Scippy

SCIP

Solving Constraint Integer Programs

prop_obbt.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) 2002-2024 Zuse Institute Berlin (ZIB) */
7 /* */
8 /* Licensed under the Apache License, Version 2.0 (the "License"); */
9 /* you may not use this file except in compliance with the License. */
10 /* You may obtain a copy of the License at */
11 /* */
12 /* http://www.apache.org/licenses/LICENSE-2.0 */
13 /* */
14 /* Unless required by applicable law or agreed to in writing, software */
15 /* distributed under the License is distributed on an "AS IS" BASIS, */
16 /* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. */
17 /* See the License for the specific language governing permissions and */
18 /* limitations under the License. */
19 /* */
20 /* You should have received a copy of the Apache-2.0 license */
21 /* along with SCIP; see the file LICENSE. If not visit scipopt.org. */
22 /* */
23 /* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
24 
25 /**@file prop_obbt.c
26  * @ingroup DEFPLUGINS_PROP
27  * @brief optimization-based bound tightening propagator
28  * @author Stefan Weltge
29  * @author Benjamin Mueller
30  */
31 
32 /**@todo if bound tightenings of other propagators are the reason for lpsolstat != SCIP_LPSOLSTAT_OPTIMAL, resolve LP */
33 /**@todo only run more than once in root node if primal bound improved or many cuts were added to the LP */
34 /**@todo filter bounds of a variable already if SCIPisLbBetter()/SCIPisUbBetter() would return FALSE */
35 /**@todo improve warmstarting of LP solving */
36 /**@todo include bound value (finite/infinite) into getScore() function */
37 /**@todo use unbounded ray in filtering */
38 /**@todo do we want to run if the LP is unbounded, maybe for infinite variable bounds? */
39 /**@todo add first filter round in direction of objective function */
40 /**@todo implement conflict resolving callback by calling public method of genvbounds propagator, since the reason are
41  * exactly the variable bounds with nonnegative reduced costs stored in the right-hand side of the generated
42  * generalized variable bound (however, this only makes sense if we run locally)
43  */
44 
45 /*---+----1----+----2----+----3----+----4----+----5----+----6----+----7----+----8----+----9----+----0----+----1----+----2*/
46 
47 #include <assert.h>
48 #include <string.h>
49 
50 #include "scip/cons_indicator.h"
51 #include "scip/cons_linear.h"
52 #include "scip/cons_nonlinear.h"
53 #include "scip/nlhdlr_bilinear.h"
54 #include "scip/prop_genvbounds.h"
55 #include "scip/prop_obbt.h"
56 #include "scip/pub_cons.h"
57 #include "scip/pub_lp.h"
58 #include "scip/pub_message.h"
59 #include "scip/pub_misc.h"
60 #include "scip/pub_misc_sort.h"
61 #include "scip/pub_nlp.h"
62 #include "scip/pub_prop.h"
63 #include "scip/pub_tree.h"
64 #include "scip/pub_var.h"
65 #include "scip/scip_cons.h"
66 #include "scip/scip_copy.h"
67 #include "scip/scip_cut.h"
68 #include "scip/scip_general.h"
69 #include "scip/scip_lp.h"
70 #include "scip/scip_mem.h"
71 #include "scip/scip_message.h"
72 #include "scip/scip_nlp.h"
73 #include "scip/scip_numerics.h"
74 #include "scip/scip_param.h"
75 #include "scip/scip_prob.h"
76 #include "scip/scip_probing.h"
77 #include "scip/scip_prop.h"
78 #include "scip/scip_randnumgen.h"
79 #include "scip/scip_solvingstats.h"
80 #include "scip/scip_tree.h"
81 #include "scip/scip_var.h"
82 
83 #define PROP_NAME "obbt"
84 #define PROP_DESC "optimization-based bound tightening propagator"
85 #define PROP_TIMING SCIP_PROPTIMING_AFTERLPLOOP
86 #define PROP_PRIORITY -1000000 /**< propagator priority */
87 #define PROP_FREQ 0 /**< propagator frequency */
88 #define PROP_DELAY TRUE /**< should propagation method be delayed, if other propagators
89  * found reductions? */
90 
91 #define DEFAULT_CREATE_GENVBOUNDS TRUE /**< should obbt try to provide genvbounds if possible? */
92 #define DEFAULT_FILTERING_NORM TRUE /**< should coefficients in filtering be normalized w.r.t. the
93  * domains sizes? */
94 #define DEFAULT_APPLY_FILTERROUNDS FALSE /**< try to filter bounds in so-called filter rounds by solving
95  * auxiliary LPs? */
96 #define DEFAULT_APPLY_TRIVIALFITLERING TRUE /**< should obbt try to use the LP solution to filter some bounds? */
97 #define DEFAULT_GENVBDSDURINGFILTER TRUE /**< try to genrate genvbounds during trivial and aggressive filtering? */
98 #define DEFAULT_DUALFEASTOL 1e-9 /**< feasibility tolerance for reduced costs used in obbt; this value
99  * is used if SCIP's dual feastol is greater */
100 #define DEFAULT_CONDITIONLIMIT -1.0 /**< maximum condition limit used in LP solver (-1.0: no limit) */
101 #define DEFAULT_BOUNDSTREPS 0.001 /**< minimal relative improve for strengthening bounds */
102 #define DEFAULT_FILTERING_MIN 2 /**< minimal number of filtered bounds to apply another filter
103  * round */
104 #define DEFAULT_ITLIMITFACTOR 10.0 /**< multiple of root node LP iterations used as total LP iteration
105  * limit for obbt (<= 0: no limit ) */
106 #define DEFAULT_MINITLIMIT 5000L /**< minimum LP iteration limit */
107 #define DEFAULT_ONLYNONCONVEXVARS TRUE /**< only apply obbt on non-convex variables */
108 #define DEFAULT_INDICATORS FALSE /**< apply obbt on variables of indicator constraints? (independent of convexity) */
109 #define DEFAULT_INDICATORTHRESHOLD 1e6 /**< variables of indicator constraints with smaller upper bound are not considered
110  * and upper bound is tightened only if new bound is smaller */
111 #define DEFAULT_TIGHTINTBOUNDSPROBING TRUE /**< should bounds of integral variables be tightened during
112  * the probing mode? */
113 #define DEFAULT_TIGHTCONTBOUNDSPROBING FALSE /**< should bounds of continuous variables be tightened during
114  * the probing mode? */
115 #define DEFAULT_ORDERINGALGO 1 /**< which type of ordering algorithm should we use?
116  * (0: no, 1: greedy, 2: greedy reverse) */
117 #define OBBT_SCOREBASE 5 /**< base that is used to calculate a bounds score value */
118 #define GENVBOUND_PROP_NAME "genvbounds"
119 
120 #define DEFAULT_SEPARATESOL FALSE /**< should the obbt LP solution be separated? note that that by
121  * separating solution OBBT will apply all bound tightenings
122  * immediatly */
123 #define DEFAULT_SEPAMINITER 0 /**< minimum number of iteration spend to separate an obbt LP solution */
124 #define DEFAULT_SEPAMAXITER 10 /**< maximum number of iteration spend to separate an obbt LP solution */
125 #define DEFAULT_GENVBDSDURINGSEPA TRUE /**< try to create genvbounds during separation process? */
126 #define DEFAULT_PROPAGATEFREQ 0 /**< trigger a propagation round after that many bound tightenings
127  * (0: no propagation) */
128 #define DEFAULT_CREATE_BILININEQS TRUE /**< solve auxiliary LPs in order to find valid inequalities for bilinear terms? */
129 #define DEFAULT_CREATE_LINCONS FALSE /**< create linear constraints from inequalities for bilinear terms? */
130 #define DEFAULT_ITLIMITFAC_BILININEQS 3.0 /**< multiple of OBBT LP limit used as total LP iteration limit for solving bilinear inequality LPs (< 0 for no limit) */
131 #define DEFAULT_MINNONCONVEXITY 1e-1 /**< minimum nonconvexity for choosing a bilinear term */
132 #define DEFAULT_RANDSEED 149 /**< initial random seed */
133 
134 /*
135  * Data structures
136  */
138 /** bound data */
139 struct Bound
140 {
141  SCIP_VAR* var; /**< variable */
142  SCIP_Real newval; /**< stores a probably tighter value for this bound */
143  SCIP_BOUNDTYPE boundtype; /**< type of bound */
144  unsigned int score; /**< score value that is used to group bounds */
145  unsigned int filtered:1; /**< thrown out during pre-filtering step */
146  unsigned int found:1; /**< stores whether a probably tighter value for this bound was found */
147  unsigned int done:1; /**< has this bound been processed already? */
148  unsigned int nonconvex:1; /**< is this bound affecting a nonconvex term? */
149  unsigned int indicator:1; /**< is this bound affecting an indicator constraint? */
150  int index; /**< unique index */
151 };
152 typedef struct Bound BOUND;
153 
154 /* all possible corners of a rectangular domain */
155 enum Corner
156 {
159  RIGHTTOP = 4,
160  LEFTTOP = 8,
161  FILTERED = 15
162 };
163 typedef enum Corner CORNER;
164 
165 /** bilinear bound data */
166 struct BilinBound
167 {
168  SCIP_EXPR* expr; /**< product expression */
169  int filtered; /**< corners that could be thrown out during pre-filtering step */
170  unsigned int done:1; /**< has this bilinear term been processed already? */
171  SCIP_Real score; /**< score value that is used to group bilinear term bounds */
172 };
173 typedef struct BilinBound BILINBOUND;
175 /** propagator data */
176 struct SCIP_PropData
177 {
178  BOUND** bounds; /**< array of interesting bounds */
179  BILINBOUND** bilinbounds; /**< array of interesting bilinear bounds */
180  SCIP_ROW* cutoffrow; /**< pointer to current objective cutoff row */
181  SCIP_PROP* genvboundprop; /**< pointer to genvbound propagator */
182  SCIP_RANDNUMGEN* randnumgen; /**< random number generator */
183  SCIP_Longint lastnode; /**< number of last node where obbt was performed */
184  SCIP_Longint npropagatedomreds; /**< number of domain reductions found during propagation */
185  SCIP_Longint nprobingiterations; /**< number of LP iterations during the probing mode */
186  SCIP_Longint nfilterlpiters; /**< number of LP iterations spend for filtering */
187  SCIP_Longint minitlimit; /**< minimum LP iteration limit */
188  SCIP_Longint itlimitbilin; /**< total LP iterations limit for solving bilinear inequality LPs */
189  SCIP_Longint itusedbilin; /**< total LP iterations used for solving bilinear inequality LPs */
190  SCIP_Real dualfeastol; /**< feasibility tolerance for reduced costs used in obbt; this value is
191  * used if SCIP's dual feastol is greater */
192  SCIP_Real conditionlimit; /**< maximum condition limit used in LP solver (-1.0: no limit) */
193  SCIP_Real boundstreps; /**< minimal relative improve for strengthening bounds */
194  SCIP_Real itlimitfactor; /**< LP iteration limit for obbt will be this factor times total LP
195  * iterations in root node */
196  SCIP_Real itlimitfactorbilin; /**< multiple of OBBT LP limit used as total LP iteration limit for solving bilinear inequality LPs (< 0 for no limit) */
197  SCIP_Real minnonconvexity; /**< lower bound on minimum absolute value of nonconvex eigenvalues for a bilinear term */
198  SCIP_Real indicatorthreshold; /**< threshold whether upper bounds of vars of indicator conss are considered or tightened */
199  SCIP_Bool applyfilterrounds; /**< apply filter rounds? */
200  SCIP_Bool applytrivialfilter; /**< should obbt try to use the LP solution to filter some bounds? */
201  SCIP_Bool genvbdsduringfilter;/**< should we try to generate genvbounds during trivial and aggressive
202  * filtering? */
203  SCIP_Bool genvbdsduringsepa; /**< try to create genvbounds during separation process? */
204  SCIP_Bool creategenvbounds; /**< should obbt try to provide genvbounds if possible? */
205  SCIP_Bool normalize; /**< should coefficients in filtering be normalized w.r.t. the domains
206  * sizes? */
207  SCIP_Bool onlynonconvexvars; /**< only apply obbt on non-convex variables */
208  SCIP_Bool indicators; /**< apply obbt on variables of indicator constraints? (independent of convexity) */
209  SCIP_Bool tightintboundsprobing; /**< should bounds of integral variables be tightened during
210  * the probing mode? */
211  SCIP_Bool tightcontboundsprobing;/**< should bounds of continuous variables be tightened during
212  * the probing mode? */
213  SCIP_Bool separatesol; /**< should the obbt LP solution be separated? note that that by
214  * separating solution OBBT will apply all bound tightenings
215  * immediatly */
216  SCIP_Bool createbilinineqs; /**< solve auxiliary LPs in order to find valid inequalities for bilinear terms? */
217  SCIP_Bool createlincons; /**< create linear constraints from inequalities for bilinear terms? */
218  int orderingalgo; /**< which type of ordering algorithm should we use?
