Scippy

SCIP

Solving Constraint Integer Programs

presol_domcol.c
Go to the documentation of this file.
1 /* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
2 /* */
3 /* This file is part of the program and library */
4 /* SCIP --- Solving Constraint Integer Programs */
5 /* */
6 /* Copyright (C) 2002-2020 Konrad-Zuse-Zentrum */
7 /* fuer Informationstechnik Berlin */
8 /* */
9 /* SCIP is distributed under the terms of the ZIB Academic License. */
10 /* */
11 /* You should have received a copy of the ZIB Academic License */
12 /* along with SCIP; see the file COPYING. If not visit scipopt.org. */
13 /* */
14 /* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
15 
16 /**@file presol_domcol.c
17  * @ingroup DEFPLUGINS_PRESOL
18  * @brief dominated column presolver
19  * @author Dieter Weninger
20  * @author Gerald Gamrath
21  *
22  * This presolver looks for dominance relations between variable pairs.
23  * From a dominance relation and certain bound/clique-constellations
24  * variable fixings mostly at the lower bound of the dominated variable can be derived.
25  * Additionally it is possible to improve bounds by predictive bound strengthening.
26  *
27  * @todo Also run on general CIPs, if the number of locks of the investigated variables comes only from (upgraded)
28  * linear constraints.
29  *
30  * @todo Instead of choosing variables for comparison out of one row, we should try to use 'hashvalues' for columns that
31  * indicate in which constraint type (<=, >=, or ranged row / ==) they are existing. Then sort the variables (and
32  * corresponding data) after the ranged row/equation hashvalue and only try to derive dominance on variables with
33  * the same hashvalue on ranged row/equation and fitting hashvalues for the other constraint types.
34  * @todo run on incomplete matrices; in order to do so, check at the time when dominance is detected that the locks are
35  * consistent; probably, it is sufficient to check one lock direction for each of the two variables
36  *
37  */
38 
39 /*---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8----+----9----+----0----+----1----+----2*/
40 
41 #include "blockmemshell/memory.h"
42 #include "scip/presol_domcol.h"
43 #include "scip/pub_matrix.h"
44 #include "scip/pub_message.h"
45 #include "scip/pub_misc_sort.h"
46 #include "scip/pub_presol.h"
47 #include "scip/pub_var.h"
48 #include "scip/scip_general.h"
49 #include "scip/scip_mem.h"
50 #include "scip/scip_message.h"
51 #include "scip/scip_nlp.h"
52 #include "scip/scip_numerics.h"
53 #include "scip/scip_param.h"
54 #include "scip/scip_presol.h"
55 #include "scip/scip_pricer.h"
56 #include "scip/scip_prob.h"
57 #include "scip/scip_probing.h"
58 #include "scip/scip_var.h"
59 #include <string.h>
60 
61 #define PRESOL_NAME "domcol"
62 #define PRESOL_DESC "dominated column presolver"
63 #define PRESOL_PRIORITY -1000 /**< priority of the presolver (>= 0: before, < 0: after constraint handlers) */
64 #define PRESOL_MAXROUNDS -1 /**< maximal number of presolving rounds the presolver participates in (-1: no limit) */
65 #define PRESOL_TIMING SCIP_PRESOLTIMING_EXHAUSTIVE /* timing of the presolver (fast, medium, or exhaustive) */
66 
67 #define DEFAULT_NUMMINPAIRS 1024 /**< minimal number of pair comparisons */
68 #define DEFAULT_NUMMAXPAIRS 1048576 /**< maximal number of pair comparisons */
69 
70 #define DEFAULT_PREDBNDSTR FALSE /**< should predictive bound strengthening be applied? */
71 #define DEFAULT_CONTINUOUS_RED TRUE /**< should reductions for continuous variables be carried out? */
72 
73 
74 
75 /*
76  * Data structures
77  */
78 
79 /** control parameters */
80 struct SCIP_PresolData
81 {
82  int numminpairs; /**< minimal number of pair comparisons */
83  int nummaxpairs; /**< maximal number of pair comparisons */
84  int numcurrentpairs; /**< current number of pair comparisons */
85  SCIP_Bool predbndstr; /**< flag indicating if predictive bound strengthening should be applied */
86  SCIP_Bool continuousred; /**< flag indicating if reductions for continuous variables should be performed */
87 };
88 
89 /** type of fixing direction */
91 {
92  FIXATLB = -1, /**< fix variable at lower bound */
93  NOFIX = 0, /**< do not fix variable */
94  FIXATUB = 1 /**< fix variable at upper bound */
95 };
97 
98 
99 /*
100  * Local methods
101  */
102 #ifdef SCIP_DEBUG
103 /** print a row from the constraint matrix */
104 static
105 void printRow(
106  SCIP* scip, /**< SCIP main data structure */
107  SCIP_MATRIX* matrix, /**< matrix containing the constraints */
108  int row /**< row index for printing */
109  )
110 {
111  int* rowpnt;
112  int* rowend;
113  int c;
114  SCIP_Real val;
115  SCIP_Real* valpnt;
116  char relation;
117  SCIP_VAR* var;
118 
119  relation='-';
120  if( !SCIPmatrixIsRowRhsInfinity(matrix, row) &&
121  SCIPisEQ(scip, SCIPmatrixGetRowLhs(matrix, row), SCIPmatrixGetRowRhs(matrix, row)) )
122  {
123  relation='e';
124  }
125  else if( !SCIPmatrixIsRowRhsInfinity(matrix, row) &&
126  !SCIPisEQ(scip, SCIPmatrixGetRowLhs(matrix, row), SCIPmatrixGetRowRhs(matrix, row)) )
127  {
128  relation='r';
129  }
130  else
131  {
132  relation='g';
133  }
134 
135  rowpnt = SCIPmatrixGetRowIdxPtr(matrix, row);
136  rowend = rowpnt + SCIPmatrixGetRowNNonzs(matrix, row);
137  valpnt = SCIPmatrixGetRowValPtr(matrix, row);
138 
139  SCIPdebugMsgPrint(scip, "\n(L:%g) [%c] %g <=",
140  (SCIPmatrixGetRowNMinActPosInf(matrix, row) + SCIPmatrixGetRowNMinActNegInf(matrix,row) > 0) ?
141  -SCIPinfinity(scip) :
142  SCIPmatrixGetRowMinActivity(matrix, row), relation, SCIPmatrixGetRowLhs(matrix, row));
143  for(; (rowpnt < rowend); rowpnt++, valpnt++)
144  {
145  c = *rowpnt;
146  val = *valpnt;
147  var = SCIPmatrixGetVar(matrix, c);
148  SCIPdebugMsgPrint(scip, " %g{%s[idx:%d][bnd:%g,%g]}",
149  val, SCIPvarGetName(var), c, SCIPvarGetLbGlobal(var), SCIPvarGetUbGlobal(var));
150  }
151  SCIPdebugMsgPrint(scip, " <= %g (U:%g)", (SCIPmatrixGetRowNMaxActPosInf(matrix, row) + SCIPmatrixGetRowNMaxActNegInf(matrix, row) > 0) ?
152  SCIPinfinity(scip) :
153  SCIPmatrixGetRowRhs(matrix, row), SCIPmatrixGetRowMaxActivity(matrix , row));
154 }
155 
156 /** print all rows from the constraint matrix containing a variable */
157 static
158 SCIP_RETCODE printRowsOfCol(
159  SCIP* scip, /**< SCIP main data structure */
160  SCIP_MATRIX* matrix, /**< matrix containing the constraints */
161  int col /**< column index for printing */
162  )
163 {
164  int numrows;
165  int* rows;
166  int i;
167  int* colpnt;
168  int* colend;
169 
170  numrows = SCIPmatrixGetColNNonzs(matrix, col);
171 
172  SCIP_CALL( SCIPallocBufferArray(scip, &rows, numrows) );
173 
174  colpnt = SCIPmatrixGetColIdxPtr(matrix, col);
175  colend = colpnt + SCIPmatrixGetColNNonzs(matrix, col);
176  for( i = 0; (colpnt < colend); colpnt++, i++ )
177  {
178  rows[i] = *colpnt;
179  }
180 
181  SCIPdebugMsgPrint(scip, "\n-------");
182  SCIPdebugMsgPrint(scip, "\ncol %d number rows: %d",col,numrows);
183  for( i = 0; i < numrows; i++ )
184  {
185  printRow(scip, matrix, rows[i]);
186  }
187  SCIPdebugMsgPrint(scip, "\n-------");
188 
189  SCIPfreeBufferArray(scip, &rows);
190 
191  return SCIP_OKAY;
192 }
193 
194 /** print information about a dominance relation */
195 static
196 SCIP_RETCODE printDomRelInfo(
197  SCIP* scip, /**< SCIP main data structure */
198  SCIP_MATRIX* matrix, /**< matrix containing the constraints */
199  SCIP_VAR* dominatingvar, /**< dominating variable */
200  int dominatingidx, /**< index of dominating variable */
201  SCIP_VAR* dominatedvar, /**< dominated variable */
202  int dominatedidx, /**< index of dominated variable */
203  SCIP_Real dominatingub, /**< predicted upper bound of dominating variable */
204  SCIP_Real dominatingwclb /**< worst case lower bound of dominating variable */
205  )
206 {
207  char type;
208 
209  assert(SCIPvarGetType(dominatingvar)==SCIPvarGetType(dominatedvar));
210 
211  switch(SCIPvarGetType(dominatingvar))
212  {
214  type='C';
215  break;
216  case SCIP_VARTYPE_BINARY:
217  type='B';
218  break;
221  type='I';
222  break;
223  default:
224  SCIPerrorMessage("unknown variable type\n");
225  SCIPABORT();
226  return SCIP_INVALIDDATA; /*lint !e527*/
227  }
228 
229  SCIPdebugMsgPrint(scip, "\n\n### [%c], obj:%g->%g,\t%s[idx:%d](nrows:%d)->%s[idx:%d](nrows:%d)\twclb=%g, ub'=%g, ub=%g",
230  type, SCIPvarGetObj(dominatingvar), SCIPvarGetObj(dominatedvar),
231  SCIPvarGetName(dominatingvar), dominatingidx, SCIPmatrixGetColNNonzs(matrix, dominatingidx),
232  SCIPvarGetName(dominatedvar), dominatedidx, SCIPmatrixGetColNNonzs(matrix, dominatedidx),
233  dominatingwclb, dominatingub, SCIPvarGetUbGlobal(dominatingvar));
234 
235  SCIP_CALL( printRowsOfCol(scip, matrix, dominatingidx) );
236 
237  return SCIP_OKAY;
238 }
239 #endif
240 
241 
242 /** get minimum/maximum residual activity for the specified variable and with another variable set to its upper bound */
243 static
245  SCIP* scip,
246  SCIP_MATRIX* matrix,
247  int row,
248  int col,
249  SCIP_Real coef,
250  int upperboundcol,
251  SCIP_Real upperboundcoef,
252  SCIP_Real* minresactivity,
253  SCIP_Real* maxresactivity,
254  SCIP_Bool* success
255  )
256 {
257  SCIP_VAR* var;
258  SCIP_VAR* upperboundvar;
259  SCIP_Real lb;
260  SCIP_Real ub;
261  SCIP_Real lbupperboundvar;
262  SCIP_Real ubupperboundvar;
263  SCIP_Real tmpmaxact;
264  SCIP_Real tmpminact;
265  int maxactinf;
266  int minactinf;
267 
268  assert(scip != NULL);
269  assert(matrix != NULL);
270  assert(row < SCIPmatrixGetNRows(matrix));
271  assert(minresactivity != NULL);
272  assert(maxresactivity != NULL);
273 
274  var = SCIPmatrixGetVar(matrix, col);
275  upperboundvar = SCIPmatrixGetVar(matrix, upperboundcol);
276  assert(var != NULL);
277  assert(upperboundvar != NULL);
278 
279  lb = SCIPvarGetLbGlobal(var);
280  ub = SCIPvarGetUbGlobal(var);
281 
282  maxactinf = SCIPmatrixGetRowNMaxActPosInf(matrix, row) + SCIPmatrixGetRowNMaxActNegInf(matrix, row);
283  minactinf = SCIPmatrixGetRowNMinActPosInf(matrix, row) + SCIPmatrixGetRowNMinActNegInf(matrix, row);
284 
285  assert(!SCIPisInfinity(scip, lb));
286  assert(!SCIPisInfinity(scip, -ub));
287 
288  lbupperboundvar = SCIPvarGetLbGlobal(upperboundvar);
289  ubupperboundvar = SCIPvarGetUbGlobal(upperboundvar);
290 
291  assert(!SCIPisInfinity(scip, lbupperboundvar));
292  assert(!SCIPisInfinity(scip, -ubupperboundvar));
293 
294  tmpminact = SCIPmatrixGetRowMinActivity(matrix, row);
295  tmpmaxact = SCIPmatrixGetRowMaxActivity(matrix, row);
296 
297  /* first, adjust min and max activity such that upperboundvar is always set to its upper bound */
298 
299  /* abort if the upper bound is infty */
300  if( SCIPisInfinity(scip, ubupperboundvar) )
301  {
302  *success = FALSE;
303  return;
304  }
305  else
306  {
307  /* coef > 0 --> the lower bound is used for the minactivity --> update the minactivity */
308  if( upperboundcoef > 0.0 )
309  {
310  if( SCIPisInfinity(scip, -lbupperboundvar) )
311  {
312  assert(minactinf > 0);
313  minactinf--;
314  tmpminact += (upperboundcoef * ubupperboundvar);
315  }
316  else
317  {
318  tmpminact = tmpminact - (upperboundcoef * lbupperboundvar) + (upperboundcoef * ubupperboundvar);
319  }
320  }
321  /* coef < 0 --> the lower bound is used for the maxactivity --> update the maxactivity */
322  else
323  {
324  if( SCIPisInfinity(scip, -lbupperboundvar) )
325  {
326  assert(maxactinf > 0);
327  maxactinf--;
328  tmpmaxact += (upperboundcoef * ubupperboundvar);
329  }
330  else
331  {
332  tmpmaxact = tmpmaxact - (upperboundcoef * lbupperboundvar) + (upperboundcoef * ubupperboundvar);
333  }
334  }
335  }
336 
337  *success = TRUE;
338 
339  /* the coefficient is positive, so the upper bound contributed to the maxactivity and the lower bound to the minactivity */
340  if( coef >= 0.0 )
341  {
342  if( SCIPisInfinity(scip, ub) )
343  {
344  assert(maxactinf >= 1);
345  if( maxactinf == 1 )
346  {
347  *maxresactivity = tmpmaxact;
348  }
349  else
350  *maxresactivity = SCIPinfinity(scip);
351  }
352  else
353  {
354  if( maxactinf > 0 )
355  *maxresactivity = SCIPinfinity(scip);
356  else
357  *maxresactivity = tmpmaxact - coef * ub;
358  }
359 
360  if( SCIPisInfinity(scip, -lb) )
