting \c LargeCost
310 318
    /// type, which is used for internal computations in the algorithm.
311 319
    /// \c Cost must be convertible to \c LargeCost.
312 320
    template <typename T>
313 321
    struct SetLargeCost
314 322
      : public CostScaling<GR, V, C, SetLargeCostTraits<T> > {
315 323
      typedef  CostScaling<GR, V, C, SetLargeCostTraits<T> > Create;
316 324
    };
317 325

	
318 326
    /// @}
319 327

	
320 328
  public:
321 329

	
322 330
    /// \brief Constructor.
323 331
    ///
324 332
    /// The constructor of the class.
325 333
    ///
326 334
    /// \param graph The digraph the algorithm runs on.
327 335
    CostScaling(const GR& graph) :
328 336
      _graph(graph), _node_id(graph), _arc_idf(graph), _arc_idb(graph),
329 337
      _cost_map(_cost_vec), _pi_map(_pi),
330 338
      INF(std::numeric_limits<Value>::has_infinity ?
331 339
          std::numeric_limits<Value>::infinity() :
332 340
          std::numeric_limits<Value>::max())
333 341
    {
334 342
      // Check the number types
335 343
      LEMON_ASSERT(std::numeric_limits<Value>::is_signed,
336 344
        "The flow type of CostScaling must be signed");
337 345
      LEMON_ASSERT(std::numeric_limits<Cost>::is_signed,
338 346
        "The cost type of CostScaling must be signed");
339 347
      
340 348
      // Reset data structures
341 349
      reset();
342 350
    }
343 351

	
344 352
    /// \name Parameters
345 353
    /// The parameters of the algorithm can be specified using these
346 354
    /// functions.
347 355

	
348 356
    /// @{
349 357

	
350 358
    /// \brief Set the lower bounds on the arcs.
351 359
    ///
352 360
    /// This function sets the lower bounds on the arcs.
353 361
    /// If it is not used before calling \ref run(), the lower bounds
354 362
    /// will be set to zero on all arcs.
355 363
    ///
356 364
    /// \param map An arc map storing the lower bounds.
357 365
    /// Its \c Value type must be convertible to the \c Value type
358 366
    /// of the algorithm.
359 367
    ///
360 368
    /// \return <tt>(*this)</tt>
361 369
    template <typename LowerMap>
362 370
    CostScaling& lowerMap(const LowerMap& map) {
363 371
      _have_lower = true;
364 372
      for (ArcIt a(_graph); a != INVALID; ++a) {
365 373
        _lower[_arc_idf[a]] = map[a];
366 374
        _lower[_arc_idb[a]] = map[a];
367 375
      }
368 376
      return *this;
369 377
    }
370 378

	
371 379
    /// \brief Set the upper bounds (capacities) on the arcs.
372 380
    ///
373 381
    /// This function sets the upper bounds (capacities) on the arcs.
374 382
    /// If it is not used before calling \ref run(), the upper bounds
375 383
    /// will be set to \ref INF on all arcs (i.e. the flow value will be
376 384
    /// unbounded from above).
377 385
    ///
378 386
    /// \param map An arc map storing the upper bounds.
379 387
    /// Its \c Value type must be convertible to the \c Value type
380 388
    /// of the algorithm.
381 389
    ///
382 390
    /// \return <tt>(*this)</tt>
383 391
    template<typename UpperMap>
384 392
    CostScaling& upperMap(const UpperMap& map) {
385 393
      for (ArcIt a(_graph); a != INVALID; ++a) {
386 394
        _upper[_arc_idf[a]] = map[a];
387 395
      }
388 396
      return *this;
389 397
    }
390 398

	
391 399
    /// \brief Set the costs of the arcs.
392 400
    ///
393 401
    /// This function sets the costs of the arcs.
394 402
    /// If it is not used before calling \ref run(), the costs
395 403
    /// will be set to \c 1 on all arcs.
396 404
    ///
397 405
    /// \param map An arc map storing the costs.
398 406
    /// Its \c Value type must be convertible to the \c Cost type
399 407
    /// of the algorithm.
400 408
    ///
401 409
    /// \return <tt>(*this)</tt>
402 410
    template<typename CostMap>
403 411
    CostScaling& costMap(const CostMap& map) {
404 412
      for (ArcIt a(_graph); a != INVALID; ++a) {
405 413
        _scost[_arc_idf[a]] =  map[a];
406 414
        _scost[_arc_idb[a]] = -map[a];
407 415
      }
408 416
      return *this;
409 417
    }
410 418

	
411 419
    /// \brief Set the supply values of the nodes.
412 420
    ///
413 421
    /// This function sets the supply values of the nodes.
414 422
    /// If neither this function nor \ref stSupply() is used before
415 423
    /// calling \ref run(), the supply of each node will be set to zero.
416 424
    ///
417 425
    /// \param map A node map storing the supply values.
418 426
    /// Its \c Value type must be convertible to the \c Value type
419 427
    /// of the algorithm.
420 428
    ///
421 429
    /// \return <tt>(*this)</tt>
422 430
    template<typename SupplyMap>
423 431
    CostScaling& supplyMap(const SupplyMap& map) {
424 432
      for (NodeIt n(_graph); n != INVALID; ++n) {
425 433
        _supply[_node_id[n]] = map[n];
426 434
      }
427 435
      return *this;
428 436
    }
429 437

	
430 438
    /// \brief Set single source and target nodes and a supply value.
431 439
    ///
432 440
    /// This function sets a single source node and a single target node
433 441
    /// and the required flow value.
434 442
    /// If neither this function nor \ref supplyMap() is used before
435 443
    /// calling \ref run(), the supply of each node will be set to zero.
436 444
    ///
437 445
    /// Using this function has the same effect as using \ref supplyMap()
438 446
    /// with such a map in which \c k is assigned to \c s, \c -k is
439 447
    /// assigned to \c t and all other nodes have zero supply value.
440 448
    ///
441 449
    /// \param s The source node.
442 450
    /// \param t The target node.
443 451
    /// \param k The required amount of flow from node \c s to node \c t
444 452
    /// (i.e. the supply of \c s and the demand of \c t).
445 453
    ///
446 454
    /// \return <tt>(*this)</tt>
447 455
    CostScaling& stSupply(const Node& s, const Node& t, Value k) {
448 456
      for (int i = 0; i != _res_node_num; ++i) {
449 457
        _supply[i] = 0;
450 458
      }
451 459
      _supply[_node_id[s]] =  k;
452 460
      _supply[_node_id[t]] = -k;
453 461
      return *this;
454 462
    }
455 463
    
456 464
    /// @}
457 465

	
458 466
    /// \name Execution control
459 467
    /// The algorithm can be executed using \ref run().
460 468

	
461 469
    /// @{
462 470

	
463 471
    /// \brief Run the algorithm.
464 472
    ///
465 473
    /// This function runs the algorithm.
466 474
    /// The paramters can be specified using functions \ref lowerMap(),
467 475
    /// \ref upperMap(), \ref costMap(), \ref supplyMap(), \ref stSupply().
468 476
    /// For example,
469 477
    /// \code
470 478
    ///   CostScaling<ListDigraph> cs(graph);
471 479
    ///   cs.lowerMap(lower).upperMap(upper).costMap(cost)
472 480
    ///     .supplyMap(sup).run();
473 481
    /// \endcode
474 482
    ///
475 483
    /// This function can be called more than once. All the given parameters
476 484
    /// are kept for the next call, unless \ref resetParams() or \ref reset()
477 485
    /// is used, thus only the modified parameters have to be set again.
478 486
    /// If the underlying digraph was also modified after the construction
479 487
    /// of the class (or the last \ref reset() call), then the \ref reset()
480 488
    /// function must be called.
481 489
    ///
482 490
    /// \param method The internal method that will be used in the
483 491
    /// algorithm. For more information, see \ref Method.
484 492
    /// \param factor The cost scaling factor. It must be larger than one.
485 493
    ///
486 494
    /// \return \c INFEASIBLE if no feasible flow exists,
487 495
    /// \n \c OPTIMAL if the problem has optimal solution
488 496
    /// (i.e. it is feasible and bounded), and the algorithm has found
489 497
    /// optimal flow and node potentials (primal and dual solutions),
490 498
    /// \n \c UNBOUNDED if the digraph contains an arc of negative cost
491 499
    /// and infinite upper bound. It means that the objective function
492 500
    /// is unbounded on that arc, however, note that it could actually be
493 501
    /// bounded over the feasible flows, but this algroithm cannot handle
494 502
    /// these cases.
495 503
    ///
496 504
    /// \see ProblemType, Method
497 505
    /// \see resetParams(), reset()
498 506
    ProblemType run(Method method = PARTIAL_AUGMENT, int factor = 8) {
499 507
      _alpha = factor;
500 508
      ProblemType pt = init();
501 509
      if (pt != OPTIMAL) return pt;
502 510
      start(method);
503 511
      return OPTIMAL;
504 512
    }
505 513

	
506 514
    /// \brief Reset all the parameters that have been given before.
507 515
    ///
508 516
    /// This function resets all the paramaters that have been given
509 517
    /// before using functions \ref lowerMap(), \ref upperMap(),
510 518
    /// \ref costMap(), \ref supplyMap(), \ref stSupply().
511 519
    ///
512 520
    /// It is useful for multiple \ref run() calls. Basically, all the given
513 521
    /// parameters are kept for the next \ref run() call, unless
514 522
    /// \ref resetParams() or \ref reset() is used.
515 523
    /// If the underlying digraph was also modified after the construction
516 524
    /// of the class or the last \ref reset() call, then the \ref reset()
517 525
    /// function must be used, otherwise \ref resetParams() is sufficient.
518 526
    ///
519 527
    /// For example,
520 528
    /// \code
521 529
    ///   CostScaling<ListDigraph> cs(graph);
522 530
    ///
523 531
    ///   // First run
524 532
    ///   cs.lowerMap(lower).upperMap(upper).costMap(cost)
525 533
    ///     .supplyMap(sup).run();
526 534
    ///
527 535
    ///   // Run again with modified cost map (resetParams() is not called,
528 536
    ///   // so only the cost map have to be set again)
529 537
    ///   cost[e] += 100;
530 538
    ///   cs.costMap(cost).run();
531 539
    ///
532 540
    ///   // Run again from scratch using resetParams()
533 541
    ///   // (the lower bounds will be set to zero on all arcs)
534 542
    ///   cs.resetParams();
535 543
    ///   cs.upperMap(capacity).costMap(cost)
536 544
    ///     .supplyMap(sup).run();
537 545
    /// \endcode
538 546
    ///
539 547
    /// \return <tt>(*this)</tt>
540 548
    ///
541 549
    /// \see reset(), run()
542 550
    CostScaling& resetParams() {
543 551
      for (int i = 0; i != _res_node_num; ++i) {
544 552
        _supply[i] = 0;
545 553
      }
546 554
      int limit = _first_out[_root];
547 555
      for (int j = 0; j != limit; ++j) {
548 556
        _lower[j] = 0;
549 557
        _upper[j] = INF;
550 558
        _scost[j] = _forward[j] ? 1 : -1;
551 559
      }
552 560
      for (int j = limit; j != _res_arc_num; ++j) {
553 561
        _lower[j] = 0;
554 562
        _upper[j] = INF;
555 563
        _scost[j] = 0;
556 564
        _scost[_reverse[j]] = 0;
557 565
      }      
558 566
      _have_lower = false;
559 567
      return *this;
560 568
    }
561 569

	
562 570
    /// \brief Reset all the parameters that have been given before.
563 571
    ///
564 572
    /// This function resets all the paramaters that have been given
565 573
    /// before using functions \ref lowerMap(), \ref upperMap(),
566 574
    /// \ref costMap(), \ref supplyMap(), \ref stSupply().
567 575
    ///
568 576
    /// It is useful for multiple run() calls. If this function is not
569 577
    /// used, all the parameters given before are kept for the next
570 578
    /// \ref run() call.
571 579
    /// However, the underlying digraph must not be modified after this
572 580
    /// class have been constructed, since it copies and extends the graph.
573 581
    /// \return <tt>(*this)</tt>
574 582
    CostScaling& reset() {
575 583
      // Resize vectors
576 584
      _node_num = countNodes(_graph);
577 585
      _arc_num = countArcs(_graph);
578 586
      _res_node_num = _node_num + 1;
579 587
      _res_arc_num = 2 * (_arc_num + _node_num);
580 588
      _root = _node_num;
581 589

	
582 590
      _first_out.resize(_res_node_num + 1);
583 591
      _forward.resize(_res_arc_num);
584 592
      _source.resize(_res_arc_num);
585 593
      _target.resize(_res_arc_num);
586 594
      _reverse.resize(_res_arc_num);
587 595

	
588 596
      _lower.resize(_res_arc_num);
589 597
      _upper.resize(_res_arc_num);
590 598
      _scost.resize(_res_arc_num);
591 599
      _supply.resize(_res_node_num);
592 600
      
593 601
      _res_cap.resize(_res_arc_num);
594 602
      _cost.resize(_res_arc_num);
595 603
      _pi.resize(_res_node_num);
596 604
      _excess.resize(_res_node_num);
597 605
      _next_out.resize(_res_node_num);
598 606

	
599 607
      _arc_vec.reserve(_res_arc_num);
600 608
      _cost_vec.reserve(_res_arc_num);
601 609

	
602 610
      // Copy the graph
603 611
      int i = 0, j = 0, k = 2 * _arc_num + _node_num;
604 612
      for (NodeIt n(_graph); n != INVALID; ++n, ++i) {
605 613
        _node_id[n] = i;
606 614
      }
607 615
      i = 0;
608 616
      for (NodeIt n(_graph); n != INVALID; ++n, ++i) {
609 617
        _first_out[i] = j;
610 618
        for (OutArcIt a(_graph, n); a != INVALID; ++a, ++j) {
611 619
          _arc_idf[a] = j;
612 620
          _forward[j] = true;
613 621
          _source[j] = i;
614 622
          _target[j] = _node_id[_graph.runningNode(a)];
615 623
        }
616 624
        for (InArcIt a(_graph, n); a != INVALID; ++a, ++j) {
617 625
          _arc_idb[a] = j;
618 626
          _forward[j] = false;
619 627
          _source[j] = i;
620 628
          _target[j] = _node_id[_graph.runningNode(a)];
621 629
        }
622 630
        _forward[j] = false;
623 631
        _source[j] = i;
624 632
        _target[j] = _root;
625 633
        _reverse[j] = k;
626 634
        _forward[k] = true;
627 635
        _source[k] = _root;
628 636
        _target[k] = i;
629 637
        _reverse[k] = j;
630 638
        ++j; ++k;
631 639
      }
632 640
      _first_out[i] = j;
633 641
      _first_out[_res_node_num] = k;
634 642
      for (ArcIt a(_graph); a != INVALID; ++a) {
635 643
        int fi = _arc_idf[a];
636 644
        int bi = _arc_idb[a];
637 645
        _reverse[fi] = bi;
638 646
        _reverse[bi] = fi;
639 647
      }
640 648
      
641 649
      // Reset parameters
642 650
      resetParams();
643 651
      return *this;
644 652
    }
645 653

	
646 654
    /// @}
647 655

	
648 656
    /// \name Query Functions
649 657
    /// The results of the algorithm can be obtained using these
650 658
    /// functions.\n
651 659
    /// The \ref run() function must be called before using them.
652 660

	
653 661
    /// @{
654 662

	
655 663
    /// \brief Return the total cost of the found flow.
656 664
    ///
657 665
    /// This function returns the total cost of the found flow.
658 666
    /// Its complexity is O(e).
659 667
    ///
660 668
    /// \note The return type of the function can be specified as a
661 669
    /// template parameter. For example,
662 670
    /// \code
663 671
    ///   cs.totalCost<double>();
664 672
    /// \endcode
665 673
    /// It is useful if the total cost cannot be stored in the \c Cost
666 674
    /// type of the algorithm, which is the default return type of the
667 675
    /// function.
668 676
    ///
669 677
    /// \pre \ref run() must be called before using this function.
670 678
    template <typename Number>
671 679
    Number totalCost() const {
672 680
      Number c = 0;
673 681
      for (ArcIt a(_graph); a != INVALID; ++a) {
674 682
        int i = _arc_idb[a];
675 683
        c += static_cast<Number>(_res_cap[i]) *
676 684
             (-static_cast<Number>(_scost[i]));
677 685
      }
678 686
      return c;
679 687
    }
680 688

	
681 689
#ifndef DOXYGEN
682 690
    Cost totalCost() const {
683 691
      return totalCost<Cost>();
684 692
    }
685 693
#endif
686 694

	
687 695
    /// \brief Return the flow on the given arc.
688 696
    ///
689 697
    /// This function returns the flow on the given arc.
690 698
    ///
691 699
    /// \pre \ref run() must be called before using this function.
692 700
    Value flow(const Arc& a) const {
693 701
      return _res_cap[_arc_idb[a]];
694 702
    }
695 703

	
696 704
    /// \brief Return the flow map (the primal solution).
697 705
    ///
698 706
    /// This function copies the flow value on each arc into the given
699 707
    /// map. The \c Value type of the algorithm must be convertible to
700 708
    /// the \c Value type of the map.
701 709
    ///
702 710
    /// \pre \ref run() must be called before using this function.
703 711
    template <typename FlowMap>
704 712
    void flowMap(FlowMap &map) const {
705 713
      for (ArcIt a(_graph); a != INVALID; ++a) {
706 714
        map.set(a, _res_cap[_arc_idb[a]]);
707 715
      }
708 716
    }
709 717

	
710 718
    /// \brief Return the potential (dual value) of the given node.
711 719
    ///
712 720
    /// This function returns the potential (dual value) of the
713 721
    /// given node.
714 722
    ///
715 723
    /// \pre \ref run() must be called before using this function.
716 724
    Cost potential(const Node& n) const {
717 725
      return static_cast<Cost>(_pi[_node_id[n]]);
718 726
    }
719 727

	
720 728
    /// \brief Return the potential map (the dual solution).
721 729
    ///
722 730
    /// This function copies the potential (dual value) of each node
723 731
    /// into the given map.
724 732
    /// The \c Cost type of the algorithm must be convertible to the
725 733
    /// \c Value type of the map.
726 734
    ///
727 735
    /// \pre \ref run() must be called before using this function.
728 736
    template <typename PotentialMap>
729 737
    void potentialMap(PotentialMap &map) const {
730 738
      for (NodeIt n(_graph); n != INVALID; ++n) {
731 739
        map.set(n, static_cast<Cost>(_pi[_node_id[n]]));
732 740
      }
733 741
    }
734 742

	
735 743
    /// @}
736 744

	
737 745
  private:
738 746

	
739 747
    // Initialize the algorithm
740 748
    ProblemType init() {
741 749
      if (_res_node_num <= 1) return INFEASIBLE;
742 750

	
743 751
      // Check the sum of supply values
744 752
      _sum_supply = 0;
745 753
      for (int i = 0; i != _root; ++i) {
746 754
        _sum_supply += _supply[i];
747 755
      }
748 756
      if (_sum_supply > 0) return INFEASIBLE;
749 757
      
750 758

	
751 759
      // Initialize vectors
752 760
      for (int i = 0; i != _res_node_num; ++i) {
753 761
        _pi[i] = 0;
754 762
        _excess[i] = _supply[i];
755 763
      }
756 764
      
757 765
      // Remove infinite upper bounds and check negative arcs
758 766
      const Value MAX = std::numeric_limits<Value>::max();
759 767
      int last_out;
760 768
      if (_have_lower) {
761 769
        for (int i = 0; i != _root; ++i) {
762 770
          last_out = _first_out[i+1];
763 771
          for (int j = _first_out[i]; j != last_out; ++j) {
764 772
            if (_forward[j]) {
765 773
              Value c = _scost[j] < 0 ? _upper[j] : _lower[j];
766 774
              if (c >= MAX) return UNBOUNDED;
767 775
              _excess[i] -= c;
768 776
              _excess[_target[j]] += c;
769 777
            }
770 778
          }
771 779
        }
772 780
      } else {
773 781
        for (int i = 0; i != _root; ++i) {
774 782
          last_out = _first_out[i+1];
775 783
          for (int j = _first_out[i]; j != last_out; ++j) {
776 784
            if (_forward[j] && _scost[j] < 0) {
777 785
              Value c = _upper[j];
778 786
              if (c >= MAX) return UNBOUNDED;
779 787
              _excess[i] -= c;
780 788
              _excess[_target[j]] += c;
781 789
            }
782 790
          }
783 791
        }
784 792
      }
785 793
      Value ex, max_cap = 0;
786 794
      for (int i = 0; i != _res_node_num; ++i) {
787 795
        ex = _excess[i];
788 796
        _excess[i] = 0;
789 797
        if (ex < 0) max_cap -= ex;
790 798
      }
791 799
      for (int j = 0; j != _res_arc_num; ++j) {
792 800
        if (_upper[j] >= MAX) _upper[j] = max_cap;
793 801
      }
794 802

	
795 803
      // Initialize the large cost vector and the epsilon parameter
796 804
      _epsilon = 0;
797 805
      LargeCost lc;
798 806
      for (int i = 0; i != _root; ++i) {
799 807
        last_out = _first_out[i+1];
800 808
        for (int j = _first_out[i]; j != last_out; ++j) {
801 809
          lc = static_cast<LargeCost>(_scost[j]) * _res_node_num * _alpha;
802 810
          _cost[j] = lc;
803 811
          if (lc > _epsilon) _epsilon = lc;
804 812
        }
805 813
      }
806 814
      _epsilon /= _alpha;
807 815