219  * (0: no, 1: greedy, 2: greedy reverse) */
220  int nbounds; /**< length of interesting bounds array */
221  int nbilinbounds; /**< length of interesting bilinear bounds array */
222  int bilinboundssize; /**< size of bilinear bounds array */
223  int boundssize; /**< size of bounds array */
224  int nminfilter; /**< minimal number of filtered bounds to apply another filter round */
225  int nfiltered; /**< number of filtered bounds by solving auxiliary variables */
226  int ntrivialfiltered; /**< number of filtered bounds because the LP value was equal to the bound */
227  int nsolvedbounds; /**< number of solved bounds during the loop in applyObbt() */
228  int ngenvboundsprobing; /**< number of non-trivial genvbounds generated and added during obbt */
229  int ngenvboundsaggrfil; /**< number of non-trivial genvbounds found during aggressive filtering */
230  int ngenvboundstrivfil; /**< number of non-trivial genvbounds found during trivial filtering */
231  int lastidx; /**< index to store the last undone and unfiltered bound */
232  int lastbilinidx; /**< index to store the last undone and unfiltered bilinear bound */
233  int sepaminiter; /**< minimum number of iteration spend to separate an obbt LP solution */
234  int sepamaxiter; /**< maximum number of iteration spend to separate an obbt LP solution */
235  int propagatefreq; /**< trigger a propagation round after that many bound tightenings
236  * (0: no propagation) */
237  int propagatecounter; /**< number of bound tightenings since the last propagation round */
238 };
239 
240 
241 /*
242  * Local methods
243  */
244 
245 /** solves the LP and handles errors */
246 static
248  SCIP* scip, /**< SCIP data structure */
249  int itlimit, /**< maximal number of LP iterations to perform, or -1 for no limit */
250  SCIP_Bool* error, /**< pointer to store whether an unresolved LP error occurred */
251  SCIP_Bool* optimal /**< was the LP solved to optimalilty? */
252  )
253 {
254  SCIP_LPSOLSTAT lpsolstat;
255  SCIP_RETCODE retcode;
256 
257  assert(scip != NULL);
258  assert(itlimit == -1 || itlimit >= 0);
259  assert(error != NULL);
260  assert(optimal != NULL);
261 
262  *optimal = FALSE;
263  *error = FALSE;
264 
265  retcode = SCIPsolveProbingLP(scip, itlimit, error, NULL);
266 
267  lpsolstat = SCIPgetLPSolstat(scip);
268 
269  /* an error should not kill the overall solving process */
270  if( retcode != SCIP_OKAY )
271  {
272  SCIPwarningMessage(scip, " error while solving LP in obbt propagator; LP solve terminated with code <%d>\n", retcode);
273  SCIPwarningMessage(scip, " this does not affect the remaining solution procedure --> continue\n");
274 
275  *error = TRUE;
276 
277  return SCIP_OKAY;
278  }
279 
280  if( lpsolstat == SCIP_LPSOLSTAT_OPTIMAL )
281  {
282  assert(!*error);
283  *optimal = TRUE;
284  }
285 #ifdef SCIP_DEBUG
286  else
287  {
288  switch( lpsolstat )
289  {
291  SCIPdebugMsg(scip, " reached lp iteration limit\n");
292  break;
294  SCIPdebugMsg(scip, " reached time limit while solving lp\n");
295  break;
297  SCIPdebugMsg(scip, " lp was unbounded\n");
298  break;
300  SCIPdebugMsg(scip, " lp was not solved\n");
301  break;
303  SCIPdebugMsg(scip, " an error occured during solving lp\n");
304  break;
307  case SCIP_LPSOLSTAT_OPTIMAL: /* should not appear because it is handled earlier */
308  default:
309  SCIPdebugMsg(scip, " received an unexpected solstat during solving lp: %d\n", lpsolstat);
310  }
311  }
312 #endif
313 
314  return SCIP_OKAY;
315 }
316 
317 /** adds the objective cutoff to the LP; must be in probing mode */
318 static
320  SCIP* scip, /**< SCIP data structure */
321  SCIP_PROPDATA* propdata /**< data of the obbt propagator */
322  )
323 {
324  SCIP_ROW* row;
325  SCIP_VAR** vars;
326  char rowname[SCIP_MAXSTRLEN];
327 
328  int nvars;
329  int i;
330 
331  assert(scip != NULL);
332  assert(SCIPinProbing(scip));
333  assert(propdata != NULL);
334  assert(propdata->cutoffrow == NULL);
335 
336  if( SCIPisInfinity(scip, SCIPgetCutoffbound(scip)) )
337  {
338  SCIPdebugMsg(scip, "no objective cutoff since there is no cutoff bound\n");
339  return SCIP_OKAY;
340  }
341 
342  SCIPdebugMsg(scip, "create objective cutoff and add it to the LP\n");
343 
344  /* get variables data */
345  SCIP_CALL( SCIPgetVarsData(scip, &vars, &nvars, NULL, NULL, NULL, NULL) );
346 
347  /* create objective cutoff row; set local flag to FALSE since primal cutoff is globally valid */
348  (void) SCIPsnprintf(rowname, SCIP_MAXSTRLEN, "obbt_objcutoff");
349  SCIP_CALL( SCIPcreateEmptyRowUnspec(scip, &row, rowname, -SCIPinfinity(scip), SCIPgetCutoffbound(scip), FALSE, FALSE, FALSE) );
350  SCIP_CALL( SCIPcacheRowExtensions(scip, row) );
351 
352  for( i = 0; i < nvars; i++ )
353  {
354  SCIP_CALL( SCIPaddVarToRow(scip, row, vars[i], SCIPvarGetObj(vars[i])) );
355  }
356  SCIP_CALL( SCIPflushRowExtensions(scip, row) );
357 
358  /* add row to the LP */
359  SCIP_CALL( SCIPaddRowProbing(scip, row) );
360 
361  propdata->cutoffrow = row;
362  assert(SCIProwIsInLP(propdata->cutoffrow));
363 
364  return SCIP_OKAY;
365 }
366 
367 /** determines, whether a variable is already locally fixed */
368 static
370  SCIP* scip, /**< SCIP data structure */
371  SCIP_VAR* var /**< variable to check */
372  )
373 {
374  return SCIPisFeasEQ(scip, SCIPvarGetLbLocal(var), SCIPvarGetUbLocal(var));
375 }
376 
377 /** sets objective to minimize or maximize a single variable */
378 static
380  SCIP* scip,
381  SCIP_PROPDATA* propdata,
383  SCIP_Real coef
384  )
385 {
386 #ifdef SCIP_DEBUG
387  SCIP_VAR** vars;
388  int nvars;
389  int counter;
390  int i;
391 #endif
393  assert( scip != NULL );
394  assert( propdata != NULL );
395  assert( bound != NULL );
396 
397  /* set the objective for bound->var */
398  if( bound->boundtype == SCIP_BOUNDTYPE_LOWER )
399  {
400  SCIP_CALL( SCIPchgVarObjProbing(scip, bound->var, coef) );
401  }
402  else
403  {
404  SCIP_CALL( SCIPchgVarObjProbing(scip, bound->var, -coef) );
405  }
406 
407 #ifdef SCIP_DEBUG
408  vars = SCIPgetVars(scip);
409  nvars = SCIPgetNVars(scip);
410  counter = 0;
411 
412  for( i = 0; i < nvars; ++i )
413  {
414  if( SCIPgetVarObjProbing(scip, vars[i]) != 0.0 )
415  ++counter;
416  }
417 
418  assert((counter == 0 && coef == 0.0) || (counter == 1 && coef != 0.0));
419 #endif
420 
421  return SCIP_OKAY;
422 }
423 
424 /** determines whether variable should be included in the right-hand side of the generalized variable bound */
425 static
427  SCIP* scip, /**< SCIP data structure */
428  SCIP_VAR* var /**< variable to check */
429  )
430 {
431  SCIP_Real redcost;
432 
433  assert(scip != NULL);
434  assert(var != NULL);
435 
437  return FALSE;
438 
439  redcost = SCIPgetVarRedcost(scip, var);
440  assert(redcost != SCIP_INVALID); /*lint !e777 */
441 
442  if( redcost == SCIP_INVALID ) /*lint !e777 */
443  return FALSE;
444 
445  if( redcost < SCIPdualfeastol(scip) && redcost > -SCIPdualfeastol(scip) )
446  return FALSE;
447 
448  return TRUE;
449 }
450 
451 /** returns number of LP iterations left (-1: no limit ) */
452 static
454  SCIP* scip, /**< SCIP data structure */
455  SCIP_Longint nolditerations, /**< iterations count at the beginning of the corresponding function */
456  SCIP_Longint itlimit /**< LP iteration limit (-1: no limit) */
457  )
458 {
459  SCIP_Longint itsleft;
460 
461  assert(scip != NULL);
462  assert(nolditerations >= 0);
463  assert(itlimit == -1 || itlimit >= 0);
464 
465  if( itlimit == -1 )
466  {
467  SCIPdebugMsg(scip, "iterations left: unlimited\n");
468  return -1;
469  }
470  else
471  {
472  itsleft = itlimit - ( SCIPgetNLPIterations(scip) - nolditerations );
473  itsleft = MAX(itsleft, 0);
474  itsleft = MIN(itsleft, INT_MAX);
475 
476  SCIPdebugMsg(scip, "iterations left: %d\n", (int) itsleft);
477  return (int) itsleft;
478  }
479 }
480 
481 /** returns the objective coefficient for a variable's bound that will be chosen during filtering */
482 static
484  SCIP* scip, /**< SCIP data structure */
485  SCIP_PROPDATA* propdata, /**< data of the obbt propagator */
486  SCIP_VAR* var, /**< variable */
487  SCIP_BOUNDTYPE boundtype /**< boundtype to be filtered? */
488  )
489 {
490  SCIP_Real lb;
491  SCIP_Real ub;
492 
493  assert(scip != NULL);
494  assert(propdata != NULL);
495  assert(var != NULL);
497  lb = SCIPvarGetLbLocal(var);
498  ub = SCIPvarGetUbLocal(var);
499 
500  /* this function should not be called for fixed variables */
501  assert(!varIsFixedLocal(scip, var));
502 
503  /* infinite bounds will not be reached */
504  if( boundtype == SCIP_BOUNDTYPE_LOWER && SCIPisInfinity(scip, -lb) )
505  return 0.0;
506  if( boundtype == SCIP_BOUNDTYPE_UPPER && SCIPisInfinity(scip, ub) )
507  return 0.0;
508 
509  if( propdata->normalize )
510  {
511  /* if the length of the domain is too large then the coefficient should be set to +/- 1.0 */
512  if( boundtype == SCIP_BOUNDTYPE_LOWER && SCIPisInfinity(scip, ub) )
513  return 1.0;
514  if( boundtype == SCIP_BOUNDTYPE_UPPER && SCIPisInfinity(scip, -lb) )
515  return -1.0;
516 
517  /* otherwise the coefficient is +/- 1.0 / ( ub - lb ) */
518  return boundtype == SCIP_BOUNDTYPE_LOWER ? 1.0 / (ub - lb) : -1.0 / (ub - lb);
519  }
520  else
521  {
522  return boundtype == SCIP_BOUNDTYPE_LOWER ? 1.0 : -1.0;
523  }
524 }
525 
526 /** creates a genvbound if the dual LP solution provides such information
527  *
528  * Consider the problem
529  *
530  * min { +/- x_i : obj * x <= z, lb <= Ax <= ub, l <= x <= u },
531  *
532  * where z is the current cutoff bound. Let (mu, nu, gamma, alpha, beta) >= 0 be the optimal solution of the dual of
533  * problem (P), where the variables correspond to the primal inequalities in the following way:
534  *
535  * Ax >= lb <-> mu
536  * -Ax >= -ub <-> nu
537  * -obj * x >= -z <-> gamma
538  * x >= l <-> alpha
539  * -x >= -u <-> beta
540  *
541  * Fixing these multipliers, by weak duality, we obtain the inequality
542  *
543  * +/- x_i >= lb*mu - ub*nu - z*gamma + l*alpha - u*beta
544  *
545  * that holds for all primal feasible points x with objective value at least z. Setting
546  *
547  * c = lb*mu - ub*nu, redcost_k = alpha_k - beta_k
548  *
549  * we obtain the inequality
550  *
551  * +/- x_i >= sum ( redcost_k * x_k ) + (-gamma) * cutoff_bound + c,
552  *
553  * that holds for all primal feasible points with objective value at least cutoff_bound. Therefore, the latter
554  * inequality can be added as a generalized variable bound.