361  {
362  assert(minactinf >= 1);
363  if( minactinf == 1 )
364  {
365  *minresactivity = tmpminact;
366  }
367  else
368  *minresactivity = -SCIPinfinity(scip);
369  }
370  else
371  {
372  if( minactinf > 0 )
373  *minresactivity = -SCIPinfinity(scip);
374  else
375  *minresactivity = tmpminact - coef * lb;
376  }
377  }
378  /* the coefficient is negative, so the lower bound contributed to the maxactivity and the upper bound to the minactivity */
379  else
380  {
381  if( SCIPisInfinity(scip, -lb) )
382  {
383  assert(maxactinf >= 1);
384  if( maxactinf == 1 )
385  {
386  *maxresactivity = tmpmaxact;
387  }
388  else
389  *maxresactivity = SCIPinfinity(scip);
390  }
391  else
392  {
393  if( maxactinf > 0 )
394  *maxresactivity = SCIPinfinity(scip);
395  else
396  *maxresactivity = tmpmaxact - coef * lb;
397  }
398 
399  if( SCIPisInfinity(scip, ub) )
400  {
401  assert(minactinf >= 1);
402  if( minactinf == 1 )
403  {
404  *minresactivity = tmpminact;
405  }
406  else
407  *minresactivity = -SCIPinfinity(scip);
408  }
409  else
410  {
411  if( minactinf > 0 )
412  *minresactivity = -SCIPinfinity(scip);
413  else
414  *minresactivity = tmpminact - coef * ub;
415  }
416  }
417 }
418 
419 /** get minimum/maximum residual activity for the specified variable and with another variable set to its lower bound */
420 static
422  SCIP* scip, /**< SCIP main data structure */
423  SCIP_MATRIX* matrix, /**< matrix containing the constraints */
424  int row, /**< row index */
425  int col, /**< column index */
426  SCIP_Real coef, /**< coefficient of the column in this row */
427  int lowerboundcol, /**< column index of variable to set to its lower bound */
428  SCIP_Real lowerboundcoef, /**< coefficient in this row of the column to be set to its lower bound */
429  SCIP_Real* minresactivity, /**< minimum residual activity of this row */
430  SCIP_Real* maxresactivity, /**< maximum residual activity of this row */
431  SCIP_Bool* success /**< pointer to store whether the computation was successful */
432  )
433 {
434  SCIP_VAR* var;
435  SCIP_VAR* lowerboundvar;
436  SCIP_Real lb;
437  SCIP_Real ub;
438  SCIP_Real lblowerboundvar;
439  SCIP_Real ublowerboundvar;
440  SCIP_Real tmpmaxact;
441  SCIP_Real tmpminact;
442  int maxactinf;
443  int minactinf;
444 
445  assert(scip != NULL);
446  assert(matrix != NULL);
447  assert(row < SCIPmatrixGetNRows(matrix));
448  assert(minresactivity != NULL);
449  assert(maxresactivity != NULL);
450 
451  var = SCIPmatrixGetVar(matrix, col);;
452  lowerboundvar = SCIPmatrixGetVar(matrix, lowerboundcol);
453  assert(var != NULL);
454  assert(lowerboundvar != NULL);
455 
456  lb = SCIPvarGetLbGlobal(var);
457  ub = SCIPvarGetUbGlobal(var);
458 
459  maxactinf = SCIPmatrixGetRowNMaxActPosInf(matrix, row) + SCIPmatrixGetRowNMaxActNegInf(matrix, row);
460  minactinf = SCIPmatrixGetRowNMinActPosInf(matrix, row) + SCIPmatrixGetRowNMinActNegInf(matrix, row);
461 
462  assert(!SCIPisInfinity(scip, lb));
463  assert(!SCIPisInfinity(scip, -ub));
464 
465  lblowerboundvar = SCIPvarGetLbGlobal(lowerboundvar);
466  ublowerboundvar = SCIPvarGetUbGlobal(lowerboundvar);
467 
468  assert(!SCIPisInfinity(scip, lblowerboundvar));
469  assert(!SCIPisInfinity(scip, -ublowerboundvar));
470 
471  tmpminact = SCIPmatrixGetRowMinActivity(matrix, row);
472  tmpmaxact = SCIPmatrixGetRowMaxActivity(matrix, row);
473 
474  /* first, adjust min and max activity such that lowerboundvar is always set to its lower bound */
475 
476  /* abort if the lower bound is -infty */
477  if( SCIPisInfinity(scip, -lblowerboundvar) )
478  {
479  *success = FALSE;
480  return;
481  }
482  else
483  {
484  /* coef > 0 --> the upper bound is used for the maxactivity --> update the maxactivity */
485  if( lowerboundcoef > 0.0 )
486  {
487  if( SCIPisInfinity(scip, ublowerboundvar) )
488  {
489  assert(maxactinf > 0);
490  maxactinf--;
491  tmpmaxact += (lowerboundcoef * lblowerboundvar);
492  }
493  else
494  {
495  tmpmaxact = tmpmaxact - (lowerboundcoef * ublowerboundvar) + (lowerboundcoef * lblowerboundvar);
496  }
497  }
498  /* coef < 0 --> the upper bound is used for the minactivity --> update the minactivity */
499  else
500  {
501  if( SCIPisInfinity(scip, ublowerboundvar) )
502  {
503  assert(minactinf > 0);
504  minactinf--;
505  tmpminact += (lowerboundcoef * lblowerboundvar);
506  }
507  else
508  {
509  tmpminact = tmpminact - (lowerboundcoef * ublowerboundvar) + (lowerboundcoef * lblowerboundvar);
510  }
511  }
512  }
513 
514  *success = TRUE;
515 
516  /* the coefficient is positive, so the upper bound contributed to the maxactivity and the lower bound to the minactivity */
517  if( coef >= 0.0 )
518  {
519  if( SCIPisInfinity(scip, ub) )
520  {
521  assert(maxactinf >= 1);
522  if( maxactinf == 1 )
523  {
524  *maxresactivity = tmpmaxact;
525  }
526  else
527  *maxresactivity = SCIPinfinity(scip);
528  }
529  else
530  {
531  if( maxactinf > 0 )
532  *maxresactivity = SCIPinfinity(scip);
533  else
534  *maxresactivity = tmpmaxact - coef * ub;
535  }
536 
537  if( SCIPisInfinity(scip, -lb) )
538  {
539  assert(minactinf >= 1);
540  if( minactinf == 1 )
541  {
542  *minresactivity = tmpminact;
543  }
544  else
545  *minresactivity = -SCIPinfinity(scip);
546  }
547  else
548  {
549  if( minactinf > 0 )
550  *minresactivity = -SCIPinfinity(scip);
551  else
552  *minresactivity = tmpminact - coef * lb;
553  }
554  }
555  /* the coefficient is negative, so the lower bound contributed to the maxactivity and the upper bound to the minactivity */
556  else
557  {
558  if( SCIPisInfinity(scip, -lb) )
559  {
560  assert(maxactinf >= 1);
561  if( maxactinf == 1 )
562  {
563  *maxresactivity = tmpmaxact;
564  }
565  else
566  *maxresactivity = SCIPinfinity(scip);
567  }
568  else
569  {
570  if( maxactinf > 0 )
571  *maxresactivity = SCIPinfinity(scip);
572  else
573  *maxresactivity = tmpmaxact - coef * lb;
574  }
575 
576  if( SCIPisInfinity(scip, ub) )
577  {
578  assert(minactinf >= 1);
579  if( minactinf == 1 )
580  {
581  *minresactivity = tmpminact;
582  }
583  else
584  *minresactivity = -SCIPinfinity(scip);
585  }
586  else
587  {
588  if( minactinf > 0 )
589  *minresactivity = -SCIPinfinity(scip);
590  else
591  *minresactivity = tmpminact - coef * ub;
592  }
593  }
594 }
595 
596 /** Calculate bounds of the dominated variable by rowbound analysis.
597  * We use a special kind of predictive rowbound analysis by first setting the dominating variable to its upper bound.
598  */
599 static
601  SCIP* scip, /**< SCIP main data structure */
602  SCIP_MATRIX* matrix, /**< matrix containing the constraints */
603  int row, /**< current row index */
604  int coldominating, /**< column index of dominating variable */
605  SCIP_Real valdominating, /**< row coefficient of dominating variable */
606  int coldominated, /**< column index of dominated variable */
607  SCIP_Real valdominated, /**< row coefficient of dominated variable */
608  SCIP_Bool* ubcalculated, /**< was a upper bound calculated? */
609  SCIP_Real* calculatedub, /**< predicted upper bound */
610  SCIP_Bool* wclbcalculated, /**< was a lower worst case bound calculated? */
611  SCIP_Real* calculatedwclb, /**< predicted worst case lower bound */
612  SCIP_Bool* lbcalculated, /**< was a lower bound calculated? */
613  SCIP_Real* calculatedlb, /**< predicted lower bound */
614  SCIP_Bool* wcubcalculated, /**< was a worst case upper bound calculated? */
615  SCIP_Real* calculatedwcub /**< calculated worst case upper bound */
616  )
617 {
618  SCIP_Real minresactivity;
619  SCIP_Real maxresactivity;
620  SCIP_Real lhs;
621  SCIP_Real rhs;
622  SCIP_Bool success;
623 
624  assert(scip != NULL);
625  assert(matrix != NULL);
626  assert(0 <= row && row < SCIPmatrixGetNRows(matrix));
627  assert(0 <= coldominating && coldominating < SCIPmatrixGetNColumns(matrix));
628  assert(0 <= coldominated && coldominated < SCIPmatrixGetNColumns(matrix));
629 
630  assert(ubcalculated != NULL);
631  assert(calculatedub != NULL);
632  assert(wclbcalculated != NULL);
633  assert(calculatedwclb != NULL);
634  assert(lbcalculated != NULL);
635  assert(calculatedlb != NULL);
636  assert(wcubcalculated != NULL);
637  assert(calculatedwcub != NULL);
638 
639  assert(!SCIPisZero(scip, valdominated));
640  assert(SCIPmatrixGetVar(matrix, coldominated) != NULL);
641 
642  *ubcalculated = FALSE;
643  *wclbcalculated = FALSE;
644  *lbcalculated = FALSE;
645  *wcubcalculated = FALSE;
646 
647  /* no rowbound analysis for multiaggregated variables, which should not exist, because the matrix only consists of
648  * active variables
649  */
650  assert(SCIPvarGetStatus(SCIPmatrixGetVar(matrix, coldominating)) != SCIP_VARSTATUS_MULTAGGR);
651  assert(SCIPvarGetStatus(SCIPmatrixGetVar(matrix, coldominated)) != SCIP_VARSTATUS_MULTAGGR);
652 
653  lhs = SCIPmatrixGetRowLhs(matrix, row);
654  rhs = SCIPmatrixGetRowRhs(matrix, row);
655  assert(!SCIPisInfinity(scip, lhs));
656  assert(!SCIPisInfinity(scip, -rhs));
657 
658  /* get minimum/maximum activity of this row without the dominated variable */
659  getActivityResidualsUpperBound(scip, matrix, row, coldominated, valdominated, coldominating, valdominating,
660  &minresactivity, &maxresactivity, &success);
661 
662  if( !success )
663  return SCIP_OKAY;
664 
665  assert(!SCIPisInfinity(scip, minresactivity));
666  assert(!SCIPisInfinity(scip, -maxresactivity));
667 
668  *calculatedub = SCIPinfinity(scip);
669  *calculatedwclb = -SCIPinfinity(scip);
670  *calculatedlb = -SCIPinfinity(scip);
671  *calculatedwcub = SCIPinfinity(scip);
672 
673  /* predictive rowbound analysis */
674 
675  if( valdominated > 0.0 )
676  {
677  /* lower bound calculation */
678  if( !SCIPisInfinity(scip, maxresactivity) )
679  {
680  *calculatedlb = (lhs - maxresactivity)/valdominated;
681  *lbcalculated = TRUE;
682  }
683 
684  /* worst case calculation of lower bound */
685  if( !SCIPisInfinity(scip, -minresactivity) )
686  {
687  *calculatedwclb = (lhs - minresactivity)/valdominated;
688  *wclbcalculated = TRUE;
689  }
690  else
691  {
692  /* worst case lower bound is infinity */
693  *calculatedwclb = SCIPinfinity(scip);
694  *wclbcalculated = TRUE;
695  }
696 
697  /* we can only calculate upper bounds, of the right hand side is finite */
698  if( !SCIPmatrixIsRowRhsInfinity(matrix, row) )
699  {
700  /* upper bound calculation */
701  if( !SCIPisInfinity(scip, -minresactivity) )
702  {
703  *calculatedub = (rhs - minresactivity)/valdominated;
704  *ubcalculated = TRUE;
705  }
706 
707  /* worst case calculation of upper bound */
708  if( !SCIPisInfinity(scip, maxresactivity) )
709  {
710  *calculatedwcub = (rhs - maxresactivity)/valdominated;
711  *wcubcalculated = TRUE;
712  }
713  else
714  {
715  /* worst case upper bound is -infinity */
716  *calculatedwcub = -SCIPinfinity(scip);
717  *wcubcalculated = TRUE;
718  }
719  }
720  }
721  else
722  {
723  /* upper bound calculation */
724  if( !SCIPisInfinity(scip, maxresactivity) )
725  {
726  *calculatedub = (lhs - maxresactivity)/valdominated;
727  *ubcalculated = TRUE;
728  }
729 
730  /* worst case calculation of upper bound */
731  if( !SCIPisInfinity(scip, -minresactivity) )
732  {
733  *calculatedwcub = (lhs - minresactivity)/valdominated;
734  *wcubcalculated = TRUE;
735  }
736  else
737  {
738  /* worst case upper bound is -infinity */
739  *calculatedwcub = -SCIPinfinity(scip);
740  *wcubcalculated = TRUE;
741  }
742 
743  /* we can only calculate lower bounds, of the right hand side is finite */
744  if( !SCIPmatrixIsRowRhsInfinity(matrix, row) )
745  {
746  /* lower bound calculation */
747  if( !SCIPisInfinity(scip, -minresactivity) )
748  {
749  *calculatedlb = (rhs - minresactivity)/valdominated;
750  *lbcalculated = TRUE;
751  }
752 
753  /* worst case calculation of lower bound */
754  if( !SCIPisInfinity(scip, maxresactivity) )
755  {
756  *calculatedwclb = (rhs - maxresactivity)/valdominated;
757  *wclbcalculated = TRUE;
758  }
759  else
760  {
761  /* worst case lower bound is infinity */
762  *calculatedwclb = SCIPinfinity(scip);
763  *wclbcalculated = TRUE;
764  }
765  }
766  }
767 
768  return SCIP_OKAY;
769 }
770 
771 /** Calculate bounds of the dominating variable by rowbound analysis.