	
808 816
      // Initialize maps for Circulation and remove non-zero lower bounds
809 817
      ConstMap<Arc, Value> low(0);
810 818
      typedef typename Digraph::template ArcMap<Value> ValueArcMap;
811 819
      typedef typename Digraph::template NodeMap<Value> ValueNodeMap;
812 820
      ValueArcMap cap(_graph), flow(_graph);
813 821
      ValueNodeMap sup(_graph);
814 822
      for (NodeIt n(_graph); n != INVALID; ++n) {
815 823
        sup[n] = _supply[_node_id[n]];
816 824
      }
817 825
      if (_have_lower) {
818 826
        for (ArcIt a(_graph); a != INVALID; ++a) {
819 827
          int j = _arc_idf[a];
820 828
          Value c = _lower[j];
821 829
          cap[a] = _upper[j] - c;
822 830
          sup[_graph.source(a)] -= c;
823 831
          sup[_graph.target(a)] += c;
824 832
        }
825 833
      } else {
826 834
        for (ArcIt a(_graph); a != INVALID; ++a) {
827 835
          cap[a] = _upper[_arc_idf[a]];
828 836
        }
829 837
      }
830 838

	
839
      _sup_node_num = 0;
840
      for (NodeIt n(_graph); n != INVALID; ++n) {
841
        if (sup[n] > 0) ++_sup_node_num;
842
      }
843

	
831 844
      // Find a feasible flow using Circulation
832 845
      Circulation<Digraph, ConstMap<Arc, Value>, ValueArcMap, ValueNodeMap>
833 846
        circ(_graph, low, cap, sup);
834 847
      if (!circ.flowMap(flow).run()) return INFEASIBLE;
835 848

	
836 849
      // Set residual capacities and handle GEQ supply type
837 850
      if (_sum_supply < 0) {
838 851
        for (ArcIt a(_graph); a != INVALID; ++a) {
839 852
          Value fa = flow[a];
840 853
          _res_cap[_arc_idf[a]] = cap[a] - fa;
841 854
          _res_cap[_arc_idb[a]] = fa;
842 855
          sup[_graph.source(a)] -= fa;
843 856
          sup[_graph.target(a)] += fa;
844 857
        }
845 858
        for (NodeIt n(_graph); n != INVALID; ++n) {
846 859
          _excess[_node_id[n]] = sup[n];
847 860
        }
848 861
        for (int a = _first_out[_root]; a != _res_arc_num; ++a) {
849 862
          int u = _target[a];
850 863
          int ra = _reverse[a];
851 864
          _res_cap[a] = -_sum_supply + 1;
852 865
          _res_cap[ra] = -_excess[u];
853 866
          _cost[a] = 0;
854 867
          _cost[ra] = 0;
855 868
          _excess[u] = 0;
856 869
        }
857 870
      } else {
858 871
        for (ArcIt a(_graph); a != INVALID; ++a) {
859 872
          Value fa = flow[a];
860 873
          _res_cap[_arc_idf[a]] = cap[a] - fa;
861 874
          _res_cap[_arc_idb[a]] = fa;
862 875
        }
863 876
        for (int a = _first_out[_root]; a != _res_arc_num; ++a) {
864 877
          int ra = _reverse[a];
865
          _res_cap[a] = 1;
878
          _res_cap[a] = 0;
866 879
          _res_cap[ra] = 0;
867 880
          _cost[a] = 0;
868 881
          _cost[ra] = 0;
869 882
        }
870 883
      }
871 884
      
872 885
      return OPTIMAL;
873 886
    }
874 887

	
875 888
    // Execute the algorithm and transform the results
876 889
    void start(Method method) {
877 890
      // Maximum path length for partial augment
878 891
      const int MAX_PATH_LENGTH = 4;
879
      
892

	
893
      // Initialize data structures for buckets      
894
      _max_rank = _alpha * _res_node_num;
895
      _buckets.resize(_max_rank);
896
      _bucket_next.resize(_res_node_num + 1);
897
      _bucket_prev.resize(_res_node_num + 1);
898
      _rank.resize(_res_node_num + 1);
899
  
880 900
      // Execute the algorithm
881 901
      switch (method) {
882 902
        case PUSH:
883 903
          startPush();
884 904
          break;
885 905
        case AUGMENT:
886 906
          startAugment();
887 907
          break;
888 908
        case PARTIAL_AUGMENT:
889 909
          startAugment(MAX_PATH_LENGTH);
890 910
          break;
891 911
      }
892 912

	
893 913
      // Compute node potentials for the original costs
894 914
      _arc_vec.clear();
895 915
      _cost_vec.clear();
896 916
      for (int j = 0; j != _res_arc_num; ++j) {
897 917
        if (_res_cap[j] > 0) {
898 918
          _arc_vec.push_back(IntPair(_source[j], _target[j]));
899 919
          _cost_vec.push_back(_scost[j]);
900 920
        }
901 921
      }
902 922
      _sgr.build(_res_node_num, _arc_vec.begin(), _arc_vec.end());
903 923

	
904 924
      typename BellmanFord<StaticDigraph, LargeCostArcMap>
905 925
        ::template SetDistMap<LargeCostNodeMap>::Create bf(_sgr, _cost_map);
906 926
      bf.distMap(_pi_map);
907 927
      bf.init(0);
908 928
      bf.start();
909 929

	
910 930
      // Handle non-zero lower bounds
911 931
      if (_have_lower) {
912 932
        int limit = _first_out[_root];
913 933
        for (int j = 0; j != limit; ++j) {
914 934
          if (!_forward[j]) _res_cap[j] += _lower[j];
915 935
        }
916 936
      }
917 937
    }
938
    
939
    // Initialize a cost scaling phase
940
    void initPhase() {
941
      // Saturate arcs not satisfying the optimality condition
942
      for (int u = 0; u != _res_node_num; ++u) {
943
        int last_out = _first_out[u+1];
944
        LargeCost pi_u = _pi[u];
945
        for (int a = _first_out[u]; a != last_out; ++a) {
946
          int v = _target[a];
947
          if (_res_cap[a] > 0 && _cost[a] + pi_u - _pi[v] < 0) {
948
            Value delta = _res_cap[a];
949
            _excess[u] -= delta;
950
            _excess[v] += delta;
951
            _res_cap[a] = 0;
952
            _res_cap[_reverse[a]] += delta;
953
          }
954
        }
955
      }
956
      
957
      // Find active nodes (i.e. nodes with positive excess)
958
      for (int u = 0; u != _res_node_num; ++u) {
959
        if (_excess[u] > 0) _active_nodes.push_back(u);
960
      }
961

	
962
      // Initialize the next arcs
963
      for (int u = 0; u != _res_node_num; ++u) {
964
        _next_out[u] = _first_out[u];
965
      }
966
    }
967
    
968
    // Early termination heuristic
969
    bool earlyTermination() {
970
      const double EARLY_TERM_FACTOR = 3.0;
971

	
972
      // Build a static residual graph
973
      _arc_vec.clear();
974
      _cost_vec.clear();
975
      for (int j = 0; j != _res_arc_num; ++j) {
976
        if (_res_cap[j] > 0) {
977
          _arc_vec.push_back(IntPair(_source[j], _target[j]));
978
          _cost_vec.push_back(_cost[j] + 1);
979
        }
980
      }
981
      _sgr.build(_res_node_num, _arc_vec.begin(), _arc_vec.end());
982

	
983
      // Run Bellman-Ford algorithm to check if the current flow is optimal
984
      BellmanFord<StaticDigraph, LargeCostArcMap> bf(_sgr, _cost_map);
985
      bf.init(0);
986
      bool done = false;
987
      int K = int(EARLY_TERM_FACTOR * std::sqrt(double(_res_node_num)));
988
      for (int i = 0; i < K && !done; ++i) {
989
        done = bf.processNextWeakRound();
990
      }
991
      return done;
992
    }
993

	
994
    // Global potential update heuristic
995
    void globalUpdate() {
996
      int bucket_end = _root + 1;
997
    
998
      // Initialize buckets
999
      for (int r = 0; r != _max_rank; ++r) {
1000
        _buckets[r] = bucket_end;
1001
      }
1002
      Value total_excess = 0;
1003
      for (int i = 0; i != _res_node_num; ++i) {
1004
        if (_excess[i] < 0) {
1005
          _rank[i] = 0;
1006
          _bucket_next[i] = _buckets[0];
1007
          _bucket_prev[_buckets[0]] = i;
1008
          _buckets[0] = i;
1009
        } else {
1010
          total_excess += _excess[i];
1011
          _rank[i] = _max_rank;
1012
        }
1013
      }
1014
      if (total_excess == 0) return;
1015

	
1016
      // Search the buckets
1017
      int r = 0;
1018
      for ( ; r != _max_rank; ++r) {
1019
        while (_buckets[r] != bucket_end) {
1020
          // Remove the first node from the current bucket
1021
          int u = _buckets[r];
1022
          _buckets[r] = _bucket_next[u];
1023
          
1024
          // Search the incomming arcs of u
1025
          LargeCost pi_u = _pi[u];
1026
          int last_out = _first_out[u+1];
1027
          for (int a = _first_out[u]; a != last_out; ++a) {
1028
            int ra = _reverse[a];
1029
            if (_res_cap[ra] > 0) {
1030
              int v = _source[ra];
1031
              int old_rank_v = _rank[v];
1032
              if (r < old_rank_v) {
1033
                // Compute the new rank of v
1034
                LargeCost nrc = (_cost[ra] + _pi[v] - pi_u) / _epsilon;
1035
                int new_rank_v = old_rank_v;
1036
                if (nrc < LargeCost(_max_rank))
1037
                  new_rank_v = r + 1 + int(nrc);
1038
                  
1039
                // Change the rank of v
1040
                if (new_rank_v < old_rank_v) {
1041
                  _rank[v] = new_rank_v;
1042
                  _next_out[v] = _first_out[v];
1043
                  
1044
                  // Remove v from its old bucket
1045
                  if (old_rank_v < _max_rank) {
1046
                    if (_buckets[old_rank_v] == v) {
1047
                      _buckets[old_rank_v] = _bucket_next[v];
1048
                    } else {
1049
                      _bucket_next[_bucket_prev[v]] = _bucket_next[v];
1050
                      _bucket_prev[_bucket_next[v]] = _bucket_prev[v];
1051
                    }
1052
                  }
1053
                  
1054
                  // Insert v to its new bucket
1055
                  _bucket_next[v] = _buckets[new_rank_v];
1056
                  _bucket_prev[_buckets[new_rank_v]] = v;
1057
                  _buckets[new_rank_v] = v;
1058
                }
1059
              }
1060
            }
1061
          }
1062

	
1063
          // Finish search if there are no more active nodes
1064
          if (_excess[u] > 0) {
1065
            total_excess -= _excess[u];
1066
            if (total_excess <= 0) break;
1067
          }
1068
        }
1069
        if (total_excess <= 0) break;
1070
      }
1071
      
1072
      // Relabel nodes
1073
      for (int u = 0; u != _res_node_num; ++u) {
1074
        int k = std::min(_rank[u], r);
1075
        if (k > 0) {
1076
          _pi[u] -= _epsilon * k;
1077
          _next_out[u] = _first_out[u];
1078
        }
1079
      }
1080
    }
918 1081

	
919 1082
    /// Execute the algorithm performing augment and relabel operations
920 1083
    void startAugment(int max_length = std::numeric_limits<int>::max()) {
921 1084
      // Paramters for heuristics
922
      const int BF_HEURISTIC_EPSILON_BOUND = 1000;
923
      const int BF_HEURISTIC_BOUND_FACTOR  = 3;
1085
      const int EARLY_TERM_EPSILON_LIMIT = 1000;
1086
      const double GLOBAL_UPDATE_FACTOR = 3.0;
924 1087

	
1088
      const int global_update_freq = int(GLOBAL_UPDATE_FACTOR *
1089
        (_res_node_num + _sup_node_num * _sup_node_num));
1090
      int next_update_limit = global_update_freq;
1091
      
1092
      int relabel_cnt = 0;
1093
      
925 1094
      // Perform cost scaling phases
926
      IntVector pred_arc(_res_node_num);
927
      std::vector<int> path_nodes;
1095
      std::vector<int> path;
928 1096
      for ( ; _epsilon >= 1; _epsilon = _epsilon < _alpha && _epsilon > 1 ?
929 1097
                                        1 : _epsilon / _alpha )
930 1098
      {
931
        // "Early Termination" heuristic: use Bellman-Ford algorithm
932
        // to check if the current flow is optimal
933
        if (_epsilon <= BF_HEURISTIC_EPSILON_BOUND) {
934
          _arc_vec.clear();
935
          _cost_vec.clear();
936
          for (int j = 0; j != _res_arc_num; ++j) {
937
            if (_res_cap[j] > 0) {
938
              _arc_vec.push_back(IntPair(_source[j], _target[j]));
939
              _cost_vec.push_back(_cost[j] + 1);
940
            }
941
          }
942
          _sgr.build(_res_node_num, _arc_vec.begin(), _arc_vec.end());
943

	
944
          BellmanFord<StaticDigraph, LargeCostArcMap> bf(_sgr, _cost_map);
945
          bf.init(0);
946
          bool done = false;
947
          int K = int(BF_HEURISTIC_BOUND_FACTOR * sqrt(_res_node_num));
948
          for (int i = 0; i < K && !done; ++i)
949
            done = bf.processNextWeakRound();
950
          if (done) break;
951
        }
952

	
953
        // Saturate arcs not satisfying the optimality condition
954
        for (int a = 0; a != _res_arc_num; ++a) {
955
          if (_res_cap[a] > 0 &&
956
              _cost[a] + _pi[_source[a]] - _pi[_target[a]] < 0) {
957
            Value delta = _res_cap[a];
958
            _excess[_source[a]] -= delta;
959
            _excess[_target[a]] += delta;
960
            _res_cap[a] = 0;
961
            _res_cap[_reverse[a]] += delta;
962
          }
1099
        // Early termination heuristic
1100
        if (_epsilon <= EARLY_TERM_EPSILON_LIMIT) {
1101
          if (earlyTermination()) break;
963 1102
        }
964 1103
        
965
        // Find active nodes (i.e. nodes with positive excess)
966
        for (int u = 0; u != _res_node_num; ++u) {
967
          if (_excess[u] > 0) _active_nodes.push_back(u);
968
        }
969

	
970
        // Initialize the next arcs
971
        for (int u = 0; u != _res_node_num; ++u) {
972
          _next_out[u] = _first_out[u];
973
        }
974

	
1104
        // Initialize current phase
1105
        initPhase();
1106
        
975 1107
        // Perform partial augment and relabel operations
976 1108
        while (true) {
977 1109
          // Select an active node (FIFO selection)
978 1110
          while (_active_nodes.size() > 0 &&
979 1111
                 _excess[_active_nodes.front()] <= 0) {
980 1112
            _active_nodes.pop_front();
981 1113
          }
982 1114
          if (_active_nodes.size() == 0) break;
983 1115
          int start = _active_nodes.front();
984
          path_nodes.clear();
985
          path_nodes.push_back(start);
986 1116

	
987 1117
          // Find an augmenting path from the start node
1118
          path.clear();
988 1119
          int tip = start;
989
          while (_excess[tip] >= 0 &&
990
                 int(path_nodes.size()) <= max_length) {
1120
          while (_excess[tip] >= 0 && int(path.size()) < max_length) {
991 1121
            int u;
992
            LargeCost min_red_cost, rc;
993
            int last_out = _sum_supply < 0 ?
994
              _first_out[tip+1] : _first_out[tip+1] - 1;
1122
            LargeCost min_red_cost, rc, pi_tip = _pi[tip];
1123
            int last_out = _first_out[tip+1];
995 1124
            for (int a = _next_out[tip]; a != last_out; ++a) {
996
              if (_res_cap[a] > 0 &&
997
                  _cost[a] + _pi[_source[a]] - _pi[_target[a]] < 0) {
998
                u = _target[a];
999
                pred_arc[u] = a;
1125
              u = _target[a];
1126
              if (_res_cap[a] > 0 && _cost[a] + pi_tip - _pi[u] < 0) {
1127
                path.push_back(a);
1000 1128
                _next_out[tip] = a;
1001 1129
                tip = u;
1002
                path_nodes.push_back(tip);
1003 1130
                goto next_step;
1004 1131
              }
1005 1132
            }
1006 1133

	
1007 1134
            // Relabel tip node
1008
            min_red_cost = std::numeric_limits<LargeCost>::max() / 2;
1135
            min_red_cost = std::numeric_limits<LargeCost>::max();
1136
            if (tip != start) {
1137
              int ra = _reverse[path.back()];
1138
              min_red_cost = _cost[ra] + pi_tip - _pi[_target[ra]];
1139
            }
1009 1140
            for (int a = _first_out[tip]; a != last_out; ++a) {
1010
              rc = _cost[a] + _pi[_source[a]] - _pi[_target[a]];
1141
              rc = _cost[a] + pi_tip - _pi[_target[a]];
1011 1142
              if (_res_cap[a] > 0 && rc < min_red_cost) {
1012 1143
                min_red_cost = rc;
1013 1144
              }
1014 1145
            }
1015 1146
            _pi[tip] -= min_red_cost + _epsilon;
1016

	
1017
            // Reset the next arc of tip
1018 1147
            _next_out[tip] = _first_out[tip];
1148
            ++relabel_cnt;
1019 1149

	
1020 1150
            // Step back
1021 1151
            if (tip != start) {
1022
              path_nodes.pop_back();
1023
              tip = path_nodes.back();
1152
              tip = _source[path.back()];
1153
              path.pop_back();
1024 1154
            }
1025 1155

	
1026 1156
          next_step: ;
1027 1157
          }
1028 1158

	
1029 1159
          // Augment along the found path (as much flow as possible)
1030 1160
          Value delta;
1031
          int u, v = path_nodes.front(), pa;
1032
          for (int i = 1; i < int(path_nodes.size()); ++i) {
1161
          int pa, u, v = start;
1162
          for (int i = 0; i != int(path.size()); ++i) {
1163
            pa = path[i];
1033 1164
            u = v;
1034
            v = path_nodes[i];
1035
            pa = pred_arc[v];
1165
            v = _target[pa];
1036 1166
            delta = std::min(_res_cap[pa], _excess[u]);
1037 1167
            _res_cap[pa] -= delta;
1038 1168
            _res_cap[_reverse[pa]] += delta;
1039 1169
            _excess[u] -= delta;
1040 1170
            _excess[v] += delta;
1041 1171
            if (_excess[v] > 0 && _excess[v] <= delta)
1042 1172
              _active_nodes.push_back(v);
1043 1173
          }
1174

	
1175
          // Global update heuristic
1176
          if (relabel_cnt >= next_update_limit) {
1177
            globalUpdate();
1178
            next_update_limit += global_update_freq;
1179
          }
1044 1180
        }
1045 1181
      }
1046 1182
    }
1047 1183

	
1048 1184
    /// Execute the algorithm performing push and relabel operations
1049 1185
    void startPush() {
1050 1186
      // Paramters for heuristics
1051
      const int BF_HEURISTIC_EPSILON_BOUND = 1000;
1052
      const int BF_HEURISTIC_BOUND_FACTOR  = 3;
1187
      const int EARLY_TERM_EPSILON_LIMIT = 1000;
1188
      const double GLOBAL_UPDATE_FACTOR = 2.0;
1053 1189

	
1190
      const int global_update_freq = int(GLOBAL_UPDATE_FACTOR *
1191
        (_res_node_num + _sup_node_num * _sup_node_num));
1192
      int next_update_limit = global_update_freq;
1193

	
1194
      int relabel_cnt = 0;
1195
      
1054 1196
      // Perform cost scaling phases
1055 1197
      BoolVector hyper(_res_node_num, false);
1198
      LargeCostVector hyper_cost(_res_node_num);
1056 1199
      for ( ; _epsilon >= 1; _epsilon = _epsilon < _alpha && _epsilon > 1 ?
1057 1200
                                        1 : _epsilon / _alpha )
1058 1201
      {
1059
        // "Early Termination" heuristic: use Bellman-Ford algorithm
1060
        // to check if the current flow is optimal
1061
        if (_epsilon <= BF_HEURISTIC_EPSILON_BOUND) {
1062
          _arc_vec.clear();
1063
          _cost_vec.clear();
1064
          for (int j = 0; j != _res_arc_num; ++j) {
1065
            if (_res_cap[j] > 0) {
1066
              _arc_vec.push_back(IntPair(_source[j], _target[j]));
1067
              _cost_vec.push_back(_cost[j] + 1);
1068
            }
1069
          }
1070
          _sgr.build(_res_node_num, _arc_vec.begin(), _arc_vec.end());
1071

	
1072
          BellmanFord<StaticDigraph, LargeCostArcMap> bf(_sgr, _cost_map);
1073
          bf.init(0);
1074
          bool done = false;
1075
          int K = int(BF_HEURISTIC_BOUND_FACTOR * sqrt(_res_node_num));
1076
          for (int i = 0; i < K && !done; ++i)
1077
            done = bf.processNextWeakRound();
1078
          if (done) break;
1202
        // Early termination heuristic
1203
        if (_epsilon <= EARLY_TERM_EPSILON_LIMIT) {
1204
          if (earlyTermination()) break;
1079 1205
        }
1080