555  */
556 static
558  SCIP* scip, /**< SCIP data structure */
559  SCIP_PROPDATA* propdata, /**< data of the obbt propagator */
560  BOUND* bound, /**< bound of x_i */
561  SCIP_Bool* found /**< pointer to store if we have found a non-trivial genvbound */
562  )
563 {
564  assert(scip != NULL);
565  assert(bound != NULL);
566  assert(propdata != NULL);
567  assert(propdata->genvboundprop != NULL);
568  assert(found != NULL);
569 
570  *found = FALSE;
571 
572  /* make sure we are in probing mode having an optimal LP solution */
573  assert(SCIPinProbing(scip));
574 
575  assert(SCIPgetLPSolstat(scip) == SCIP_LPSOLSTAT_OPTIMAL);
576 
577  /* only genvbounds created in the root node are globally valid
578  *
579  * note: depth changes to one if we use the probing mode to solve the obbt LPs
580  */
581  assert(SCIPgetDepth(scip) == 0 || (SCIPinProbing(scip) && SCIPgetDepth(scip) == 1));
582 
583  SCIPdebugMsg(scip, " try to create a genvbound for <%s>...\n", SCIPvarGetName(bound->var));
584 
585  /* a genvbound with a multiplier for x_i would not help us */
586  if( SCIPisZero(scip, SCIPgetVarRedcost(scip, bound->var)) )
587  {
588  SCIP_VAR** vars; /* global variables array */
589  SCIP_VAR** genvboundvars; /* genvbound variables array */
590 
591  SCIP_VAR* xi; /* variable x_i */
592 
593  SCIP_Real* genvboundcoefs; /* genvbound coefficients array */
594 
595  SCIP_Real gamma_dual; /* dual multiplier of objective cutoff */
596 
597  int k; /* variable for indexing global variables array */
598  int ncoefs; /* number of nonzero coefficients in genvbound */
599  int nvars; /* number of global variables */
600 
601  /* set x_i */
602  xi = bound->var;
603 
604  /* get variable data */
605  SCIP_CALL( SCIPgetVarsData(scip, &vars, &nvars, NULL, NULL, NULL, NULL) );
606 
607  /* count nonzero coefficients in genvbound */
608  ncoefs = 0;
609  for( k = 0; k < nvars; k++ )
610  {
611  if( includeVarGenVBound(scip, vars[k]) )
612  {
613  assert(vars[k] != xi);
614  ncoefs++;
615  }
616  }
617 
618  /* get dual multiplier for the objective cutoff (set to zero if there is no) */
619  if( propdata->cutoffrow == NULL )
620  {
621  gamma_dual = 0.0;
622  }
623  else
624  {
625  assert(!SCIPisInfinity(scip, SCIPgetCutoffbound(scip)));
626 
627  /* note that the objective cutoff is of the form
628  * -inf <= obj * x <= cutoff_bound
629  * but we want the positive dual multiplier!
630  */
631  gamma_dual = -SCIProwGetDualsol(propdata->cutoffrow);
632 
633  /* we need to treat gamma to be exactly 0 if it is below the dual feasibility tolerance, see #2914 */
634  if( EPSZ(gamma_dual, SCIPdualfeastol(scip)) )
635  gamma_dual = 0.0;
636  }
637 
638  /* we need at least one nonzero coefficient or a nonzero dual multiplier for the objective cutoff */
639  if( ncoefs > 0 || gamma_dual != 0.0 )
640  {
641  SCIP_Bool addgenvbound; /* if everything is fine with the redcosts and the bounds, add the genvbound */
642  SCIP_Real c; /* helper variable to calculate constant term in genvbound */
643  int idx; /* variable for indexing genvbound's coefficients array */
644 
645  /* add the bound if the bool is still TRUE after the loop */
646  addgenvbound = TRUE;
647 
648  /* there should be no coefficient for x_i */
649  assert(SCIPisZero(scip, SCIPgetVarRedcost(scip, xi)));
650 
651  /* allocate memory for storing the genvbounds right-hand side variables and coefficients */
652  SCIP_CALL( SCIPallocBufferArray(scip, &(genvboundvars), ncoefs) );
653  SCIP_CALL( SCIPallocBufferArray(scip, &(genvboundcoefs), ncoefs) );
654 
655  /* set c = lb*mu - ub*nu - z*gamma + l*alpha - u*beta */
656  c = SCIPgetLPObjval(scip);
657 
658  /* subtract ( - z * gamma ) from c */
659  c += SCIPgetCutoffbound(scip) * gamma_dual;
660 
661  /* subtract ( l*alpha - u*beta ) from c and set the coefficients of the variables */
662  idx = 0;
663  for( k = 0; k < nvars; k++ )
664  {
665  SCIP_VAR* xk;
666 
667  xk = vars[k];
668 
669  if( includeVarGenVBound(scip, xk) )
670  {
671  SCIP_Real redcost;
672 
673  redcost = SCIPgetVarRedcost(scip, xk);
674 
675  assert(redcost != SCIP_INVALID); /*lint !e777 */
676  assert(xk != xi);
677 
678  /* in this case dont add a genvbound */
679  if( ( (redcost > SCIPdualfeastol(scip)) && SCIPisInfinity(scip, -SCIPvarGetLbLocal(xk)) ) ||
680  ( (redcost < -SCIPdualfeastol(scip)) && SCIPisInfinity(scip, SCIPvarGetUbLocal(xk)) ) )
681  {
682  addgenvbound = FALSE;
683  break;
684  }
685 
686  /* store coefficients */
687  assert(idx < ncoefs);
688  genvboundvars[idx] = xk;
689  genvboundcoefs[idx] = redcost;
690  idx++;
691 
692  /* if redcost > 0, then redcost = alpha_k, otherwise redcost = - beta_k */
693  assert(redcost <= 0 || !SCIPisInfinity(scip, -SCIPvarGetLbLocal(xk)));
694  assert(redcost >= 0 || !SCIPisInfinity(scip, SCIPvarGetUbLocal(xk)));
695  c -= redcost > 0 ? redcost * SCIPvarGetLbLocal(xk) : redcost * SCIPvarGetUbLocal(xk);
696  }
697  }
698 
699  assert(!addgenvbound || idx == ncoefs);
700 
701  /* add genvbound */
702  if( addgenvbound && !SCIPisInfinity(scip, -c) )
703  {
704 #ifndef NDEBUG
705  /* check whether the activity of the LVB in the optimal solution of the LP is equal to the LP objective value */
706  SCIP_Real activity = c - gamma_dual * SCIPgetCutoffbound(scip);
707 
708  for( k = 0; k < ncoefs; ++k )
709  activity += genvboundcoefs[k] * SCIPvarGetLPSol(genvboundvars[k]);
710 
711  SCIPdebugMsg(scip, "LVB activity = %g lpobj = %g\n", activity, SCIPgetLPObjval(scip));
712  assert(EPSZ(SCIPrelDiff(activity, SCIPgetLPObjval(scip)), 18.0 * SCIPdualfeastol(scip)));
713 #endif
714 
715  SCIPdebugMsg(scip, " adding genvbound\n");
716  SCIP_CALL( SCIPgenVBoundAdd(scip, propdata->genvboundprop, genvboundvars, xi, genvboundcoefs, ncoefs,
717  gamma_dual < SCIPdualfeastol(scip) ? 0.0 : -gamma_dual, c, bound->boundtype) );
718  *found = TRUE;
719  }
720 
721  /* free arrays */
722  SCIPfreeBufferArray(scip, &genvboundcoefs);
723  SCIPfreeBufferArray(scip, &genvboundvars);
724  }
725  else
726  {
727  SCIPdebugMsg(scip, " trivial genvbound, skipping\n");
728  }
729  }
730  else
731  {
732  SCIPdebugMsg(scip, " found multiplier for <%s>: %g, skipping\n",
733  SCIPvarGetName(bound->var), SCIPgetVarRedcost(scip, bound->var));
734  }
735 
736  return SCIP_OKAY;
737 }
738 
739 /** exchange a bound which has been processed and updates the last undone and unfiltered bound index
740  * NOTE: this method has to be called after filtering or processing a bound
741  */
742 static
743 void exchangeBounds(
744  SCIP_PROPDATA* propdata, /**< propagator data */
745  int i /**< bound that was filtered or processed */
746  )
747 {
748  assert(i >= 0 && i < propdata->nbounds);
749  assert(propdata->lastidx >= 0 && propdata->lastidx < propdata->nbounds);
750 
751  /* exchange the bounds */
752  if( propdata->lastidx != i )
753  {
754  BOUND* tmp;
755 
756  tmp = propdata->bounds[i];
757  propdata->bounds[i] = propdata->bounds[propdata->lastidx];
758  propdata->bounds[propdata->lastidx] = tmp;
759  }
760 
761  propdata->lastidx -= 1;
762 }
763 
764 /** helper function to return a corner of the domain of two variables */
765 static
766 void getCorner(
767  SCIP_VAR* x, /**< first variable */
768  SCIP_VAR* y, /**< second variable */
769  CORNER corner, /**< corner */
770  SCIP_Real* px, /**< buffer to store point for x */
771  SCIP_Real* py /**< buffer to store point for y */
772  )
773 {
774  assert(x != NULL);
775  assert(y != NULL);
776  assert(px != NULL);
777  assert(py != NULL);
778 
779  switch( corner )
780  {
781  case LEFTBOTTOM:
782  *px = SCIPvarGetLbGlobal(x);
783  *py = SCIPvarGetLbGlobal(y);
784  break;
785  case RIGHTBOTTOM:
786  *px = SCIPvarGetUbGlobal(x);
787  *py = SCIPvarGetLbGlobal(y);
788  break;
789  case LEFTTOP:
790  *px = SCIPvarGetLbGlobal(x);
791  *py = SCIPvarGetUbGlobal(y);
792  break;
793  case RIGHTTOP:
794  *px = SCIPvarGetUbGlobal(x);
795  *py = SCIPvarGetUbGlobal(y);
796  break;
797  case FILTERED:
798  SCIPABORT();
799  }
800 }
801 
802 /** helper function to return the two end points of a diagonal */
803 static
804 void getCorners(
805  SCIP_VAR* x, /**< first variable */
806  SCIP_VAR* y, /**< second variable */
807  CORNER corner, /**< corner */
808  SCIP_Real* xs, /**< buffer to store start point for x */
809  SCIP_Real* ys, /**< buffer to store start point for y */
810  SCIP_Real* xt, /**< buffer to store end point for x */
811  SCIP_Real* yt /**< buffer to store end point for y */
812  )
813 {
814  assert(x != NULL);
815  assert(y != NULL);
816  assert(xs != NULL);
817  assert(ys != NULL);
818  assert(xt != NULL);
819  assert(yt != NULL);
820 
821  /* get end point */
822  getCorner(x,y, corner, xt, yt);
823 
824  /* get start point */
825  switch( corner )
826  {
827  case LEFTBOTTOM:
828  getCorner(x,y, RIGHTTOP, xs, ys);
829  break;
830  case RIGHTBOTTOM:
831  getCorner(x,y, LEFTTOP, xs, ys);
832  break;
833  case LEFTTOP:
834  getCorner(x,y, RIGHTBOTTOM, xs, ys);
835  break;
836  case RIGHTTOP:
837  getCorner(x,y, LEFTBOTTOM, xs, ys);
838  break;
839  case FILTERED:
840  SCIPABORT();
841  }
842 }
843 
844 /** returns the first variable of a bilinear bound */
845 static
847  BILINBOUND* bilinbound /**< bilinear bound */
848  )
849 {
850  assert(bilinbound->expr != NULL);
851  assert(SCIPexprGetNChildren(bilinbound->expr) == 2);
852 
853  return SCIPgetExprAuxVarNonlinear(SCIPexprGetChildren(bilinbound->expr)[0]);
854 }
855 
856 /** returns the second variable of a bilinear bound */
857 static
859  BILINBOUND* bilinbound /**< bilinear bound */
860  )
861 {
862  assert(bilinbound->expr != NULL);
863  assert(SCIPexprGetNChildren(bilinbound->expr) == 2);
864 
865  return SCIPgetExprAuxVarNonlinear(SCIPexprGetChildren(bilinbound->expr)[1]);
866 }
867 
868 /** returns the negative locks of the expression in a bilinear bound */
869 static
871  BILINBOUND* bilinbound /**< bilinear bound */
872  )
873 {
874  assert(bilinbound->expr != NULL);
875 
876  return SCIPgetExprNLocksNegNonlinear(bilinbound->expr);
877 }
878 
879 /** returns the positive locks of the expression in a bilinear bound */
880 static
882  BILINBOUND* bilinbound /**< bilinear bound */
883  )
884 {
885  assert(bilinbound->expr != NULL);
886 
887  return SCIPgetExprNLocksPosNonlinear(bilinbound->expr);
888 }
889 
890 /** computes the score of a bilinear term bound */
891 static
893  SCIP* scip, /**< SCIP data structure */
894  SCIP_RANDNUMGEN* randnumgen, /**< random number generator */
895  BILINBOUND* bilinbound /**< bilinear bound */
896  )
897 {
898  SCIP_VAR* x = bilinboundGetX(bilinbound);
899  SCIP_VAR* y = bilinboundGetY(bilinbound);
900  SCIP_Real lbx = SCIPvarGetLbLocal(x);
901  SCIP_Real ubx = SCIPvarGetUbLocal(x);
902  SCIP_Real lby = SCIPvarGetLbLocal(y);
903  SCIP_Real uby = SCIPvarGetUbLocal(y);
906  assert(scip != NULL);
907  assert(randnumgen != NULL);
908  assert(bilinbound != NULL);
909 
910  /* consider how often a bilinear term is present in the problem */
911  score = bilinboundGetLocksNeg(bilinbound) + bilinboundGetLocksPos(bilinbound);
912 
913  /* penalize small variable domains; TODO tune the factor in the logarithm, maybe add a parameter for it */
914  if( ubx - lbx < 0.5 )
915  score += log(2.0*(ubx-lbx) + SCIPepsilon(scip));
916  if( uby - lby < 0.5 )
917  score += log(2.0*(uby-lby) + SCIPepsilon(scip));
918 
919  /* consider interiority of variables in the LP solution */
921  {
922  SCIP_Real solx = SCIPvarGetLPSol(x);
923  SCIP_Real soly = SCIPvarGetLPSol(y);
924  SCIP_Real interiorityx = MIN(solx-lbx, ubx-solx) / MAX(ubx-lbx, SCIPepsilon(scip)); /*lint !e666*/
925  SCIP_Real interiorityy = MIN(soly-lby, uby-soly) / MAX(uby-lby, SCIPepsilon(scip)); /*lint !e666*/
926 
927  score += interiorityx + interiorityy;
928  }
929 
930  /* randomize score */
931  score *= 1.0 + SCIPrandomGetReal(randnumgen, -SCIPepsilon(scip), SCIPepsilon(scip));
932 
933  return score;
934 }
935 
936 /** determines whether a variable of an indicator constraint is (still) interesting
937  *
938  * A variable is interesting if it is not only part of indicator constraints or if the upper bound is greater than the given threshold.