772  * We use a special kind of predictive rowbound analysis by first setting the dominated variable to its lower bound.
773  */
774 static
776  SCIP* scip, /**< SCIP main data structure */
777  SCIP_MATRIX* matrix, /**< matrix containing the constraints */
778  int row, /**< current row index */
779  int coldominating, /**< column index of dominating variable */
780  SCIP_Real valdominating, /**< row coefficient of dominating variable */
781  int coldominated, /**< column index of dominated variable */
782  SCIP_Real valdominated, /**< row coefficient of dominated variable */
783  SCIP_Bool* ubcalculated, /**< was a upper bound calculated? */
784  SCIP_Real* calculatedub, /**< predicted upper bound */
785  SCIP_Bool* wclbcalculated, /**< was a lower worst case bound calculated? */
786  SCIP_Real* calculatedwclb, /**< predicted worst case lower bound */
787  SCIP_Bool* lbcalculated, /**< was a lower bound calculated? */
788  SCIP_Real* calculatedlb, /**< predicted lower bound */
789  SCIP_Bool* wcubcalculated, /**< was a worst case upper bound calculated? */
790  SCIP_Real* calculatedwcub /**< calculated worst case upper bound */
791  )
792 {
793  SCIP_Real minresactivity;
794  SCIP_Real maxresactivity;
795  SCIP_Real lhs;
796  SCIP_Real rhs;
797  SCIP_Bool success;
798 
799  assert(scip != NULL);
800  assert(matrix != NULL);
801  assert(0 <= row && row < SCIPmatrixGetNRows(matrix) );
802  assert(0 <= coldominating && coldominating < SCIPmatrixGetNColumns(matrix));
803  assert(0 <= coldominated && coldominated < SCIPmatrixGetNColumns(matrix));
804 
805  assert(ubcalculated != NULL);
806  assert(calculatedub != NULL);
807  assert(wclbcalculated != NULL);
808  assert(calculatedwclb != NULL);
809  assert(lbcalculated != NULL);
810  assert(calculatedlb != NULL);
811  assert(wcubcalculated != NULL);
812  assert(calculatedwcub != NULL);
813 
814  assert(!SCIPisZero(scip, valdominating));
815  assert(SCIPmatrixGetVar(matrix, coldominating) != NULL);
816 
817  *ubcalculated = FALSE;
818  *wclbcalculated = FALSE;
819  *lbcalculated = FALSE;
820  *wcubcalculated = FALSE;
821 
822  /* no rowbound analysis for multiaggregated variables, which should not exist, because the matrix only consists of
823  * active variables
824  */
825  assert(SCIPvarGetStatus(SCIPmatrixGetVar(matrix, coldominating)) != SCIP_VARSTATUS_MULTAGGR);
826  assert(SCIPvarGetStatus(SCIPmatrixGetVar(matrix, coldominated)) != SCIP_VARSTATUS_MULTAGGR);
827 
828  lhs = SCIPmatrixGetRowLhs(matrix, row);
829  rhs = SCIPmatrixGetRowRhs(matrix, row);
830  assert(!SCIPisInfinity(scip, lhs));
831  assert(!SCIPisInfinity(scip, -rhs));
832 
833  /* get minimum/maximum activity of this row without the dominating variable */
834  getActivityResidualsLowerBound(scip, matrix, row, coldominating, valdominating, coldominated, valdominated,
835  &minresactivity, &maxresactivity, &success);
836 
837  if( !success )
838  return SCIP_OKAY;
839 
840  assert(!SCIPisInfinity(scip, minresactivity));
841  assert(!SCIPisInfinity(scip, -maxresactivity));
842 
843  *calculatedub = SCIPinfinity(scip);
844  *calculatedwclb = -SCIPinfinity(scip);
845  *calculatedlb = -SCIPinfinity(scip);
846  *calculatedwcub = SCIPinfinity(scip);
847 
848  /* predictive rowbound analysis */
849 
850  if( valdominating > 0.0 )
851  {
852  /* lower bound calculation */
853  if( !SCIPisInfinity(scip, maxresactivity) )
854  {
855  *calculatedlb = (lhs - maxresactivity)/valdominating;
856  *lbcalculated = TRUE;
857  }
858 
859  /* worst case calculation of lower bound */
860  if( !SCIPisInfinity(scip, -minresactivity) )
861  {
862  *calculatedwclb = (lhs - minresactivity)/valdominating;
863  *wclbcalculated = TRUE;
864  }
865  else
866  {
867  /* worst case lower bound is infinity */
868  *calculatedwclb = SCIPinfinity(scip);
869  *wclbcalculated = TRUE;
870  }
871 
872  /* we can only calculate upper bounds, of the right hand side is finite */
873  if( !SCIPmatrixIsRowRhsInfinity(matrix, row) )
874  {
875  /* upper bound calculation */
876  if( !SCIPisInfinity(scip, -minresactivity) )
877  {
878  *calculatedub = (rhs - minresactivity)/valdominating;
879  *ubcalculated = TRUE;
880  }
881 
882  /* worst case calculation of upper bound */
883  if( !SCIPisInfinity(scip, maxresactivity) )
884  {
885  *calculatedwcub = (rhs - maxresactivity)/valdominating;
886  *wcubcalculated = TRUE;
887  }
888  else
889  {
890  /* worst case upper bound is -infinity */
891  *calculatedwcub = -SCIPinfinity(scip);
892  *wcubcalculated = TRUE;
893  }
894  }
895  }
896  else
897  {
898  /* upper bound calculation */
899  if( !SCIPisInfinity(scip, maxresactivity) )
900  {
901  *calculatedub = (lhs - maxresactivity)/valdominating;
902  *ubcalculated = TRUE;
903  }
904 
905  /* worst case calculation of upper bound */
906  if( !SCIPisInfinity(scip, -minresactivity) )
907  {
908  *calculatedwcub = (lhs - minresactivity)/valdominating;
909  *wcubcalculated = TRUE;
910  }
911  else
912  {
913  /* worst case upper bound is -infinity */
914  *calculatedwcub = -SCIPinfinity(scip);
915  *wcubcalculated = TRUE;
916  }
917 
918  /* we can only calculate lower bounds, of the right hand side is finite */
919  if( !SCIPmatrixIsRowRhsInfinity(matrix, row) )
920  {
921  /* lower bound calculation */
922  if( !SCIPisInfinity(scip, -minresactivity) )
923  {
924  *calculatedlb = (rhs - minresactivity)/valdominating;
925  *lbcalculated = TRUE;
926  }
927 
928  /* worst case calculation of lower bound */
929  if( !SCIPisInfinity(scip, maxresactivity) )
930  {
931  *calculatedwclb = (rhs - maxresactivity)/valdominating;
932  *wclbcalculated = TRUE;
933  }
934  else
935  {
936  /* worst case lower bound is infinity */
937  *calculatedwclb = SCIPinfinity(scip);
938  *wclbcalculated = TRUE;
939  }
940  }
941  }
942 
943  return SCIP_OKAY;
944 }
945 
946 /** try to find new variable bounds and update them when they are better then the old bounds */
947 static
949  SCIP* scip, /**< SCIP main data structure */
950  SCIP_MATRIX* matrix, /**< matrix containing the constraints */
951  int row, /**< row index */
952  int col1, /**< dominating variable index */
953  SCIP_Real val1, /**< dominating variable coefficient */
954  int col2, /**< dominated variable index */
955  SCIP_Real val2, /**< dominated variable coefficient */
956  SCIP_Bool predictdominating, /**< flag indicating if bounds of dominating or dominated variable are predicted */
957  SCIP_Real* upperbound, /**< predicted upper bound */
958  SCIP_Real* wclowerbound, /**< predicted worst case lower bound */
959  SCIP_Real* lowerbound, /**< predicted lower bound */
960  SCIP_Real* wcupperbound /**< predicted worst case upper bound */
961  )
962 {
963  SCIP_Bool ubcalculated;
964  SCIP_Bool wclbcalculated;
965  SCIP_Bool lbcalculated;
966  SCIP_Bool wcubcalculated;
967  SCIP_Real newub;
968  SCIP_Real newwclb;
969  SCIP_Real newlb;
970  SCIP_Real newwcub;
971 
972  assert(scip != NULL);
973  assert(matrix != NULL);
974  assert(upperbound != NULL);
975  assert(wclowerbound != NULL);
976  assert(lowerbound != NULL);
977  assert(wcupperbound != NULL);
978 
979  newub = SCIPinfinity(scip);
980  newlb = -SCIPinfinity(scip);
981  newwcub = newub;
982  newwclb = newlb;
983 
984  if( predictdominating )
985  {
986  /* do predictive rowbound analysis for the dominating variable */
987  SCIP_CALL( calcVarBoundsDominating(scip, matrix, row, col1, val1, col2, val2,
988  &ubcalculated, &newub, &wclbcalculated, &newwclb,
989  &lbcalculated, &newlb, &wcubcalculated, &newwcub) );
990  }
991  else
992  {
993  /* do predictive rowbound analysis for the dominated variable */
994  SCIP_CALL( calcVarBoundsDominated(scip, matrix, row, col1, val1, col2, val2,
995  &ubcalculated, &newub, &wclbcalculated, &newwclb,
996  &lbcalculated, &newlb, &wcubcalculated, &newwcub) );
997  }
998 
999  /* update bounds in case if they are better */
1000  if( ubcalculated )
1001  {
1002  if( newub < *upperbound )
1003  *upperbound = newub;
1004  }
1005  if( wclbcalculated )
1006  {
1007  if( newwclb > *wclowerbound )
1008  *wclowerbound = newwclb;
1009  }
1010  if( lbcalculated )
1011  {
1012  if( newlb > *lowerbound )
1013  *lowerbound = newlb;
1014  }
1015  if( wcubcalculated )
1016  {
1017  if( newwcub < *wcupperbound )
1018  *wcupperbound = newwcub;
1019  }
1020 
1021  return SCIP_OKAY;
1022 }
1023 
1024 /** detect parallel columns by using the algorithm of Bixby and Wagner
1025  * see paper: "A note on Detecting Simple Redundancies in Linear Systems", June 1986
1026  */
1027 static
1029  SCIP* scip, /**< SCIP main data structure */
1030  SCIP_MATRIX* matrix, /**< matrix containing the constraints */
1031  int* pclass, /**< parallel column classes */
1032  SCIP_Bool* varineq /**< indicating if variable is within an equation */
1033  )
1034 {
1035  SCIP_Real* valpnt;
1036  SCIP_Real* values;
1037  SCIP_Real* scale;
1038  int* classsizes;
1039  int* pcset;
1040  int* rowpnt;
1041  int* rowend;
1042  int* colindices;
1043  int* pcs;
1044  SCIP_Real startval;
1045  SCIP_Real aij;
1046  int startpc;
1047  int startk;
1048  int startt;
1049  int pcsetfill;
1050  int colidx;
1051  int k;
1052  int t;
1053  int m;
1054  int i;
1055  int r;
1056  int newpclass;
1057  int pc;
1058  int nrows;
1059  int ncols;
1060 
1061  assert(scip != NULL);
1062  assert(matrix != NULL);
1063  assert(pclass != NULL);
1064  assert(varineq != NULL);
1065 
1066  nrows = SCIPmatrixGetNRows(matrix);
1067  ncols = SCIPmatrixGetNColumns(matrix);
1068 
1069  SCIP_CALL( SCIPallocBufferArray(scip, &classsizes, ncols) );
1070  SCIP_CALL( SCIPallocBufferArray(scip, &scale, ncols) );
1071  SCIP_CALL( SCIPallocBufferArray(scip, &pcset, ncols) );
1072  SCIP_CALL( SCIPallocBufferArray(scip, &values, ncols) );
1073  SCIP_CALL( SCIPallocBufferArray(scip, &colindices, ncols) );
1074  SCIP_CALL( SCIPallocBufferArray(scip, &pcs, ncols) );
1075 
1076  BMSclearMemoryArray(scale, ncols);
1077  BMSclearMemoryArray(pclass, ncols);
1078  BMSclearMemoryArray(classsizes, ncols);
1079 
1080  classsizes[0] = ncols;
1081  pcsetfill = 0;
1082  for( t = 1; t < ncols; ++t )
1083  pcset[pcsetfill++] = t;
1084 
1085  /* loop over all rows */
1086  for( r = 0; r < nrows; ++r )
1087  {
1088  /* we consider only non-empty equations or ranged rows */
1089  if( !SCIPmatrixIsRowRhsInfinity(matrix, r) && SCIPmatrixGetRowNNonzs(matrix, r) > 0 )
1090  {
1091  rowpnt = SCIPmatrixGetRowIdxPtr(matrix, r);
1092  rowend = rowpnt + SCIPmatrixGetRowNNonzs(matrix, r);
1093  valpnt = SCIPmatrixGetRowValPtr(matrix, r);
1094 
1095  i = 0;
1096  for( ; (rowpnt < rowend); rowpnt++, valpnt++ )
1097  {
1098  aij = *valpnt;
1099  colidx = *rowpnt;
1100 
1101  /* remember variable was part of an equation or ranged row */
1102  varineq[colidx] = TRUE;
1103 
1104  if( scale[colidx] == 0.0 )
1105  scale[colidx] = aij;
1106  assert(scale[colidx] != 0.0);
1107 
1108  colindices[i] = colidx;
1109  values[i] = aij / scale[colidx];
1110  pc = pclass[colidx];
1111  assert(pc < ncols);
1112 
1113  /* update class sizes and pclass set */
1114  assert(classsizes[pc] > 0);
1115  classsizes[pc]--;
1116  if( classsizes[pc] == 0 )
1117  {
1118  assert(pcsetfill < ncols);
1119  pcset[pcsetfill++] = pc;
1120  }
1121  pcs[i] = pc;
1122 
1123  i++;
1124  }
1125 
1126  assert(i > 0);
1127 
1128  /* sort on the pclass values */
1129  if( i > 1 )
1130  {
1131  SCIPsortIntIntReal(pcs, colindices, values, i);
1132  }
1133 
1134  k = 0;
1135  while( TRUE ) /*lint !e716*/
1136  {
1137  assert(k < i);
1138  startpc = pcs[k];
1139  startk = k;
1140 
1141  /* find pclass-sets */
1142  while( k < i && pcs[k] == startpc )
1143  k++;
1144 
1145  /* sort on the A values which have equal pclass values */
1146  if( k - startk > 1 )
1147  SCIPsortRealInt(&(values[startk]), &(colindices[startk]), k - startk);
1148 
1149  t = 0;
1150  while( TRUE ) /*lint !e716*/
1151  {
1152  assert(startk + t < i);
1153  startval = values[startk + t];
1154  startt = t;
1155 
1156  /* find A-sets */
1157  while( t < k - startk && SCIPisEQ(scip, startval, values[startk + t]) )
1158  t++;
1159 
1160  /* get new pclass */
1161  newpclass = pcset[0];
1162  assert(pcsetfill > 0);
1163  pcset[0] = pcset[--pcsetfill];
1164 
1165  /* renumbering */
1166  for( m = startk + startt; m < startk + t; m++ )
1167  {
1168  assert(m < i);
1169  assert(colindices[m] < ncols);
1170  assert(newpclass < ncols);
1171 
1172  pclass[colindices[m]] = newpclass;
1173  classsizes[newpclass]++;
1174  }
1175 
1176  if( t == k - startk )
1177  break;
1178  }
1179 
1180  if( k == SCIPmatrixGetRowNNonzs(matrix, r) )
1181  break;
1182  }
1183  }
1184  }
1185 
1186  SCIPfreeBufferArray(scip, &pcs);
1187  SCIPfreeBufferArray(scip, &colindices);
1188  SCIPfreeBufferArray(scip, &values);
1189  SCIPfreeBufferArray(scip, &pcset);
1190  SCIPfreeBufferArray(scip, &scale);
1191  SCIPfreeBufferArray(scip, &classsizes);
1192 
1193  return SCIP_OKAY;
1194 }
1195 
1196 
1197 /** try to improve variable bounds by predictive bound strengthening */
1198 static
1200  SCIP* scip, /**< SCIP main data structure */
1201  SCIP_VAR* dominatingvar, /**< dominating variable */
1202  int dominatingidx, /**< column index of the dominating variable */
1203  SCIP_Real dominatingub, /**< predicted upper bound of the dominating variable */
1204  SCIP_Real dominatinglb, /**< predicted lower bound of the dominating variable */
1205  SCIP_Real dominatingwcub, /**< predicted worst case upper bound of the dominating variable */
1206  SCIP_VAR* dominatedvar, /**< dominated variable */
1207  int dominatedidx, /**< column index of the dominated variable */
1208  SCIP_Real dominatedub, /**< predicted upper bound of the dominated variable */
1209  SCIP_Real dominatedwclb, /**< predicted worst case lower bound of the dominated variable */
1210  SCIP_Real dominatedlb, /**< predicted lower bound of the dominated variable */
1211  FIXINGDIRECTION* varstofix, /**< array holding fixing information */
1212  int* nchgbds /**< count number of bound changes */
1213  )
1214 {
1215  /* we compare only variables from the same type */
1216  if( !(SCIPvarGetType(dominatingvar) == SCIPvarGetType(dominatedvar) ||
1217  SCIPvarIsBinary(dominatingvar) == SCIPvarIsBinary(dominatedvar) ||
1218  (SCIPvarGetType(dominatingvar) == SCIP_VARTYPE_INTEGER && SCIPvarGetType(dominatedvar) == SCIP_VARTYPE_IMPLINT) ||
1219  (SCIPvarGetType(dominatedvar) == SCIP_VARTYPE_INTEGER && SCIPvarGetType(dominatingvar) == SCIP_VARTYPE_IMPLINT)) )
1220  {
1221  return SCIP_OKAY;
1222  }
1223 
1224  if( varstofix[dominatingidx] == NOFIX )
1225  {
1226  /* assume x dominates y (x->y). we get this bound from a positive coefficient
1227  * of the dominating variable. because x->y the dominated variable y has
1228  * a positive coefficient too. thus y contributes to the minactivity with its
1229  * lower bound. but this case is considered within predictive bound analysis.