	
1081
        // Saturate arcs not satisfying the optimality condition
1082
        for (int a = 0; a != _res_arc_num; ++a) {
1083
          if (_res_cap[a] > 0 &&
1084
              _cost[a] + _pi[_source[a]] - _pi[_target[a]] < 0) {
1085
            Value delta = _res_cap[a];
1086
            _excess[_source[a]] -= delta;
1087
            _excess[_target[a]] += delta;
1088
            _res_cap[a] = 0;
1089
            _res_cap[_reverse[a]] += delta;
1090
          }
1091
        }
1092

	
1093
        // Find active nodes (i.e. nodes with positive excess)
1094
        for (int u = 0; u != _res_node_num; ++u) {
1095
          if (_excess[u] > 0) _active_nodes.push_back(u);
1096
        }
1097

	
1098
        // Initialize the next arcs
1099
        for (int u = 0; u != _res_node_num; ++u) {
1100
          _next_out[u] = _first_out[u];
1101
        }
1206
        
1207
        // Initialize current phase
1208
        initPhase();
1102 1209

	
1103 1210
        // Perform push and relabel operations
1104 1211
        while (_active_nodes.size() > 0) {
1105
          LargeCost min_red_cost, rc;
1212
          LargeCost min_red_cost, rc, pi_n;
1106 1213
          Value delta;
1107 1214
          int n, t, a, last_out = _res_arc_num;
1108 1215

	
1216
        next_node:
1109 1217
          // Select an active node (FIFO selection)
1110
        next_node:
1111 1218
          n = _active_nodes.front();
1112
          last_out = _sum_supply < 0 ?
1113
            _first_out[n+1] : _first_out[n+1] - 1;
1114

	
1219
          last_out = _first_out[n+1];
1220
          pi_n = _pi[n];
1221
          
1115 1222
          // Perform push operations if there are admissible arcs
1116 1223
          if (_excess[n] > 0) {
1117 1224
            for (a = _next_out[n]; a != last_out; ++a) {
1118 1225
              if (_res_cap[a] > 0 &&
1119
                  _cost[a] + _pi[_source[a]] - _pi[_target[a]] < 0) {
1226
                  _cost[a] + pi_n - _pi[_target[a]] < 0) {
1120 1227
                delta = std::min(_res_cap[a], _excess[n]);
1121 1228
                t = _target[a];
1122 1229

	
1123 1230
                // Push-look-ahead heuristic
1124 1231
                Value ahead = -_excess[t];
1125
                int last_out_t = _sum_supply < 0 ?
1126
                  _first_out[t+1] : _first_out[t+1] - 1;
1232
                int last_out_t = _first_out[t+1];
1233
                LargeCost pi_t = _pi[t];
1127 1234
                for (int ta = _next_out[t]; ta != last_out_t; ++ta) {
1128 1235
                  if (_res_cap[ta] > 0 && 
1129
                      _cost[ta] + _pi[_source[ta]] - _pi[_target[ta]] < 0)
1236
                      _cost[ta] + pi_t - _pi[_target[ta]] < 0)
1130 1237
                    ahead += _res_cap[ta];
1131 1238
                  if (ahead >= delta) break;
1132 1239
                }
1133 1240
                if (ahead < 0) ahead = 0;
1134 1241

	
1135 1242
                // Push flow along the arc
1136
                if (ahead < delta) {
1243
                if (ahead < delta && !hyper[t]) {
1137 1244
                  _res_cap[a] -= ahead;
1138 1245
                  _res_cap[_reverse[a]] += ahead;
1139 1246
                  _excess[n] -= ahead;
1140 1247
                  _excess[t] += ahead;
1141 1248
                  _active_nodes.push_front(t);
1142 1249
                  hyper[t] = true;
1250
                  hyper_cost[t] = _cost[a] + pi_n - pi_t;
1143 1251
                  _next_out[n] = a;
1144 1252
                  goto next_node;
1145 1253
                } else {
1146 1254
                  _res_cap[a] -= delta;
1147 1255
                  _res_cap[_reverse[a]] += delta;
1148 1256
                  _excess[n] -= delta;
1149 1257
                  _excess[t] += delta;
1150 1258
                  if (_excess[t] > 0 && _excess[t] <= delta)
1151 1259
                    _active_nodes.push_back(t);
1152 1260
                }
1153 1261

	
1154 1262
                if (_excess[n] == 0) {
1155 1263
                  _next_out[n] = a;
1156 1264
                  goto remove_nodes;
1157 1265
                }
1158 1266
              }
1159 1267
            }
1160 1268
            _next_out[n] = a;
1161 1269
          }
1162 1270

	
1163 1271
          // Relabel the node if it is still active (or hyper)
1164 1272
          if (_excess[n] > 0 || hyper[n]) {
1165
            min_red_cost = std::numeric_limits<LargeCost>::max() / 2;
1273
             min_red_cost = hyper[n] ? -hyper_cost[n] :
1274
               std::numeric_limits<LargeCost>::max();
1166 1275
            for (int a = _first_out[n]; a != last_out; ++a) {
1167
              rc = _cost[a] + _pi[_source[a]] - _pi[_target[a]];
1276
              rc = _cost[a] + pi_n - _pi[_target[a]];
1168 1277
              if (_res_cap[a] > 0 && rc < min_red_cost) {
1169 1278
                min_red_cost = rc;
1170 1279
              }
1171 1280
            }
1172 1281
            _pi[n] -= min_red_cost + _epsilon;
1282
            _next_out[n] = _first_out[n];
1173 1283
            hyper[n] = false;
1174

	
1175
            // Reset the next arc
1176
            _next_out[n] = _first_out[n];
1284
            ++relabel_cnt;
1177 1285
          }
1178 1286
        
1179 1287
          // Remove nodes that are not active nor hyper
1180 1288
        remove_nodes:
1181 1289
          while ( _active_nodes.size() > 0 &&
1182 1290
                  _excess[_active_nodes.front()] <= 0 &&
1183 1291
                  !hyper[_active_nodes.front()] ) {
1184 1292
            _active_nodes.pop_front();
1185 1293
          }
1294
          
1295
          // Global update heuristic
1296
          if (relabel_cnt >= next_update_limit) {
1297
            globalUpdate();
1298
            for (int u = 0; u != _res_node_num; ++u)
1299
              hyper[u] = false;
1300
            next_update_limit += global_update_freq;
1301
          }
1186 1302
        }
1187 1303
      }
1188 1304
    }
1189 1305

	
1190 1306
  }; //class CostScaling
1191 1307

	
1192 1308
  ///@}
1193 1309

	
1194 1310
} //namespace lemon
1195 1311

	
1196 1312
#endif //LEMON_COST_SCALING_H
Ignore white space 6 line context
1 1
/* -*- C++ -*-
2 2
 *
3 3
 * This file is a part of LEMON, a generic C++ optimization library
4 4
 *
5 5
 * Copyright (C) 2003-2008
6 6
 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
7 7
 * (Egervary Research Group on Combinatorial Optimization, EGRES).
8 8
 *
9 9
 * Permission to use, modify and distribute this software is granted
10 10
 * provided that this copyright notice appears in all copies. For
11 11
 * precise terms see the accompanying LICENSE file.
12 12
 *
13 13
 * This software is provided "AS IS" with no warranty of any kind,
14 14
 * express or implied, and with no claim as to its suitability for any
15 15
 * purpose.
16 16
 *
17 17
 */
18 18

	
19 19
#ifndef LEMON_CYCLE_CANCELING_H
20 20
#define LEMON_CYCLE_CANCELING_H
21 21

	
22 22
/// \ingroup min_cost_flow_algs
23 23
/// \file
24 24
/// \brief Cycle-canceling algorithms for finding a minimum cost flow.
25 25

	
26 26
#include <vector>
27 27
#include <limits>
28 28

	
29 29
#include <lemon/core.h>
30 30
#include <lemon/maps.h>
31 31
#include <lemon/path.h>
32 32
#include <lemon/math.h>
33 33
#include <lemon/static_graph.h>
34 34
#include <lemon/adaptors.h>
35 35
#include <lemon/circulation.h>
36 36
#include <lemon/bellman_ford.h>
37 37
#include <lemon/howard.h>
38 38

	
39 39
namespace lemon {
40 40

	
41 41
  /// \addtogroup min_cost_flow_algs
42 42
  /// @{
43 43

	
44 44
  /// \brief Implementation of cycle-canceling algorithms for
45 45
  /// finding a \ref min_cost_flow "minimum cost flow".
46 46
  ///
47 47
  /// \ref CycleCanceling implements three different cycle-canceling
48 48
  /// algorithms for finding a \ref min_cost_flow "minimum cost flow"
49 49
  /// \ref amo93networkflows, \ref klein67primal,
50 50
  /// \ref goldberg89cyclecanceling.
51 51
  /// The most efficent one (both theoretically and practically)
52 52
  /// is the \ref CANCEL_AND_TIGHTEN "Cancel and Tighten" algorithm,
53 53
  /// thus it is the default method.
54 54
  /// It is strongly polynomial, but in practice, it is typically much
55 55
  /// slower than the scaling algorithms and NetworkSimplex.
56 56
  ///
57 57
  /// Most of the parameters of the problem (except for the digraph)
58 58
  /// can be given using separate functions, and the algorithm can be
59 59
  /// executed using the \ref run() function. If some parameters are not
60 60
  /// specified, then default values will be used.
61 61
  ///
62 62
  /// \tparam GR The digraph type the algorithm runs on.
63 63
  /// \tparam V The number type used for flow amounts, capacity bounds
64 64
  /// and supply values in the algorithm. By default, it is \c int.
65 65
  /// \tparam C The number type used for costs and potentials in the
66 66
  /// algorithm. By default, it is the same as \c V.
67 67
  ///
68 68
  /// \warning Both number types must be signed and all input data must
69 69
  /// be integer.
70 70
  /// \warning This algorithm does not support negative costs for such
71 71
  /// arcs that have infinite upper bound.
72 72
  ///
73 73
  /// \note For more information about the three available methods,
74 74
  /// see \ref Method.
75 75
#ifdef DOXYGEN
76 76
  template <typename GR, typename V, typename C>
77 77
#else
78 78
  template <typename GR, typename V = int, typename C = V>
79 79
#endif
80 80
  class CycleCanceling
81 81
  {
82 82
  public:
83 83

	
84 84
    /// The type of the digraph
85 85
    typedef GR Digraph;
86 86
    /// The type of the flow amounts, capacity bounds and supply values
87 87
    typedef V Value;
88 88
    /// The type of the arc costs
89 89
    typedef C Cost;
90 90

	
91 91
  public:
92 92

	
93 93
    /// \brief Problem type constants for the \c run() function.
94 94
    ///
95 95
    /// Enum type containing the problem type constants that can be
96 96
    /// returned by the \ref run() function of the algorithm.
97 97
    enum ProblemType {
98 98
      /// The problem has no feasible solution (flow).
99 99
      INFEASIBLE,
100 100
      /// The problem has optimal solution (i.e. it is feasible and
101 101
      /// bounded), and the algorithm has found optimal flow and node
102 102
      /// potentials (primal and dual solutions).
103 103
      OPTIMAL,
104 104
      /// The digraph contains an arc of negative cost and infinite
105 105
      /// upper bound. It means that the objective function is unbounded
106 106
      /// on that arc, however, note that it could actually be bounded
107 107
      /// over the feasible flows, but this algroithm cannot handle
108 108
      /// these cases.
109 109
      UNBOUNDED
110 110
    };
111 111

	
112 112
    /// \brief Constants for selecting the used method.
113 113
    ///
114 114
    /// Enum type containing constants for selecting the used method
115 115
    /// for the \ref run() function.
116 116
    ///
117 117
    /// \ref CycleCanceling provides three different cycle-canceling
118 118
    /// methods. By default, \ref CANCEL_AND_TIGHTEN "Cancel and Tighten"
119 119
    /// is used, which proved to be the most efficient and the most robust
120 120
    /// on various test inputs.
121 121
    /// However, the other methods can be selected using the \ref run()
122 122
    /// function with the proper parameter.
123 123
    enum Method {
124 124
      /// A simple cycle-canceling method, which uses the
125 125
      /// \ref BellmanFord "Bellman-Ford" algorithm with limited iteration
126 126
      /// number for detecting negative cycles in the residual network.
127 127
      SIMPLE_CYCLE_CANCELING,
128 128
      /// The "Minimum Mean Cycle-Canceling" algorithm, which is a
129 129
      /// well-known strongly polynomial method
130 130
      /// \ref goldberg89cyclecanceling. It improves along a
131 131
      /// \ref min_mean_cycle "minimum mean cycle" in each iteration.
132 132
      /// Its running time complexity is O(n<sup>2</sup>m<sup>3</sup>log(n)).
133 133
      MINIMUM_MEAN_CYCLE_CANCELING,
134 134
      /// The "Cancel And Tighten" algorithm, which can be viewed as an
135 135
      /// improved version of the previous method
136 136
      /// \ref goldberg89cyclecanceling.
137 137
      /// It is faster both in theory and in practice, its running time
138 138
      /// complexity is O(n<sup>2</sup>m<sup>2</sup>log(n)).
139 139
      CANCEL_AND_TIGHTEN
140 140
    };
141 141

	
142 142
  private:
143 143

	
144 144
    TEMPLATE_DIGRAPH_TYPEDEFS(GR);
145 145
    
146 146
    typedef std::vector<int> IntVector;
147
    typedef std::vector<char> CharVector;
148 147
    typedef std::vector<double> DoubleVector;
149 148
    typedef std::vector<Value> ValueVector;
150 149
    typedef std::vector<Cost> CostVector;
150
    typedef std::vector<char> BoolVector;
151
    // Note: vector<char> is used instead of vector<bool> for efficiency reasons
151 152

	
152 153
  private:
153 154
  
154 155
    template <typename KT, typename VT>
155 156
    class StaticVectorMap {
156 157
    public:
157 158
      typedef KT Key;
158 159
      typedef VT Value;
159 160
      
160 161
      StaticVectorMap(std::vector<Value>& v) : _v(v) {}
161 162
      
162 163
      const Value& operator[](const Key& key) const {
163 164
        return _v[StaticDigraph::id(key)];
164 165
      }
165 166

	
166 167
      Value& operator[](const Key& key) {
167 168
        return _v[StaticDigraph::id(key)];
168 169
      }
169 170
      
170 171
      void set(const Key& key, const Value& val) {
171 172
        _v[StaticDigraph::id(key)] = val;
172 173
      }
173 174

	
174 175
    private:
175 176
      std::vector<Value>& _v;
176 177
    };
177 178

	
178 179
    typedef StaticVectorMap<StaticDigraph::Node, Cost> CostNodeMap;
179 180
    typedef StaticVectorMap<StaticDigraph::Arc, Cost> CostArcMap;
180 181

	
181 182
  private:
182 183

	
183 184

	
184 185
    // Data related to the underlying digraph
185 186
    const GR &_graph;
186 187
    int _node_num;
187 188
    int _arc_num;
188 189
    int _res_node_num;
189 190
    int _res_arc_num;
190 191
    int _root;
191 192

	
192 193
    // Parameters of the problem
193 194
    bool _have_lower;
194 195
    Value _sum_supply;
195 196

	
196 197
    // Data structures for storing the digraph
197 198
    IntNodeMap _node_id;
198 199
    IntArcMap _arc_idf;
199 200
    IntArcMap _arc_idb;
200 201
    IntVector _first_out;
201
    CharVector _forward;
202
    BoolVector _forward;
202 203
    IntVector _source;
203 204
    IntVector _target;
204 205
    IntVector _reverse;
205 206

	
206 207
    // Node and arc data
207 208
    ValueVector _lower;
208 209
    ValueVector _upper;
209 210
    CostVector _cost;
210 211
    ValueVector _supply;
211 212

	
212 213
    ValueVector _res_cap;
213 214
    CostVector _pi;
214 215

	
215 216
    // Data for a StaticDigraph structure
216 217
    typedef std::pair<int, int> IntPair;
217 218
    StaticDigraph _sgr;
218 219
    std::vector<IntPair> _arc_vec;
219 220
    std::vector<Cost> _cost_vec;
220 221
    IntVector _id_vec;
221 222
    CostArcMap _cost_map;
222 223
    CostNodeMap _pi_map;
223 224
  
224 225
  public:
225 226
  
226 227
    /// \brief Constant for infinite upper bounds (capacities).
227 228
    ///
228 229
    /// Constant for infinite upper bounds (capacities).
229 230
    /// It is \c std::numeric_limits<Value>::infinity() if available,
230 231
    /// \c std::numeric_limits<Value>::max() otherwise.
231 232
    const Value INF;
232 233

	
233 234
  public:
234 235

	
235 236
    /// \brief Constructor.
236 237
    ///
237 238
    /// The constructor of the class.
238 239
    ///
239 240
    /// \param graph The digraph the algorithm runs on.
240 241
    CycleCanceling(const GR& graph) :
241 242
      _graph(graph), _node_id(graph), _arc_idf(graph), _arc_idb(graph),
242 243
      _cost_map(_cost_vec), _pi_map(_pi),
243 244
      INF(std::numeric_limits<Value>::has_infinity ?
244 245
          std::numeric_limits<Value>::infinity() :
245 246
          std::numeric_limits<Value>::max())
246 247
    {
247 248
      // Check the number types
248 249
      LEMON_ASSERT(std::numeric_limits<Value>::is_signed,
249 250
        "The flow type of CycleCanceling must be signed");
250 251
      LEMON_ASSERT(std::numeric_limits<Cost>::is_signed,
251 252
        "The cost type of CycleCanceling must be signed");
252 253

	
253 254
      // Reset data structures
254 255
      reset();
255 256
    }
256 257

	
257 258
    /// \name Parameters
258 259
    /// The parameters of the algorithm can be specified using these
259 260
    /// functions.
260 261

	
261 262
    /// @{
262 263

	
263 264
    /// \brief Set the lower bounds on the arcs.
264 265
    ///
265 266
    /// This function sets the lower bounds on the arcs.
266 267
    /// If it is not used before calling \ref run(), the lower bounds
267 268
    /// will be set to zero on all arcs.
268 269
    ///
269 270
    /// \param map An arc map storing the lower bounds.
270 271
    /// Its \c Value type must be convertible to the \c Value type
271 272
    /// of the algorithm.
272 273
    ///
273 274
    /// \return <tt>(*this)</tt>
274 275
    template <typename LowerMap>
275 276
    CycleCanceling& lowerMap(const LowerMap& map) {
276 277
      _have_lower = true;
277 278
      for (ArcIt a(_graph); a != INVALID; ++a) {
278 279
        _lower[_arc_idf[a]] = map[a];
279 280
        _lower[_arc_idb[a]] = map[a];
280 281
      }
281 282
      return *this;
282 283
    }
283 284

	
284 285
    /// \brief Set the upper bounds (capacities) on the arcs.
285 286
    ///
286 287
    /// This function sets the upper bounds (capacities) on the arcs.
287 288
    /// If it is not used before calling \ref run(), the upper bounds
288 289
    /// will be set to \ref INF on all arcs (i.e. the flow value will be
289 290
    /// unbounded from above).
290 291
    ///
291 292
    /// \param map An arc map storing the upper bounds.
292 293
    /// Its \c Value type must be convertible to the \c Value type
293 294
    /// of the algorithm.
294 295
    ///
295 296
    /// \return <tt>(*this)</tt>
296 297
    template<typename UpperMap>
297 298
    CycleCanceling& upperMap(const UpperMap& map) {
298 299
      for (ArcIt a(_graph); a != INVALID; ++a) {
299 300
        _upper[_arc_idf[a]] = map[a];
300 301
      }
301 302
      return *this;
302 303
    }
303 304

	
304 305
    /// \brief Set the costs of the arcs.
305 306
    ///
306 307
    /// This function sets the costs of the arcs.
307 308
    /// If it is not used before calling \ref run(), the costs
308 309
    /// will be set to \c 1 on all arcs.
309 310
    ///
310 311
    /// \param map An arc map storing the costs.
311 312
    /// Its \c Value type must be convertible to the \c Cost type
312 313
    /// of the algorithm.
313 314
    ///
314 315
    /// \return <tt>(*this)</tt>
315 316
    template<typename CostMap>
316 317
    CycleCanceling& costMap(const CostMap& map) {
317 318
      for (ArcIt a(_graph); a != INVALID; ++a) {
318 319
        _cost[_arc_idf[a]] =  map[a];
319 320
        _cost[_arc_idb[a]] = -map[a];
320 321
      }
321 322
      return *this;
322 323
    }
323 324

	
324 325
    /// \brief Set the supply values of the nodes.
325 326
    ///
326 327
    /// This function sets the supply values of the nodes.
327 328
    /// If neither this function nor \ref stSupply() is used before
328 329
    /// calling \ref run(), the supply of each node will be set to zero.
329 330
    ///
330 331
    /// \param map A node map storing the supply values.
331 332
    /// Its \c Value type must be convertible to the \c Value type
332 333
    /// of the algorithm.
333 334
    ///
334 335
    /// \return <tt>(*this)</tt>
335 336
    template<typename SupplyMap>
336 337
    CycleCanceling& supplyMap(const SupplyMap& map) {
337 338
      for (NodeIt n(_graph); n != INVALID; ++n) {
338 339
        _supply[_node_id[n]] = map[n];
339 340
      }
340 341
      return *this;
341 342
    }
342 343