939  */
940 static
942  SCIP* scip, /**< SCIP data structure */
943  SCIP_VAR* var, /**< variable to check */
944  int nlcount, /**< number of nonlinear constraints containing the variable
945  * or number of non-convex terms containing the variable
946  * (depends on propdata->onlynonconvexvars) */
947  int nindcount, /**< number of indicator constraints containing the variable
948  * or 0 (depends on propdata->indicators) */
949  SCIP_Real threshold /**< variables with smaller upper bound are not interesting */
950  )
951 {
952  /* if variable is only part of indicator constraints, consider current upper bound */
953  if( nlcount == 0 && nindcount > 0 )
954  {
955  if( SCIPisLE(scip, SCIPvarGetUbLocal(var), threshold) )
956  return FALSE;
957  }
958 
959  return TRUE;
960 }
961 
962 /** trying to filter some bounds using the existing LP solution */
963 static
965  SCIP* scip, /**< original SCIP data structure */
966  SCIP_PROPDATA* propdata, /**< data of the obbt propagator */
967  int* nfiltered, /**< how many bounds were filtered this round? */
968  BOUND* currbound /**< bound for which OBBT LP was solved (Note: might be NULL) */
969  )
970 {
971  int i;
972 
973  assert(scip != NULL);
974  assert(propdata != NULL);
975  assert(nfiltered != NULL);
976 
977  *nfiltered = 0;
978 
979  /* only apply filtering if an LP solution is at hand */
981  {
982  SCIPdebugMsg(scip, "can't filter using existing lp solution since it was not solved to optimality\n");
983  return SCIP_OKAY;
984  }
985 
986  /* check if a bound is tight */
987  for( i = propdata->nbounds - 1; i >= 0; --i )
988  {
989  BOUND* bound; /* shortcut for current bound */
990 
991  SCIP_Real solval; /* the variables value in the current solution */
992  SCIP_Real boundval; /* current local bound for the variable */
993 
994  bound = propdata->bounds[i];
995  if( bound->filtered || bound->done )
996  continue;
997 
998  boundval = bound->boundtype == SCIP_BOUNDTYPE_UPPER ?
999  SCIPvarGetUbLocal(bound->var) : SCIPvarGetLbLocal(bound->var);
1000  solval = SCIPvarGetLPSol(bound->var);
1001 
1002  /* bound is tight; since this holds for all fixed variables, those are filtered here automatically; if the lp solution
1003  * is infinity, then also the bound is tight */
1004  if( (bound->boundtype == SCIP_BOUNDTYPE_UPPER &&
1005  (SCIPisInfinity(scip, solval) || SCIPisFeasGE(scip, solval, boundval)))
1006  || (bound->boundtype == SCIP_BOUNDTYPE_LOWER &&
1007  (SCIPisInfinity(scip, -solval) || SCIPisFeasLE(scip, solval, boundval))) )
1008  {
1009  SCIP_BASESTAT basestat;
1010 
1011  /* mark bound as filtered */
1012  bound->filtered = TRUE;
1013  SCIPdebugMsg(scip, "trivial filtered var: %s boundval=%e solval=%e\n", SCIPvarGetName(bound->var), boundval, solval);
1014 
1015  /* get the basis status of the variable */
1016  basestat = SCIPcolGetBasisStatus(SCIPvarGetCol(bound->var));
1017 
1018  /* solve corresponding OBBT LP and try to generate a nontrivial genvbound */
1019  if( propdata->genvbdsduringfilter && currbound != NULL && basestat == SCIP_BASESTAT_BASIC )
1020  {
1021 #ifndef NDEBUG
1022  int j;
1023 #endif
1024  SCIP_Bool optimal;
1025  SCIP_Bool error;
1026 
1027  /* set objective coefficient of the bound */
1028  SCIP_CALL( SCIPchgVarObjProbing(scip, currbound->var, 0.0) );
1029  SCIP_CALL( setObjProbing(scip, propdata, bound, 1.0) );
1030 
1031 #ifndef NDEBUG
1032  for( j = 0; j < SCIPgetNVars(scip); ++j )
1033  {
1034  SCIP_VAR* var;
1035 
1036  var = SCIPgetVars(scip)[j];
1037  assert(var != NULL);
1038  assert(SCIPisZero(scip, SCIPgetVarObjProbing(scip, var)) || var == bound->var);
1039  }
1040 #endif
1041 
1042  /* solve the OBBT LP */
1043  propdata->nprobingiterations -= SCIPgetNLPIterations(scip);
1044  SCIP_CALL( solveLP(scip, -1, &error, &optimal) );
1045  propdata->nprobingiterations += SCIPgetNLPIterations(scip);
1046  assert(propdata->nprobingiterations >= 0);
1047 
1048  /* try to generate a genvbound if we have solved the OBBT LP */
1049  if( optimal && propdata->genvboundprop != NULL
1050  && (SCIPgetDepth(scip) == 0 || (SCIPinProbing(scip) && SCIPgetDepth(scip) == 1)) )
1051  {
1052  SCIP_Bool found;
1053 
1054  assert(!error);
1055  SCIP_CALL( createGenVBound(scip, propdata, bound, &found) );
1056 
1057  if( found )
1058  {
1059  propdata->ngenvboundstrivfil += 1;
1060  SCIPdebugMsg(scip, "found genvbound during trivial filtering\n");
1061  }
1062  }
1063 
1064  /* restore objective function */
1065  SCIP_CALL( setObjProbing(scip, propdata, bound, 0.0) );
1066  SCIP_CALL( setObjProbing(scip, propdata, currbound, 1.0) );
1067  }
1068 
1069  /* exchange bound i with propdata->bounds[propdata->lastidx] */
1070  if( propdata->lastidx >= 0 )
1071  exchangeBounds(propdata, i);
1072 
1073  /* increase number of filtered variables */
1074  (*nfiltered)++;
1075  }
1076  }
1077 
1078  /* try to filter bilinear bounds */
1079  for( i = propdata->lastbilinidx; i < propdata->nbilinbounds; ++i )
1080  {
1081  CORNER corners[4] = {LEFTTOP, LEFTBOTTOM, RIGHTTOP, RIGHTBOTTOM};
1082  BILINBOUND* bilinbound = propdata->bilinbounds[i];
1083  SCIP_Real solx;
1084  SCIP_Real soly;
1085  SCIPdebug(int oldfiltered;)
1086  int j;
1087 
1088  /* skip processed and filtered bounds */
1089  if( bilinbound->done || bilinbound->filtered == FILTERED ) /*lint !e641*/
1090  continue;
1091 
1092  SCIPdebug(oldfiltered = bilinbound->filtered;)
1093  solx = SCIPvarGetLPSol(bilinboundGetX(bilinbound));
1094  soly = SCIPvarGetLPSol(bilinboundGetY(bilinbound));
1095 
1096  /* check cases of unbounded solution values */
1097  if( SCIPisInfinity(scip, solx) )
1098  bilinbound->filtered = bilinbound->filtered | RIGHTTOP | RIGHTBOTTOM; /*lint !e641*/
1099  else if( SCIPisInfinity(scip, -solx) )
1100  bilinbound->filtered = bilinbound->filtered | LEFTTOP | LEFTBOTTOM; /*lint !e641*/
1101 
1102  if( SCIPisInfinity(scip, soly) )
1103  bilinbound->filtered = bilinbound->filtered | RIGHTTOP | LEFTTOP; /*lint !e641*/
1104  else if( SCIPisInfinity(scip, -soly) )
1105  bilinbound->filtered = bilinbound->filtered | RIGHTBOTTOM | LEFTBOTTOM; /*lint !e641*/
1106 
1107  /* check all corners */
1108  for( j = 0; j < 4; ++j )
1109  {
1110  SCIP_Real xt = SCIP_INVALID;
1111  SCIP_Real yt = SCIP_INVALID;
1112 
1113  getCorner(bilinboundGetX(bilinbound), bilinboundGetY(bilinbound), corners[j], &xt, &yt);
1114 
1115  if( (SCIPisInfinity(scip, REALABS(solx)) || SCIPisFeasEQ(scip, xt, solx))
1116  && (SCIPisInfinity(scip, REALABS(soly)) || SCIPisFeasEQ(scip, yt, soly)) )
1117  bilinbound->filtered = bilinbound->filtered | corners[j]; /*lint !e641*/
1118  }
1119 
1120 #ifdef SCIP_DEBUG
1121  if( oldfiltered != bilinbound->filtered )
1122  {
1123  SCIP_VAR* x = bilinboundGetX(bilinbound);
1124  SCIP_VAR* y = bilinboundGetY(bilinbound);
1125  SCIPdebugMessage("filtered corners %d for (%s,%s) = (%g,%g) in [%g,%g]x[%g,%g]\n",
1126  bilinbound->filtered - oldfiltered, SCIPvarGetName(x), SCIPvarGetName(y), solx, soly,
1128  }
1129 #endif
1130  }
1131 
1132  return SCIP_OKAY;
1133 }
1134 
1135 /** enforces one round of filtering */
1136 static
1138  SCIP* scip, /**< SCIP data structure */
1139  SCIP_PROPDATA* propdata, /**< data of the obbt propagator */
1140  int itlimit, /**< LP iteration limit (-1: no limit) */
1141  int* nfiltered, /**< how many bounds were filtered this round */
1142  SCIP_Real* objcoefs, /**< array to store the nontrivial objective coefficients */
1143  int* objcoefsinds, /**< array to store bound indices for which their corresponding variables
1144  * has a nontrivial objective coefficient */
1145  int nobjcoefs /**< number of nontrivial objective coefficients */
1146  )
1147 {
1148  SCIP_VAR** vars; /* array of the problems variables */
1149  SCIP_Bool error;
1150  SCIP_Bool optimal;
1151 
1152  int nvars; /* number of the problems variables */
1153  int i;
1154 
1155  assert(scip != NULL);
1156  assert(SCIPinProbing(scip));
1157  assert(propdata != NULL);
1158  assert(itlimit == -1 || itlimit >= 0);
1159  assert(nfiltered != NULL);
1160  assert(objcoefs != NULL);
1161  assert(objcoefsinds != NULL);
1162  assert(nobjcoefs >= 0);
1163 
1164  *nfiltered = 0;
1165 
1166  /* get variable data */
1167  SCIP_CALL( SCIPgetVarsData(scip, &vars, &nvars, NULL, NULL, NULL, NULL) );
1168 
1169  /* solve LP */
1170  propdata->nfilterlpiters -= (int) SCIPgetNLPIterations(scip);
1171  SCIP_CALL( solveLP(scip, itlimit, &error, &optimal) );
1172  propdata->nfilterlpiters += (int) SCIPgetNLPIterations(scip);
1173  assert(propdata->nfilterlpiters >= 0);
1174 
1175  if( !optimal )
1176  {
1177  SCIPdebugMsg(scip, "skipping filter round since the LP was not solved to optimality\n");
1178  return SCIP_OKAY;
1179  }
1180 
1181  assert(!error);
1182 
1183  /* check if a bound is tight */
1184  for( i = 0; i < propdata->nbounds; i++ )
1185  {
1186  BOUND* bound; /* shortcut for current bound */
1187 
1188  SCIP_Real solval; /* the variables value in the current solution */
1189  SCIP_Real boundval; /* current local bound for the variable */
1190 
1191  bound = propdata->bounds[i];
1192 
1193  /* if bound is filtered it was handled already before */
1194  if( bound->filtered )
1195  continue;
1196 
1197  boundval = bound->boundtype == SCIP_BOUNDTYPE_UPPER ?