1230  * thus the dominating upper bound is a common upper bound.
1231  */
1232  if( !SCIPisInfinity(scip, dominatingub) )
1233  {
1234  SCIP_Real newub;
1235  SCIP_Real oldub;
1236  SCIP_Real lb;
1237 
1238  newub = dominatingub;
1239  oldub = SCIPvarGetUbGlobal(dominatingvar);
1240  lb = SCIPvarGetLbGlobal(dominatingvar);
1241 
1242  /* if( SCIPvarGetType(dominatingvar) != SCIP_VARTYPE_CONTINUOUS )
1243  newub = SCIPceil(scip, newub); */
1244 
1245  if( SCIPisLE(scip, lb, newub) && SCIPisLT(scip, newub, oldub) )
1246  {
1247  SCIPdebugMsg(scip, "[ub]\tupper bound for dominating variable <%s> changed: [%.17f,%.17f] -> [%.17f,%.17f]\n",
1248  SCIPvarGetName(dominatingvar), lb, oldub, lb, newub);
1249  SCIP_CALL( SCIPchgVarUb(scip, dominatingvar, newub) );
1250  (*nchgbds)++;
1251  }
1252  }
1253 
1254  /* assume x dominates y (x->y). we get this lower bound of the dominating variable
1255  * from a negative coefficient within a <= relation. if y has a positive coefficient
1256  * we get a common lower bound of x from predictive bound analysis. if y has a
1257  * negative coefficient we get an improved lower bound of x because the minactivity
1258  * is greater. for discrete variables we have to round down the lower bound.
1259  */
1260  if( !SCIPisInfinity(scip, -dominatinglb) )
1261  {
1262  SCIP_Real newlb;
1263  SCIP_Real oldlb;
1264  SCIP_Real ub;
1265 
1266  newlb = dominatinglb;
1267  oldlb = SCIPvarGetLbGlobal(dominatingvar);
1268  ub = SCIPvarGetUbGlobal(dominatingvar);
1269 
1270  if( SCIPvarGetType(dominatingvar) != SCIP_VARTYPE_CONTINUOUS )
1271  newlb = SCIPfloor(scip, newlb);
1272 
1273  if( SCIPisLT(scip, oldlb, newlb) && SCIPisLE(scip, newlb, ub) )
1274  {
1275  SCIPdebugMsg(scip, "[lb]\tlower bound of dominating variable <%s> changed: [%.17f,%.17f] -> [%.17f,%.17f]\n",
1276  SCIPvarGetName(dominatingvar), oldlb, ub, newlb, ub);
1277  SCIP_CALL( SCIPchgVarLb(scip, dominatingvar, newlb) );
1278  (*nchgbds)++;
1279  }
1280  }
1281 
1282  /* assume x dominates y (x->y). we get this bound from a positive coefficient
1283  * of x within a <= relation. from x->y it follows, that y has a positive
1284  * coefficient in this row too. the worst case upper bound of x is an estimation
1285  * of how great x can be in every case. if the objective coefficient of x is
1286  * negative we get thus a lower bound of x. for discrete variables we have
1287  * to round down the lower bound before.
1288  */
1289  if( !SCIPisInfinity(scip, dominatingwcub) && SCIPisNegative(scip, SCIPvarGetObj(dominatingvar)))
1290  {
1291  SCIP_Real newlb;
1292  SCIP_Real oldlb;
1293  SCIP_Real ub;
1294 
1295  newlb = dominatingwcub;
1296  oldlb = SCIPvarGetLbGlobal(dominatingvar);
1297  ub = SCIPvarGetUbGlobal(dominatingvar);
1298 
1299  if( SCIPvarGetType(dominatingvar) != SCIP_VARTYPE_CONTINUOUS )
1300  newlb = SCIPfloor(scip, newlb);
1301 
1302  if( SCIPisLT(scip, oldlb, newlb) && SCIPisLE(scip, newlb, ub) )
1303  {
1304  SCIPdebugMsg(scip, "[wcub]\tlower bound of dominating variable <%s> changed: [%.17f,%.17f] -> [%.17f,%.17f]\n",
1305  SCIPvarGetName(dominatingvar), oldlb, ub, newlb, ub);
1306  SCIP_CALL( SCIPchgVarLb(scip, dominatingvar, newlb) );
1307  (*nchgbds)++;
1308  }
1309  }
1310  }
1311 
1312  if( varstofix[dominatedidx] == NOFIX )
1313  {
1314  /* assume x dominates y (x->y). we get this bound for a positive coefficient of y
1315  * within a <= relation. if x has a negative coefficient we get a common upper
1316  * bound of y. if x has a positive coefficient we get an improved upper bound
1317  * of y because the minactivity is greater.
1318  */
1319  if( !SCIPisInfinity(scip, dominatedub) )
1320  {
1321  SCIP_Real newub;
1322  SCIP_Real oldub;
1323  SCIP_Real lb;
1324 
1325  newub = dominatedub;
1326  oldub = SCIPvarGetUbGlobal(dominatedvar);
1327  lb = SCIPvarGetLbGlobal(dominatedvar);
1328 
1329  if( SCIPisLE(scip, lb, newub) && SCIPisLT(scip, newub, oldub) )
1330  {
1331  SCIPdebugMsg(scip, "[ub]\tupper bound of dominated variable <%s> changed: [%.17f,%.17f] -> [%.17f,%.17f]\n",
1332  SCIPvarGetName(dominatedvar), lb, oldub, lb, newub);
1333  SCIP_CALL( SCIPchgVarUb(scip, dominatedvar, newub) );
1334  (*nchgbds)++;
1335  }
1336  }
1337 
1338  /* assume x dominates y (x->y). we get this bound only from a negative
1339  * coefficient of y within a <= relation. because of x->y then x has a negative
1340  * coefficient too. the worst case lower bound is an estimation what property
1341  * the dominated variable must have if the dominating variable is at its upper bound.
1342  * to get an upper bound of the dominated variable we need to consider a positive
1343  * objective coefficient. for discrete variables we have to round up the upper bound.
1344  */
1345  if( !SCIPisInfinity(scip, -dominatedwclb) && SCIPisPositive(scip, SCIPvarGetObj(dominatedvar)) )
1346  {
1347  SCIP_Real newub;
1348  SCIP_Real oldub;
1349  SCIP_Real lb;
1350 
1351  newub = dominatedwclb;
1352  oldub = SCIPvarGetUbGlobal(dominatedvar);
1353  lb = SCIPvarGetLbGlobal(dominatedvar);
1354 
1355  if( SCIPvarGetType(dominatedvar) != SCIP_VARTYPE_CONTINUOUS )
1356  newub = SCIPceil(scip, newub);
1357 
1358  if( SCIPisLE(scip, lb, newub) && SCIPisLT(scip, newub, oldub) )
1359  {
1360  SCIPdebugMsg(scip, "[wclb]\tupper bound of dominated variable <%s> changed: [%.17f,%.17f] -> [%.17f,%.17f]\n",
1361  SCIPvarGetName(dominatedvar), lb, oldub, lb, newub);
1362  SCIP_CALL( SCIPchgVarUb(scip, dominatedvar, newub) );
1363  (*nchgbds)++;
1364  }
1365  }
1366 
1367  /* assume x dominates y (x->y). we get a lower bound of y from a negative
1368  * coefficient within a <= relation. but if x->y then x has a negative
1369  * coefficient too and contributes with its upper bound to the minactivity.
1370  * thus in all we have a common lower bound calculation and no rounding is necessary.
1371  */
1372  if( !SCIPisInfinity(scip, -dominatedlb) )
1373  {
1374  SCIP_Real newlb;
1375  SCIP_Real oldlb;
1376  SCIP_Real ub;
1377 
1378  newlb = dominatedlb;
1379  oldlb = SCIPvarGetLbGlobal(dominatedvar);
1380  ub = SCIPvarGetUbGlobal(dominatedvar);
1381 
1382  if( SCIPisLT(scip, oldlb, newlb) && SCIPisLE(scip, newlb, ub) )
1383  {
1384  SCIPdebugMsg(scip, "[lb]\tlower bound of dominated variable <%s> changed: [%.17f,%.17f] -> [%.17f,%.17f]\n",
1385  SCIPvarGetName(dominatedvar), oldlb, ub, newlb, ub);
1386  SCIP_CALL( SCIPchgVarLb(scip, dominatedvar, newlb) );
1387  (*nchgbds)++;
1388  }
1389  }
1390  }
1391 
1392  return SCIP_OKAY;
1393 }
1394 
1395 /** try to find variable fixings */
1396 static
1398  SCIP* scip, /**< SCIP main data structure */
1399  SCIP_MATRIX* matrix, /**< constraint matrix structure */
1400  SCIP_VAR* dominatingvar, /**< dominating variable */
1401  int dominatingidx, /**< column index of the dominating variable */
1402  SCIP_Real dominatingub, /**< predicted upper bound of the dominating variable */
1403  SCIP_Real dominatingwclb, /**< predicted worst case lower bound of the dominating variable */
1404  SCIP_Real dominatinglb, /**< predicted lower bound of the dominating variable */
1405  SCIP_Real dominatingwcub, /**< predicted worst case upper bound of the dominating variable */
1406  SCIP_VAR* dominatedvar, /**< dominated variable */
1407  int dominatedidx, /**< column index of the dominated variable */
1408  FIXINGDIRECTION* varstofix, /**< array holding fixing information */
1409  SCIP_Bool onlybinvars, /**< flag indicating only binary variables are present */
1410  SCIP_Bool onlyoneone, /**< when onlybinvars is TRUE, flag indicates if both binary variables are in clique */
1411  int* nfixings /**< counter for possible fixings */
1412  )
1413 {
1414  /* we compare only variables from the same type */
1415  if( !(SCIPvarGetType(dominatingvar) == SCIPvarGetType(dominatedvar) ||
1416  SCIPvarIsBinary(dominatingvar) == SCIPvarIsBinary(dominatedvar) ||
1417  (SCIPvarGetType(dominatingvar) == SCIP_VARTYPE_INTEGER && SCIPvarGetType(dominatedvar) == SCIP_VARTYPE_IMPLINT) ||
1418  (SCIPvarGetType(dominatedvar) == SCIP_VARTYPE_INTEGER && SCIPvarGetType(dominatingvar) == SCIP_VARTYPE_IMPLINT)) )
1419  {
1420  return SCIP_OKAY;
1421  }
1422 
1423  if( varstofix[dominatedidx] == NOFIX && SCIPmatrixGetColNNonzs(matrix, dominatingidx) == 1
1424  && SCIPmatrixGetColNNonzs(matrix, dominatedidx) == 1 )
1425  {
1426  /* We have a x->y dominance relation and only one equality constraint
1427  * where the dominating variable x with an infinity upper bound and the
1428  * dominated variable y are present. Then the dominated variable y
1429  * can be fixed at its lower bound.