	
343 344
    /// \brief Set single source and target nodes and a supply value.
344 345
    ///
345 346
    /// This function sets a single source node and a single target node
346 347
    /// and the required flow value.
347 348
    /// If neither this function nor \ref supplyMap() is used before
348 349
    /// calling \ref run(), the supply of each node will be set to zero.
349 350
    ///
350 351
    /// Using this function has the same effect as using \ref supplyMap()
351 352
    /// with such a map in which \c k is assigned to \c s, \c -k is
352 353
    /// assigned to \c t and all other nodes have zero supply value.
353 354
    ///
354 355
    /// \param s The source node.
355 356
    /// \param t The target node.
356 357
    /// \param k The required amount of flow from node \c s to node \c t
357 358
    /// (i.e. the supply of \c s and the demand of \c t).
358 359
    ///
359 360
    /// \return <tt>(*this)</tt>
360 361
    CycleCanceling& stSupply(const Node& s, const Node& t, Value k) {
361 362
      for (int i = 0; i != _res_node_num; ++i) {
362 363
        _supply[i] = 0;
363 364
      }
364 365
      _supply[_node_id[s]] =  k;
365 366
      _supply[_node_id[t]] = -k;
366 367
      return *this;
367 368
    }
368 369
    
369 370
    /// @}
370 371

	
371 372
    /// \name Execution control
372 373
    /// The algorithm can be executed using \ref run().
373 374

	
374 375
    /// @{
375 376

	
376 377
    /// \brief Run the algorithm.
377 378
    ///
378 379
    /// This function runs the algorithm.
379 380
    /// The paramters can be specified using functions \ref lowerMap(),
380 381
    /// \ref upperMap(), \ref costMap(), \ref supplyMap(), \ref stSupply().
381 382
    /// For example,
382 383
    /// \code
383 384
    ///   CycleCanceling<ListDigraph> cc(graph);
384 385
    ///   cc.lowerMap(lower).upperMap(upper).costMap(cost)
385 386
    ///     .supplyMap(sup).run();
386 387
    /// \endcode
387 388
    ///
388 389
    /// This function can be called more than once. All the given parameters
389 390
    /// are kept for the next call, unless \ref resetParams() or \ref reset()
390 391
    /// is used, thus only the modified parameters have to be set again.
391 392
    /// If the underlying digraph was also modified after the construction
392 393
    /// of the class (or the last \ref reset() call), then the \ref reset()
393 394
    /// function must be called.
394 395
    ///
395 396
    /// \param method The cycle-canceling method that will be used.
396 397
    /// For more information, see \ref Method.
397 398
    ///
398 399
    /// \return \c INFEASIBLE if no feasible flow exists,
399 400
    /// \n \c OPTIMAL if the problem has optimal solution
400 401
    /// (i.e. it is feasible and bounded), and the algorithm has found
401 402
    /// optimal flow and node potentials (primal and dual solutions),
402 403
    /// \n \c UNBOUNDED if the digraph contains an arc of negative cost
403 404
    /// and infinite upper bound. It means that the objective function
404 405
    /// is unbounded on that arc, however, note that it could actually be
405 406
    /// bounded over the feasible flows, but this algroithm cannot handle
406 407
    /// these cases.
407 408
    ///
408 409
    /// \see ProblemType, Method
409 410
    /// \see resetParams(), reset()
410 411
    ProblemType run(Method method = CANCEL_AND_TIGHTEN) {
411 412
      ProblemType pt = init();
412 413
      if (pt != OPTIMAL) return pt;
413 414
      start(method);
414 415
      return OPTIMAL;
415 416
    }
416 417

	
417 418
    /// \brief Reset all the parameters that have been given before.
418 419
    ///
419 420
    /// This function resets all the paramaters that have been given
420 421
    /// before using functions \ref lowerMap(), \ref upperMap(),
421 422
    /// \ref costMap(), \ref supplyMap(), \ref stSupply().
422 423
    ///
423 424
    /// It is useful for multiple \ref run() calls. Basically, all the given
424 425
    /// parameters are kept for the next \ref run() call, unless
425 426
    /// \ref resetParams() or \ref reset() is used.
426 427
    /// If the underlying digraph was also modified after the construction
427 428
    /// of the class or the last \ref reset() call, then the \ref reset()
428 429
    /// function must be used, otherwise \ref resetParams() is sufficient.
429 430
    ///
430 431
    /// For example,
431 432
    /// \code
432 433
    ///   CycleCanceling<ListDigraph> cs(graph);
433 434
    ///
434 435
    ///   // First run
435 436
    ///   cc.lowerMap(lower).upperMap(upper).costMap(cost)
436 437
    ///     .supplyMap(sup).run();
437 438
    ///
438 439
    ///   // Run again with modified cost map (resetParams() is not called,
439 440
    ///   // so only the cost map have to be set again)
440 441
    ///   cost[e] += 100;
441 442
    ///   cc.costMap(cost).run();
442 443
    ///
443 444
    ///   // Run again from scratch using resetParams()
444 445
    ///   // (the lower bounds will be set to zero on all arcs)
445 446
    ///   cc.resetParams();
446 447
    ///   cc.upperMap(capacity).costMap(cost)
447 448
    ///     .supplyMap(sup).run();
448 449
    /// \endcode
449 450
    ///
450 451
    /// \return <tt>(*this)</tt>
451 452
    ///
452 453
    /// \see reset(), run()
453 454
    CycleCanceling& resetParams() {
454 455
      for (int i = 0; i != _res_node_num; ++i) {
455 456
        _supply[i] = 0;
456 457
      }
457 458
      int limit = _first_out[_root];
458 459
      for (int j = 0; j != limit; ++j) {
459 460
        _lower[j] = 0;
460 461
        _upper[j] = INF;
461 462
        _cost[j] = _forward[j] ? 1 : -1;
462 463
      }
463 464
      for (int j = limit; j != _res_arc_num; ++j) {
464 465
        _lower[j] = 0;
465 466
        _upper[j] = INF;
466 467
        _cost[j] = 0;
467 468
        _cost[_reverse[j]] = 0;
468 469
      }      
469 470
      _have_lower = false;
470 471
      return *this;
471 472
    }
472 473

	
473 474
    /// \brief Reset the internal data structures and all the parameters
474 475
    /// that have been given before.
475 476
    ///
476 477
    /// This function resets the internal data structures and all the
477 478
    /// paramaters that have been given before using functions \ref lowerMap(),
478 479
    /// \ref upperMap(), \ref costMap(), \ref supplyMap(), \ref stSupply().
479 480
    ///
480 481
    /// It is useful for multiple \ref run() calls. Basically, all the given
481 482
    /// parameters are kept for the next \ref run() call, unless
482 483
    /// \ref resetParams() or \ref reset() is used.
483 484
    /// If the underlying digraph was also modified after the construction
484 485
    /// of the class or the last \ref reset() call, then the \ref reset()
485 486
    /// function must be used, otherwise \ref resetParams() is sufficient.
486 487
    ///
487 488
    /// See \ref resetParams() for examples.
488 489
    ///
489 490
    /// \return <tt>(*this)</tt>
490 491
    ///
491 492
    /// \see resetParams(), run()
492 493
    CycleCanceling& reset() {
493 494
      // Resize vectors
494 495
      _node_num = countNodes(_graph);
495 496
      _arc_num = countArcs(_graph);
496 497
      _res_node_num = _node_num + 1;
497 498
      _res_arc_num = 2 * (_arc_num + _node_num);
498 499
      _root = _node_num;
499 500

	
500 501
      _first_out.resize(_res_node_num + 1);
501 502
      _forward.resize(_res_arc_num);
502 503
      _source.resize(_res_arc_num);
503 504
      _target.resize(_res_arc_num);
504 505
      _reverse.resize(_res_arc_num);
505 506

	
506 507
      _lower.resize(_res_arc_num);
507 508
      _upper.resize(_res_arc_num);
508 509
      _cost.resize(_res_arc_num);
509 510
      _supply.resize(_res_node_num);
510 511
      
511 512
      _res_cap.resize(_res_arc_num);
512 513
      _pi.resize(_res_node_num);
513 514

	
514 515
      _arc_vec.reserve(_res_arc_num);
515 516
      _cost_vec.reserve(_res_arc_num);
516 517
      _id_vec.reserve(_res_arc_num);
517 518

	
518 519
      // Copy the graph
519 520
      int i = 0, j = 0, k = 2 * _arc_num + _node_num;
520 521
      for (NodeIt n(_graph); n != INVALID; ++n, ++i) {
521 522
        _node_id[n] = i;
522 523
      }
523 524
      i = 0;
524 525
      for (NodeIt n(_graph); n != INVALID; ++n, ++i) {
525 526
        _first_out[i] = j;
526 527
        for (OutArcIt a(_graph, n); a != INVALID; ++a, ++j) {
527 528
          _arc_idf[a] = j;
528 529
          _forward[j] = true;
529 530
          _source[j] = i;
530 531
          _target[j] = _node_id[_graph.runningNode(a)];
531 532
        }
532 533
        for (InArcIt a(_graph, n); a != INVALID; ++a, ++j) {
533 534
          _arc_idb[a] = j;
534 535
          _forward[j] = false;
535 536
          _source[j] = i;
536 537
          _target[j] = _node_id[_graph.runningNode(a)];
537 538
        }
538 539
        _forward[j] = false;
539 540
        _source[j] = i;
540 541
        _target[j] = _root;
541 542
        _reverse[j] = k;
542 543
        _forward[k] = true;
543 544
        _source[k] = _root;
544 545
        _target[k] = i;
545 546
        _reverse[k] = j;
546 547
        ++j; ++k;
547 548
      }
548 549
      _first_out[i] = j;
549 550
      _first_out[_res_node_num] = k;
550 551
      for (ArcIt a(_graph); a != INVALID; ++a) {
551 552
        int fi = _arc_idf[a];
552 553
        int bi = _arc_idb[a];
553 554
        _reverse[fi] = bi;
554 555
        _reverse[bi] = fi;
555 556
      }
556 557
      
557 558
      // Reset parameters
558 559
      resetParams();
559 560
      return *this;
560 561
    }
561 562

	
562 563
    /// @}
563 564

	
564 565
    /// \name Query Functions
565 566
    /// The results of the algorithm can be obtained using these
566 567
    /// functions.\n
567 568
    /// The \ref run() function must be called before using them.
568 569

	
569 570
    /// @{
570 571

	
571 572
    /// \brief Return the total cost of the found flow.
572 573
    ///
573 574
    /// This function returns the total cost of the found flow.
574 575
    /// Its complexity is O(e).
575 576
    ///
576 577
    /// \note The return type of the function can be specified as a
577 578
    /// template parameter. For example,
578 579
    /// \code
579 580
    ///   cc.totalCost<double>();
580 581
    /// \endcode
581 582
    /// It is useful if the total cost cannot be stored in the \c Cost
582 583
    /// type of the algorithm, which is the default return type of the
583 584
    /// function.
584 585
    ///
585 586
    /// \pre \ref run() must be called before using this function.
586 587
    template <typename Number>
587 588
    Number totalCost() const {
588 589
      Number c = 0;
589 590
      for (ArcIt a(_graph); a != INVALID; ++a) {
590 591
        int i = _arc_idb[a];
591 592
        c += static_cast<Number>(_res_cap[i]) *
592 593
             (-static_cast<Number>(_cost[i]));
593 594
      }
594 595
      return c;
595 596
    }
596 597

	
597 598
#ifndef DOXYGEN
598 599
    Cost totalCost() const {
599 600
      return totalCost<Cost>();
600 601
    }
601 602
#endif
602 603

	
603 604
    /// \brief Return the flow on the given arc.
604 605
    ///
605 606
    /// This function returns the flow on the given arc.
606 607
    ///
607 608
    /// \pre \ref run() must be called before using this function.
608 609
    Value flow(const Arc& a) const {
609 610
      return _res_cap[_arc_idb[a]];
610 611
    }
611 612

	
612 613
    /// \brief Return the flow map (the primal solution).
613 614
    ///
614 615
    /// This function copies the flow value on each arc into the given
615 616
    /// map. The \c Value type of the algorithm must be convertible to
616 617
    /// the \c Value type of the map.
617 618
    ///
618 619
    /// \pre \ref run() must be called before using this function.
619 620
    template <typename FlowMap>
620 621
    void flowMap(FlowMap &map) const {
621 622
      for (ArcIt a(_graph); a != INVALID; ++a) {
622 623
        map.set(a, _res_cap[_arc_idb[a]]);
623 624
      }
624 625
    }
625 626

	
626 627
    /// \brief Return the potential (dual value) of the given node.
627 628
    ///
628 629
    /// This function returns the potential (dual value) of the
629 630
    /// given node.
630 631
    ///
631 632
    /// \pre \ref run() must be called before using this function.
632 633
    Cost potential(const Node& n) const {
633 634
      return static_cast<Cost>(_pi[_node_id[n]]);
634 635
    }
635 636

	
636 637
    /// \brief Return the potential map (the dual solution).
637 638
    ///
638 639
    /// This function copies the potential (dual value) of each node
639 640
    /// into the given map.
640 641
    /// The \c Cost type of the algorithm must be convertible to the
641 642
    /// \c Value type of the map.
642 643
    ///
643 644
    /// \pre \ref run() must be called before using this function.
644 645
    template <typename PotentialMap>
645 646
    void potentialMap(PotentialMap &map) const {
646 647
      for (NodeIt n(_graph); n != INVALID; ++n) {
647 648
        map.set(n, static_cast<Cost>(_pi[_node_id[n]]));
648 649
      }
649 650
    }
650 651

	
651 652
    /// @}
652 653

	
653 654
  private:
654 655

	
655 656
    // Initialize the algorithm
656 657
    ProblemType init() {
657 658
      if (_res_node_num <= 1) return INFEASIBLE;
658 659

	
659 660
      // Check the sum of supply values
660 661
      _sum_supply = 0;
661 662
      for (int i = 0; i != _root; ++i) {
662 663
        _sum_supply += _supply[i];
663 664
      }
664 665
      if (_sum_supply > 0) return INFEASIBLE;
665 666
      
666 667

	
667 668
      // Initialize vectors
668 669
      for (int i = 0; i != _res_node_num; ++i) {
669 670
        _pi[i] = 0;
670 671
      }
671 672
      ValueVector excess(_supply);
672 673
      
673 674
      // Remove infinite upper bounds and check negative arcs
674 675
      const Value MAX = std::numeric_limits<Value>::max();
675 676
      int last_out;
676 677
      if (_have_lower) {
677 678
        for (int i = 0; i != _root; ++i) {
678 679
          last_out = _first_out[i+1];
679 680
          for (int j = _first_out[i]; j != last_out; ++j) {
680 681
            if (_forward[j]) {
681 682
              Value c = _cost[j] < 0 ? _upper[j] : _lower[j];
682 683
              if (c >= MAX) return UNBOUNDED;
683 684
              excess[i] -= c;
684 685
              excess[_target[j]] += c;
685 686
            }
686 687
          }
687 688
        }
688 689
      } else {
689 690
        for (int i = 0; i != _root; ++i) {
690 691
          last_out = _first_out[i+1];
691 692
          for (int j = _first_out[i]; j != last_out; ++j) {
692 693
            if (_forward[j] && _cost[j] < 0) {
693 694
              Value c = _upper[j];
694 695
              if (c >= MAX) return UNBOUNDED;
695 696
              excess[i] -= c;
696 697
              excess[_target[j]] += c;
697 698
            }
698 699
          }
699 700
        }
700 701
      }
701 702
      Value ex, max_cap = 0;
702 703
      for (int i = 0; i != _res_node_num; ++i) {
703 704
        ex = excess[i];
704 705
        if (ex < 0) max_cap -= ex;
705 706
      }
706 707
      for (int j = 0; j != _res_arc_num; ++j) {
707 708
        if (_upper[j] >= MAX) _upper[j] = max_cap;
708 709
      }
709 710

	
710 711
      // Initialize maps for Circulation and remove non-zero lower bounds
711 712
      ConstMap<Arc, Value> low(0);
712 713
      typedef typename Digraph::template ArcMap<Value> ValueArcMap;
713 714
      typedef typename Digraph::template NodeMap<Value> ValueNodeMap;
714 715
      ValueArcMap cap(_graph), flow(_graph);
715 716
      ValueNodeMap sup(_graph);
716 717
      for (NodeIt n(_graph); n != INVALID; ++n) {
717 718
        sup[n] = _supply[_node_id[n]];
718 719
      }
719 720
      if (_have_lower) {
720 721
        for (ArcIt a(_graph); a != INVALID; ++a) {
721 722
          int j = _arc_idf[a];
722 723
          Value c = _lower[j];
723 724
          cap[a] = _upper[j] - c;
724 725
          sup[_graph.source(a)] -= c;
725 726
          sup[_graph.target(a)] += c;
726 727
        }
727 728
      } else {
728 729
        for (ArcIt a(_graph); a != INVALID; ++a) {
729 730
          cap[a] = _upper[_arc_idf[a]];
730 731
        }
731 732
      }
732 733

	
733 734
      // Find a feasible flow using Circulation
734 735
      Circulation<Digraph, ConstMap<Arc, Value>, ValueArcMap, ValueNodeMap>
735 736
        circ(_graph, low, cap, sup);
736 737
      if (!circ.flowMap(flow).run()) return INFEASIBLE;
737 738

	
738 739
      // Set residual capacities and handle GEQ supply type
739 740
      if (_sum_supply < 0) {
740 741
        for (ArcIt a(_graph); a != INVALID; ++a) {
741 742
          Value fa = flow[a];
742 743
          _res_cap[_arc_idf[a]] = cap[a] - fa;
743 744
          _res_cap[_arc_idb[a]] = fa;
744 745
          sup[_graph.source(a)] -= fa;
745 746
          sup[_graph.target(a)] += fa;
746 747
        }
747 748
        for (NodeIt n(_graph); n != INVALID; ++n) {
748 749
          excess[_node_id[n]] = sup[n];
749 750
        }
750 751
        for (int a = _first_out[_root]; a != _res_arc_num; ++a) {
751 752
          int u = _target[a];
752 753
          int ra = _reverse[a];
753 754
          _res_cap[a] = -_sum_supply + 1;
754 755
          _res_cap[ra] = -excess[u];
755 756
          _cost[a] = 0;
756 757
          _cost[ra] = 0;
757 758
        }
758 759
      } else {
759 760
        for (ArcIt a(_graph); a != INVALID; ++a) {
760 761
          Value fa = flow[a];
761 762
          _res_cap[_arc_idf[a]] = cap[a] - fa;
762 763
          _res_cap[_arc_idb[a]] = fa;
763 764
        }
764 765
        for (int a = _first_out[_root]; a != _res_arc_num; ++a) {
765 766
          int ra = _reverse[a];
766 767
          _res_cap[a] = 1;
767 768
          _res_cap[ra] = 0;
768 769
          _cost[a] = 0;
769 770
          _cost[ra] = 0;
770 771
        }
771 772
      }
772 773
      
773 774
      return OPTIMAL;
774 775
    }
775 776
    
776 777
    // Build a StaticDigraph structure containing the current
777 778
    // residual network
778 779
    void buildResidualNetwork() {
779 780
      _arc_vec.clear();
780 781
      _cost_vec.clear();
781 782
      _id_vec.clear();
782 783
      for (int j = 0; j != _res_arc_num; ++j) {
783 784
        if (_res_cap[j] > 0) {
784 785
          _arc_vec.push_back(IntPair(_source[j], _target[j]));
785 786
          _cost_vec.push_back(_cost[j]);
786 787
          _id_vec.push_back(j);
787 788
        }
788 789
      }
789 790
      _sgr.build(_res_node_num, _arc_vec.begin(), _arc_vec.end());
790 791
    }
791 792

	
792 793
    // Execute the algorithm and transform the results
793 794
    void start(Method method) {
794 795
      // Execute the algorithm
795 796
      switch (method) {
796 797
        case SIMPLE_CYCLE_CANCELING:
797 798
          startSimpleCycleCanceling();
798 799
          break;
799 800
        case MINIMUM_MEAN_CYCLE_CANCELING:
800 801
          startMinMeanCycleCanceling();
801 802
          break;
802 803
        case CANCEL_AND_TIGHTEN:
803 804
          startCancelAndTighten();
804 805
          break;
805 806
      }
806 807

	
807 808
      // Compute node potentials
808 809
      if (method != SIMPLE_CYCLE_CANCELING) {
809 810
        buildResidualNetwork();
810 811
        typename BellmanFord<StaticDigraph, CostArcMap>
811 812
          ::template SetDistMap<CostNodeMap>::Create bf(_sgr, _cost_map);
812 813
        bf.distMap(_pi_map);
813 814
        bf.init(0);
814 815
        bf.start();
815 816
      }
816 817

	
817 818
      // Handle non-zero lower bounds
818 819
      if (_have_lower) {
819 820
        int limit = _first_out[_root];
820 821
        for (int j = 0; j != limit; ++j) {
821 822
          if (!_forward[j]) _res_cap[j] += _lower[j];
822 823
        }
823 824
      }
824 825
    }
825 826

	
826 827
    // Execute the "Simple Cycle Canceling" method
827 828
    void startSimpleCycleCanceling() {
828 829
      // Constants for computing the iteration limits
829 830
      const int BF_FIRST_LIMIT  = 2;
830 831
      const double BF_LIMIT_FACTOR = 1.5;
831 832
      