1198  SCIPvarGetUbLocal(bound->var) : SCIPvarGetLbLocal(bound->var);
1199  solval = SCIPvarGetLPSol(bound->var);
1200 
1201  /* bound is tight */
1202  if( (bound->boundtype == SCIP_BOUNDTYPE_UPPER && SCIPisFeasGE(scip, solval, boundval))
1203  || (bound->boundtype == SCIP_BOUNDTYPE_LOWER && SCIPisFeasLE(scip, solval, boundval)) )
1204  {
1205  SCIP_Real objcoef;
1206  SCIP_BASESTAT basestat;
1207 
1208  /* mark bound as filtered */
1209  bound->filtered = TRUE;
1210 
1211  /* get the basis status of the variable */
1212  basestat = SCIPcolGetBasisStatus(SCIPvarGetCol(bound->var));
1213 
1214  /* increase number of filtered variables */
1215  (*nfiltered)++;
1216 
1217  /* solve corresponding OBBT LP and try to generate a nontrivial genvbound */
1218  if( propdata->genvbdsduringfilter && basestat == SCIP_BASESTAT_BASIC )
1219  {
1220  int j;
1221 
1222  /* set all objective coefficients to zero */
1223  for( j = 0; j < nobjcoefs; ++j )
1224  {
1225  BOUND* filterbound;
1226 
1227  filterbound = propdata->bounds[ objcoefsinds[j] ];
1228  assert(filterbound != NULL);
1229 
1230  SCIP_CALL( SCIPchgVarObjProbing(scip, filterbound->var, 0.0) );
1231  }
1232 
1233 #ifndef NDEBUG
1234  for( j = 0; j < nvars; ++j )
1235  assert(SCIPisZero(scip, SCIPgetVarObjProbing(scip, vars[j])));
1236 #endif
1237 
1238  /* set objective coefficient of the bound */
1239  SCIP_CALL( setObjProbing(scip, propdata, bound, 1.0) );
1240 
1241  /* solve the OBBT LP */
1242  propdata->nfilterlpiters -= (int) SCIPgetNLPIterations(scip);
1243  SCIP_CALL( solveLP(scip, -1, &error, &optimal) );
1244  propdata->nfilterlpiters += (int) SCIPgetNLPIterations(scip);
1245  assert(propdata->nfilterlpiters >= 0);
1246 
1247  /* try to generate a genvbound if we have solved the OBBT LP */
1248  if( optimal && propdata->genvboundprop != NULL
1249  && (SCIPgetDepth(scip) == 0 || (SCIPinProbing(scip) && SCIPgetDepth(scip) == 1)) )
1250  {
1251  SCIP_Bool found;
1252 
1253  assert(!error);
1254  SCIP_CALL( createGenVBound(scip, propdata, bound, &found) );
1255 
1256  if( found )
1257  {
1258  propdata->ngenvboundsaggrfil += 1;
1259  SCIPdebugMsg(scip, "found genvbound during aggressive filtering\n");
1260  }
1261  }
1262 
1263  /* restore objective function */
1264  for( j = 0; j < nobjcoefs; ++j )
1265  {
1266  BOUND* filterbound;
1267 
1268  filterbound = propdata->bounds[ objcoefsinds[j] ];
1269  assert(filterbound != NULL);
1270 
1271  /* NOTE: only restore coefficients of nonfiltered bounds */
1272  if( !filterbound->filtered )
1273  {
1274  assert(!SCIPisZero(scip, objcoefs[j]));
1275  SCIP_CALL( SCIPchgVarObjProbing(scip, propdata->bounds[ objcoefsinds[j] ]->var, objcoefs[j]) );
1276  }
1277  }
1278  }
1279 
1280  /* get the corresponding variable's objective coefficient */
1281  objcoef = SCIPgetVarObjProbing(scip, bound->var);
1282 
1283  /* change objective coefficient if it was set up for this bound */
1284  if( (bound->boundtype == SCIP_BOUNDTYPE_UPPER && SCIPisNegative(scip, objcoef))
1285  || (bound->boundtype == SCIP_BOUNDTYPE_LOWER && SCIPisPositive(scip, objcoef)) )
1286  {
1287  SCIP_CALL( SCIPchgVarObjProbing(scip, bound->var, 0.0) );
1288  }
1289  }
1290  }
1291 
1292  return SCIP_OKAY;
1293 }
1294 
1295 /** filter some bounds that are not improvable by solving auxiliary LPs */
1296 static
1298  SCIP* scip, /**< SCIP data structure */
1299  SCIP_PROPDATA* propdata, /**< data of the obbt propagator */
1300  SCIP_Longint itlimit /**< LP iteration limit (-1: no limit) */
1301  )
1302 {
1303  SCIP_VAR** vars;
1304  SCIP_Longint nolditerations;
1305  SCIP_Real* objcoefs; /* array to store the nontrivial objective coefficients */
1306  int* objcoefsinds; /* array to store bound indices for which the corresponding variable
1307  * has a nontrivial objective coefficient */
1308  int nobjcoefs; /* number of nontrivial objective coefficients */
1309  int nleftiterations;
1310  int i;
1311  int nfiltered;
1312  int ntotalfiltered;
1313  int nvars;
1314 
1315  assert(scip != NULL);
1316  assert(SCIPinProbing(scip));
1317  assert(propdata != NULL);
1318  assert(itlimit == -1 || itlimit >= 0);
1319 
1320  ntotalfiltered = 0;
1321  nolditerations = SCIPgetNLPIterations(scip);
1322  nleftiterations = getIterationsLeft(scip, nolditerations, itlimit);
1323 
1324  /* get variable data */
1325  SCIP_CALL( SCIPgetVarsData(scip, &vars, &nvars, NULL, NULL, NULL, NULL) );
1326 
1327  SCIPdebugMsg(scip, "start filter rounds\n");
1328 
1329  SCIP_CALL( SCIPallocBufferArray(scip, &objcoefs, propdata->nbounds) );
1330  SCIP_CALL( SCIPallocBufferArray(scip, &objcoefsinds, propdata->nbounds) );
1331  nobjcoefs = 0;
1332 
1333  /*
1334  * 1.) Filter bounds of variables that are only part of indicator constraints if they are not interesting any more
1335  */
1336  for( i = 0; i < propdata->nbounds; i++ )
1337  {
1338  if( !propdata->bounds[i]->filtered && !propdata->bounds[i]->done && propdata->bounds[i]->indicator && !propdata->bounds[i]->nonconvex )
1339  {
1340  if( !indicatorVarIsInteresting(scip, vars[i], (int)propdata->bounds[i]->nonconvex, (int)propdata->bounds[i]->indicator, propdata->indicatorthreshold) )
1341  {
1342  /* mark bound as filtered */
1343  propdata->bounds[i]->filtered = TRUE;
1344 
1345  /* increase number of filtered variables */
1346  ntotalfiltered++;
1347  }
1348  }
1349  }
1350 
1351  /*
1352  * 2.) Try first to filter lower bounds of interesting variables, whose bounds are not already filtered
1353  */
1354 
1355  for( i = 0; i < nvars; i++ )
1356  {
1357  SCIP_CALL( SCIPchgVarObjProbing(scip, vars[i], 0.0) );
1358  }
1359 
1360  for( i = 0; i < propdata->nbounds; i++ )
1361  {
1362  if( propdata->bounds[i]->boundtype == SCIP_BOUNDTYPE_LOWER && !propdata->bounds[i]->filtered
1363  && !propdata->bounds[i]->done )
1364  {
1365  SCIP_Real objcoef;
1366 
1367  objcoef = getFilterCoef(scip, propdata, propdata->bounds[i]->var, SCIP_BOUNDTYPE_LOWER);
1368 
1369  if( !SCIPisZero(scip, objcoef) )
1370  {
1371  SCIP_CALL( SCIPchgVarObjProbing(scip, propdata->bounds[i]->var, objcoef) );
1372 
1373  /* store nontrivial objective coefficients */
1374  objcoefs[nobjcoefs] = objcoef;
1375  objcoefsinds[nobjcoefs] = i;
1376  ++nobjcoefs;
1377  }
1378  }
1379  }
1380 
1381  do
1382  {
1383  SCIPdebugMsg(scip, "doing a lower bounds round\n");
1384  SCIP_CALL( filterRound(scip, propdata, nleftiterations, &nfiltered, objcoefs, objcoefsinds, nobjcoefs) );
1385  ntotalfiltered += nfiltered;
1386  SCIPdebugMsg(scip, "filtered %d more bounds in lower bounds round\n", nfiltered);
1387 
1388  /* update iterations left */
1389  nleftiterations = getIterationsLeft(scip, nolditerations, itlimit);
1390  }
1391  while( nfiltered >= propdata->nminfilter && ( nleftiterations == -1 || nleftiterations > 0 ) );
1392 
1393  /*
1394  * 3.) Now try to filter the remaining upper bounds of interesting variables, whose bounds are not already filtered
1395  */
1396 
1397  /* set all objective coefficients to zero */
1398  for( i = 0; i < nobjcoefs; i++ )
1399  {
1400  BOUND* bound;
1401 
1402  assert(objcoefsinds[i] >= 0 && objcoefsinds[i] < propdata->nbounds);
1403  bound = propdata->bounds[ objcoefsinds[i] ];
1404  assert(bound != NULL);
1405  SCIP_CALL( SCIPchgVarObjProbing(scip, bound->var, 0.0) );
1406  }
1407 
1408  /* reset number of nontrivial objective coefficients */
1409  nobjcoefs = 0;
1410 
1411 #ifndef NDEBUG
1412  for( i = 0; i < nvars; ++i )
1413  assert(SCIPisZero(scip, SCIPgetVarObjProbing(scip, vars[i])));
1414 #endif
1415 
1416  for( i = 0; i < propdata->nbounds; i++ )
1417  {
1418  if( propdata->bounds[i]->boundtype == SCIP_BOUNDTYPE_UPPER && !propdata->bounds[i]->filtered )
1419  {
1420  SCIP_Real objcoef;
1421 
1422  objcoef = getFilterCoef(scip, propdata, propdata->bounds[i]->var, SCIP_BOUNDTYPE_UPPER);
1423 
1424  if( !SCIPisZero(scip, objcoef) )
1425  {
1426  SCIP_CALL( SCIPchgVarObjProbing(scip, propdata->bounds[i]->var, objcoef) );
1427 
1428  /* store nontrivial objective coefficients */
1429  objcoefs[nobjcoefs] = objcoef;
1430  objcoefsinds[nobjcoefs] = i;
1431  ++nobjcoefs;
1432  }
1433  }
1434  }
1435 
1436  do
1437  {
1438  SCIPdebugMsg(scip, "doing an upper bounds round\n");
1439  SCIP_CALL( filterRound(scip, propdata, nleftiterations, &nfiltered, objcoefs, objcoefsinds, nobjcoefs) );
1440  SCIPdebugMsg(scip, "filtered %d more bounds in upper bounds round\n", nfiltered);
1441  ntotalfiltered += nfiltered;
1442  /* update iterations left */
1443  nleftiterations = getIterationsLeft(scip, nolditerations, itlimit);
1444  }
1445  while( nfiltered >= propdata->nminfilter && ( nleftiterations == -1 || nleftiterations > 0 ) );
1446 
1447  SCIPdebugMsg(scip, "filtered %d this round\n", ntotalfiltered);
1448  propdata->nfiltered += ntotalfiltered;
1449 
1450  /* free array */
1451  SCIPfreeBufferArray(scip, &objcoefsinds);
1452  SCIPfreeBufferArray(scip, &objcoefs);
1453 
1454  return SCIP_OKAY;
1455 }
1456 
1457 /** applies possible bound changes that were found */
1458 static
1460  SCIP* scip, /**< SCIP data structure */
1461  SCIP_PROPDATA* propdata, /**< data of the obbt propagator */
1462  SCIP_RESULT* result /**< result pointer */
1463  )
1464 {
1465 #ifdef SCIP_DEBUG
1466  int ntightened; /* stores the number of successful bound changes */
1467 #endif
1468  int i;
1469 
1470  assert(scip != NULL);
1471  assert(!SCIPinProbing(scip));
1472  assert(propdata != NULL);
1473  assert(result != NULL);
1474  assert(*result == SCIP_DIDNOTFIND);
1475 
1476  SCIPdebug( ntightened = 0 );
1477 
1478  for( i = 0; i < propdata->nbounds; i++ )
1479  {
1480  BOUND* bound; /* shortcut to the current bound */
1481  SCIP_Bool infeas; /* stores wether a tightening approach forced an infeasibilty */
1482  SCIP_Bool tightened; /* stores wether a tightening approach was successful */
1483 
1484  bound = propdata->bounds[i];
1485  infeas = FALSE;
1486 
1487  if( bound->found )
1488  {
1489  SCIPdebug( double oldbound = (bound->boundtype == SCIP_BOUNDTYPE_LOWER)
1490  ? SCIPvarGetLbLocal(bound->var)
1491  : SCIPvarGetUbLocal(bound->var) );
1492 
1493  if( bound->boundtype == SCIP_BOUNDTYPE_LOWER )
1494  {
1495  SCIP_CALL( SCIPtightenVarLb(scip, bound->var, bound->newval, FALSE, &infeas, &tightened) );
1496  }
1497  else
1498  {
1499  /* tighten only if new bound is small enough due to numerical reasons */
1500  if( SCIPisLE(scip, bound->newval, propdata->indicatorthreshold) )
1501  {
1502  SCIP_CALL( SCIPtightenVarUb(scip, bound->var, bound->newval, FALSE, &infeas, &tightened) );
1503  }
1504  else
1505  tightened = FALSE;
1506  }
1507 
1508  /* handle information about the success */
1509  if( infeas )
1510  {
1511  *result = SCIP_CUTOFF;
1512  SCIPdebugMsg(scip, "cut off\n");
1513  break;
1514  }
1515 