1430  */
1431  int row;
1432  row = *(SCIPmatrixGetColIdxPtr(matrix, dominatedidx));
1433 
1434  if( SCIPisEQ(scip, SCIPmatrixGetRowLhs(matrix, row), SCIPmatrixGetRowRhs(matrix, row)) &&
1435  SCIPisInfinity(scip, SCIPvarGetUbGlobal(dominatingvar)) )
1436  {
1437  varstofix[dominatedidx] = FIXATLB;
1438  (*nfixings)++;
1439 
1440  return SCIP_OKAY;
1441  }
1442  }
1443 
1444  if( varstofix[dominatedidx] == NOFIX && !SCIPisNegative(scip, SCIPvarGetObj(dominatedvar)) )
1445  {
1446  if( !SCIPisInfinity(scip, -dominatingwclb) &&
1447  SCIPisLE(scip, dominatingwclb, SCIPvarGetUbGlobal(dominatingvar)) )
1448  {
1449  /* we have a x->y dominance relation with a positive obj coefficient
1450  * of the dominated variable y. we need to secure feasibility
1451  * by testing if the predicted lower worst case bound is less equal the
1452  * current upper bound. it is possible, that the lower worst case bound
1453  * is infinity and the upper bound of the dominating variable x is
1454  * infinity too.
1455  */
1456  varstofix[dominatedidx] = FIXATLB;
1457  (*nfixings)++;
1458  }
1459  }
1460 
1461  if( varstofix[dominatedidx] == NOFIX && !SCIPisInfinity(scip, dominatingub) &&
1462  SCIPisLE(scip, dominatingub, SCIPvarGetUbGlobal(dominatingvar)) )
1463  {
1464  /* we have a x->y dominance relation with an arbitrary obj coefficient
1465  * of the dominating variable x. in all cases we have to look
1466  * if the predicted upper bound of the dominating variable is great enough.
1467  * by testing, that the predicted upper bound is not infinity we avoid problems
1468  * with x->y e.g.
1469  * min -x -y
1470  * s.t. -x -y <= -1
1471  * 0<=x<=1, 0<=y<=1
1472  * where y is not at their lower bound.
1473  */
1474  varstofix[dominatedidx] = FIXATLB;
1475  (*nfixings)++;
1476  }
1477 
1478  if( varstofix[dominatingidx] == NOFIX && !SCIPisPositive(scip, SCIPvarGetObj(dominatingvar)) )
1479  {
1480  /* we have a x->y dominance relation with a negative obj coefficient
1481  * of the dominating variable x. if the worst case upper bound is
1482  * greater equal than upper bound, we fix x at the upper bound
1483  */
1484  if( !SCIPisInfinity(scip, dominatingwcub) &&
1485  SCIPisGE(scip, dominatingwcub, SCIPvarGetUbGlobal(dominatingvar)) )
1486  {
1487  varstofix[dominatingidx] = FIXATUB;
1488  (*nfixings)++;
1489  }
1490  }
1491 
1492  if( varstofix[dominatingidx] == NOFIX && !SCIPisInfinity(scip, -dominatinglb) &&
1493  SCIPisGE(scip, dominatinglb, SCIPvarGetUbGlobal(dominatingvar)) )
1494  {
1495  /* we have a x->y dominance relation with an arbitrary obj coefficient
1496  * of the dominating variable x. if the predicted lower bound is greater
1497  * equal than upper bound, we fix x at the upper bound.
1498  */
1499  varstofix[dominatingidx] = FIXATUB;
1500  (*nfixings)++;
1501  }
1502 
1503  if( onlybinvars )
1504  {
1505  if( varstofix[dominatedidx] == NOFIX && (onlyoneone || SCIPvarsHaveCommonClique(dominatingvar, TRUE, dominatedvar, TRUE, TRUE)) )
1506  {
1507  /* We have a (1->1)-clique with dominance relation (x->y) (x dominates y).
1508  * From this dominance relation, we know (1->0) is possible and not worse than (0->1)
1509  * concerning the objective function. It follows that only (1->0) or (0->0) are possible,
1510  * but in both cases y has the value 0 => y=0.
1511  */
1512  varstofix[dominatedidx] = FIXATLB;
1513  (*nfixings)++;
1514  }
1515 
1516  if( varstofix[dominatingidx] == NOFIX && SCIPvarsHaveCommonClique(dominatingvar, FALSE, dominatedvar, FALSE, TRUE) )
1517  {
1518  /* We have a (0->0)-clique with dominance relation x->y (x dominates y).
1519  * From this dominance relation, we know (1->0) is possible and not worse than (0->1)
1520  * concerning the objective function. It follows that only (1->0) or (1->1) are possible,
1521  * but in both cases x has the value 1 => x=1
1522  */
1523  varstofix[dominatingidx] = FIXATUB;
1524  (*nfixings)++;
1525  }
1526  }
1527  else
1528  assert(!onlyoneone);
1529 
1530  return SCIP_OKAY;
1531 }
1532 
1533 /** find dominance relation between variable pairs */
1534 static
1536  SCIP* scip, /**< SCIP main data structure */
1537  SCIP_MATRIX* matrix, /**< matrix containing the constraints */
1538  SCIP_PRESOLDATA* presoldata, /**< presolver data */
1539  int* searchcols, /**< indexes of variables for pair comparisons */
1540  int searchsize, /**< number of variables for pair comparisons */
1541  SCIP_Bool onlybinvars, /**< flag indicating searchcols contains only binary variable indexes */
1542  FIXINGDIRECTION* varstofix, /**< array holding information for later upper/lower bound fixing */
1543  int* nfixings, /**< found number of possible fixings */
1544  SCIP_Longint* ndomrelations, /**< found number of dominance relations */
1545  int* nchgbds /**< number of changed bounds */
1546  )
1547 {
1548  SCIP_Real* vals1;
1549  SCIP_Real* vals2;
1550  SCIP_Real tmpupperbounddominatingcol1;
1551  SCIP_Real tmpupperbounddominatingcol2;
1552  SCIP_Real tmpwclowerbounddominatingcol1;
1553  SCIP_Real tmpwclowerbounddominatingcol2;
1554  SCIP_Real tmplowerbounddominatingcol1;
1555  SCIP_Real tmplowerbounddominatingcol2;
1556  SCIP_Real tmpwcupperbounddominatingcol1;
1557  SCIP_Real tmpwcupperbounddominatingcol2;
1558  int* rows1;
1559  int* rows2;
1560  int nrows1;
1561  int nrows2;
1562  SCIP_Real tmpupperbounddominatedcol1;
1563  SCIP_Real tmpupperbounddominatedcol2;
1564  SCIP_Real tmpwclowerbounddominatedcol1;
1565  SCIP_Real tmpwclowerbounddominatedcol2;
1566  SCIP_Real tmplowerbounddominatedcol1;
1567  SCIP_Real tmplowerbounddominatedcol2;
1568  SCIP_Real tmpwcupperbounddominatedcol1;
1569  SCIP_Real tmpwcupperbounddominatedcol2;
1570  SCIP_Real obj1;
1571  SCIP_Bool col1domcol2;
1572  SCIP_Bool col2domcol1;
1573  SCIP_Bool onlyoneone;
1574  int cnt1;
1575  int cnt2;
1576  int col1;
1577  int col2;
1578  int r1;
1579  int r2;
1580  int paircnt;
1581  int oldnfixings;
1582 
1583  assert(scip != NULL);
1584  assert(matrix != NULL);
1585  assert(presoldata != NULL);
1586  assert(searchcols != NULL);
1587  assert(varstofix != NULL);
1588  assert(nfixings != NULL);
1589  assert(ndomrelations != NULL);
1590  assert(nchgbds != NULL);
1591 
1592  paircnt = 0;
1593  oldnfixings = *nfixings;
1594 
1595  /* pair comparisons */
1596  for( cnt1 = 0; cnt1 < searchsize; cnt1++ )
1597  {
1598  SCIP_VAR* varcol1;
1599  SCIP_VAR* varcol2;
1600 
1601  /* get index of the first variable */
1602  col1 = searchcols[cnt1];
1603 
1604  if( varstofix[col1] == FIXATLB )
1605  continue;
1606 
1607  varcol1 = SCIPmatrixGetVar(matrix, col1);
1608  obj1 = SCIPvarGetObj(varcol1);
1609 
1610  for( cnt2 = cnt1+1; cnt2 < searchsize; cnt2++ )
1611  {
1612  /* get index of the second variable */
1613  col2 = searchcols[cnt2];
1614  varcol2 = SCIPmatrixGetVar(matrix, col2);
1615  onlyoneone = FALSE;
1616 
1617  /* we always have minimize as obj sense */
1618 
1619  /* column 1 dominating column 2 */
1620  col1domcol2 = (obj1 <= SCIPvarGetObj(varcol2));
1621 
1622  /* column 2 dominating column 1 */
1623  col2domcol1 = (SCIPvarGetObj(varcol2) <= obj1);
1624 
1625  /* search only if nothing was found yet */
1626  col1domcol2 = col1domcol2 && (varstofix[col2] == NOFIX);
1627  col2domcol1 = col2domcol1 && (varstofix[col1] == NOFIX);
1628 
1629  /* we only search for a dominance relation if the lower bounds are not negative */
1630  if( !onlybinvars )
1631  {
1632  if( SCIPisLT(scip, SCIPvarGetLbGlobal(varcol1), 0.0) ||
1633  SCIPisLT(scip, SCIPvarGetLbGlobal(varcol2), 0.0) )
1634  {
1635  col1domcol2 = FALSE;
1636  col2domcol1 = FALSE;
1637  }
1638  }
1639 
1640  /* pair comparison control */
1641  if( paircnt == presoldata->numcurrentpairs )
1642  {
1643  assert(*nfixings >= oldnfixings);
1644  if( *nfixings == oldnfixings )
1645  {
1646  /* not enough fixings found, decrement number of comparisons */
1647  presoldata->numcurrentpairs >>= 1; /*lint !e702*/
1648  if( presoldata->numcurrentpairs < presoldata->numminpairs )
1649  presoldata->numcurrentpairs = presoldata->numminpairs;
1650 
1651  /* stop searching in this row */
1652  return SCIP_OKAY;
1653  }
1654  oldnfixings = *nfixings;
1655  paircnt = 0;
1656 
1657  /* increment number of comparisons */
1658  presoldata->numcurrentpairs <<= 1; /*lint !e701*/
1659  if( presoldata->numcurrentpairs > presoldata->nummaxpairs )
1660  presoldata->numcurrentpairs = presoldata->nummaxpairs;
1661  }
1662  paircnt++;
1663 
1664  if( !col1domcol2 && !col2domcol1 )
1665  continue;
1666 
1667  /* get the data for both columns */
1668  vals1 = SCIPmatrixGetColValPtr(matrix, col1);
1669  rows1 = SCIPmatrixGetColIdxPtr(matrix, col1);
1670  nrows1 = SCIPmatrixGetColNNonzs(matrix, col1);
1671  r1 = 0;
1672  vals2 = SCIPmatrixGetColValPtr(matrix, col2);
1673  rows2 = SCIPmatrixGetColIdxPtr(matrix, col2);
1674  nrows2 = SCIPmatrixGetColNNonzs(matrix, col2);
1675  r2 = 0;
1676 
1677  /* do we have a obj constant? */
1678  if( nrows1 == 0 || nrows2 == 0 )
1679  continue;
1680 
1681  /* initialize temporary bounds of dominating variable */
1682  tmpupperbounddominatingcol1 = SCIPinfinity(scip);
1683  tmpupperbounddominatingcol2 = tmpupperbounddominatingcol1;
1684  tmpwclowerbounddominatingcol1 = -SCIPinfinity(scip);
1685  tmpwclowerbounddominatingcol2 = tmpwclowerbounddominatingcol1;
1686  tmplowerbounddominatingcol1 = -SCIPinfinity(scip);
1687  tmplowerbounddominatingcol2 = tmplowerbounddominatingcol1;
1688  tmpwcupperbounddominatingcol1 = SCIPinfinity(scip);
1689  tmpwcupperbounddominatingcol2 = tmpwcupperbounddominatingcol1;
1690 
1691  /* initialize temporary bounds of dominated variable */
1692  tmpupperbounddominatedcol1 = SCIPinfinity(scip);
1693  tmpupperbounddominatedcol2 = tmpupperbounddominatedcol1;
1694  tmpwclowerbounddominatedcol1 = -SCIPinfinity(scip);
1695  tmpwclowerbounddominatedcol2 = tmpwclowerbounddominatedcol1;
1696  tmplowerbounddominatedcol1 = -SCIPinfinity(scip);
1697  tmplowerbounddominatedcol2 = tmplowerbounddominatedcol1;
1698  tmpwcupperbounddominatedcol1 = SCIPinfinity(scip);
1699  tmpwcupperbounddominatedcol2 = tmpwcupperbounddominatedcol1;
1700 
1701  /* compare rows of this column pair */
1702  while( (col1domcol2 || col2domcol1) && (r1 < nrows1 || r2 < nrows2) )
1703  {
1704  assert((r1 >= nrows1-1) || (rows1[r1] < rows1[r1+1]));
1705  assert((r2 >= nrows2-1) || (rows2[r2] < rows2[r2+1]));
1706 
1707  /* there is a nonredundant row containing column 1 but not column 2 */
1708  if( r1 < nrows1 && (r2 == nrows2 || rows1[r1] < rows2[r2]) )
1709  {
1710  /* dominance depends on the type of relation */
1711  if( !SCIPmatrixIsRowRhsInfinity(matrix, rows1[r1]) )
1712  {
1713  /* no dominance relation for equations or ranged rows */
1714  col2domcol1 = FALSE;
1715  col1domcol2 = FALSE;
1716  }
1717  else
1718  {
1719  /* >= relation, larger coefficients dominate smaller ones */
1720  if( vals1[r1] > 0.0 )
1721  col2domcol1 = FALSE;
1722  else
1723  col1domcol2 = FALSE;
1724  }
1725 
1726  r1++;
1727  }