832 833
      typedef StaticVectorMap<StaticDigraph::Arc, Value> FilterMap;
833 834
      typedef FilterArcs<StaticDigraph, FilterMap> ResDigraph;
834 835
      typedef StaticVectorMap<StaticDigraph::Node, StaticDigraph::Arc> PredMap;
835 836
      typedef typename BellmanFord<ResDigraph, CostArcMap>
836 837
        ::template SetDistMap<CostNodeMap>
837 838
        ::template SetPredMap<PredMap>::Create BF;
838 839
      
839 840
      // Build the residual network
840 841
      _arc_vec.clear();
841 842
      _cost_vec.clear();
842 843
      for (int j = 0; j != _res_arc_num; ++j) {
843 844
        _arc_vec.push_back(IntPair(_source[j], _target[j]));
844 845
        _cost_vec.push_back(_cost[j]);
845 846
      }
846 847
      _sgr.build(_res_node_num, _arc_vec.begin(), _arc_vec.end());
847 848

	
848 849
      FilterMap filter_map(_res_cap);
849 850
      ResDigraph rgr(_sgr, filter_map);
850 851
      std::vector<int> cycle;
851 852
      std::vector<StaticDigraph::Arc> pred(_res_arc_num);
852 853
      PredMap pred_map(pred);
853 854
      BF bf(rgr, _cost_map);
854 855
      bf.distMap(_pi_map).predMap(pred_map);
855 856

	
856 857
      int length_bound = BF_FIRST_LIMIT;
857 858
      bool optimal = false;
858 859
      while (!optimal) {
859 860
        bf.init(0);
860 861
        int iter_num = 0;
861 862
        bool cycle_found = false;
862 863
        while (!cycle_found) {
863 864
          // Perform some iterations of the Bellman-Ford algorithm
864 865
          int curr_iter_num = iter_num + length_bound <= _node_num ?
865 866
            length_bound : _node_num - iter_num;
866 867
          iter_num += curr_iter_num;
867 868
          int real_iter_num = curr_iter_num;
868 869
          for (int i = 0; i < curr_iter_num; ++i) {
869 870
            if (bf.processNextWeakRound()) {
870 871
              real_iter_num = i;
871 872
              break;
872 873
            }
873 874
          }
874 875
          if (real_iter_num < curr_iter_num) {
875 876
            // Optimal flow is found
876 877
            optimal = true;
877 878
            break;
878 879
          } else {
879 880
            // Search for node disjoint negative cycles
880 881
            std::vector<int> state(_res_node_num, 0);
881 882
            int id = 0;
882 883
            for (int u = 0; u != _res_node_num; ++u) {
883 884
              if (state[u] != 0) continue;
884 885
              ++id;
885 886
              int v = u;
886 887
              for (; v != -1 && state[v] == 0; v = pred[v] == INVALID ?
887 888
                   -1 : rgr.id(rgr.source(pred[v]))) {
888 889
                state[v] = id;
889 890
              }
890 891
              if (v != -1 && state[v] == id) {
891 892
                // A negative cycle is found
892 893
                cycle_found = true;
893 894
                cycle.clear();
894 895
                StaticDigraph::Arc a = pred[v];
895 896
                Value d, delta = _res_cap[rgr.id(a)];
896 897
                cycle.push_back(rgr.id(a));
897 898
                while (rgr.id(rgr.source(a)) != v) {
898 899
                  a = pred_map[rgr.source(a)];
899 900
                  d = _res_cap[rgr.id(a)];
900 901
                  if (d < delta) delta = d;
901 902
                  cycle.push_back(rgr.id(a));
902 903
                }
903 904

	
904 905
                // Augment along the cycle
905 906
                for (int i = 0; i < int(cycle.size()); ++i) {
906 907
                  int j = cycle[i];
907 908
                  _res_cap[j] -= delta;
908 909
                  _res_cap[_reverse[j]] += delta;
909 910
                }
910 911
              }
911 912
            }
912 913
          }
913 914

	
914 915
          // Increase iteration limit if no cycle is found
915 916
          if (!cycle_found) {
916 917
            length_bound = static_cast<int>(length_bound * BF_LIMIT_FACTOR);
917 918
          }
918 919
        }
919 920
      }
920 921
    }
921 922

	
922 923
    // Execute the "Minimum Mean Cycle Canceling" method
923 924
    void startMinMeanCycleCanceling() {
924 925
      typedef SimplePath<StaticDigraph> SPath;
925 926
      typedef typename SPath::ArcIt SPathArcIt;
926 927
      typedef typename Howard<StaticDigraph, CostArcMap>
927 928
        ::template SetPath<SPath>::Create MMC;
928 929
      
929 930
      SPath cycle;
930 931
      MMC mmc(_sgr, _cost_map);
931 932
      mmc.cycle(cycle);
932 933
      buildResidualNetwork();
933 934
      while (mmc.findMinMean() && mmc.cycleLength() < 0) {
934 935
        // Find the cycle
935 936
        mmc.findCycle();
936 937

	
937 938
        // Compute delta value
938 939
        Value delta = INF;
939 940
        for (SPathArcIt a(cycle); a != INVALID; ++a) {
940 941
          Value d = _res_cap[_id_vec[_sgr.id(a)]];
941 942
          if (d < delta) delta = d;
942 943
        }
943 944

	
944 945
        // Augment along the cycle
945 946
        for (SPathArcIt a(cycle); a != INVALID; ++a) {
946 947
          int j = _id_vec[_sgr.id(a)];
947 948
          _res_cap[j] -= delta;
948 949
          _res_cap[_reverse[j]] += delta;
949 950
        }
950 951

	
951 952
        // Rebuild the residual network        
952 953
        buildResidualNetwork();
953 954
      }
954 955
    }
955 956

	
956 957
    // Execute the "Cancel And Tighten" method
957 958
    void startCancelAndTighten() {
958 959
      // Constants for the min mean cycle computations
959 960
      const double LIMIT_FACTOR = 1.0;
960 961
      const int MIN_LIMIT = 5;
961 962

	
962 963
      // Contruct auxiliary data vectors
963 964
      DoubleVector pi(_res_node_num, 0.0);
964 965
      IntVector level(_res_node_num);
965
      CharVector reached(_res_node_num);
966
      CharVector processed(_res_node_num);
966
      BoolVector reached(_res_node_num);
967
      BoolVector processed(_res_node_num);
967 968
      IntVector pred_node(_res_node_num);
968 969
      IntVector pred_arc(_res_node_num);
969 970
      std::vector<int> stack(_res_node_num);
970 971
      std::vector<int> proc_vector(_res_node_num);
971 972

	
972 973
      // Initialize epsilon
973 974
      double epsilon = 0;
974 975
      for (int a = 0; a != _res_arc_num; ++a) {
975 976
        if (_res_cap[a] > 0 && -_cost[a] > epsilon)
976 977
          epsilon = -_cost[a];
977 978
      }
978 979

	
979 980
      // Start phases
980 981
      Tolerance<double> tol;
981 982
      tol.epsilon(1e-6);
982 983
      int limit = int(LIMIT_FACTOR * std::sqrt(double(_res_node_num)));
983 984
      if (limit < MIN_LIMIT) limit = MIN_LIMIT;
984 985
      int iter = limit;
985 986
      while (epsilon * _res_node_num >= 1) {
986 987
        // Find and cancel cycles in the admissible network using DFS
987 988
        for (int u = 0; u != _res_node_num; ++u) {
988 989
          reached[u] = false;
989 990
          processed[u] = false;
990 991
        }
991 992
        int stack_head = -1;
992 993
        int proc_head = -1;
993 994
        for (int start = 0; start != _res_node_num; ++start) {
994 995
          if (reached[start]) continue;
995 996

	
996 997
          // New start node
997 998
          reached[start] = true;
998 999
          pred_arc[start] = -1;
999 1000
          pred_node[start] = -1;
1000 1001

	
1001 1002
          // Find the first admissible outgoing arc
1002 1003
          double p = pi[start];
1003 1004
          int a = _first_out[start];
1004 1005
          int last_out = _first_out[start+1];
1005 1006
          for (; a != last_out && (_res_cap[a] == 0 ||
1006 1007
               !tol.negative(_cost[a] + p - pi[_target[a]])); ++a) ;
1007 1008
          if (a == last_out) {
1008 1009
            processed[start] = true;
1009 1010
            proc_vector[++proc_head] = start;
1010 1011
            continue;
1011 1012
          }
1012 1013
          stack[++stack_head] = a;
1013 1014

	
1014 1015
          while (stack_head >= 0) {
1015 1016
            int sa = stack[stack_head];
1016 1017
            int u = _source[sa];
1017 1018
            int v = _target[sa];
1018 1019

	
1019 1020
            if (!reached[v]) {
1020 1021
              // A new node is reached
1021 1022
              reached[v] = true;
1022 1023
              pred_node[v] = u;
1023 1024
              pred_arc[v] = sa;
1024 1025
              p = pi[v];
1025 1026
              a = _first_out[v];
1026 1027
              last_out = _first_out[v+1];
1027 1028
              for (; a != last_out && (_res_cap[a] == 0 ||
1028 1029
                   !tol.negative(_cost[a] + p - pi[_target[a]])); ++a) ;
1029 1030
              stack[++stack_head] = a == last_out ? -1 : a;
1030 1031
            } else {
1031 1032
              if (!processed[v]) {
1032 1033
                // A cycle is found
1033 1034
                int n, w = u;
1034 1035
                Value d, delta = _res_cap[sa];
1035 1036
                for (n = u; n != v; n = pred_node[n]) {
1036 1037
                  d = _res_cap[pred_arc[n]];
1037 1038
                  if (d <= delta) {
1038 1039
                    delta = d;
1039 1040
                    w = pred_node[n];
1040 1041
                  }
1041 1042
                }
1042 1043

	
1043 1044
                // Augment along the cycle
1044 1045
                _res_cap[sa] -= delta;
1045 1046
                _res_cap[_reverse[sa]] += delta;
1046 1047
                for (n = u; n != v; n = pred_node[n]) {
1047 1048
                  int pa = pred_arc[n];
1048 1049
                  _res_cap[pa] -= delta;
1049 1050
                  _res_cap[_reverse[pa]] += delta;
1050 1051
                }
1051 1052
                for (n = u; stack_head > 0 && n != w; n = pred_node[n]) {
1052 1053
                  --stack_head;
1053 1054
                  reached[n] = false;
1054 1055
                }
1055 1056
                u = w;
1056 1057
              }
1057 1058
              v = u;
1058 1059

	
1059 1060
              // Find the next admissible outgoing arc
1060 1061
              p = pi[v];
1061 1062
              a = stack[stack_head] + 1;
1062 1063
              last_out = _first_out[v+1];
1063 1064
              for (; a != last_out && (_res_cap[a] == 0 ||
1064 1065
                   !tol.negative(_cost[a] + p - pi[_target[a]])); ++a) ;
1065 1066
              stack[stack_head] = a == last_out ? -1 : a;
1066 1067
            }
1067 1068

	
1068 1069
            while (stack_head >= 0 && stack[stack_head] == -1) {
1069 1070
              processed[v] = true;
1070 1071
              proc_vector[++proc_head] = v;
1071 1072
              if (--stack_head >= 0) {
1072 1073
                // Find the next admissible outgoing arc
1073 1074
                v = _source[stack[stack_head]];
1074 1075
                p = pi[v];
1075 1076
                a = stack[stack_head] + 1;
1076 1077
                last_out = _first_out[v+1];
1077 1078
                for (; a != last_out && (_res_cap[a] == 0 ||
1078 1079
                     !tol.negative(_cost[a] + p - pi[_target[a]])); ++a) ;
1079 1080
                stack[stack_head] = a == last_out ? -1 : a;
1080 1081
              }
1081 1082
            }
1082 1083
          }
1083 1084
        }
1084 1085

	
1085 1086
        // Tighten potentials and epsilon
1086 1087
        if (--iter > 0) {
1087 1088
          for (int u = 0; u != _res_node_num; ++u) {
1088 1089
            level[u] = 0;
1089 1090
          }
1090 1091
          for (int i = proc_head; i > 0; --i) {
1091 1092
            int u = proc_vector[i];
1092 1093
            double p = pi[u];
1093 1094
            int l = level[u] + 1;
1094 1095
            int last_out = _first_out[u+1];
1095 1096
            for (int a = _first_out[u]; a != last_out; ++a) {
1096 1097
              int v = _target[a];
1097 1098
              if (_res_cap[a] > 0 && tol.negative(_cost[a] + p - pi[v]) &&
1098 1099
                  l > level[v]) level[v] = l;
1099 1100
            }
1100 1101
          }
1101 1102

	
1102 1103
          // Modify potentials
1103 1104
          double q = std::numeric_limits<double>::max();
1104 1105
          for (int u = 0; u != _res_node_num; ++u) {
1105 1106
            int lu = level[u];
1106 1107
            double p, pu = pi[u];
1107 1108
            int last_out = _first_out[u+1];
1108 1109
            for (int a = _first_out[u]; a != last_out; ++a) {
1109 1110
              if (_res_cap[a] == 0) continue;
1110 1111
              int v = _target[a];
1111 1112
              int ld = lu - level[v];
1112 1113
              if (ld > 0) {
1113 1114
                p = (_cost[a] + pu - pi[v] + epsilon) / (ld + 1);
1114 1115
                if (p < q) q = p;
1115 1116
              }
1116 1117
            }
1117 1118
          }
1118 1119
          for (int u = 0; u != _res_node_num; ++u) {
1119 1120
            pi[u] -= q * level[u];
1120 1121
          }
1121 1122

	
1122 1123
          // Modify epsilon
1123 1124
          epsilon = 0;
1124 1125
          for (int u = 0; u != _res_node_num; ++u) {
1125 1126
            double curr, pu = pi[u];
1126 1127
            int last_out = _first_out[u+1];
1127 1128
            for (int a = _first_out[u]; a != last_out; ++a) {
1128 1129
              if (_res_cap[a] == 0) continue;
1129 1130
              curr = _cost[a] + pu - pi[_target[a]];
1130 1131
              if (-curr > epsilon) epsilon = -curr;
1131 1132
            }
1132 1133
          }
1133 1134
        } else {
1134 1135
          typedef Howard<StaticDigraph, CostArcMap> MMC;
1135 1136
          typedef typename BellmanFord<StaticDigraph, CostArcMap>
1136 1137
            ::template SetDistMap<CostNodeMap>::Create BF;
1137 1138

	
1138 1139
          // Set epsilon to the minimum cycle mean
1139 1140
          buildResidualNetwork();
1140 1141
          MMC mmc(_sgr, _cost_map);
1141 1142
          mmc.findMinMean();
1142 1143
          epsilon = -mmc.cycleMean();
1143 1144
          Cost cycle_cost = mmc.cycleLength();
1144 1145
          int cycle_size = mmc.cycleArcNum();
1145 1146
          
1146 1147
          // Compute feasible potentials for the current epsilon
1147 1148
          for (int i = 0; i != int(_cost_vec.size()); ++i) {
1148 1149
            _cost_vec[i] = cycle_size * _cost_vec[i] - cycle_cost;
1149 1150
          }
1150 1151
          BF bf(_sgr, _cost_map);
1151 1152
          bf.distMap(_pi_map);
1152 1153
          bf.init(0);
1153 1154
          bf.start();
1154 1155
          for (int u = 0; u != _res_node_num; ++u) {
1155 1156
            pi[u] = static_cast<double>(_pi[u]) / cycle_size;
1156 1157
          }
1157 1158
        
1158 1159
          iter = limit;
1159 1160
        }
1160 1161
      }
1161 1162
    }
1162 1163

	
1163 1164
  }; //class CycleCanceling
1164 1165

	
1165 1166
  ///@}
1166 1167

	
1167 1168
} //namespace lemon
1168 1169

	
1169 1170
#endif //LEMON_CYCLE_CANCELING_H
Ignore white space 768 line context
1 1
/* -*- mode: C++; indent-tabs-mode: nil; -*-
2 2
 *
3 3
 * This file is a part of LEMON, a generic C++ optimization library.
4 4
 *
5 5
 * Copyright (C) 2003-2009
6 6
 * Egervary Jeno Kombinatorikus Optimalizalasi Kutatocsoport
7 7
 * (Egervary Research Group on Combinatorial Optimization, EGRES).
8 8
 *
9 9
 * Permission to use, modify and distribute this software is granted
10 10
 * provided that this copyright notice appears in all copies. For
11 11
 * precise terms see the accompanying LICENSE file.
12 12
 *
13 13
 * This software is provided "AS IS" with no warranty of any kind,
14 14
 * express or implied, and with no claim as to its suitability for any
15 15
 * purpose.
16 16
 *
17 17
 */
18 18

	
19 19
#ifndef LEMON_NETWORK_SIMPLEX_H
20 20
#define LEMON_NETWORK_SIMPLEX_H
21 21

	
22 22
/// \ingroup min_cost_flow_algs
23 23
///
24 24
/// \file
25 25
/// \brief Network Simplex algorithm for finding a minimum cost flow.
26 26

	
27 27
#include <vector>
28 28
#include <limits>
29 29
#include <algorithm>
30 30

	
31 31
#include <lemon/core.h>
32 32
#include <lemon/math.h>
33 33

	
34 34
namespace lemon {
35 35

	
36 36
  /// \addtogroup min_cost_flow_algs
37 37
  /// @{
38 38

	
39 39
  /// \brief Implementation of the primal Network Simplex algorithm
40 40
  /// for finding a \ref min_cost_flow "minimum cost flow".
41 41
  ///
42 42
  /// \ref NetworkSimplex implements the primal Network Simplex algorithm
43 43
  /// for finding a \ref min_cost_flow "minimum cost flow"
44 44
  /// \ref amo93networkflows, \ref dantzig63linearprog,
45 45
  /// \ref kellyoneill91netsimplex.
46 46
  /// This algorithm is a highly efficient specialized version of the
47 47
  /// linear programming simplex method directly for the minimum cost
48 48
  /// flow problem.
49 49
  ///
50 50
  /// In general, %NetworkSimplex is the fastest implementation available
51 51
  /// in LEMON for this problem.
52 52
  /// Moreover, it supports both directions of the supply/demand inequality
53 53
  /// constraints. For more information, see \ref SupplyType.
54 54
  ///
55 55
  /// Most of the parameters of the problem (except for the digraph)
56 56
  /// can be given using separate functions, and the algorithm can be
57 57
  /// executed using the \ref run() function. If some parameters are not
58 58
  /// specified, then default values will be used.
59 59
  ///
60 60
  /// \tparam GR The digraph type the algorithm runs on.
61 61
  /// \tparam V The number type used for flow amounts, capacity bounds
62 62
  /// and supply values in the algorithm. By default, it is \c int.
63 63
  /// \tparam C The number type used for costs and potentials in the
64 64
  /// algorithm. By default, it is the same as \c V.
65 65
  ///
66 66
  /// \warning Both number types must be signed and all input data must
67 67
  /// be integer.
68 68
  ///
69 69
  /// \note %NetworkSimplex provides five different pivot rule
70 70
  /// implementations, from which the most efficient one is used
71 71
  /// by default. For more information, see \ref PivotRule.
72 72
  template <typename GR, typename V = int, typename C = V>
73 73
  class NetworkSimplex
74 74
  {
75 75
  public:
76 76

	
77 77
    /// The type of the flow amounts, capacity bounds and supply values
78 78
    typedef V Value;
79 79
    /// The type of the arc costs
80 80
    typedef C Cost;
81 81

	
82 82
  public:
83 83

	
84 84
    /// \brief Problem type constants for the \c run() function.
85 85
    ///
86 86
    /// Enum type containing the problem type constants that can be
87 87
    /// returned by the \ref run() function of the algorithm.
88 88
    enum ProblemType {
89 89
      /// The problem has no feasible solution (flow).
90 90
      INFEASIBLE,
91 91
      /// The problem has optimal solution (i.e. it is feasible and
92 92
      /// bounded), and the algorithm has found optimal flow and node
93 93
      /// potentials (primal and dual solutions).
94 94
      OPTIMAL,
95 95
      /// The objective function of the problem is unbounded, i.e.
96 96
      /// there is a directed cycle having negative total cost and
97 97
      /// infinite upper bound.
98 98
      UNBOUNDED
99 99
    };
100 100
    
101 101
    /// \brief Constants for selecting the type of the supply constraints.
102 102
    ///
103 103
    /// Enum type containing constants for selecting the supply type,
104 104
    /// i.e. the direction of the inequalities in the supply/demand
105 105
    /// constraints of the \ref min_cost_flow "minimum cost flow problem".
106 106
    ///
107 107
    /// The default supply type is \c GEQ, the \c LEQ type can be
108 108
    /// selected using \ref supplyType().
109 109
    /// The equality form is a special case of both supply types.
110 110
    enum SupplyType {
111 111
      /// This option means that there are <em>"greater or equal"</em>
112 112
      /// supply/demand constraints in the definition of the problem.
113 113
      GEQ,
114 114
      /// This option means that there are <em>"less or equal"</em>
115 115
      /// supply/demand constraints in the definition of the problem.
116 116
      LEQ
117 117
    };
118 118
    