1516  if( tightened )
1517  {
1518  SCIPdebug( SCIPdebugMsg(scip, "tightended: %s old: %e new: %e\n" , SCIPvarGetName(bound->var), oldbound,
1519  bound->newval) );
1520 
1521  *result = SCIP_REDUCEDDOM;
1522  SCIPdebug( ntightened++ );
1523  }
1524  }
1525  }
1526 
1527  SCIPdebug( SCIPdebugMsg(scip, "tightened bounds: %d\n", ntightened) );
1528 
1529  return SCIP_OKAY;
1530 }
1531 
1532 /** tries to tighten a bound in probing mode */
1533 static
1535  SCIP* scip, /**< SCIP data structure */
1536  BOUND* bound, /**< bound that could be tightened */
1537  SCIP_Real newval, /**< new bound value */
1538  SCIP_Bool* tightened /**< was tightening successful? */
1539  )
1540 {
1541  SCIP_Real lb;
1542  SCIP_Real ub;
1543 
1544  assert(scip != NULL);
1545  assert(SCIPinProbing(scip));
1546  assert(bound != NULL);
1547  assert(tightened != NULL);
1548 
1549  *tightened = FALSE;
1550 
1551  /* get old bounds */
1552  lb = SCIPvarGetLbLocal(bound->var);
1553  ub = SCIPvarGetUbLocal(bound->var);
1554 
1555  if( bound->boundtype == SCIP_BOUNDTYPE_LOWER )
1556  {
1557  /* round bounds new value if variable is integral */
1558  if( SCIPvarIsIntegral(bound->var) )
1559  newval = SCIPceil(scip, newval);
1560 
1561  /* ensure that we give consistent bounds to the LP solver */
1562  if( newval > ub )
1563  newval = ub;
1564 
1565  /* tighten if really better */
1566  if( SCIPisLbBetter(scip, newval, lb, ub) )
1567  {
1568  SCIP_CALL( SCIPchgVarLbProbing(scip, bound->var, newval) );
1569  *tightened = TRUE;
1570  }
1571  }
1572  else
1573  {
1574  /* round bounds new value if variable is integral */
1575  if( SCIPvarIsIntegral(bound->var) )
1576  newval = SCIPfloor(scip, newval);
1577 
1578  /* ensure that we give consistent bounds to the LP solver */
1579  if( newval < lb )
1580  newval = lb;
1581 
1582  /* tighten if really better */
1583  if( SCIPisUbBetter(scip, newval, lb, ub) )
1584  {
1585  SCIP_CALL( SCIPchgVarUbProbing(scip, bound->var, newval) );
1586  *tightened = TRUE;
1587  }
1588  }
1589 
1590  return SCIP_OKAY;
1591 }
1592 
1593 /** comparison method for two bounds w.r.t. their scores */
1594 static
1595 SCIP_DECL_SORTPTRCOMP(compBoundsScore)
1596 {
1597  BOUND* bound1 = (BOUND*) elem1;
1598  BOUND* bound2 = (BOUND*) elem2;
1599 
1600  return bound1->score == bound2->score ? 0 : ( bound1->score > bound2->score ? 1 : -1 );
1601 }
1602 
1603 /** comparison method for two bilinear term bounds w.r.t. their scores */
1604 static
1605 SCIP_DECL_SORTPTRCOMP(compBilinboundsScore)
1606 {
1607  BILINBOUND* bound1 = (BILINBOUND*) elem1;
1608  BILINBOUND* bound2 = (BILINBOUND*) elem2;
1609 
1610  return bound1->score == bound2->score ? 0 : ( bound1->score > bound2->score ? 1 : -1 ); /*lint !e777*/
1611 }
1612 
1613 /** comparison method for two bounds w.r.t. their boundtype */
1614 static
1615 SCIP_DECL_SORTPTRCOMP(compBoundsBoundtype)
1616 {
1617  int diff;
1618  BOUND* bound1 = (BOUND*) elem1;
1619  BOUND* bound2 = (BOUND*) elem2;
1620 
1621  /* prioritize undone bounds */
1622  diff = (!bound1->done ? 1 : 0) - (!bound2->done ? 1 : 0);
1623  if( diff != 0 )
1624  return diff;
1625 
1626  /* prioritize unfiltered bounds */
1627  diff = (!bound1->filtered ? 1 : 0) - (!bound2->filtered ? 1 : 0);
1628  if( diff != 0 )
1629  return diff;
1630 
1631  diff = (bound1->boundtype == SCIP_BOUNDTYPE_LOWER ? 1 : 0) - (bound2->boundtype == SCIP_BOUNDTYPE_LOWER ? 1 : 0);
1632 
1633  if( diff == 0 )
1634  return (bound1->score == bound2->score) ? 0 : (bound1->score > bound2->score ? 1 : -1);
1635  else
1636  return diff;
1637 }
1638 
1639 /** sort the propdata->bounds array with their distance or their boundtype key */
1640 static
1642  SCIP* scip, /**< SCIP data structure */
1643  SCIP_PROPDATA* propdata /**< propagator data */
1644  )
1645 {
1646  assert(scip != NULL);
1647  assert(propdata != NULL);
1648 
1649  SCIPdebugMsg(scip, "sort bounds\n");
1650  SCIPsortDownPtr((void**) propdata->bounds, compBoundsBoundtype, propdata->nbounds);
1651 
1652  return SCIP_OKAY;
1653 }
1655 /** evaluates a bound for the current LP solution */
1656 static
1658  SCIP* scip,
1659  BOUND* bound
1660  )
1661 {
1662  assert(scip != NULL);
1663  assert(bound != NULL);
1664 
1665  if( bound->boundtype == SCIP_BOUNDTYPE_LOWER )
1666  return REALABS( SCIPvarGetLPSol(bound->var) - SCIPvarGetLbLocal(bound->var) );
1667  else
1668  return REALABS( SCIPvarGetUbLocal(bound->var) - SCIPvarGetLPSol(bound->var) );
1669 }
1671 /** returns the index of the next undone and unfiltered bound with the smallest distance */
1672 static
1673 int nextBound(
1674  SCIP* scip, /**< SCIP data structure */
1675  SCIP_PROPDATA* propdata, /**< data of the obbt propagator */
1676  SCIP_Bool convexphase /**< consider only convex variables? */
1677  )
1678 {
1679  SCIP_Real bestval;
1680  int bestidx;
1681  int k;
1682 
1683  assert(scip != NULL);
1684  assert(propdata != NULL);
1685 
1686  bestidx = -1;
1687  bestval = SCIPinfinity(scip);
1688 
1689  for( k = 0; k <= propdata->lastidx; ++k )
1690  {
1691  BOUND* tmpbound;
1692  tmpbound = propdata->bounds[k];
1693 
1694  assert(tmpbound != NULL);
1695 
1696  /* variables of indicator constraints are considered as nonconvex */
1697  if( !tmpbound->filtered && !tmpbound->done && (tmpbound->nonconvex == !convexphase || tmpbound->indicator == !convexphase) )
1698  {
1699  SCIP_Real boundval;
1700 
1701  /* return the next bound which is not done or unfiltered yet */
1702  if( propdata->orderingalgo == 0 )
1703  return k;
1704 
1705  boundval = evalBound(scip, tmpbound);
1706 
1707  /* negate boundval if we use the reverse greedy algorithm */
1708  boundval = (propdata->orderingalgo == 2) ? -1.0 * boundval : boundval;
1709 
1710  if( bestidx == -1 || boundval < bestval )
1711  {
1712  bestidx = k;
1713  bestval = boundval;
1714  }
1715  }
1716  }
1717 
1718  return bestidx; /*lint !e438*/
1719 }
1720 
1721 /** try to separate the solution of the last OBBT LP in order to learn better variable bounds; we apply additional
1722  * separation rounds as long as the routine finds better bounds; because of dual degeneracy we apply a minimum number of
1723  * separation rounds
1724  */
1725 static
1727  SCIP* scip, /**< SCIP data structure */
1728  SCIP_PROPDATA* propdata, /**< data of the obbt propagator */
1729  BOUND* currbound, /**< current bound */
1730  SCIP_Longint* nleftiterations, /**< number of left iterations (-1 for no limit) */
1731  SCIP_Bool* success /**< pointer to store if we have found a better bound */
1732  )
1733 {
1734  SCIP_Bool inroot;
1735  int i;
1736 
1737  assert(nleftiterations != NULL);
1738  assert(success != NULL);
1739  assert(SCIPinProbing(scip));
1740 
1741  *success = FALSE;
1742 
1743  /* check if we are originally in the root node */
1744  inroot = SCIPgetDepth(scip) == 1;
1745 
1746  for( i = 0; i <= propdata->sepamaxiter; ++i )
1747  {
1748  SCIPdebug( SCIP_Longint nlpiter; )
1749  SCIP_Real oldval;
1750  SCIP_Bool cutoff;
1751  SCIP_Bool delayed;
1752  SCIP_Bool error;
1753  SCIP_Bool optimal;
1754  SCIP_Bool tightened;
1755 
1756  oldval = SCIPvarGetLPSol(currbound->var);
1757 
1758  /* find and store cuts to separate the current LP solution */
1759  SCIP_CALL( SCIPseparateSol(scip, NULL, inroot, TRUE, FALSE, &delayed, &cutoff) );
1760  SCIPdebugMsg(scip, "applySeparation() - ncuts = %d\n", SCIPgetNCuts(scip));
1761 
1762  /* leave if we did not found any cut */
1763  if( SCIPgetNCuts(scip) == 0 )
1764  break;
1765 
1766  /* apply cuts and resolve LP */
1767  SCIP_CALL( SCIPapplyCutsProbing(scip, &cutoff) );
1768  assert(SCIPgetNCuts(scip) == 0);
1769  SCIPdebug( nlpiter = SCIPgetNLPIterations(scip); )
1770  SCIP_CALL( solveLP(scip, (int) *nleftiterations, &error, &optimal) );
1771  SCIPdebug( nlpiter = SCIPgetNLPIterations(scip) - nlpiter; )
1772  SCIPdebug( SCIPdebugMsg(scip, "applySeparation() - optimal=%u error=%u lpiter=%" SCIP_LONGINT_FORMAT "\n", optimal, error, nlpiter); )
1773  SCIPdebugMsg(scip, "oldval = %e newval = %e\n", oldval, SCIPvarGetLPSol(currbound->var));
1774 
1775  /* leave if we did not solve the LP to optimality or an error occured */
1776  if( error || !optimal )
1777  break;
1778 
1779  /* try to generate a genvbound */
1780  if( inroot && propdata->genvboundprop != NULL && propdata->genvbdsduringsepa )
1781  {
1782  SCIP_Bool found;
1783  SCIP_CALL( createGenVBound(scip, propdata, currbound, &found) );
1784  propdata->ngenvboundsprobing += found ? 1 : 0;
1785  }
1786 
1787  /* try to tight the variable bound */
1788  tightened = FALSE;
1789  if( !SCIPisEQ(scip, oldval, SCIPvarGetLPSol(currbound->var)) )
1790  {
1791  SCIP_CALL( tightenBoundProbing(scip, currbound, SCIPvarGetLPSol(currbound->var), &tightened) );
1792  SCIPdebugMsg(scip, "apply separation - tightened=%u oldval=%e newval=%e\n", tightened, oldval,
1793  SCIPvarGetLPSol(currbound->var));
1794 
1795  *success |= tightened;
1796  }
1797 
1798  /* leave the separation if we did not tighten the bound and proceed at least propdata->sepaminiter iterations */
1799  if( !tightened && i >= propdata->sepaminiter )
1800  break;
1801  }
1802 
1803  return SCIP_OKAY;
1804 }
1805 
1806 /** finds new variable bounds until no iterations left or all bounds have been checked */
1807 static
1809  SCIP* scip, /**< SCIP data structure */
1810  SCIP_PROPDATA* propdata, /**< data of the obbt propagator */
1811  SCIP_Longint* nleftiterations, /**< pointer to store the number of left iterations */
1812  SCIP_Bool convexphase /**< consider only convex variables? */
1813  )
1814 {
1815  SCIP_Longint nolditerations;
1816  SCIP_Bool iterationsleft;
1817  BOUND* currbound;
1818  SCIP_Longint itlimit;
1819  int nextboundidx;
1820 
1821  assert(scip != NULL);
1822  assert(propdata != NULL);
1823  assert(nleftiterations != NULL);
1824 
1825  /* update the number of left iterations */
1826  nolditerations = SCIPgetNLPIterations(scip);
1827  itlimit = *nleftiterations;
1828  assert(*nleftiterations == getIterationsLeft(scip, nolditerations, itlimit));
1829  iterationsleft = (*nleftiterations == -1) || (*nleftiterations > 0);
1830 
1831  /* To improve the performance we sort the bound in such a way that the undone and
1832  * unfiltered bounds are at the end of propdata->bounds. We calculate and update
1833  * the position of the last unfiltered and undone bound in propdata->lastidx
1834  */
1835  if( !convexphase )
1836  {
1837  /* sort bounds */
1838  SCIP_CALL( sortBounds(scip, propdata) );
1839 
1840  /* if the first bound is filtered or done then there is no bound left */
1841  if( propdata->bounds[0]->done || propdata->bounds[0]->filtered )
1842  {
1843  SCIPdebugMsg(scip, "no unprocessed/unfiltered bound left\n");
1844  return SCIP_OKAY;
1845  }
1846 
1847  /* compute the last undone and unfiltered node */