1728  /* there is a nonredundant row containing column 2, but not column 1 */
1729  else if( r2 < nrows2 && (r1 == nrows1 || rows1[r1] > rows2[r2]) )
1730  {
1731  /* dominance depends on the type of relation */
1732  if( !SCIPmatrixIsRowRhsInfinity(matrix, rows2[r2]) )
1733  {
1734  /* no dominance relation for equations or ranged rows */
1735  col2domcol1 = FALSE;
1736  col1domcol2 = FALSE;
1737  }
1738  else
1739  {
1740  /* >= relation, larger coefficients dominate smaller ones */
1741  if( vals2[r2] < 0.0 )
1742  col2domcol1 = FALSE;
1743  else
1744  col1domcol2 = FALSE;
1745  }
1746 
1747  r2++;
1748  }
1749  /* if both columns appear in a common row, compare the coefficients */
1750  else
1751  {
1752  assert(r1 < nrows1 && r2 < nrows2);
1753  assert(rows1[r1] == rows2[r2]);
1754 
1755  /* if both columns are binary variables we check if they have a common clique
1756  * and do not calculate any bounds
1757  */
1758  if( onlybinvars && !onlyoneone )
1759  {
1760  if( vals1[r1] < 0 && vals2[r2] < 0 )
1761  {
1762  if( (SCIPmatrixGetRowNMaxActPosInf(matrix, rows1[r1]) + SCIPmatrixGetRowNMaxActNegInf(matrix, rows1[r1]) == 0)
1763  && SCIPisFeasLE(scip, SCIPmatrixGetRowMaxActivity(matrix, rows1[r1]) + MAX(vals1[r1], vals2[r2]),
1764  SCIPmatrixGetRowLhs(matrix, rows1[r1])) )
1765  {
1766  onlyoneone = TRUE;
1767  }
1768  }
1769 
1770  if( !onlyoneone && !SCIPmatrixIsRowRhsInfinity(matrix, rows1[r1]) )
1771  {
1772  if ( vals1[r1] > 0 && vals2[r2] > 0 )
1773  {
1774  if( (SCIPmatrixGetRowNMinActPosInf(matrix, rows1[r1]) + SCIPmatrixGetRowNMinActNegInf(matrix, rows1[r1]) == 0)
1775  && SCIPisFeasGE(scip, SCIPmatrixGetRowMinActivity(matrix, rows1[r1]) + MIN(vals1[r1], vals2[r2]),
1776  SCIPmatrixGetRowRhs(matrix, rows1[r1])) )
1777  {
1778  onlyoneone = TRUE;
1779  }
1780  }
1781  }
1782 
1783  if( onlyoneone )
1784  {
1785  /* reset bounds */
1786  tmpupperbounddominatingcol1 = SCIPinfinity(scip);
1787  tmpupperbounddominatingcol2 = tmpupperbounddominatingcol1;
1788  tmpwclowerbounddominatingcol1 = -SCIPinfinity(scip);
1789  tmpwclowerbounddominatingcol2 = tmpwclowerbounddominatingcol1;
1790  tmplowerbounddominatingcol1 = -SCIPinfinity(scip);
1791  tmplowerbounddominatingcol2 = tmplowerbounddominatingcol1;
1792  tmpwcupperbounddominatingcol1 = SCIPinfinity(scip);
1793  tmpwcupperbounddominatingcol2 = tmpwcupperbounddominatingcol1;
1794 
1795  tmpupperbounddominatedcol1 = SCIPinfinity(scip);
1796  tmpupperbounddominatedcol2 = tmpupperbounddominatedcol1;
1797  tmpwclowerbounddominatedcol1 = -SCIPinfinity(scip);
1798  tmpwclowerbounddominatedcol2 = tmpwclowerbounddominatedcol1;
1799  tmplowerbounddominatedcol1 = -SCIPinfinity(scip);
1800  tmplowerbounddominatedcol2 = tmplowerbounddominatedcol1;
1801  tmpwcupperbounddominatedcol1 = SCIPinfinity(scip);
1802  tmpwcupperbounddominatedcol2 = tmpwcupperbounddominatedcol1;
1803  }
1804  }
1805 
1806  /* dominance depends on the type of inequality */
1807  if( !SCIPmatrixIsRowRhsInfinity(matrix, rows1[r1]) )
1808  {
1809  /* no dominance relation if coefficients differ for equations or ranged rows */
1810  if( !SCIPisEQ(scip, vals1[r1], vals2[r2]) )
1811  {
1812  col2domcol1 = FALSE;
1813  col1domcol2 = FALSE;
1814  }
1815  }
1816  else
1817  {
1818  /* >= relation, larger coefficients dominate smaller ones */
1819  if( vals1[r1] > vals2[r2] )
1820  col2domcol1 = FALSE;
1821  else if( vals1[r1] < vals2[r2] )
1822  col1domcol2 = FALSE;
1823  }
1824 
1825  /* we do not use bound calulations if two binary variable are in one common clique.
1826  * for the other cases we claim the same sign for the coefficients to
1827  * achieve monotonically decreasing predictive bound functions.
1828  */
1829  if( !onlyoneone &&
1830  ((vals1[r1] < 0 && vals2[r2] < 0) || (vals1[r1] > 0 && vals2[r2] > 0)) )
1831  {
1832  if( col1domcol2 )
1833  {
1834  /* update bounds of dominating variable for column 1 */
1835  SCIP_CALL( updateBounds(scip, matrix, rows1[r1],
1836  col1, vals1[r1], col2, vals2[r2], TRUE,
1837  &tmpupperbounddominatingcol1, &tmpwclowerbounddominatingcol1,
1838  &tmplowerbounddominatingcol1, &tmpwcupperbounddominatingcol1) );
1839 
1840  /* update bounds of dominated variable for column 1 */
1841  SCIP_CALL( updateBounds(scip, matrix, rows1[r1],
1842  col1, vals1[r1], col2, vals2[r2], FALSE,
1843  &tmpupperbounddominatedcol1, &tmpwclowerbounddominatedcol1,
1844  &tmplowerbounddominatedcol1, &tmpwcupperbounddominatedcol1) );
1845  }
1846 
1847  if( col2domcol1 )
1848  {
1849  /* update bounds of dominating variable for column 2 */
1850  SCIP_CALL( updateBounds(scip, matrix, rows2[r2],
1851  col2, vals2[r2], col1, vals1[r1], TRUE,
1852  &tmpupperbounddominatingcol2, &tmpwclowerbounddominatingcol2,
1853  &tmplowerbounddominatingcol2, &tmpwcupperbounddominatingcol2) );
1854 
1855  /* update bounds of dominated variable for column 2 */
1856  SCIP_CALL( updateBounds(scip, matrix, rows2[r2],
1857  col2, vals2[r2], col1, vals1[r1], FALSE,
1858  &tmpupperbounddominatedcol2, &tmpwclowerbounddominatedcol2,
1859  &tmplowerbounddominatedcol2, &tmpwcupperbounddominatedcol2) );
1860  }
1861  }
1862 
1863  r1++;
1864  r2++;
1865  }
1866  }
1867 
1868  /* a column is only dominated, if there are no more rows in which it is contained */
1869  col1domcol2 = col1domcol2 && r2 == nrows2;
1870  col2domcol1 = col2domcol1 && r1 == nrows1;
1871 
1872  if( !col1domcol2 && !col2domcol1 )
1873  continue;
1874 
1875  /* no dominance relation for left equations or ranged rows */
1876  while( r1 < nrows1 )
1877  {
1878  if( !SCIPmatrixIsRowRhsInfinity(matrix, rows1[r1]) )
1879  {
1880  col2domcol1 = FALSE;
1881  col1domcol2 = FALSE;
1882  break;
1883  }
1884  r1++;
1885  }
1886  if( !col1domcol2 && !col2domcol1 )
1887  continue;
1888  while( r2 < nrows2 )
1889  {
1890  if( !SCIPmatrixIsRowRhsInfinity(matrix, rows2[r2]) )
1891  {
1892  col2domcol1 = FALSE;
1893  col1domcol2 = FALSE;
1894  break;
1895  }
1896  r2++;
1897  }
1898 
1899  if( col1domcol2 || col2domcol1 )
1900  (*ndomrelations)++;
1901 
1902  if( col1domcol2 && col2domcol1 )
1903  {
1904  /* prefer the variable as dominating variable with the greater upper bound */
1905  if( SCIPisGE(scip, SCIPvarGetUbGlobal(varcol1), SCIPvarGetUbGlobal(varcol2)) )
1906  col2domcol1 = FALSE;
1907  else
1908  col1domcol2 = FALSE;
1909  }
1910 
1911  /* use dominance relation and clique/bound-information
1912  * to find variable fixings. additionally try to strengthen
1913  * variable bounds by predictive bound strengthening.
1914  */
1915  if( col1domcol2 )
1916  {
1917  SCIP_CALL( findFixings(scip, matrix, varcol1, col1,
1918  tmpupperbounddominatingcol1, tmpwclowerbounddominatingcol1,
1919  tmplowerbounddominatingcol1, tmpwcupperbounddominatingcol1,
1920  varcol2, col2,
1921  varstofix, onlybinvars, onlyoneone, nfixings) );
1922 
1923  if( presoldata->predbndstr )
1924  {
1925  SCIP_CALL( predBndStr(scip, varcol1, col1,
1926  tmpupperbounddominatingcol1,
1927  tmplowerbounddominatingcol1, tmpwcupperbounddominatingcol1,
1928  varcol2, col2,
1929  tmpupperbounddominatedcol1, tmpwclowerbounddominatedcol1,
1930  tmplowerbounddominatedcol1,
1931  varstofix, nchgbds) );
1932  }
1933  }
1934  else if( col2domcol1 )
1935  {
1936  SCIP_CALL( findFixings(scip, matrix, varcol2, col2,
1937  tmpupperbounddominatingcol2, tmpwclowerbounddominatingcol2,
1938  tmplowerbounddominatingcol2, tmpwcupperbounddominatingcol2,
1939  varcol1, col1,
1940  varstofix, onlybinvars, onlyoneone, nfixings) );
1941 
1942  if( presoldata->predbndstr )
1943  {
1944  SCIP_CALL( predBndStr(scip, varcol2, col2,
1945  tmpupperbounddominatingcol2,
1946  tmplowerbounddominatingcol2, tmpwcupperbounddominatingcol2,
1947  varcol1, col1,
1948  tmpupperbounddominatedcol2, tmpwclowerbounddominatedcol2,
1949  tmplowerbounddominatedcol2,
1950  varstofix, nchgbds) );
1951  }
1952  }
1953  if( varstofix[col1] == FIXATLB )
1954  break;
1955  }
1956  }
1957 
1958  return SCIP_OKAY;
1959 }
1960 
1961 
1962 /*
1963  * Callback methods of presolver
1964  */
1965 
1966 
1967 /** copy method for constraint handler plugins (called when SCIP copies plugins) */
1968 static
1969 SCIP_DECL_PRESOLCOPY(presolCopyDomcol)
1970 { /*lint --e{715}*/
1971  assert(scip != NULL);
1972  assert(presol != NULL);
1973  assert(strcmp(SCIPpresolGetName(presol), PRESOL_NAME) == 0);
1974 
1975  /* call inclusion method of presolver */
1977 
1978  return SCIP_OKAY;
1979 }
1980 
1981 /** destructor of presolver to free user data (called when SCIP is exiting) */
1982 static
1983 SCIP_DECL_PRESOLFREE(presolFreeDomcol)
1984 { /*lint --e{715}*/
1985  SCIP_PRESOLDATA* presoldata;
1986 
1987  /* free presolver data */
1988  presoldata = SCIPpresolGetData(presol);
1989  assert(presoldata != NULL);
1990 
1991  SCIPfreeBlockMemory(scip, &presoldata);
1992  SCIPpresolSetData(presol, NULL);
1993 
1994  return SCIP_OKAY;
1995 }
1996 
1997 /** execution method of presolver */
1998 static
1999 SCIP_DECL_PRESOLEXEC(presolExecDomcol)
2000 { /*lint --e{715}*/
2001  SCIP_PRESOLDATA* presoldata;
2002  SCIP_MATRIX* matrix;
2003  SCIP_Bool initialized;
2004  SCIP_Bool complete;
2005  SCIP_Bool infeasible;
2006  int nfixings;
2007  SCIP_Longint ndomrelations;
2008  int v;
2009  int r;
2010  FIXINGDIRECTION* varstofix;
2011  SCIP_Bool* varsprocessed;
2012  int nrows;
2013  int ncols;
2014  int* rowidxsorted;
2015  int* rowsparsity;
2016  int varcount;
2017  int* consearchcols;
2018  int* intsearchcols;
2019  int* binsearchcols;
2020  int nconfill;
2021  int nintfill;
2022  int nbinfill;
2023 #ifdef SCIP_DEBUG
2024  int nconvarsfixed = 0;
2025  int nintvarsfixed = 0;
2026  int nbinvarsfixed = 0;
2027 #endif
2028  int* pclass;
2029  int* colidx;
2030  int pclassstart;
2031  int pc;
2032  SCIP_Bool* varineq;
2033 
2034  assert(result != NULL);
2035  *result = SCIP_DIDNOTRUN;
2036 
2037  if( !SCIPallowStrongDualReds(scip) || SCIPisStopped(scip) )
2038  return SCIP_OKAY;
2039 
2040  presoldata = SCIPpresolGetData(presol);
2041  assert(presoldata != NULL);
2042 
2043  /* don't run for pure LPs */
2044  if( !presoldata->continuousred && (SCIPgetNBinVars(scip) + SCIPgetNIntVars(scip) == 0) )
2045  return SCIP_OKAY;
2046 
2047  *result = SCIP_DIDNOTFIND;
2048 
2049  matrix = NULL;
2050  SCIP_CALL( SCIPmatrixCreate(scip, &matrix, TRUE, &initialized, &complete, &infeasible,
2051  naddconss, ndelconss, nchgcoefs, nchgbds, nfixedvars) );
2052 
2053  /* if infeasibility was detected during matrix creation, return here */
2054  if( infeasible )
2055  {
2056  if( initialized )
2057  SCIPmatrixFree(scip, &matrix);
2058 
2059  *result = SCIP_CUTOFF;
2060  return SCIP_OKAY;
2061  }
2062 
2063  if( !initialized )
2064  return SCIP_OKAY;
2065 
2066  nfixings = 0;
2067  ndomrelations = 0;
2068  nrows = SCIPmatrixGetNRows(matrix);
2069  ncols = SCIPmatrixGetNColumns(matrix);
2070  assert(SCIPgetNVars(scip) == ncols);
2071 
2072  SCIP_CALL( SCIPallocBufferArray(scip, &varstofix, ncols) );
2073  BMSclearMemoryArray(varstofix, ncols);