119 119
    /// \brief Constants for selecting the pivot rule.
120 120
    ///
121 121
    /// Enum type containing constants for selecting the pivot rule for
122 122
    /// the \ref run() function.
123 123
    ///
124 124
    /// \ref NetworkSimplex provides five different pivot rule
125 125
    /// implementations that significantly affect the running time
126 126
    /// of the algorithm.
127 127
    /// By default, \ref BLOCK_SEARCH "Block Search" is used, which
128 128
    /// proved to be the most efficient and the most robust on various
129 129
    /// test inputs.
130 130
    /// However, another pivot rule can be selected using the \ref run()
131 131
    /// function with the proper parameter.
132 132
    enum PivotRule {
133 133

	
134 134
      /// The \e First \e Eligible pivot rule.
135 135
      /// The next eligible arc is selected in a wraparound fashion
136 136
      /// in every iteration.
137 137
      FIRST_ELIGIBLE,
138 138

	
139 139
      /// The \e Best \e Eligible pivot rule.
140 140
      /// The best eligible arc is selected in every iteration.
141 141
      BEST_ELIGIBLE,
142 142

	
143 143
      /// The \e Block \e Search pivot rule.
144 144
      /// A specified number of arcs are examined in every iteration
145 145
      /// in a wraparound fashion and the best eligible arc is selected
146 146
      /// from this block.
147 147
      BLOCK_SEARCH,
148 148

	
149 149
      /// The \e Candidate \e List pivot rule.
150 150
      /// In a major iteration a candidate list is built from eligible arcs
151 151
      /// in a wraparound fashion and in the following minor iterations
152 152
      /// the best eligible arc is selected from this list.
153 153
      CANDIDATE_LIST,
154 154

	
155 155
      /// The \e Altering \e Candidate \e List pivot rule.
156 156
      /// It is a modified version of the Candidate List method.
157 157
      /// It keeps only the several best eligible arcs from the former
158 158
      /// candidate list and extends this list in every iteration.
159 159
      ALTERING_LIST
160 160
    };
161 161
    
162 162
  private:
163 163

	
164 164
    TEMPLATE_DIGRAPH_TYPEDEFS(GR);
165 165

	
166 166
    typedef std::vector<int> IntVector;
167
    typedef std::vector<char> CharVector;
168 167
    typedef std::vector<Value> ValueVector;
169 168
    typedef std::vector<Cost> CostVector;
169
    typedef std::vector<char> BoolVector;
170
    // Note: vector<char> is used instead of vector<bool> for efficiency reasons
170 171

	
171 172
    // State constants for arcs
172 173
    enum ArcStateEnum {
173 174
      STATE_UPPER = -1,
174 175
      STATE_TREE  =  0,
175 176
      STATE_LOWER =  1
176 177
    };
177 178

	
178 179
  private:
179 180

	
180 181
    // Data related to the underlying digraph
181 182
    const GR &_graph;
182 183
    int _node_num;
183 184
    int _arc_num;
184 185
    int _all_arc_num;
185 186
    int _search_arc_num;
186 187

	
187 188
    // Parameters of the problem
188 189
    bool _have_lower;
189 190
    SupplyType _stype;
190 191
    Value _sum_supply;
191 192

	
192 193
    // Data structures for storing the digraph
193 194
    IntNodeMap _node_id;
194 195
    IntArcMap _arc_id;
195 196
    IntVector _source;
196 197
    IntVector _target;
197 198
    bool _arc_mixing;
198 199

	
199 200
    // Node and arc data
200 201
    ValueVector _lower;
201 202
    ValueVector _upper;
202 203
    ValueVector _cap;
203 204
    CostVector _cost;
204 205
    ValueVector _supply;
205 206
    ValueVector _flow;
206 207
    CostVector _pi;
207 208

	
208 209
    // Data for storing the spanning tree structure
209 210
    IntVector _parent;
210 211
    IntVector _pred;
211 212
    IntVector _thread;
212 213
    IntVector _rev_thread;
213 214
    IntVector _succ_num;
214 215
    IntVector _last_succ;
215 216
    IntVector _dirty_revs;
216
    CharVector _forward;
217
    CharVector _state;
217
    BoolVector _forward;
218
    BoolVector _state;
218 219
    int _root;
219 220

	
220 221
    // Temporary data used in the current pivot iteration
221 222
    int in_arc, join, u_in, v_in, u_out, v_out;
222 223
    int first, second, right, last;
223 224
    int stem, par_stem, new_stem;
224 225
    Value delta;
225 226
    
226 227
    const Value MAX;
227 228

	
228 229
  public:
229 230
  
230 231
    /// \brief Constant for infinite upper bounds (capacities).
231 232
    ///
232 233
    /// Constant for infinite upper bounds (capacities).
233 234
    /// It is \c std::numeric_limits<Value>::infinity() if available,
234 235
    /// \c std::numeric_limits<Value>::max() otherwise.
235 236
    const Value INF;
236 237

	
237 238
  private:
238 239

	
239 240
    // Implementation of the First Eligible pivot rule
240 241
    class FirstEligiblePivotRule
241 242
    {
242 243
    private:
243 244

	
244 245
      // References to the NetworkSimplex class
245 246
      const IntVector  &_source;
246 247
      const IntVector  &_target;
247 248
      const CostVector &_cost;
248
      const CharVector &_state;
249
      const BoolVector &_state;
249 250
      const CostVector &_pi;
250 251
      int &_in_arc;
251 252
      int _search_arc_num;
252 253

	
253 254
      // Pivot rule data
254 255
      int _next_arc;
255 256

	
256 257
    public:
257 258

	
258 259
      // Constructor
259 260
      FirstEligiblePivotRule(NetworkSimplex &ns) :
260 261
        _source(ns._source), _target(ns._target),
261 262
        _cost(ns._cost), _state(ns._state), _pi(ns._pi),
262 263
        _in_arc(ns.in_arc), _search_arc_num(ns._search_arc_num),
263 264
        _next_arc(0)
264 265
      {}
265 266

	
266 267
      // Find next entering arc
267 268
      bool findEnteringArc() {
268 269
        Cost c;
269
        for (int e = _next_arc; e < _search_arc_num; ++e) {
270
        for (int e = _next_arc; e != _search_arc_num; ++e) {
270 271
          c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]);
271 272
          if (c < 0) {
272 273
            _in_arc = e;
273 274
            _next_arc = e + 1;
274 275
            return true;
275 276
          }
276 277
        }
277
        for (int e = 0; e < _next_arc; ++e) {
278
        for (int e = 0; e != _next_arc; ++e) {
278 279
          c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]);
279 280
          if (c < 0) {
280 281
            _in_arc = e;
281 282
            _next_arc = e + 1;
282 283
            return true;
283 284
          }
284 285
        }
285 286
        return false;
286 287
      }
287 288

	
288 289
    }; //class FirstEligiblePivotRule
289 290

	
290 291

	
291 292
    // Implementation of the Best Eligible pivot rule
292 293
    class BestEligiblePivotRule
293 294
    {
294 295
    private:
295 296

	
296 297
      // References to the NetworkSimplex class
297 298
      const IntVector  &_source;
298 299
      const IntVector  &_target;
299 300
      const CostVector &_cost;
300
      const CharVector &_state;
301
      const BoolVector &_state;
301 302
      const CostVector &_pi;
302 303
      int &_in_arc;
303 304
      int _search_arc_num;
304 305

	
305 306
    public:
306 307

	
307 308
      // Constructor
308 309
      BestEligiblePivotRule(NetworkSimplex &ns) :
309 310
        _source(ns._source), _target(ns._target),
310 311
        _cost(ns._cost), _state(ns._state), _pi(ns._pi),
311 312
        _in_arc(ns.in_arc), _search_arc_num(ns._search_arc_num)
312 313
      {}
313 314

	
314 315
      // Find next entering arc
315 316
      bool findEnteringArc() {
316 317
        Cost c, min = 0;
317
        for (int e = 0; e < _search_arc_num; ++e) {
318
        for (int e = 0; e != _search_arc_num; ++e) {
318 319
          c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]);
319 320
          if (c < min) {
320 321
            min = c;
321 322
            _in_arc = e;
322 323
          }
323 324
        }
324 325
        return min < 0;
325 326
      }
326 327

	
327 328
    }; //class BestEligiblePivotRule
328 329

	
329 330

	
330 331
    // Implementation of the Block Search pivot rule
331 332
    class BlockSearchPivotRule
332 333
    {
333 334
    private:
334 335

	
335 336
      // References to the NetworkSimplex class
336 337
      const IntVector  &_source;
337 338
      const IntVector  &_target;
338 339
      const CostVector &_cost;
339
      const CharVector &_state;
340
      const BoolVector &_state;
340 341
      const CostVector &_pi;
341 342
      int &_in_arc;
342 343
      int _search_arc_num;
343 344

	
344 345
      // Pivot rule data
345 346
      int _block_size;
346 347
      int _next_arc;
347 348

	
348 349
    public:
349 350

	
350 351
      // Constructor
351 352
      BlockSearchPivotRule(NetworkSimplex &ns) :
352 353
        _source(ns._source), _target(ns._target),
353 354
        _cost(ns._cost), _state(ns._state), _pi(ns._pi),
354 355
        _in_arc(ns.in_arc), _search_arc_num(ns._search_arc_num),
355 356
        _next_arc(0)
356 357
      {
357 358
        // The main parameters of the pivot rule
358
        const double BLOCK_SIZE_FACTOR = 0.5;
359
        const double BLOCK_SIZE_FACTOR = 1.0;
359 360
        const int MIN_BLOCK_SIZE = 10;
360 361

	
361 362
        _block_size = std::max( int(BLOCK_SIZE_FACTOR *
362 363
                                    std::sqrt(double(_search_arc_num))),
363 364
                                MIN_BLOCK_SIZE );
364 365
      }
365 366

	
366 367
      // Find next entering arc
367 368
      bool findEnteringArc() {
368 369
        Cost c, min = 0;
369 370
        int cnt = _block_size;
370 371
        int e;
371
        for (e = _next_arc; e < _search_arc_num; ++e) {
372
        for (e = _next_arc; e != _search_arc_num; ++e) {
372 373
          c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]);
373 374
          if (c < min) {
374 375
            min = c;
375 376
            _in_arc = e;
376 377
          }
377 378
          if (--cnt == 0) {
378 379
            if (min < 0) goto search_end;
379 380
            cnt = _block_size;
380 381
          }
381 382
        }
382
        for (e = 0; e < _next_arc; ++e) {
383
        for (e = 0; e != _next_arc; ++e) {
383 384
          c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]);
384 385
          if (c < min) {
385 386
            min = c;
386 387
            _in_arc = e;
387 388
          }
388 389
          if (--cnt == 0) {
389 390
            if (min < 0) goto search_end;
390 391
            cnt = _block_size;
391 392
          }
392 393
        }
393 394
        if (min >= 0) return false;
394 395

	
395 396
      search_end:
396 397
        _next_arc = e;
397 398
        return true;
398 399
      }
399 400

	
400 401
    }; //class BlockSearchPivotRule
401 402

	
402 403

	
403 404
    // Implementation of the Candidate List pivot rule
404 405
    class CandidateListPivotRule
405 406
    {
406 407
    private:
407 408

	
408 409
      // References to the NetworkSimplex class
409 410
      const IntVector  &_source;
410 411
      const IntVector  &_target;
411 412
      const CostVector &_cost;
412
      const CharVector &_state;
413
      const BoolVector &_state;
413 414
      const CostVector &_pi;
414 415
      int &_in_arc;
415 416
      int _search_arc_num;
416 417

	
417 418
      // Pivot rule data
418 419
      IntVector _candidates;
419 420
      int _list_length, _minor_limit;
420 421
      int _curr_length, _minor_count;
421 422
      int _next_arc;
422 423

	
423 424
    public:
424 425

	
425 426
      /// Constructor
426 427
      CandidateListPivotRule(NetworkSimplex &ns) :
427 428
        _source(ns._source), _target(ns._target),
428 429
        _cost(ns._cost), _state(ns._state), _pi(ns._pi),
429 430
        _in_arc(ns.in_arc), _search_arc_num(ns._search_arc_num),
430 431
        _next_arc(0)
431 432
      {
432 433
        // The main parameters of the pivot rule
433 434
        const double LIST_LENGTH_FACTOR = 0.25;
434 435
        const int MIN_LIST_LENGTH = 10;
435 436
        const double MINOR_LIMIT_FACTOR = 0.1;
436 437
        const int MIN_MINOR_LIMIT = 3;
437 438

	
438 439
        _list_length = std::max( int(LIST_LENGTH_FACTOR *
439 440
                                     std::sqrt(double(_search_arc_num))),
440 441
                                 MIN_LIST_LENGTH );
441 442
        _minor_limit = std::max( int(MINOR_LIMIT_FACTOR * _list_length),
442 443
                                 MIN_MINOR_LIMIT );
443 444
        _curr_length = _minor_count = 0;
444 445
        _candidates.resize(_list_length);
445 446
      }
446 447

	
447 448
      /// Find next entering arc
448 449
      bool findEnteringArc() {
449 450
        Cost min, c;
450 451
        int e;
451 452
        if (_curr_length > 0 && _minor_count < _minor_limit) {
452 453
          // Minor iteration: select the best eligible arc from the
453 454
          // current candidate list
454 455
          ++_minor_count;
455 456
          min = 0;
456 457
          for (int i = 0; i < _curr_length; ++i) {
457 458
            e = _candidates[i];
458 459
            c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]);
459 460
            if (c < min) {
460 461
              min = c;
461 462
              _in_arc = e;
462 463
            }
463 464
            else if (c >= 0) {
464 465
              _candidates[i--] = _candidates[--_curr_length];
465 466
            }
466 467
          }
467 468
          if (min < 0) return true;
468 469
        }
469 470

	
470 471
        // Major iteration: build a new candidate list
471 472
        min = 0;
472 473
        _curr_length = 0;
473
        for (e = _next_arc; e < _search_arc_num; ++e) {
474
        for (e = _next_arc; e != _search_arc_num; ++e) {
474 475
          c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]);
475 476
          if (c < 0) {
476 477
            _candidates[_curr_length++] = e;
477 478
            if (c < min) {
478 479
              min = c;
479 480
              _in_arc = e;
480 481
            }
481 482
            if (_curr_length == _list_length) goto search_end;
482 483
          }
483 484
        }
484
        for (e = 0; e < _next_arc; ++e) {
485
        for (e = 0; e != _next_arc; ++e) {
485 486
          c = _state[e] * (_cost[e] + _pi[_source[e]] - _pi[_target[e]]);
486 487
          if (c < 0) {
487 488
            _candidates[_curr_length++] = e;
488 489
            if (c < min) {
489 490
              min = c;
490 491
              _in_arc = e;
491 492
            }
492 493
            if (_curr_length == _list_length) goto search_end;
493 494
          }
494 495
        }
495 496
        if (_curr_length == 0) return false;
496 497
      
497 498
      search_end:        
498 499
        _minor_count = 1;
499 500
        _next_arc = e;
500 501
        return true;
501 502
      }
502 503

	
503 504
    }; //class CandidateListPivotRule
504 505

	
505 506

	
506 507
    // Implementation of the Altering Candidate List pivot rule
507 508
    class AlteringListPivotRule
508 509
    {
509 510
    private:
510 511

	
511 512
      // References to the NetworkSimplex class
512 513
      const IntVector  &_source;
513 514
      const IntVector  &_target;
514 515
      const CostVector &_cost;
515
      const CharVector &_state;
516
      const BoolVector &_state;
516 517
      const CostVector &_pi;
517 518
      int &_in_arc;
518 519
      int _search_arc_num;
519 520

	
520 521
      // Pivot rule data
521 522
      int _block_size, _head_length, _curr_length;
522 523
      int _next_arc;
523 524
      IntVector _candidates;
524 525
      CostVector _cand_cost;
525 526

	
526 527
      // Functor class to compare arcs during sort of the candidate list
527 528
      class SortFunc
528 529
      {
529 530
      private:
530 531
        const CostVector &_map;
531 532
      public:
532 533
        SortFunc(const CostVector &map) : _map(map) {}
533 534
        bool operator()(int left, int right) {
534 535
          return _map[left] > _map[right];
535 536
        }
536 537
      };
537 538

	
538 539
      SortFunc _sort_func;
539 540

	
540 541
    public:
541 542

	
542 543
      // Constructor
543 544
      AlteringListPivotRule(NetworkSimplex &ns) :
544 545
        _source(ns._source), _target(ns._target),
545 546
        _cost(ns._cost), _state(ns._state), _pi(ns._pi),
546 547
        _in_arc(ns.in_arc), _search_arc_num(ns._search_arc_num),
547 548
        _next_arc(0), _cand_cost(ns._search_arc_num), _sort_func(_cand_cost)
548 549
      {
549 550
        // The main parameters of the pivot rule
550 551
        const double BLOCK_SIZE_FACTOR = 1.0;
551 552
        const int MIN_BLOCK_SIZE = 10;
552 553
        const double HEAD_LENGTH_FACTOR = 0.1;
553 554
        const int MIN_HEAD_LENGTH = 3;
554 555

	
555 556
        _block_size = std::max( int(BLOCK_SIZE_FACTOR *
556 557
                                    std::sqrt(double(_search_arc_num))),
557 558
                                MIN_BLOCK_SIZE );
558 559
        _head_length = std::max( int(HEAD_LENGTH_FACTOR * _block_size),
559 560
                                 MIN_HEAD_LENGTH );
560 561
        _candidates.resize(_head_length + _block_size);
561 562
        _curr_length = 0;
562 563
      }
563 564

	
564 565
      // Find next entering arc
565 566
      bool findEnteringArc() {
566 567
        // Check the current candidate list
567 568
        int e;
568
        for (int i = 0; i < _curr_length; ++i) {
569
        for (int i = 0; i != _curr_length; ++i) {
569 570
          e = _candidates[i];
570 571
          _cand_cost[e] = _state[e] *
571 572
            (_cost[e] + _pi[_source[e]] - _pi[_target[e]]);
572 573
          if (_cand_cost[e] >= 0) {
573 574
            _candidates[i--] = _candidates[--_curr_length];
574 575
          }
575 576
        }
576 577

	
577 578
        // Extend the list
578 579
        int cnt = _block_size;
579 580
        int limit = _head_length;
580 581

	
581
        for (e = _next_arc; e < _search_arc_num; ++e) {
582
        for (e = _next_arc; e != _search_arc_num; ++e) {
582 583
          _cand_cost[e] = _state[e] *
583 584
            (_cost[e] + _pi[_source[e]] - _pi[_target[e]]);
584 585
          if (_cand_cost[e] < 0) {
585 586
            _candidates[_curr_length++] = e;
586 587
          }
587 588
          if (--cnt == 0) {
588 589
            if (_curr_length > limit) goto search_end;
589 590
            limit = 0;
590 591
            cnt = _block_size;
591 592
          }
592 593
        }
593
        for (e = 0; e < _next_arc; ++e) {
594
        for (e = 0; e != _next_arc; ++e) {
594 595
          _cand_cost[e] = _state[e] *
595 596
            (_cost[e] + _pi[_source[e]] - _pi[_target[e]]);
596 597
          if (_cand_cost[e] < 0) {
597 598
            _candidates[_curr_length++] = e;
598 599
          }
599 600
          if (--cnt == 0) {
600 601
            if (_curr_length > limit) goto search_end;
601 602
            limit = 0;
602 603
            cnt = _block_size;
603 604
          }
604 605
        }
605 606
        if (_curr_length == 0) return false;
606 607
        
607 608
      search_end:
608 609

	
609 610
        // Make heap of the candidate list (approximating a partial sort)
610 611
        make_heap( _candidates.begin(), _candidates.begin() + _curr_length,
611 612
                   _sort_func );
612 613

	
613 614
        // Pop the first element of the heap
614 615
        _in_arc = _candidates[0];
615 616
        _next_arc = e;
616 617
        pop_heap( _candidates.begin(), _candidates.begin() + _curr_length,
617 618
                  _sort_func );
618 619
        _curr_length = std::min(_head_length, _curr_length - 1);
619 620
        return true;
620 621
      }
621 622

	
622 623
    }; //class AlteringListPivotRule
623 624

	
624 625
  public:
625 626

	
626 627
    /// \brief Constructor.
627 628
    ///
628 629
    /// The constructor of the class.
629 630
    ///
630 631
    /// \param graph The digraph the algorithm runs on.
631 632
    /// \param arc_mixing Indicate if the arcs have to be stored in a
632 633
    /// mixed order in the internal data structure. 
633 634
    /// In special cases, it could lead to better overall performance,
634 635
    /// but it is usually slower. Therefore it is disabled by default.
635 636
    NetworkSimplex(const GR& graph, bool arc_mixing = false) :
636 637
      _graph(graph), _node_id(graph), _arc_id(graph),
637 638
      _arc_mixing(arc_mixing),
638 639
      MAX(std::numeric_limits<Value>::max()),
639 640
      INF(std::numeric_limits<Value>::has_infinity ?
640 641
          std::numeric_limits<Value>::infinity() : MAX)
641 642
    {
642 643
      // Check the number types
643 644
      LEMON_ASSERT(std::numeric_limits<Value>::is_signed,
644 645
        "The flow type of NetworkSimplex must be signed");
645 646
      LEMON_ASSERT(std::numeric_limits<Cost>::is_signed,
646 647
        "The cost type of NetworkSimplex must be signed");
647 648
        
648 649
      // Reset data structures
649 650
      reset();
650 651
    }
651 652

	
652 653
    /// \name Parameters
653 654
    /// The parameters of the algorithm can be specified using these
654 655
    /// functions.
655 656