1848  propdata->lastidx = 0;
1849  while( propdata->lastidx < propdata->nbounds - 1 && !propdata->bounds[propdata->lastidx]->done &&
1850  !propdata->bounds[propdata->lastidx]->filtered )
1851  ++propdata->lastidx;
1852 
1853  SCIPdebugMsg(scip, "lastidx = %d\n", propdata->lastidx);
1854  }
1855 
1856  /* find the first unprocessed bound */
1857  nextboundidx = nextBound(scip, propdata, convexphase);
1858 
1859  /* skip if there is no bound left */
1860  if( nextboundidx == -1 )
1861  {
1862  SCIPdebugMsg(scip, "no unprocessed/unfiltered bound left\n");
1863  return SCIP_OKAY;
1864  }
1865 
1866  currbound = propdata->bounds[nextboundidx];
1867  assert(!currbound->done && !currbound->filtered);
1868 
1869  /* main loop */
1870  while( iterationsleft && !SCIPisStopped(scip) )
1871  {
1872  SCIP_Bool optimal;
1873  SCIP_Bool error;
1874  int nfiltered;
1875 
1876  assert(currbound != NULL);
1877  assert(currbound->done == FALSE);
1878  assert(currbound->filtered == FALSE);
1879 
1880  /* do not visit currbound more than once */
1881  currbound->done = TRUE;
1882  exchangeBounds(propdata, nextboundidx);
1883 
1884  /* set objective for curr */
1885  SCIP_CALL( setObjProbing(scip, propdata, currbound, 1.0) );
1886 
1887  SCIPdebugMsg(scip, "before solving Boundtype: %d , LB: %e , UB: %e\n",
1888  currbound->boundtype == SCIP_BOUNDTYPE_LOWER, SCIPvarGetLbLocal(currbound->var),
1889  SCIPvarGetUbLocal(currbound->var) );
1890  SCIPdebugMsg(scip, "before solving var <%s>, LP value: %f\n",
1891  SCIPvarGetName(currbound->var), SCIPvarGetLPSol(currbound->var));
1892 
1893  SCIPdebugMsg(scip, "probing iterations before solve: %lld \n", SCIPgetNLPIterations(scip));
1894 
1895  propdata->nprobingiterations -= SCIPgetNLPIterations(scip);
1896 
1897  /* now solve the LP */
1898  SCIP_CALL( solveLP(scip, (int) *nleftiterations, &error, &optimal) );
1899 
1900  propdata->nprobingiterations += SCIPgetNLPIterations(scip);
1901  propdata->nsolvedbounds++;
1902 
1903  SCIPdebugMsg(scip, "probing iterations after solve: %lld \n", SCIPgetNLPIterations(scip));
1904  SCIPdebugMsg(scip, "OPT: %u ERROR: %u\n" , optimal, error);
1905  SCIPdebugMsg(scip, "after solving Boundtype: %d , LB: %e , UB: %e\n",
1906  currbound->boundtype == SCIP_BOUNDTYPE_LOWER, SCIPvarGetLbLocal(currbound->var),
1907  SCIPvarGetUbLocal(currbound->var) );
1908  SCIPdebugMsg(scip, "after solving var <%s>, LP value: %f\n",
1909  SCIPvarGetName(currbound->var), SCIPvarGetLPSol(currbound->var));
1910 
1911  /* update nleftiterations */
1912  *nleftiterations = getIterationsLeft(scip, nolditerations, itlimit);
1913  iterationsleft = (*nleftiterations == -1) || (*nleftiterations > 0);
1914 
1915  if( error )
1916  {
1917  SCIPdebugMsg(scip, "ERROR during LP solving\n");
1918 
1919  /* set the objective of currbound to zero to null the whole objective; otherwise the objective is wrong when
1920  * we call findNewBounds() for the convex phase
1921  */
1922  SCIP_CALL( SCIPchgVarObjProbing(scip, currbound->var, 0.0) );
1923 
1924  return SCIP_OKAY;
1925  }
1926 
1927  if( optimal )
1928  {
1929  SCIP_Bool success;
1930 
1931  currbound->newval = SCIPvarGetLPSol(currbound->var);
1932  currbound->found = TRUE;
1933 
1934  /* in root node we may want to create a genvbound (independent of tightening success) */
1935  if( (SCIPgetDepth(scip) == 0 || (SCIPinProbing(scip) && SCIPgetDepth(scip) == 1))
1936  && propdata->genvboundprop != NULL )
1937  {
1938  SCIP_Bool found;
1939 
1940  SCIP_CALL( createGenVBound(scip, propdata, currbound, &found) );
1941 
1942  if( found )
1943  propdata->ngenvboundsprobing += 1;
1944  }
1945 
1946  /* try to tighten bound in probing mode */
1947  success = FALSE;
1948  if( propdata->tightintboundsprobing && SCIPvarIsIntegral(currbound->var) )
1949  {
1950  SCIPdebugMsg(scip, "tightening bound %s = %e bounds: [%e, %e]\n", SCIPvarGetName(currbound->var),
1951  currbound->newval, SCIPvarGetLbLocal(currbound->var), SCIPvarGetUbLocal(currbound->var) );
1952  SCIP_CALL( tightenBoundProbing(scip, currbound, currbound->newval, &success) );
1953  SCIPdebugMsg(scip, "tightening bound %s\n", success ? "successful" : "not successful");
1954  }
1955  else if( propdata->tightcontboundsprobing && !SCIPvarIsIntegral(currbound->var) )
1956  {
1957  SCIPdebugMsg(scip, "tightening bound %s = %e bounds: [%e, %e]\n", SCIPvarGetName(currbound->var),
1958  currbound->newval, SCIPvarGetLbLocal(currbound->var), SCIPvarGetUbLocal(currbound->var) );
1959  SCIP_CALL( tightenBoundProbing(scip, currbound, currbound->newval, &success) );
1960  SCIPdebugMsg(scip, "tightening bound %s\n", success ? "successful" : "not successful");
1961  }
1962 
1963  /* separate current OBBT LP solution */
1964  if( iterationsleft && propdata->separatesol )
1965  {
1966  propdata->nprobingiterations -= SCIPgetNLPIterations(scip);
1967  SCIP_CALL( applySeparation(scip, propdata, currbound, nleftiterations, &success) );
1968  propdata->nprobingiterations += SCIPgetNLPIterations(scip);
1969 
1970  /* remember best solution value after solving additional separations LPs */
1971  if( success )
1972  {
1973 #ifndef NDEBUG
1974  SCIP_Real newval = SCIPvarGetLPSol(currbound->var);
1975 
1976  /* round new bound if the variable is integral */
1977  if( SCIPvarIsIntegral(currbound->var) )
1978  newval = currbound->boundtype == SCIP_BOUNDTYPE_LOWER ?
1979  SCIPceil(scip, newval) : SCIPfloor(scip, newval);
1980 
1981  assert((currbound->boundtype == SCIP_BOUNDTYPE_LOWER &&
1982  SCIPisGT(scip, newval, currbound->newval))
1983  || (currbound->boundtype == SCIP_BOUNDTYPE_UPPER &&
1984  SCIPisLT(scip, newval, currbound->newval)));
1985 #endif
1986 
1987  currbound->newval = SCIPvarGetLPSol(currbound->var);
1988  }
1989  }
1990 
1991  /* filter bound candidates by using the current LP solution */
1992  if( propdata->applytrivialfilter )
1993  {
1994  SCIP_CALL( filterExistingLP(scip, propdata, &nfiltered, currbound) );
1995  SCIPdebugMsg(scip, "filtered %d bounds via inspecting present LP solution\n", nfiltered);
1996  propdata->ntrivialfiltered += nfiltered;
1997  }
1998 
1999  propdata->propagatecounter += success ? 1 : 0;
2000 
2001  /* propagate if we have found enough bound tightenings */
2002  if( propdata->propagatefreq != 0 && propdata->propagatecounter >= propdata->propagatefreq )
2003  {
2004  SCIP_Longint ndomredsfound;
2005  SCIP_Bool cutoff;
2006 
2007  SCIP_CALL( SCIPpropagateProbing(scip, 0, &cutoff, &ndomredsfound) );
2008  SCIPdebugMsg(scip, "propagation - cutoff %u ndomreds %" SCIP_LONGINT_FORMAT "\n", cutoff, ndomredsfound);
2009 
2010  propdata->npropagatedomreds += ndomredsfound;
2011  propdata->propagatecounter = 0;
2012  }
2013  }
2014 
2015  /* set objective to zero */
2016  SCIP_CALL( setObjProbing(scip, propdata, currbound, 0.0) );
2017 
2018  /* find the first unprocessed bound */
2019  nextboundidx = nextBound(scip, propdata, convexphase);
2020 
2021  /* check if there is no unprocessed and unfiltered node left */
2022  if( nextboundidx == -1 )
2023  {
2024  SCIPdebugMsg(scip, "NO unvisited/unfiltered bound left!\n");
2025  break;
2026  }
2027 
2028  currbound = propdata->bounds[nextboundidx];
2029  assert(!currbound->done && !currbound->filtered);
2030  }
2031 
2032  if( iterationsleft )
2033  {
2034  SCIPdebugMsg(scip, "still iterations left: %" SCIP_LONGINT_FORMAT "\n", *nleftiterations);
2035  }
2036  else
2037  {
2038  SCIPdebugMsg(scip, "no iterations left\n");
2039  }
2040 
2041  return SCIP_OKAY;
2042 }
2043 
2044 
2045 /** main function of obbt */
2046 static
2048  SCIP* scip, /**< SCIP data structure */
2049  SCIP_PROPDATA* propdata, /**< data of the obbt propagator */
2050  SCIP_Longint itlimit, /**< LP iteration limit (-1: no limit) */
2051  SCIP_RESULT* result /**< result pointer */
2052  )
2053 {
2054  SCIP_VAR** vars;
2055  SCIP_Real* oldlbs;
2056  SCIP_Real* oldubs;
2057  SCIP_Longint lastnpropagatedomreds;
2058  SCIP_Longint nleftiterations;
2059  SCIP_Real oldconditionlimit;
2060  SCIP_Real oldboundstreps;
2061  SCIP_Real olddualfeastol;
2062  SCIP_Bool hasconditionlimit;
2063  SCIP_Bool continuenode;
2064  SCIP_Bool boundleft;
2065  int oldpolishing;
2066  int nfiltered;
2067  int nvars;
2068  int i;
2069 
2070  assert(scip != NULL);
2071  assert(propdata != NULL);
2072  assert(itlimit == -1 || itlimit >= 0);
2073 
2074  SCIPdebugMsg(scip, "apply obbt\n");
2075 
2076  oldlbs = NULL;
2077  oldubs = NULL;
2078  lastnpropagatedomreds = propdata->npropagatedomreds;
2079  nleftiterations = itlimit;
2080  continuenode = SCIPnodeGetNumber(SCIPgetCurrentNode(scip)) == propdata->lastnode;
2081  propdata->lastidx = -1;
2082  boundleft = FALSE;
2083  *result = SCIP_DIDNOTFIND;
2084 
2085  /* store old variable bounds if we use propagation during obbt */
2086  if( propdata->propagatefreq > 0 )
2087  {
2088  SCIP_CALL( SCIPallocBufferArray(scip, &oldlbs, propdata->nbounds) );
2089  SCIP_CALL( SCIPallocBufferArray(scip, &oldubs, propdata->nbounds) );
2090  }
2091 
2092  /* reset bound data structure flags; fixed variables are marked as filtered */
2093  for( i = 0; i < propdata->nbounds; i++ )
2094  {
2095  BOUND* bound = propdata->bounds[i];
2096  bound->found = FALSE;
2097 
2098  /* store old variable bounds */
2099  if( oldlbs != NULL && oldubs != NULL )
2100  {
2101  oldlbs[bound->index] = SCIPvarGetLbLocal(bound->var);
2102  oldubs[bound->index] = SCIPvarGetUbLocal(bound->var);
2103  }
2104 
2105  /* reset 'done' and 'filtered' flag in a new B&B node */
2106  if( !continuenode )
2107  {
2108  bound->done = FALSE;
2109  bound->filtered = FALSE;
2110  }
2111 
2112  /* mark fixed variables as filtered */
2113  bound->filtered |= varIsFixedLocal(scip, bound->var);
2114 
2115  /* check for an unprocessed bound */
2116  if( !bound->filtered && !bound->done )
2117  boundleft = TRUE;
2118  }
2119 
2120  /* no bound left to check */
2121  if( !boundleft )
2122  goto TERMINATE;
2123 
2124  /* filter variables via inspecting present LP solution */
2125  if( propdata->applytrivialfilter && !continuenode )
2126  {
2127  SCIP_CALL( filterExistingLP(scip, propdata, &nfiltered, NULL) );
2128  SCIPdebugMsg(scip, "filtered %d bounds via inspecting present LP solution\n", nfiltered);
2129  propdata->ntrivialfiltered += nfiltered;
2130  }
2131 
2132  /* store old dualfeasibletol */
2133  olddualfeastol = SCIPdualfeastol(scip);
2134 
2135  /* start probing */
2136  SCIP_CALL( SCIPstartProbing(scip) );
2137  SCIPdebugMsg(scip, "start probing\n");
2138 
2139  /* tighten dual feastol */
2140  if( propdata->dualfeastol < olddualfeastol )
2141  {
2142  SCIP_CALL( SCIPchgDualfeastol(scip, propdata->dualfeastol) );
2143  }
2144 
2145  /* tighten condition limit */