2074 
2075  SCIP_CALL( SCIPallocBufferArray(scip, &varsprocessed, ncols) );
2076 
2077  /* do not process columns that do not have all locks represented in the matrix */
2078  for( v = 0; v < ncols; ++v )
2079  varsprocessed[v] = SCIPmatrixUplockConflict(matrix, v) || SCIPmatrixDownlockConflict(matrix, v);
2080 
2081  SCIP_CALL( SCIPallocBufferArray(scip, &consearchcols, ncols) );
2082  SCIP_CALL( SCIPallocBufferArray(scip, &intsearchcols, ncols) );
2083  SCIP_CALL( SCIPallocBufferArray(scip, &binsearchcols, ncols) );
2084 
2085  SCIP_CALL( SCIPallocBufferArray(scip, &rowidxsorted, nrows) );
2086  SCIP_CALL( SCIPallocBufferArray(scip, &rowsparsity, nrows) );
2087  for( r = 0; r < nrows; ++r )
2088  {
2089  rowidxsorted[r] = r;
2090  rowsparsity[r] = SCIPmatrixGetRowNNonzs(matrix, r);
2091  }
2092 
2093  SCIP_CALL( SCIPallocBufferArray(scip, &pclass, ncols) );
2094  SCIP_CALL( SCIPallocBufferArray(scip, &colidx, ncols) );
2095  SCIP_CALL( SCIPallocBufferArray(scip, &varineq, ncols) );
2096  for( v = 0; v < ncols; v++ )
2097  {
2098  colidx[v] = v;
2099  varineq[v] = FALSE;
2100  }
2101 
2102  /* init pair comparision control */
2103  presoldata->numcurrentpairs = presoldata->nummaxpairs;
2104 
2105  varcount = 0;
2106 
2107  /* 1.stage: search dominance relations of parallel columns
2108  * within equalities and ranged rows
2109  */
2110  if( (presoltiming & SCIP_PRESOLTIMING_EXHAUSTIVE) != 0 )
2111  {
2112  SCIP_CALL( detectParallelCols(scip, matrix, pclass, varineq) );
2113  SCIPsortIntInt(pclass, colidx, ncols);
2114 
2115  pc = 0;
2116  while( pc < ncols )
2117  {
2118  int varidx;
2119 
2120  varidx = 0;
2121  nconfill = 0;
2122  nintfill = 0;
2123  nbinfill = 0;
2124 
2125  pclassstart = pclass[pc];
2126  while( pc < ncols && pclassstart == pclass[pc] )
2127  {
2128  SCIP_VAR* var;
2129 
2130  varidx = colidx[pc];
2131  var = SCIPmatrixGetVar(matrix, varidx);
2132 
2133  /* we only regard variables which were not processed yet and
2134  * are present within equalities or ranged rows
2135  */
2136  if( !varsprocessed[varidx] && varineq[varidx] )
2137  {
2138  /* we search only for dominance relations between the same variable type */
2140  {
2141  consearchcols[nconfill++] = varidx;
2142  }
2143  else if( SCIPvarIsBinary(var) )
2144  {
2145  binsearchcols[nbinfill++] = varidx;
2146  }
2147  else
2148  {
2150  intsearchcols[nintfill++] = varidx;
2151  }
2152  }
2153  ++pc;
2154  }
2155 
2156  /* continuous variables */
2157  if( nconfill > 1 && presoldata->continuousred )
2158  {
2159  SCIP_CALL( findDominancePairs(scip, matrix, presoldata, consearchcols, nconfill, FALSE,
2160  varstofix, &nfixings, &ndomrelations, nchgbds) );
2161 
2162  for( v = 0; v < nconfill; ++v )
2163  varsprocessed[consearchcols[v]] = TRUE;
2164 
2165  varcount += nconfill;
2166  }
2167  else if( nconfill == 1 )
2168  {
2169  if( varineq[varidx] )
2170  varsprocessed[consearchcols[0]] = TRUE;
2171  }
2172 
2173  /* integer and impl-integer variables */
2174  if( nintfill > 1 )
2175  {
2176  SCIP_CALL( findDominancePairs(scip, matrix, presoldata, intsearchcols, nintfill, FALSE,
2177  varstofix, &nfixings, &ndomrelations, nchgbds) );
2178 
2179  for( v = 0; v < nintfill; ++v )
2180  varsprocessed[intsearchcols[v]] = TRUE;
2181 
2182  varcount += nintfill;
2183  }
2184  else if( nintfill == 1 )
2185  {
2186  if( varineq[varidx] )
2187  varsprocessed[intsearchcols[0]] = TRUE;
2188  }
2189 
2190  /* binary variables */
2191  if( nbinfill > 1 )
2192  {
2193  SCIP_CALL( findDominancePairs(scip, matrix, presoldata, binsearchcols, nbinfill, TRUE,
2194  varstofix, &nfixings, &ndomrelations, nchgbds) );
2195 
2196  for( v = 0; v < nbinfill; ++v )
2197  varsprocessed[binsearchcols[v]] = TRUE;
2198 
2199  varcount += nbinfill;
2200  }
2201  else if( nbinfill == 1 )
2202  {
2203  if( varineq[varidx] )
2204  varsprocessed[binsearchcols[0]] = TRUE;
2205  }
2206 
2207  if( varcount >= ncols )
2208  break;
2209  }
2210  }
2211 
2212  /* 2.stage: search dominance relations for the remaining columns
2213  * by increasing row-sparsity
2214  */
2215  if( (presoltiming & SCIP_PRESOLTIMING_EXHAUSTIVE) != 0 )
2216  {
2217  SCIPsortIntInt(rowsparsity, rowidxsorted, nrows);
2218 
2219  for( r = 0; r < nrows; ++r )
2220  {
2221  int rowidx;
2222  int* rowpnt;
2223  int* rowend;
2224 
2225  /* break if the time limit was reached; since the check is expensive,
2226  * we only check all 1000 constraints
2227  */
2228  if( (r % 1000 == 0) && SCIPisStopped(scip) )
2229  break;
2230 
2231  rowidx = rowidxsorted[r];
2232  rowpnt = SCIPmatrixGetRowIdxPtr(matrix, rowidx);
2233  rowend = rowpnt + SCIPmatrixGetRowNNonzs(matrix, rowidx);
2234 
2235  if( SCIPmatrixGetRowNNonzs(matrix, rowidx) == 1 )
2236  continue;
2237 
2238  nconfill = 0;
2239  nintfill = 0;
2240  nbinfill = 0;
2241 
2242  for( ; rowpnt < rowend; rowpnt++ )
2243  {
2244  if( !(varsprocessed[*rowpnt]) )
2245  {
2246  int varidx;
2247  SCIP_VAR* var;
2248 
2249  varidx = *rowpnt;
2250  var = SCIPmatrixGetVar(matrix, varidx);
2251 
2252  /* we search only for dominance relations between the same variable type */
2254  {
2255  consearchcols[nconfill++] = varidx;
2256  }
2257  else if( SCIPvarIsBinary(var) )
2258  {
2259  binsearchcols[nbinfill++] = varidx;
2260  }
2261  else
2262  {
2264  intsearchcols[nintfill++] = varidx;
2265  }
2266  }
2267  }
2268 
2269  /* continuous variables */
2270  if( nconfill > 1 && presoldata->continuousred )
2271  {
2272  SCIP_CALL( findDominancePairs(scip, matrix, presoldata, consearchcols, nconfill, FALSE,
2273  varstofix, &nfixings, &ndomrelations, nchgbds) );
2274 
2275  for( v = 0; v < nconfill; ++v )
2276  varsprocessed[consearchcols[v]] = TRUE;
2277 
2278  varcount += nconfill;
2279  }
2280 
2281  /* integer and impl-integer variables */
2282  if( nintfill > 1 )
2283  {
2284  SCIP_CALL( findDominancePairs(scip, matrix, presoldata, intsearchcols, nintfill, FALSE,
2285  varstofix, &nfixings, &ndomrelations, nchgbds) );
2286 
2287  for( v = 0; v < nintfill; ++v )
2288  varsprocessed[intsearchcols[v]] = TRUE;
2289 
2290  varcount += nintfill;
2291  }
2292 
2293  /* binary variables */
2294  if( nbinfill > 1 )
2295  {
2296  SCIP_CALL( findDominancePairs(scip, matrix, presoldata, binsearchcols, nbinfill, TRUE,
2297  varstofix, &nfixings, &ndomrelations, nchgbds) );
2298 
2299  for( v = 0; v < nbinfill; ++v )
2300  varsprocessed[binsearchcols[v]] = TRUE;
2301 
2302  varcount += nbinfill;
2303  }
2304 
2305  if( varcount >= ncols )
2306  break;
2307  }
2308  }
2309 
2310  if( nfixings > 0 )
2311  {
2312  int oldnfixedvars;
2313 
2314  oldnfixedvars = *nfixedvars;
2315 
2316  for( v = ncols - 1; v >= 0; --v )
2317  {
2318  SCIP_Bool fixed;
2319  SCIP_VAR* var;
2320 
2321  var = SCIPmatrixGetVar(matrix,v);
2322 
2323  /* there should be no fixings for variables with lock conflicts,
2324  * since they where marked as processed and therefore should be skipped
2325  */
2326  assert(varstofix[v] == NOFIX || (!SCIPmatrixUplockConflict(matrix, v) && !SCIPmatrixDownlockConflict(matrix, v)) );
2327 
2328  if( varstofix[v] == FIXATLB )
2329  {
2330  SCIP_Real lb;
2331 
2332  lb = SCIPvarGetLbGlobal(var);
2333  assert(SCIPvarGetType(var) == SCIP_VARTYPE_CONTINUOUS || SCIPisFeasIntegral(scip, lb));
2334 
2335  /* fix at lower bound */
2336  SCIP_CALL( SCIPfixVar(scip, var, lb, &infeasible, &fixed) );
2337  if( infeasible )
2338  {
2339  SCIPdebugMsg(scip, " -> infeasible fixing\n");
2340  *result = SCIP_CUTOFF;
2341 
2342  break;
2343  }
2344  assert(fixed);
2345  (*nfixedvars)++;
2346 
2347 #ifdef SCIP_DEBUG
2349  nconvarsfixed++;
2350  else if( SCIPvarIsBinary(var) )
2351  nbinvarsfixed++;
2352  else
2353  nintvarsfixed++;
2354 #endif
2355  }
2356  else if( varstofix[v] == FIXATUB )
2357  {
2358  SCIP_Real ub;
2359 
2360  ub = SCIPvarGetUbGlobal(var);
2361  assert(SCIPvarGetType(var) == SCIP_VARTYPE_CONTINUOUS || SCIPisFeasIntegral(scip, ub));
2362 
2363  /* fix at upper bound */
2364  SCIP_CALL( SCIPfixVar(scip, var, ub, &infeasible, &fixed) );
2365  if( infeasible )
2366  {
2367  SCIPdebugMsg(scip, " -> infeasible fixing\n");
2368  *result = SCIP_CUTOFF;
2369 
2370  break;
2371  }
2372  assert(fixed);
2373  (*nfixedvars)++;
2374 
2375 #ifdef SCIP_DEBUG
2377  nconvarsfixed++;
2378  else if( SCIPvarIsBinary(var) )
2379  nbinvarsfixed++;
2380  else
2381  nintvarsfixed++;
2382 #endif
2383  }
2384  }
2385 
2386  if( *result != SCIP_CUTOFF && *nfixedvars > oldnfixedvars )
2387  *result = SCIP_SUCCESS;
2388  }
2389 
2390  SCIPfreeBufferArray(scip, &varineq);
2391  SCIPfreeBufferArray(scip, &colidx);
2392  SCIPfreeBufferArray(scip, &pclass);
2393  SCIPfreeBufferArray(scip, &rowsparsity);
2394  SCIPfreeBufferArray(scip, &rowidxsorted);
2395  SCIPfreeBufferArray(scip, &binsearchcols);
2396  SCIPfreeBufferArray(scip, &intsearchcols);
2397  SCIPfreeBufferArray(scip, &consearchcols);
2398  SCIPfreeBufferArray(scip, &varsprocessed);
2399  SCIPfreeBufferArray(scip, &varstofix);
2400 
2401 #ifdef SCIP_DEBUG
2402  if( (nconvarsfixed + nintvarsfixed + nbinvarsfixed) > 0 )
2403  {
2404  SCIPdebugMsg(scip, "### %d vars [%" SCIP_LONGINT_FORMAT " dom] => fixed [cont: %d, int: %d, bin: %d], %scutoff detected\n",
2405  ncols, ndomrelations, nconvarsfixed, nintvarsfixed, nbinvarsfixed, (*result != SCIP_CUTOFF) ? "no " : "");
2406  }
2407 #endif
2408 
2409  SCIPmatrixFree(scip, &matrix);
2410 
2411  return SCIP_OKAY;
2412 }
2413 
2414 /*
2415  * presolver specific interface methods
2416  */
2417 
2418 /** creates the domcol presolver and includes it in SCIP */
2420  SCIP* scip /**< SCIP data structure */
2421  )
2422 {
2423  SCIP_PRESOLDATA* presoldata;
2424  SCIP_PRESOL* presol;
2425 
2426  /* create domcol presolver data */
2427  SCIP_CALL( SCIPallocBlockMemory(scip, &presoldata) );
2428 
2429  /* include presolver */
2431  PRESOL_TIMING, presolExecDomcol, presoldata) );
2432  SCIP_CALL( SCIPsetPresolCopy(scip, presol, presolCopyDomcol) );
2433  SCIP_CALL( SCIPsetPresolFree(scip, presol, presolFreeDomcol) );
2434 
2435  SCIP_CALL( SCIPaddIntParam(scip,
2436  "presolving/domcol/numminpairs",
2437  "minimal number of pair comparisons",
2438  &presoldata->numminpairs, FALSE, DEFAULT_NUMMINPAIRS, 100, DEFAULT_NUMMAXPAIRS, NULL, NULL) );
2439 
2440  SCIP_CALL( SCIPaddIntParam(scip,
2441  "presolving/domcol/nummaxpairs",
2442  "maximal number of pair comparisons",
2443  &presoldata->nummaxpairs, FALSE, DEFAULT_NUMMAXPAIRS, DEFAULT_NUMMINPAIRS, 1000000000, NULL, NULL) );
2444 
2446  "presolving/domcol/predbndstr",
2447  "should predictive bound strengthening be applied?",
2448  &presoldata->predbndstr, FALSE, DEFAULT_PREDBNDSTR, NULL, NULL) );
2449 
2451  "presolving/domcol/continuousred",
2452  "should reductions for continuous variables be performed?",
2453  &presoldata->continuousred, FALSE, DEFAULT_CONTINUOUS_RED, NULL, NULL) );
2454 
2455  return SCIP_OKAY;
2456 }
SCIP_Bool SCIPisEQ(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
#define PRESOL_PRIORITY