	
656 657
    /// @{
657 658

	
658 659
    /// \brief Set the lower bounds on the arcs.
659 660
    ///
660 661
    /// This function sets the lower bounds on the arcs.
661 662
    /// If it is not used before calling \ref run(), the lower bounds
662 663
    /// will be set to zero on all arcs.
663 664
    ///
664 665
    /// \param map An arc map storing the lower bounds.
665 666
    /// Its \c Value type must be convertible to the \c Value type
666 667
    /// of the algorithm.
667 668
    ///
668 669
    /// \return <tt>(*this)</tt>
669 670
    template <typename LowerMap>
670 671
    NetworkSimplex& lowerMap(const LowerMap& map) {
671 672
      _have_lower = true;
672 673
      for (ArcIt a(_graph); a != INVALID; ++a) {
673 674
        _lower[_arc_id[a]] = map[a];
674 675
      }
675 676
      return *this;
676 677
    }
677 678

	
678 679
    /// \brief Set the upper bounds (capacities) on the arcs.
679 680
    ///
680 681
    /// This function sets the upper bounds (capacities) on the arcs.
681 682
    /// If it is not used before calling \ref run(), the upper bounds
682 683
    /// will be set to \ref INF on all arcs (i.e. the flow value will be
683 684
    /// unbounded from above).
684 685
    ///
685 686
    /// \param map An arc map storing the upper bounds.
686 687
    /// Its \c Value type must be convertible to the \c Value type
687 688
    /// of the algorithm.
688 689
    ///
689 690
    /// \return <tt>(*this)</tt>
690 691
    template<typename UpperMap>
691 692
    NetworkSimplex& upperMap(const UpperMap& map) {
692 693
      for (ArcIt a(_graph); a != INVALID; ++a) {
693 694
        _upper[_arc_id[a]] = map[a];
694 695
      }
695 696
      return *this;
696 697
    }
697 698

	
698 699
    /// \brief Set the costs of the arcs.
699 700
    ///
700 701
    /// This function sets the costs of the arcs.
701 702
    /// If it is not used before calling \ref run(), the costs
702 703
    /// will be set to \c 1 on all arcs.
703 704
    ///
704 705
    /// \param map An arc map storing the costs.
705 706
    /// Its \c Value type must be convertible to the \c Cost type
706 707
    /// of the algorithm.
707 708
    ///
708 709
    /// \return <tt>(*this)</tt>
709 710
    template<typename CostMap>
710 711
    NetworkSimplex& costMap(const CostMap& map) {
711 712
      for (ArcIt a(_graph); a != INVALID; ++a) {
712 713
        _cost[_arc_id[a]] = map[a];
713 714
      }
714 715
      return *this;
715 716
    }
716 717

	
717 718
    /// \brief Set the supply values of the nodes.
718 719
    ///
719 720
    /// This function sets the supply values of the nodes.
720 721
    /// If neither this function nor \ref stSupply() is used before
721 722
    /// calling \ref run(), the supply of each node will be set to zero.
722 723
    ///
723 724
    /// \param map A node map storing the supply values.
724 725
    /// Its \c Value type must be convertible to the \c Value type
725 726
    /// of the algorithm.
726 727
    ///
727 728
    /// \return <tt>(*this)</tt>
728 729
    template<typename SupplyMap>
729 730
    NetworkSimplex& supplyMap(const SupplyMap& map) {
730 731
      for (NodeIt n(_graph); n != INVALID; ++n) {
731 732
        _supply[_node_id[n]] = map[n];
732 733
      }
733 734
      return *this;
734 735
    }
735 736

	
736 737
    /// \brief Set single source and target nodes and a supply value.
737 738
    ///
738 739
    /// This function sets a single source node and a single target node
739 740
    /// and the required flow value.
740 741
    /// If neither this function nor \ref supplyMap() is used before
741 742
    /// calling \ref run(), the supply of each node will be set to zero.
742 743
    ///
743 744
    /// Using this function has the same effect as using \ref supplyMap()
744 745
    /// with such a map in which \c k is assigned to \c s, \c -k is
745 746
    /// assigned to \c t and all other nodes have zero supply value.
746 747
    ///
747 748
    /// \param s The source node.
748 749
    /// \param t The target node.
749 750
    /// \param k The required amount of flow from node \c s to node \c t
750 751
    /// (i.e. the supply of \c s and the demand of \c t).
751 752
    ///
752 753
    /// \return <tt>(*this)</tt>
753 754
    NetworkSimplex& stSupply(const Node& s, const Node& t, Value k) {
754 755
      for (int i = 0; i != _node_num; ++i) {
755 756
        _supply[i] = 0;
756 757
      }
757 758
      _supply[_node_id[s]] =  k;
758 759
      _supply[_node_id[t]] = -k;
759 760
      return *this;
760 761
    }
761 762
    
762 763
    /// \brief Set the type of the supply constraints.
763 764
    ///
764 765
    /// This function sets the type of the supply/demand constraints.
765 766
    /// If it is not used before calling \ref run(), the \ref GEQ supply
766 767
    /// type will be used.
767 768
    ///
768 769
    /// For more information, see \ref SupplyType.
769 770
    ///
770 771
    /// \return <tt>(*this)</tt>
771 772
    NetworkSimplex& supplyType(SupplyType supply_type) {
772 773
      _stype = supply_type;
773 774
      return *this;
774 775
    }
775 776

	
776 777
    /// @}
777 778

	
778 779
    /// \name Execution Control
779 780
    /// The algorithm can be executed using \ref run().
780 781

	
781 782
    /// @{
782 783

	
783 784
    /// \brief Run the algorithm.
784 785
    ///
785 786
    /// This function runs the algorithm.
786 787
    /// The paramters can be specified using functions \ref lowerMap(),
787 788
    /// \ref upperMap(), \ref costMap(), \ref supplyMap(), \ref stSupply(), 
788 789
    /// \ref supplyType().
789 790
    /// For example,
790 791
    /// \code
791 792
    ///   NetworkSimplex<ListDigraph> ns(graph);
792 793
    ///   ns.lowerMap(lower).upperMap(upper).costMap(cost)
793 794
    ///     .supplyMap(sup).run();
794 795
    /// \endcode
795 796
    ///
796 797
    /// This function can be called more than once. All the given parameters
797 798
    /// are kept for the next call, unless \ref resetParams() or \ref reset()
798 799
    /// is used, thus only the modified parameters have to be set again.
799 800
    /// If the underlying digraph was also modified after the construction
800 801
    /// of the class (or the last \ref reset() call), then the \ref reset()
801 802
    /// function must be called.
802 803
    ///
803 804
    /// \param pivot_rule The pivot rule that will be used during the
804 805
    /// algorithm. For more information, see \ref PivotRule.
805 806
    ///
806 807
    /// \return \c INFEASIBLE if no feasible flow exists,
807 808
    /// \n \c OPTIMAL if the problem has optimal solution
808 809
    /// (i.e. it is feasible and bounded), and the algorithm has found
809 810
    /// optimal flow and node potentials (primal and dual solutions),
810 811
    /// \n \c UNBOUNDED if the objective function of the problem is
811 812
    /// unbounded, i.e. there is a directed cycle having negative total
812 813
    /// cost and infinite upper bound.
813 814
    ///
814 815
    /// \see ProblemType, PivotRule
815 816
    /// \see resetParams(), reset()
816 817
    ProblemType run(PivotRule pivot_rule = BLOCK_SEARCH) {
817 818
      if (!init()) return INFEASIBLE;
818 819
      return start(pivot_rule);
819 820
    }
820 821

	
821 822
    /// \brief Reset all the parameters that have been given before.
822 823
    ///
823 824
    /// This function resets all the paramaters that have been given
824 825
    /// before using functions \ref lowerMap(), \ref upperMap(),
825 826
    /// \ref costMap(), \ref supplyMap(), \ref stSupply(), \ref supplyType().
826 827
    ///
827 828
    /// It is useful for multiple \ref run() calls. Basically, all the given
828 829
    /// parameters are kept for the next \ref run() call, unless
829 830
    /// \ref resetParams() or \ref reset() is used.
830 831
    /// If the underlying digraph was also modified after the construction
831 832
    /// of the class or the last \ref reset() call, then the \ref reset()
832 833
    /// function must be used, otherwise \ref resetParams() is sufficient.
833 834
    ///
834 835
    /// For example,
835 836
    /// \code
836 837
    ///   NetworkSimplex<ListDigraph> ns(graph);
837 838
    ///
838 839
    ///   // First run
839 840
    ///   ns.lowerMap(lower).upperMap(upper).costMap(cost)
840 841
    ///     .supplyMap(sup).run();
841 842
    ///
842 843
    ///   // Run again with modified cost map (resetParams() is not called,
843 844
    ///   // so only the cost map have to be set again)
844 845
    ///   cost[e] += 100;
845 846
    ///   ns.costMap(cost).run();
846 847
    ///
847 848
    ///   // Run again from scratch using resetParams()
848 849
    ///   // (the lower bounds will be set to zero on all arcs)
849 850
    ///   ns.resetParams();
850 851
    ///   ns.upperMap(capacity).costMap(cost)
851 852
    ///     .supplyMap(sup).run();
852 853
    /// \endcode
853 854
    ///
854 855
    /// \return <tt>(*this)</tt>
855 856
    ///
856 857
    /// \see reset(), run()
857 858
    NetworkSimplex& resetParams() {
858 859
      for (int i = 0; i != _node_num; ++i) {
859 860
        _supply[i] = 0;
860 861
      }
861 862
      for (int i = 0; i != _arc_num; ++i) {
862 863
        _lower[i] = 0;
863 864
        _upper[i] = INF;
864 865
        _cost[i] = 1;
865 866
      }
866 867
      _have_lower = false;
867 868
      _stype = GEQ;
868 869
      return *this;
869 870
    }
870 871

	
871 872
    /// \brief Reset the internal data structures and all the parameters
872 873
    /// that have been given before.
873 874
    ///
874 875
    /// This function resets the internal data structures and all the
875 876
    /// paramaters that have been given before using functions \ref lowerMap(),
876 877
    /// \ref upperMap(), \ref costMap(), \ref supplyMap(), \ref stSupply(),
877 878
    /// \ref supplyType().
878 879
    ///
879 880
    /// It is useful for multiple \ref run() calls. Basically, all the given
880 881
    /// parameters are kept for the next \ref run() call, unless
881 882
    /// \ref resetParams() or \ref reset() is used.
882 883
    /// If the underlying digraph was also modified after the construction
883 884
    /// of the class or the last \ref reset() call, then the \ref reset()
884 885
    /// function must be used, otherwise \ref resetParams() is sufficient.
885 886
    ///
886 887
    /// See \ref resetParams() for examples.
887 888
    ///
888 889
    /// \return <tt>(*this)</tt>
889 890
    ///
890 891
    /// \see resetParams(), run()
891 892
    NetworkSimplex& reset() {
892 893
      // Resize vectors
893 894
      _node_num = countNodes(_graph);
894 895
      _arc_num = countArcs(_graph);
895 896
      int all_node_num = _node_num + 1;
896 897
      int max_arc_num = _arc_num + 2 * _node_num;
897 898

	
898 899
      _source.resize(max_arc_num);
899 900
      _target.resize(max_arc_num);
900 901

	
901 902
      _lower.resize(_arc_num);
902 903
      _upper.resize(_arc_num);
903 904
      _cap.resize(max_arc_num);
904 905
      _cost.resize(max_arc_num);
905 906
      _supply.resize(all_node_num);
906 907
      _flow.resize(max_arc_num);
907 908
      _pi.resize(all_node_num);
908 909

	
909 910
      _parent.resize(all_node_num);
910 911
      _pred.resize(all_node_num);
911 912
      _forward.resize(all_node_num);
912 913
      _thread.resize(all_node_num);
913 914
      _rev_thread.resize(all_node_num);
914 915
      _succ_num.resize(all_node_num);
915 916
      _last_succ.resize(all_node_num);
916 917
      _state.resize(max_arc_num);
917 918

	
918 919
      // Copy the graph
919 920
      int i = 0;
920 921
      for (NodeIt n(_graph); n != INVALID; ++n, ++i) {
921 922
        _node_id[n] = i;
922 923
      }
923 924
      if (_arc_mixing) {
924 925
        // Store the arcs in a mixed order
925 926
        int k = std::max(int(std::sqrt(double(_arc_num))), 10);
926 927
        int i = 0, j = 0;
927 928
        for (ArcIt a(_graph); a != INVALID; ++a) {
928 929
          _arc_id[a] = i;
929 930
          _source[i] = _node_id[_graph.source(a)];
930 931
          _target[i] = _node_id[_graph.target(a)];
931 932
          if ((i += k) >= _arc_num) i = ++j;
932 933
        }
933 934
      } else {
934 935
        // Store the arcs in the original order
935 936
        int i = 0;
936 937
        for (ArcIt a(_graph); a != INVALID; ++a, ++i) {
937 938
          _arc_id[a] = i;
938 939
          _source[i] = _node_id[_graph.source(a)];
939 940
          _target[i] = _node_id[_graph.target(a)];
940 941
        }
941 942
      }
942 943
      
943 944
      // Reset parameters
944 945
      resetParams();
945 946
      return *this;
946 947
    }
947 948
    
948 949
    /// @}
949 950

	
950 951
    /// \name Query Functions
951 952
    /// The results of the algorithm can be obtained using these
952 953
    /// functions.\n
953 954
    /// The \ref run() function must be called before using them.
954 955

	
955 956
    /// @{
956 957

	
957 958
    /// \brief Return the total cost of the found flow.
958 959
    ///
959 960
    /// This function returns the total cost of the found flow.
960 961
    /// Its complexity is O(e).
961 962
    ///
962 963
    /// \note The return type of the function can be specified as a
963 964
    /// template parameter. For example,
964 965
    /// \code
965 966
    ///   ns.totalCost<double>();
966 967
    /// \endcode
967 968
    /// It is useful if the total cost cannot be stored in the \c Cost
968 969
    /// type of the algorithm, which is the default return type of the
969 970
    /// function.
970 971
    ///
971 972
    /// \pre \ref run() must be called before using this function.
972 973
    template <typename Number>
973 974
    Number totalCost() const {
974 975
      Number c = 0;
975 976
      for (ArcIt a(_graph); a != INVALID; ++a) {
976 977
        int i = _arc_id[a];
977 978
        c += Number(_flow[i]) * Number(_cost[i]);
... ...
@@ -979,541 +980,638 @@
979 980
      return c;
980 981
    }
981 982

	
982 983
#ifndef DOXYGEN
983 984
    Cost totalCost() const {
984 985
      return totalCost<Cost>();
985 986
    }
986 987
#endif
987 988

	
988 989
    /// \brief Return the flow on the given arc.
989 990
    ///
990 991
    /// This function returns the flow on the given arc.
991 992
    ///
992 993
    /// \pre \ref run() must be called before using this function.
993 994
    Value flow(const Arc& a) const {
994 995
      return _flow[_arc_id[a]];
995 996
    }
996 997

	
997 998
    /// \brief Return the flow map (the primal solution).
998 999
    ///
999 1000
    /// This function copies the flow value on each arc into the given
1000 1001
    /// map. The \c Value type of the algorithm must be convertible to
1001 1002
    /// the \c Value type of the map.
1002 1003
    ///
1003 1004
    /// \pre \ref run() must be called before using this function.
1004 1005
    template <typename FlowMap>
1005 1006
    void flowMap(FlowMap &map) const {
1006 1007
      for (ArcIt a(_graph); a != INVALID; ++a) {
1007 1008
        map.set(a, _flow[_arc_id[a]]);
1008 1009
      }
1009 1010
    }
1010 1011

	
1011 1012
    /// \brief Return the potential (dual value) of the given node.
1012 1013
    ///
1013 1014
    /// This function returns the potential (dual value) of the
1014 1015
    /// given node.
1015 1016
    ///
1016 1017
    /// \pre \ref run() must be called before using this function.
1017 1018
    Cost potential(const Node& n) const {
1018 1019
      return _pi[_node_id[n]];
1019 1020
    }
1020 1021

	
1021 1022
    /// \brief Return the potential map (the dual solution).
1022 1023
    ///
1023 1024
    /// This function copies the potential (dual value) of each node
1024 1025
    /// into the given map.
1025 1026
    /// The \c Cost type of the algorithm must be convertible to the
1026 1027
    /// \c Value type of the map.
1027 1028
    ///
1028 1029
    /// \pre \ref run() must be called before using this function.
1029 1030
    template <typename PotentialMap>
1030 1031
    void potentialMap(PotentialMap &map) const {
1031 1032
      for (NodeIt n(_graph); n != INVALID; ++n) {
1032 1033
        map.set(n, _pi[_node_id[n]]);
1033 1034
      }
1034 1035
    }
1035 1036

	
1036 1037
    /// @}
1037 1038

	
1038 1039
  private:
1039 1040

	
1040 1041
    // Initialize internal data structures
1041 1042
    bool init() {
1042 1043
      if (_node_num == 0) return false;
1043 1044

	
1044 1045
      // Check the sum of supply values
1045 1046
      _sum_supply = 0;
1046 1047
      for (int i = 0; i != _node_num; ++i) {
1047 1048
        _sum_supply += _supply[i];
1048 1049
      }
1049 1050
      if ( !((_stype == GEQ && _sum_supply <= 0) ||
1050 1051
             (_stype == LEQ && _sum_supply >= 0)) ) return false;
1051 1052

	
1052 1053
      // Remove non-zero lower bounds
1053 1054
      if (_have_lower) {
1054 1055
        for (int i = 0; i != _arc_num; ++i) {
1055 1056
          Value c = _lower[i];
1056 1057
          if (c >= 0) {
1057 1058
            _cap[i] = _upper[i] < MAX ? _upper[i] - c : INF;
1058 1059
          } else {
1059 1060
            _cap[i] = _upper[i] < MAX + c ? _upper[i] - c : INF;
1060 1061
          }
1061 1062
          _supply[_source[i]] -= c;
1062 1063
          _supply[_target[i]] += c;
1063 1064
        }
1064 1065
      } else {
1065 1066
        for (int i = 0; i != _arc_num; ++i) {
1066 1067
          _cap[i] = _upper[i];
1067 1068
        }
1068 1069
      }
1069 1070

	
1070 1071
      // Initialize artifical cost
1071 1072
      Cost ART_COST;
1072 1073
      if (std::numeric_limits<Cost>::is_exact) {
1073 1074
        ART_COST = std::numeric_limits<Cost>::max() / 2 + 1;
1074 1075
      } else {
1075 1076
        ART_COST = std::numeric_limits<Cost>::min();
1076 1077
        for (int i = 0; i != _arc_num; ++i) {
1077 1078
          if (_cost[i] > ART_COST) ART_COST = _cost[i];
1078 1079
        }
1079 1080
        ART_COST = (ART_COST + 1) * _node_num;
1080 1081
      }
1081 1082

	
1082 1083
      // Initialize arc maps
1083 1084
      for (int i = 0; i != _arc_num; ++i) {
1084 1085
        _flow[i] = 0;
1085 1086
        _state[i] = STATE_LOWER;
1086 1087
      }
1087 1088
      
1088 1089
      // Set data for the artificial root node
1089 1090
      _root = _node_num;
1090 1091
      _parent[_root] = -1;
1091 1092
      _pred[_root] = -1;
1092 1093
      _thread[_root] = 0;
1093 1094
      _rev_thread[0] = _root;
1094 1095
      _succ_num[_root] = _node_num + 1;
1095 1096
      _last_succ[_root] = _root - 1;
1096 1097
      _supply[_root] = -_sum_supply;
1097 1098
      _pi[_root] = 0;
1098 1099