Definition: presol_domcol.c:63
SCIP_Bool SCIPisPositive(SCIP *scip, SCIP_Real val)
struct SCIP_PresolData SCIP_PRESOLDATA
Definition: type_presol.h:42
SCIP_VAR * SCIPmatrixGetVar(SCIP_MATRIX *matrix, int col)
Definition: matrix.c:1620
SCIP_Bool SCIPisStopped(SCIP *scip)
Definition: scip_general.c:687
int SCIPmatrixGetNRows(SCIP_MATRIX *matrix)
Definition: matrix.c:1692
SCIP_RETCODE SCIPincludePresolDomcol(SCIP *scip)
public methods for SCIP parameter handling
static SCIP_RETCODE calcVarBoundsDominating(SCIP *scip, SCIP_MATRIX *matrix, int row, int coldominating, SCIP_Real valdominating, int coldominated, SCIP_Real valdominated, SCIP_Bool *ubcalculated, SCIP_Real *calculatedub, SCIP_Bool *wclbcalculated, SCIP_Real *calculatedwclb, SCIP_Bool *lbcalculated, SCIP_Real *calculatedlb, SCIP_Bool *wcubcalculated, SCIP_Real *calculatedwcub)
SCIP_Bool SCIPisGE(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
#define DEFAULT_PREDBNDSTR
Definition: presol_domcol.c:70
public methods for memory management
static SCIP_RETCODE updateBounds(SCIP *scip, SCIP_MATRIX *matrix, int row, int col1, SCIP_Real val1, int col2, SCIP_Real val2, SCIP_Bool predictdominating, SCIP_Real *upperbound, SCIP_Real *wclowerbound, SCIP_Real *lowerbound, SCIP_Real *wcupperbound)
void SCIPmatrixFree(SCIP *scip, SCIP_MATRIX **matrix)
Definition: matrix.c:1032
SCIP_EXPORT SCIP_Bool SCIPvarIsBinary(SCIP_VAR *var)
Definition: var.c:17197
int SCIPgetNVars(SCIP *scip)
Definition: scip_prob.c:1986
dominated column presolver
#define FALSE
Definition: def.h:73
public methods for presolving plugins
SCIP_EXPORT SCIP_Real SCIPvarGetObj(SCIP_VAR *var)
Definition: var.c:17515
SCIP_EXPORT SCIP_VARTYPE SCIPvarGetType(SCIP_VAR *var)
Definition: var.c:17182
#define TRUE
Definition: def.h:72
#define PRESOL_NAME
Definition: presol_domcol.c:61
enum SCIP_Retcode SCIP_RETCODE
Definition: type_retcode.h:54
#define SCIP_PRESOLTIMING_EXHAUSTIVE
Definition: type_timing.h:45
#define PRESOL_DESC
Definition: presol_domcol.c:62
SCIP_Bool SCIPisFeasLE(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
public methods for problem variables
SCIP_RETCODE SCIPaddBoolParam(SCIP *scip, const char *name, const char *desc, SCIP_Bool *valueptr, SCIP_Bool isadvanced, SCIP_Bool defaultvalue, SCIP_DECL_PARAMCHGD((*paramchgd)), SCIP_PARAMDATA *paramdata)
Definition: scip_param.c:48
SCIP_Real SCIPceil(SCIP *scip, SCIP_Real val)
#define SCIPfreeBlockMemory(scip, ptr)
Definition: scip_mem.h:95
static SCIP_RETCODE findFixings(SCIP *scip, SCIP_MATRIX *matrix, SCIP_VAR *dominatingvar, int dominatingidx, SCIP_Real dominatingub, SCIP_Real dominatingwclb, SCIP_Real dominatinglb, SCIP_Real dominatingwcub, SCIP_VAR *dominatedvar, int dominatedidx, FIXINGDIRECTION *varstofix, SCIP_Bool onlybinvars, SCIP_Bool onlyoneone, int *nfixings)
SCIP_EXPORT SCIP_VARSTATUS SCIPvarGetStatus(SCIP_VAR *var)
Definition: var.c:17136
#define PRESOL_TIMING
Definition: presol_domcol.c:65
#define SCIPfreeBufferArray(scip, ptr)
Definition: scip_mem.h:123
SCIP_Real SCIPmatrixGetRowMaxActivity(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1760
#define DEFAULT_NUMMINPAIRS
Definition: presol_domcol.c:67
#define SCIPallocBlockMemory(scip, ptr)
Definition: scip_mem.h:78
public methods for SCIP variables
#define SCIPdebugMsgPrint
Definition: scip_message.h:70
#define SCIPdebugMsg
Definition: scip_message.h:69
SCIP_Bool SCIPisFeasIntegral(SCIP *scip, SCIP_Real val)
SCIP_Bool SCIPisInfinity(SCIP *scip, SCIP_Real val)
int SCIPgetNIntVars(SCIP *scip)
Definition: scip_prob.c:2076
public methods for numerical tolerances
int SCIPmatrixGetRowNNonzs(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1668
SCIP_EXPORT void SCIPsortIntIntReal(int *intarray1, int *intarray2, SCIP_Real *realarray, int len)
SCIP_Bool SCIPisLE(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
SCIP_Real SCIPmatrixGetRowLhs(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1702
enum Fixingdirection FIXINGDIRECTION
Definition: presol_domcol.c:96
Fixingdirection
Definition: presol_domcol.c:90
SCIP_EXPORT const char * SCIPvarGetName(SCIP_VAR *var)
Definition: var.c:17017
SCIP_Real * SCIPmatrixGetColValPtr(SCIP_MATRIX *matrix, int col)
Definition: matrix.c:1528
static void getActivityResidualsLowerBound(SCIP *scip, SCIP_MATRIX *matrix, int row, int col, SCIP_Real coef, int lowerboundcol, SCIP_Real lowerboundcoef, SCIP_Real *minresactivity, SCIP_Real *maxresactivity, SCIP_Bool *success)
SCIP_RETCODE SCIPchgVarLb(SCIP *scip, SCIP_VAR *var, SCIP_Real newbound)
Definition: scip_var.c:4677
#define SCIPerrorMessage
Definition: pub_message.h:55
SCIP_PRESOLDATA * SCIPpresolGetData(SCIP_PRESOL *presol)
Definition: presol.c:503
int SCIPmatrixGetRowNMaxActPosInf(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1808
SCIP_Bool SCIPisZero(SCIP *scip, SCIP_Real val)
int * SCIPmatrixGetRowIdxPtr(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1656
#define NULL
Definition: lpi_spx1.cpp:155
int SCIPmatrixGetRowNMaxActNegInf(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1796
#define SCIP_CALL(x)
Definition: def.h:364
SCIP_RETCODE SCIPmatrixCreate(SCIP *scip, SCIP_MATRIX **matrixptr, SCIP_Bool onlyifcomplete, SCIP_Bool *initialized, SCIP_Bool *complete, SCIP_Bool *infeasible, int *naddconss, int *ndelconss, int *nchgcoefs, int *nchgbds, int *nfixedvars)
Definition: matrix.c:445
SCIP_Real * SCIPmatrixGetRowValPtr(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1644
SCIP_Bool SCIPmatrixDownlockConflict(SCIP_MATRIX *matrix, int col)
Definition: matrix.c:1844
SCIP_RETCODE SCIPsetPresolCopy(SCIP *scip, SCIP_PRESOL *presol, SCIP_DECL_PRESOLCOPY((*presolcopy)))
Definition: scip_presol.c:130
void SCIPpresolSetData(SCIP_PRESOL *presol, SCIP_PRESOLDATA *presoldata)
Definition: presol.c:513
static SCIP_RETCODE predBndStr(SCIP *scip, SCIP_VAR *dominatingvar, int dominatingidx, SCIP_Real dominatingub, SCIP_Real dominatinglb, SCIP_Real dominatingwcub, SCIP_VAR *dominatedvar, int dominatedidx, SCIP_Real dominatedub, SCIP_Real dominatedwclb, SCIP_Real dominatedlb, FIXINGDIRECTION *varstofix, int *nchgbds)
static SCIP_RETCODE findDominancePairs(SCIP *scip, SCIP_MATRIX *matrix, SCIP_PRESOLDATA *presoldata, int *searchcols, int searchsize, SCIP_Bool onlybinvars, FIXINGDIRECTION *varstofix, int *nfixings, SCIP_Longint *ndomrelations, int *nchgbds)
#define SCIPallocBufferArray(scip, ptr, num)
Definition: scip_mem.h:111
SCIP_Real SCIPinfinity(SCIP *scip)
SCIP_EXPORT void SCIPsortRealInt(SCIP_Real *realarray, int *intarray, int len)
#define SCIP_Bool
Definition: def.h:70
SCIP_EXPORT void SCIPsortIntInt(int *intarray1, int *intarray2, int len)
static SCIP_DECL_PRESOLFREE(presolFreeDomcol)
SCIP_EXPORT SCIP_Real SCIPvarGetUbGlobal(SCIP_VAR *var)
Definition: var.c:17677
const char * SCIPpresolGetName(SCIP_PRESOL *presol)
Definition: presol.c:590
#define MAX(x, y)
Definition: tclique_def.h:83
int * SCIPmatrixGetColIdxPtr(SCIP_MATRIX *matrix, int col)
Definition: matrix.c:1540
SCIP_Bool SCIPisNegative(SCIP *scip, SCIP_Real val)
SCIP_RETCODE SCIPfixVar(SCIP *scip, SCIP_VAR *var, SCIP_Real fixedval, SCIP_Bool *infeasible, SCIP_Bool *fixed)
Definition: scip_var.c:8257
static SCIP_RETCODE calcVarBoundsDominated(SCIP *scip, SCIP_MATRIX *matrix, int row, int coldominating, SCIP_Real valdominating, int coldominated, SCIP_Real valdominated, SCIP_Bool *ubcalculated, SCIP_Real *calculatedub, SCIP_Bool *wclbcalculated, SCIP_Real *calculatedwclb, SCIP_Bool *lbcalculated, SCIP_Real *calculatedlb, SCIP_Bool *wcubcalculated, SCIP_Real *calculatedwcub)
int SCIPmatrixGetRowNMinActNegInf(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1772
SCIP_Bool SCIPisFeasGE(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
SCIP_RETCODE SCIPchgVarUb(SCIP *scip, SCIP_VAR *var, SCIP_Real newbound)
Definition: scip_var.c:4767
static SCIP_DECL_PRESOLEXEC(presolExecDomcol)
public methods for matrix
public methods for variable pricer plugins
SCIP_RETCODE SCIPsetPresolFree(SCIP *scip, SCIP_PRESOL *presol, SCIP_DECL_PRESOLFREE((*presolfree)))
Definition: scip_presol.c:146
public methods for nonlinear relaxations
SCIP_Real * r
Definition: circlepacking.c:50
static void getActivityResidualsUpperBound(SCIP *scip, SCIP_MATRIX *matrix, int row, int col, SCIP_Real coef, int upperboundcol, SCIP_Real upperboundcoef, SCIP_Real *minresactivity, SCIP_Real *maxresactivity, SCIP_Bool *success)
SCIP_Real SCIPfloor(SCIP *scip, SCIP_Real val)
methods for sorting joint arrays of various types
public methods for presolvers
general public methods
SCIP_Real SCIPmatrixGetRowRhs(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1714
SCIP_EXPORT SCIP_Real SCIPvarGetLbGlobal(SCIP_VAR *var)
Definition: var.c:17667
public methods for the probing mode
SCIP_Bool SCIPmatrixIsRowRhsInfinity(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1726
public methods for message output
SCIP_RETCODE SCIPaddIntParam(SCIP *scip, const char *name, const char *desc, int *valueptr, SCIP_Bool isadvanced, int defaultvalue, int minvalue, int maxvalue, SCIP_DECL_PARAMCHGD((*paramchgd)), SCIP_PARAMDATA *paramdata)
Definition: scip_param.c:74
#define DEFAULT_NUMMAXPAIRS
Definition: presol_domcol.c:68
#define SCIP_Real
Definition: def.h:163
SCIP_RETCODE SCIPincludePresolBasic(SCIP *scip, SCIP_PRESOL **presolptr, const char *name, const char *desc, int priority, int maxrounds, SCIP_PRESOLTIMING timing, SCIP_DECL_PRESOLEXEC((*presolexec)), SCIP_PRESOLDATA *presoldata)
Definition: scip_presol.c:95
static SCIP_RETCODE detectParallelCols(SCIP *scip, SCIP_MATRIX *matrix, int *pclass, SCIP_Bool *varineq)
SCIP_Bool SCIPisLT(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
public methods for message handling
SCIP_Real SCIPmatrixGetRowMinActivity(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1748
SCIP_Bool SCIPmatrixUplockConflict(SCIP_MATRIX *matrix, int col)
Definition: matrix.c:1832
#define SCIP_Longint
Definition: def.h:148
SCIP_EXPORT SCIP_Bool SCIPvarsHaveCommonClique(SCIP_VAR *var1, SCIP_Bool value1, SCIP_VAR *var2, SCIP_Bool value2, SCIP_Bool regardimplics)
Definition: var.c:11249
int SCIPgetNBinVars(SCIP *scip)
Definition: scip_prob.c:2031
int SCIPmatrixGetRowNMinActPosInf(SCIP_MATRIX *matrix, int row)
Definition: matrix.c:1784
SCIP_Bool SCIPallowStrongDualReds(SCIP *scip)
Definition: scip_var.c:8597
#define BMSclearMemoryArray(ptr, num)
Definition: memory.h:122
#define SCIPABORT()
Definition: def.h:336
public methods for global and local (sub)problems
#define PRESOL_MAXROUNDS
Definition: presol_domcol.c:64
static SCIP_DECL_PRESOLCOPY(presolCopyDomcol)
static SCIP_RETCODE printRow(SCIP *scip, FZNOUTPUT *fznoutput, const char *type, SCIP_VAR **vars, SCIP_Real *vals, int nvars, SCIP_Real rhs, SCIP_Bool hasfloats)
Definition: reader_fzn.c:4012
int SCIPmatrixGetNColumns(SCIP_MATRIX *matrix)
Definition: matrix.c:1564
#define DEFAULT_CONTINUOUS_RED
Definition: presol_domcol.c:71
int SCIPmatrixGetColNNonzs(SCIP_MATRIX *matrix, int col)
Definition: matrix.c:1552
memory allocation routines