	
1099 1100
      // Add artificial arcs and initialize the spanning tree data structure
1100 1101
      if (_sum_supply == 0) {
1101 1102
        // EQ supply constraints
1102 1103
        _search_arc_num = _arc_num;
1103 1104
        _all_arc_num = _arc_num + _node_num;
1104 1105
        for (int u = 0, e = _arc_num; u != _node_num; ++u, ++e) {
1105 1106
          _parent[u] = _root;
1106 1107
          _pred[u] = e;
1107 1108
          _thread[u] = u + 1;
1108 1109
          _rev_thread[u + 1] = u;
1109 1110
          _succ_num[u] = 1;
1110 1111
          _last_succ[u] = u;
1111 1112
          _cap[e] = INF;
1112 1113
          _state[e] = STATE_TREE;
1113 1114
          if (_supply[u] >= 0) {
1114 1115
            _forward[u] = true;
1115 1116
            _pi[u] = 0;
1116 1117
            _source[e] = u;
1117 1118
            _target[e] = _root;
1118 1119
            _flow[e] = _supply[u];
1119 1120
            _cost[e] = 0;
1120 1121
          } else {
1121 1122
            _forward[u] = false;
1122 1123
            _pi[u] = ART_COST;
1123 1124
            _source[e] = _root;
1124 1125
            _target[e] = u;
1125 1126
            _flow[e] = -_supply[u];
1126 1127
            _cost[e] = ART_COST;
1127 1128
          }
1128 1129
        }
1129 1130
      }
1130 1131
      else if (_sum_supply > 0) {
1131 1132
        // LEQ supply constraints
1132 1133
        _search_arc_num = _arc_num + _node_num;
1133 1134
        int f = _arc_num + _node_num;
1134 1135
        for (int u = 0, e = _arc_num; u != _node_num; ++u, ++e) {
1135 1136
          _parent[u] = _root;
1136 1137
          _thread[u] = u + 1;
1137 1138
          _rev_thread[u + 1] = u;
1138 1139
          _succ_num[u] = 1;
1139 1140
          _last_succ[u] = u;
1140 1141
          if (_supply[u] >= 0) {
1141 1142
            _forward[u] = true;
1142 1143
            _pi[u] = 0;
1143 1144
            _pred[u] = e;
1144 1145
            _source[e] = u;
1145 1146
            _target[e] = _root;
1146 1147
            _cap[e] = INF;
1147 1148
            _flow[e] = _supply[u];
1148 1149
            _cost[e] = 0;
1149 1150
            _state[e] = STATE_TREE;
1150 1151
          } else {
1151 1152
            _forward[u] = false;
1152 1153
            _pi[u] = ART_COST;
1153 1154
            _pred[u] = f;
1154 1155
            _source[f] = _root;
1155 1156
            _target[f] = u;
1156 1157
            _cap[f] = INF;
1157 1158
            _flow[f] = -_supply[u];
1158 1159
            _cost[f] = ART_COST;
1159 1160
            _state[f] = STATE_TREE;
1160 1161
            _source[e] = u;
1161 1162
            _target[e] = _root;
1162 1163
            _cap[e] = INF;
1163 1164
            _flow[e] = 0;
1164 1165
            _cost[e] = 0;
1165 1166
            _state[e] = STATE_LOWER;
1166 1167
            ++f;
1167 1168
          }
1168 1169
        }
1169 1170
        _all_arc_num = f;
1170 1171
      }
1171 1172
      else {
1172 1173
        // GEQ supply constraints
1173 1174
        _search_arc_num = _arc_num + _node_num;
1174 1175
        int f = _arc_num + _node_num;
1175 1176
        for (int u = 0, e = _arc_num; u != _node_num; ++u, ++e) {
1176 1177
          _parent[u] = _root;
1177 1178
          _thread[u] = u + 1;
1178 1179
          _rev_thread[u + 1] = u;
1179 1180
          _succ_num[u] = 1;
1180 1181
          _last_succ[u] = u;
1181 1182
          if (_supply[u] <= 0) {
1182 1183
            _forward[u] = false;
1183 1184
            _pi[u] = 0;
1184 1185
            _pred[u] = e;
1185 1186
            _source[e] = _root;
1186 1187
            _target[e] = u;
1187 1188
            _cap[e] = INF;
1188 1189
            _flow[e] = -_supply[u];
1189 1190
            _cost[e] = 0;
1190 1191
            _state[e] = STATE_TREE;
1191 1192
          } else {
1192 1193
            _forward[u] = true;
1193 1194
            _pi[u] = -ART_COST;
1194 1195
            _pred[u] = f;
1195 1196
            _source[f] = u;
1196 1197
            _target[f] = _root;
1197 1198
            _cap[f] = INF;
1198 1199
            _flow[f] = _supply[u];
1199 1200
            _state[f] = STATE_TREE;
1200 1201
            _cost[f] = ART_COST;
1201 1202
            _source[e] = _root;
1202 1203
            _target[e] = u;
1203 1204
            _cap[e] = INF;
1204 1205
            _flow[e] = 0;
1205 1206
            _cost[e] = 0;
1206 1207
            _state[e] = STATE_LOWER;
1207 1208
            ++f;
1208 1209
          }
1209 1210
        }
1210 1211
        _all_arc_num = f;
1211 1212
      }
1212 1213

	
1213 1214
      return true;
1214 1215
    }
1215 1216

	
1216 1217
    // Find the join node
1217 1218
    void findJoinNode() {
1218 1219
      int u = _source[in_arc];
1219 1220
      int v = _target[in_arc];
1220 1221
      while (u != v) {
1221 1222
        if (_succ_num[u] < _succ_num[v]) {
1222 1223
          u = _parent[u];
1223 1224
        } else {
1224 1225
          v = _parent[v];
1225 1226
        }
1226 1227
      }
1227 1228
      join = u;
1228 1229
    }
1229 1230

	
1230 1231
    // Find the leaving arc of the cycle and returns true if the
1231 1232
    // leaving arc is not the same as the entering arc
1232 1233
    bool findLeavingArc() {
1233 1234
      // Initialize first and second nodes according to the direction
1234 1235
      // of the cycle
1235 1236
      if (_state[in_arc] == STATE_LOWER) {
1236 1237
        first  = _source[in_arc];
1237 1238
        second = _target[in_arc];
1238 1239
      } else {
1239 1240
        first  = _target[in_arc];
1240 1241
        second = _source[in_arc];
1241 1242
      }
1242 1243
      delta = _cap[in_arc];
1243 1244
      int result = 0;
1244 1245
      Value d;
1245 1246
      int e;
1246 1247

	
1247 1248
      // Search the cycle along the path form the first node to the root
1248 1249
      for (int u = first; u != join; u = _parent[u]) {
1249 1250
        e = _pred[u];
1250 1251
        d = _forward[u] ?
1251 1252
          _flow[e] : (_cap[e] >= MAX ? INF : _cap[e] - _flow[e]);
1252 1253
        if (d < delta) {
1253 1254
          delta = d;
1254 1255
          u_out = u;
1255 1256
          result = 1;
1256 1257
        }
1257 1258
      }
1258 1259
      // Search the cycle along the path form the second node to the root
1259 1260
      for (int u = second; u != join; u = _parent[u]) {
1260 1261
        e = _pred[u];
1261 1262
        d = _forward[u] ? 
1262 1263
          (_cap[e] >= MAX ? INF : _cap[e] - _flow[e]) : _flow[e];
1263 1264
        if (d <= delta) {
1264 1265
          delta = d;
1265 1266
          u_out = u;
1266 1267
          result = 2;
1267 1268
        }
1268 1269
      }
1269 1270

	
1270 1271
      if (result == 1) {
1271 1272
        u_in = first;
1272 1273
        v_in = second;
1273 1274
      } else {
1274 1275
        u_in = second;
1275 1276
        v_in = first;
1276 1277
      }
1277 1278
      return result != 0;
1278 1279
    }
1279 1280

	
1280 1281
    // Change _flow and _state vectors
1281 1282
    void changeFlow(bool change) {
1282 1283
      // Augment along the cycle
1283 1284
      if (delta > 0) {
1284 1285
        Value val = _state[in_arc] * delta;
1285 1286
        _flow[in_arc] += val;
1286 1287
        for (int u = _source[in_arc]; u != join; u = _parent[u]) {
1287 1288
          _flow[_pred[u]] += _forward[u] ? -val : val;
1288 1289
        }
1289 1290
        for (int u = _target[in_arc]; u != join; u = _parent[u]) {
1290 1291
          _flow[_pred[u]] += _forward[u] ? val : -val;
1291 1292
        }
1292 1293
      }
1293 1294
      // Update the state of the entering and leaving arcs
1294 1295
      if (change) {
1295 1296
        _state[in_arc] = STATE_TREE;
1296 1297
        _state[_pred[u_out]] =
1297 1298
          (_flow[_pred[u_out]] == 0) ? STATE_LOWER : STATE_UPPER;
1298 1299
      } else {
1299 1300
        _state[in_arc] = -_state[in_arc];
1300 1301
      }
1301 1302
    }
1302 1303

	
1303 1304
    // Update the tree structure
1304 1305
    void updateTreeStructure() {
1305 1306
      int u, w;
1306 1307
      int old_rev_thread = _rev_thread[u_out];
1307 1308
      int old_succ_num = _succ_num[u_out];
1308 1309
      int old_last_succ = _last_succ[u_out];
1309 1310
      v_out = _parent[u_out];
1310 1311

	
1311 1312
      u = _last_succ[u_in];  // the last successor of u_in
1312 1313
      right = _thread[u];    // the node after it
1313 1314

	
1314 1315
      // Handle the case when old_rev_thread equals to v_in
1315 1316
      // (it also means that join and v_out coincide)
1316 1317
      if (old_rev_thread == v_in) {
1317 1318
        last = _thread[_last_succ[u_out]];
1318 1319
      } else {
1319 1320
        last = _thread[v_in];
1320 1321
      }
1321 1322

	
1322 1323
      // Update _thread and _parent along the stem nodes (i.e. the nodes
1323 1324
      // between u_in and u_out, whose parent have to be changed)
1324 1325
      _thread[v_in] = stem = u_in;
1325 1326
      _dirty_revs.clear();
1326 1327
      _dirty_revs.push_back(v_in);
1327 1328
      par_stem = v_in;
1328 1329
      while (stem != u_out) {
1329 1330
        // Insert the next stem node into the thread list
1330 1331
        new_stem = _parent[stem];
1331 1332
        _thread[u] = new_stem;
1332 1333
        _dirty_revs.push_back(u);
1333 1334

	
1334 1335
        // Remove the subtree of stem from the thread list
1335 1336
        w = _rev_thread[stem];
1336 1337
        _thread[w] = right;
1337 1338
        _rev_thread[right] = w;
1338 1339

	
1339 1340
        // Change the parent node and shift stem nodes
1340 1341
        _parent[stem] = par_stem;
1341 1342
        par_stem = stem;
1342 1343
        stem = new_stem;
1343 1344

	
1344 1345
        // Update u and right
1345 1346
        u = _last_succ[stem] == _last_succ[par_stem] ?
1346 1347
          _rev_thread[par_stem] : _last_succ[stem];
1347 1348
        right = _thread[u];
1348 1349
      }
1349 1350
      _parent[u_out] = par_stem;
1350 1351
      _thread[u] = last;
1351 1352
      _rev_thread[last] = u;
1352 1353
      _last_succ[u_out] = u;
1353 1354

	
1354 1355
      // Remove the subtree of u_out from the thread list except for
1355 1356
      // the case when old_rev_thread equals to v_in
1356 1357
      // (it also means that join and v_out coincide)
1357 1358
      if (old_rev_thread != v_in) {
1358 1359
        _thread[old_rev_thread] = right;
1359 1360
        _rev_thread[right] = old_rev_thread;
1360 1361
      }
1361 1362

	
1362 1363
      // Update _rev_thread using the new _thread values
1363
      for (int i = 0; i < int(_dirty_revs.size()); ++i) {
1364
      for (int i = 0; i != int(_dirty_revs.size()); ++i) {
1364 1365
        u = _dirty_revs[i];
1365 1366
        _rev_thread[_thread[u]] = u;
1366 1367
      }
1367 1368

	
1368 1369
      // Update _pred, _forward, _last_succ and _succ_num for the
1369 1370
      // stem nodes from u_out to u_in
1370 1371
      int tmp_sc = 0, tmp_ls = _last_succ[u_out];
1371 1372
      u = u_out;
1372 1373
      while (u != u_in) {
1373 1374
        w = _parent[u];
1374 1375
        _pred[u] = _pred[w];
1375 1376
        _forward[u] = !_forward[w];
1376 1377
        tmp_sc += _succ_num[u] - _succ_num[w];
1377 1378
        _succ_num[u] = tmp_sc;
1378 1379
        _last_succ[w] = tmp_ls;
1379 1380
        u = w;
1380 1381
      }
1381 1382
      _pred[u_in] = in_arc;
1382 1383
      _forward[u_in] = (u_in == _source[in_arc]);
1383 1384
      _succ_num[u_in] = old_succ_num;
1384 1385

	
1385 1386
      // Set limits for updating _last_succ form v_in and v_out
1386 1387
      // towards the root
1387 1388
      int up_limit_in = -1;
1388 1389
      int up_limit_out = -1;
1389 1390
      if (_last_succ[join] == v_in) {
1390 1391
        up_limit_out = join;
1391 1392
      } else {
1392 1393
        up_limit_in = join;
1393 1394
      }
1394 1395

	
1395 1396
      // Update _last_succ from v_in towards the root
1396 1397
      for (u = v_in; u != up_limit_in && _last_succ[u] == v_in;
1397 1398
           u = _parent[u]) {
1398 1399
        _last_succ[u] = _last_succ[u_out];
1399 1400
      }
1400 1401
      // Update _last_succ from v_out towards the root
1401 1402
      if (join != old_rev_thread && v_in != old_rev_thread) {
1402 1403
        for (u = v_out; u != up_limit_out && _last_succ[u] == old_last_succ;
1403 1404
             u = _parent[u]) {
1404 1405
          _last_succ[u] = old_rev_thread;
1405 1406
        }
1406 1407
      } else {
1407 1408
        for (u = v_out; u != up_limit_out && _last_succ[u] == old_last_succ;
1408 1409
             u = _parent[u]) {
1409 1410
          _last_succ[u] = _last_succ[u_out];
1410 1411
        }
1411 1412
      }
1412 1413

	
1413 1414
      // Update _succ_num from v_in to join
1414 1415
      for (u = v_in; u != join; u = _parent[u]) {
1415 1416
        _succ_num[u] += old_succ_num;
1416 1417
      }
1417 1418
      // Update _succ_num from v_out to join
1418 1419
      for (u = v_out; u != join; u = _parent[u]) {
1419 1420
        _succ_num[u] -= old_succ_num;
1420 1421
      }
1421 1422
    }
1422 1423

	
1423 1424
    // Update potentials
1424 1425
    void updatePotential() {
1425 1426
      Cost sigma = _forward[u_in] ?
1426 1427
        _pi[v_in] - _pi[u_in] - _cost[_pred[u_in]] :
1427 1428
        _pi[v_in] - _pi[u_in] + _cost[_pred[u_in]];
1428 1429
      // Update potentials in the subtree, which has been moved
1429 1430
      int end = _thread[_last_succ[u_in]];
1430 1431
      for (int u = u_in; u != end; u = _thread[u]) {
1431 1432
        _pi[u] += sigma;
1432 1433
      }
1433 1434
    }
1434 1435

	
1436
    // Heuristic initial pivots
1437
    bool initialPivots() {
1438
      Value curr, total = 0;
1439
      std::vector<Node> supply_nodes, demand_nodes;
1440
      for (NodeIt u(_graph); u != INVALID; ++u) {
1441
        curr = _supply[_node_id[u]];
1442
        if (curr > 0) {
1443
          total += curr;
1444
          supply_nodes.push_back(u);
1445
        }
1446
        else if (curr < 0) {
1447
          demand_nodes.push_back(u);
1448
        }
1449
      }
1450
      if (_sum_supply > 0) total -= _sum_supply;
1451
      if (total <= 0) return true;
1452

	
1453
      IntVector arc_vector;
1454
      if (_sum_supply >= 0) {
1455
        if (supply_nodes.size() == 1 && demand_nodes.size() == 1) {
1456
          // Perform a reverse graph search from the sink to the source
1457
          typename GR::template NodeMap<bool> reached(_graph, false);
1458
          Node s = supply_nodes[0], t = demand_nodes[0];
1459
          std::vector<Node> stack;
1460
          reached[t] = true;
1461
          stack.push_back(t);
1462
          while (!stack.empty()) {
1463
            Node u, v = stack.back();
1464
            stack.pop_back();
1465
            if (v == s) break;
1466
            for (InArcIt a(_graph, v); a != INVALID; ++a) {
1467
              if (reached[u = _graph.source(a)]) continue;
1468
              int j = _arc_id[a];
1469
              if (_cap[j] >= total) {
1470
                arc_vector.push_back(j);
1471
                reached[u] = true;
1472
                stack.push_back(u);
1473
              }
1474
            }
1475
          }
1476
        } else {
1477
          // Find the min. cost incomming arc for each demand node
1478
          for (int i = 0; i != int(demand_nodes.size()); ++i) {
1479
            Node v = demand_nodes[i];
1480
            Cost c, min_cost = std::numeric_limits<Cost>::max();
1481
            Arc min_arc = INVALID;
1482
            for (InArcIt a(_graph, v); a != INVALID; ++a) {
1483
              c = _cost[_arc_id[a]];
1484
              if (c < min_cost) {
1485
                min_cost = c;
1486
                min_arc = a;
1487
              }
1488
            }
1489
            if (min_arc != INVALID) {
1490
              arc_vector.push_back(_arc_id[min_arc]);
1491
            }
1492
          }
1493
        }
1494
      } else {
1495
        // Find the min. cost outgoing arc for each supply node
1496
        for (int i = 0; i != int(supply_nodes.size()); ++i) {
1497
          Node u = supply_nodes[i];
1498
          Cost c, min_cost = std::numeric_limits<Cost>::max();
1499
          Arc min_arc = INVALID;
1500
          for (OutArcIt a(_graph, u); a != INVALID; ++a) {
1501
            c = _cost[_arc_id[a]];
1502
            if (c < min_cost) {
1503
              min_cost = c;
1504
              min_arc = a;
1505
            }
1506
          }
1507
          if (min_arc != INVALID) {
1508
            arc_vector.push_back(_arc_id[min_arc]);
1509
          }
1510
        }
1511
      }
1512

	
1513
      // Perform heuristic initial pivots
1514
      for (int i = 0; i != int(arc_vector.size()); ++i) {
1515
        in_arc = arc_vector[i];
1516
        if (_state[in_arc] * (_cost[in_arc] + _pi[_source[in_arc]] -
1517
            _pi[_target[in_arc]]) >= 0) continue;
1518
        findJoinNode();
1519
        bool change = findLeavingArc();
1520
        if (delta >= MAX) return false;
1521
        changeFlow(change);
1522
        if (change) {
1523
          updateTreeStructure();
1524
          updatePotential();
1525
        }
1526
      }
1527
      return true;
1528
    }
1529

	
1435 1530
    // Execute the algorithm
1436 1531
    ProblemType start(PivotRule pivot_rule) {
1437 1532
      // Select the pivot rule implementation
1438 1533
      switch (pivot_rule) {
1439 1534
        case FIRST_ELIGIBLE:
1440 1535
          return start<FirstEligiblePivotRule>();
1441 1536
        case BEST_ELIGIBLE:
1442 1537
          return start<BestEligiblePivotRule>();
1443 1538
        case BLOCK_SEARCH:
1444 1539
          return start<BlockSearchPivotRule>();
1445 1540
        case CANDIDATE_LIST:
1446 1541
          return start<CandidateListPivotRule>();
1447 1542
        case ALTERING_LIST:
1448 1543
          return start<AlteringListPivotRule>();
1449 1544
      }
1450 1545
      return INFEASIBLE; // avoid warning
1451 1546
    }
1452 1547

	
1453 1548
    template <typename PivotRuleImpl>
1454 1549
    ProblemType start() {
1455 1550
      PivotRuleImpl pivot(*this);
1456 1551

	
1552
      // Perform heuristic initial pivots
1553
      if (!initialPivots()) return UNBOUNDED;
1554

	
1457 1555
      // Execute the Network Simplex algorithm
1458 1556
      while (pivot.findEnteringArc()) {
1459 1557
        findJoinNode();
1460 1558
        bool change = findLeavingArc();
1461 1559
        if (delta >= MAX) return UNBOUNDED;
1462 1560
        changeFlow(change);
1463 1561
        if (change) {
1464 1562
          updateTreeStructure();
1465 1563
          updatePotential();
1466 1564
        }
1467 1565
      }
1468 1566
      
1469 1567
      // Check feasibility
1470 1568
      for (int e = _search_arc_num; e != _all_arc_num; ++e) {
1471 1569
        if (_flow[e] != 0) return INFEASIBLE;
1472 1570
      }
1473 1571

	
1474 1572
      // Transform the solution and the supply map to the original form
1475 1573
      if (_have_lower) {
1476 1574
        for (int i = 0; i != _arc_num; ++i) {
1477 1575
          Value c = _lower[i];
1478 1576
          if (c != 0) {
1479 1577
            _flow[i] += c;
1480 1578
            _supply[_source[i]] += c;
1481 1579
            _supply[_target[i]] -= c;
1482 1580
          }
1483 1581
        }
1484 1582
      }
1485 1583
      
1486 1584
      // Shift potentials to meet the requirements of the GEQ/LEQ type
1487 1585
      // optimality conditions
1488 1586
      if (_sum_supply == 0) {
1489 1587
        if (_stype == GEQ) {
1490 1588
          Cost max_pot = std::numeric_limits<Cost>::min();
1491 1589
          for (int i = 0; i != _node_num; ++i) {
1492 1590
            if (_pi[i] > max_pot) max_pot = _pi[i];
1493 1591
          }
1494 1592
          if (max_pot > 0) {
1495 1593
            for (int i = 0; i != _node_num; ++i)
1496 1594
              _pi[i] -= max_pot;
1497 1595
          }
1498 1596
        } else {
1499 1597
          Cost min_pot = std::numeric_limits<Cost>::max();
1500 1598
          for (int i = 0; i != _node_num; ++i) {
1501 1599
            if (_pi[i] < min_pot) min_pot = _pi[i];
1502 1600
          }
1503 1601
          if (min_pot < 0) {
1504 1602
            for (int i = 0; i != _node_num; ++i)
1505 1603
              _pi[i] -= min_pot;
1506 1604
          }
1507 1605
        }
1508 1606
      }
1509 1607

	
1510 1608
      return OPTIMAL;
1511 1609
    }
1512 1610

	
1513 1611
  }; //class NetworkSimplex
1514 1612

	
1515 1613
  ///@}
1516 1614

	
1517 1615
} //namespace lemon
1518 1616

	
1519 1617
#endif //LEMON_NETWORK_SIMPLEX_H
0 comments (0 inline)