2146  hasconditionlimit = (SCIPgetRealParam(scip, "lp/conditionlimit", &oldconditionlimit) == SCIP_OKAY);
2147  if( !hasconditionlimit )
2148  {
2149  SCIPwarningMessage(scip, "obbt propagator could not set condition limit in LP solver - running without\n");
2150  }
2151  else if( propdata->conditionlimit > 0.0 && (oldconditionlimit < 0.0 || propdata->conditionlimit < oldconditionlimit) )
2152  {
2153  SCIP_CALL( SCIPsetRealParam(scip, "lp/conditionlimit", propdata->conditionlimit) );
2154  }
2155 
2156  /* tighten relative bound improvement limit */
2157  SCIP_CALL( SCIPgetRealParam(scip, "numerics/boundstreps", &oldboundstreps) );
2158  if( !SCIPisEQ(scip, oldboundstreps, propdata->boundstreps) )
2159  {
2160  SCIP_CALL( SCIPsetRealParam(scip, "numerics/boundstreps", propdata->boundstreps) );
2161  }
2162 
2163  /* add objective cutoff */
2164  SCIP_CALL( addObjCutoff(scip, propdata) );
2165 
2166  /* deactivate LP polishing */
2167  SCIP_CALL( SCIPgetIntParam(scip, "lp/solutionpolishing", &oldpolishing) );
2168  SCIP_CALL( SCIPsetIntParam(scip, "lp/solutionpolishing", 0) );
2169 
2170  /* apply filtering */
2171  if( propdata->applyfilterrounds )
2172  {
2173  SCIP_CALL( filterBounds(scip, propdata, nleftiterations) );
2174  }
2175 
2176  /* set objective coefficients to zero */
2177  vars = SCIPgetVars(scip);
2178  nvars = SCIPgetNVars(scip);
2179  for( i = 0; i < nvars; ++i )
2180  {
2181  /* note that it is not possible to change the objective of non-column variables during probing; we have to take
2182  * care of the objective contribution of loose variables in createGenVBound()
2183  */
2184  if( SCIPvarGetObj(vars[i]) != 0.0 && SCIPvarGetStatus(vars[i]) == SCIP_VARSTATUS_COLUMN )
2185  {
2186  SCIP_CALL( SCIPchgVarObjProbing(scip, vars[i], 0.0) );
2187  }
2188  }
2189 
2190  /* find new bounds for the variables */
2191  SCIP_CALL( findNewBounds(scip, propdata, &nleftiterations, FALSE) );
2192 
2193  if( nleftiterations > 0 || itlimit < 0 )
2194  {
2195  SCIP_CALL( findNewBounds(scip, propdata, &nleftiterations, TRUE) );
2196  }
2197 
2198  /* reset dual feastol and condition limit */
2199  SCIP_CALL( SCIPchgDualfeastol(scip, olddualfeastol) );
2200  if( hasconditionlimit )
2201  {
2202  SCIP_CALL( SCIPsetRealParam(scip, "lp/conditionlimit", oldconditionlimit) );
2203  }
2204 
2205  /* update bound->newval if we have learned additional bound tightenings during SCIPpropagateProbing() */
2206  if( oldlbs != NULL && oldubs != NULL && propdata->npropagatedomreds - lastnpropagatedomreds > 0 )
2207  {
2208  assert(propdata->propagatefreq > 0);
2209  for( i = 0; i < propdata->nbounds; ++i )
2210  {
2211  BOUND* bound = propdata->bounds[i];
2212 
2213  /* it might be the case that a bound found by the additional propagation is better than the bound found after solving an OBBT
2214  * LP
2215  */
2216  if( bound->found )
2217  {
2218  if( bound->boundtype == SCIP_BOUNDTYPE_LOWER )
2219  bound->newval = MAX(bound->newval, SCIPvarGetLbLocal(bound->var)); /*lint !e666*/
2220  else
2221  bound->newval = MIN(bound->newval, SCIPvarGetUbLocal(bound->var)); /*lint !e666*/
2222  }
2223  else
2224  {
2225  SCIP_Real oldlb;
2226  SCIP_Real oldub;
2227 
2228  oldlb = oldlbs[bound->index];
2229  oldub = oldubs[bound->index];
2230 
2231  if( bound->boundtype == SCIP_BOUNDTYPE_LOWER && SCIPisLbBetter(scip, SCIPvarGetLbLocal(bound->var), oldlb, oldub) )
2232  {
2233  SCIPdebugMsg(scip, "tighter lower bound due to propagation: %d - %e -> %e\n", i, oldlb, SCIPvarGetLbLocal(bound->var));
2234  bound->newval = SCIPvarGetLbLocal(bound->var);
2235  bound->found = TRUE;
2236  }
2237 
2238  if( bound->boundtype == SCIP_BOUNDTYPE_UPPER && SCIPisUbBetter(scip, SCIPvarGetUbLocal(bound->var), oldlb, oldub) )
2239  {
2240  SCIPdebugMsg(scip, "tighter upper bound due to propagation: %d - %e -> %e\n", i, oldub, SCIPvarGetUbLocal(bound->var));
2241  bound->newval = SCIPvarGetUbLocal(bound->var);
2242  bound->found = TRUE;
2243  }
2244  }
2245  }
2246  }
2247 
2248  /* reset relative bound improvement limit */
2249  SCIP_CALL( SCIPsetRealParam(scip, "numerics/boundstreps", oldboundstreps) );
2250 
2251  /* reset original LP polishing setting */
2252  SCIP_CALL( SCIPsetIntParam(scip, "lp/solutionpolishing", oldpolishing) );
2253 
2254  /* end probing */
2255  SCIP_CALL( SCIPendProbing(scip) );
2256  SCIPdebugMsg(scip, "end probing!\n");
2257 
2258  /* release cutoff row if there is one */
2259  if( propdata->cutoffrow != NULL )
2260  {
2261  assert(!SCIProwIsInLP(propdata->cutoffrow));
2262  SCIP_CALL( SCIPreleaseRow(scip, &(propdata->cutoffrow)) );
2263  }
2264 
2265  /* apply buffered bound changes */
2266  SCIP_CALL( applyBoundChgs(scip, propdata, result) );
2267 
2268 TERMINATE:
2269  SCIPfreeBufferArrayNull(scip, &oldubs);
2270  SCIPfreeBufferArrayNull(scip, &oldlbs);
2271 
2272  return SCIP_OKAY;
2273 }
2274 
2275 /** computes a valid inequality from the current LP relaxation for a bilinear term xy only involving x and y; the
2276  * inequality is found by optimizing along the line connecting the points (xs,ys) and (xt,yt) over the currently given
2277  * linear relaxation of the problem; this optimization problem is an LP
2278  *
2279  * max lambda
2280  * s.t. Ax <= b
2281  * (x,y) = (xs,ys) + lambda ((xt,yt) - (xs,ys))
2282  * lambda in [0,1]
2283  *
2284  * which is equivalent to
2285  *
2286  * max x
2287  * s.t. (1) Ax <= b
2288  * (2) (x - xs) / (xt - xs) = (y - ys) / (yt - ys)
2289  *
2290  * Let x* be the optimal primal and (mu,theta) be the optimal dual solution of this LP. The KKT conditions imply that
2291  * the aggregation of the linear constraints mu*Ax <= mu*b can be written as
2292  *
2293  * x * (1 - theta) / (xt - xs) + y * theta / (yt - ys) = mu * Ax <= mu * b
2294  *
2295  * <=> alpha * x + beta * y <= mu * b = alpha * (x*) + beta * (y*)
2296  *
2297  * which is a valid inequality in the (x,y)-space; in order to avoid numerical difficulties when (xs,ys) is too close
2298  * to (xt,yt), we scale constraint (1) by max{1,|xt-xs|,|yt-ys|} beforehand
2299  */
2300 static
2302  SCIP* scip, /**< SCIP data structure */
2303  SCIP_VAR* x, /**< first variable */
2304  SCIP_VAR* y, /**< second variable */
2305  SCIP_Real xs, /**< x-coordinate of the first point */
2306  SCIP_Real ys, /**< y-coordinate of the first point */
2307  SCIP_Real xt, /**< x-coordinate of the second point */
2308  SCIP_Real yt, /**< y-coordinate of the second point */
2309  SCIP_Real* xcoef, /**< pointer to store the coefficient of x */
2310  SCIP_Real* ycoef, /**< pointer to store the coefficient of y */
2311  SCIP_Real* constant, /**< pointer to store the constant */
2312  SCIP_Longint iterlim, /**< iteration limit (-1: for no limit) */
2313  int* nnonzduals /**< buffer to store the number of non-zero dual multipliers except for
2314  * the auxiliary row (NULL if not needed) */
2315  )
2316 {
2317  SCIP_ROW* row;
2318  SCIP_Real signx;
2319  SCIP_Real scale;
2320  SCIP_Real side;
2321  SCIP_Bool lperror;
2322 
2323  assert(xcoef != NULL);
2324  assert(ycoef != NULL);
2325  assert(constant != NULL);
2326  assert(SCIPinProbing(scip));
2327 
2328  *xcoef = SCIP_INVALID;
2329  *ycoef = SCIP_INVALID;
2330  *constant= SCIP_INVALID;
2331  if( nnonzduals != NULL )
2332  *nnonzduals = 0;
2333 
2334  SCIPdebugMsg(scip, " solve bilinear LP for (%s,%s) from (%g,%g) to (%g,%g)\n", SCIPvarGetName(x), SCIPvarGetName(y), xs,
2335  ys, xt, yt);
2336 
2337  /* skip computations if (xs,ys) and (xt,yt) are too close to each other or contain too large values */
2338  if( SCIPisFeasEQ(scip, xs, xt) || SCIPisFeasEQ(scip, ys, yt)
2339  || SCIPisHugeValue(scip, REALABS(xs)) || SCIPisHugeValue(scip, REALABS(xt))
2340  || SCIPisHugeValue(scip, REALABS(ys)) || SCIPisHugeValue(scip, REALABS(yt)) )
2341  {
2342  SCIPdebugMsg(scip, " -> skip: bounds are too close/large\n");
2343  return SCIP_OKAY;
2344  }
2345 
2346  /* compute scaler for the additional linear constraint */
2347  scale = MIN(MAX3(1.0, REALABS(xt-xs), REALABS(yt-ys)), 100.0); /*lint !e666*/
2348 
2349  /* set objective function */
2350  signx = (xs > xt) ? 1.0 : -1.0;
2351  SCIP_CALL( SCIPchgVarObjProbing(scip, x, signx) );
2352 
2353  /* create new probing node to remove the added LP row afterwards */
2354  SCIP_CALL( SCIPnewProbingNode(scip) );
2355 
2356  /* create row */
2357  side = scale * (xs/(xt-xs) - ys/(yt-ys));
2358  SCIP_CALL( SCIPcreateEmptyRowUnspec(scip, &row, "bilinrow", side, side, FALSE, FALSE, TRUE) );
2359  SCIP_CALL( SCIPaddVarToRow(scip, row, x, scale/(xt-xs)) );
2360  SCIP_CALL( SCIPaddVarToRow(scip, row, y, -scale/(yt-ys)) );
2361  SCIP_CALL( SCIPaddRowProbing(scip, row) );
2362 
2363  /* solve probing LP */
2364 #ifdef NDEBUG
2365  {
2366  SCIP_RETCODE retstat;
2367  retstat = SCIPsolveProbingLP(scip, iterlim, &lperror, NULL);
2368  if( retstat != SCIP_OKAY )
2369  {
2370  SCIPwarningMessage(scip, "Error while solving LP in quadratic constraint handler; LP solve terminated with" \
2371  "code <%d>\n", retstat);
2372  }
2373  }
2374 #else
2375  SCIP_CALL( SCIPsolveProbingLP(scip, (int)iterlim, &lperror, NULL) ); /*lint !e712*/
2376 #endif
2377 
2378  SCIPdebugMsg(scip, " solved probing LP -> lperror=%u lpstat=%d\n", lperror, SCIPgetLPSolstat(scip));
2379 
2380  /* collect dual and primal solution entries */
2381  if( !lperror && SCIPgetLPSolstat(scip) == SCIP_LPSOLSTAT_OPTIMAL )
2382  {
2383  SCIP_Real xval = SCIPvarGetLPSol(x);
2384  SCIP_Real yval = SCIPvarGetLPSol(y);
2385  SCIP_Real mu = -SCIProwGetDualsol(row);
2386 
2387  SCIPdebugMsg(scip, " primal=(%g,%g) dual=%g\n", xval, yval, mu);
2388 
2389  /* xcoef x + ycoef y <= constant */
2390  *xcoef = -signx - (mu * scale) / (xt - xs);
2391  *ycoef = (mu * scale) / (yt - ys);
2392  *constant = (*xcoef) * xval + (*ycoef) * yval;
2393 
2394  /* xcoef x <= -ycoef y + constant */
2395  *ycoef = -(*ycoef);
2396 
2397  /* inequality is only useful when both coefficients are different from zero; normalize inequality if possible */
2398  if( !SCIPisFeasZero(scip, *xcoef) && !SCIPisFeasZero(scip, *ycoef) )
2399  {
2400  SCIP_Real val = REALABS(*xcoef);
2401  int r;
2402 
2403  *xcoef /= val;
2404  *ycoef /= val;
2405  *constant /= val;
2406 
2407  if( SCIPisZero(scip, *constant) )
2408  *constant = 0.0;
2409 
2410  if( nnonzduals != NULL )
2411  {
2412  /* count the number of non-zero dual multipliers except for the added row */
2413  for( r = 0; r < SCIPgetNLPRows(scip); ++r )
2414  {
2415  if( SCIPgetLPRows(scip)[r] != row && !SCIPisFeasZero(scip, SCIProwGetDualsol(SCIPgetLPRows(scip)[r])) )
2416  ++(*nnonzduals);
2417  }
2418  }
2419  }
2420  else
2421  {
2422  *xcoef = SCIP_INVALID;
2423  *ycoef = SCIP_INVALID;
2424  *constant = SCIP_INVALID;
2425  }
2426  }
2427 
2428  /* release row and backtrack probing node */
2429  SCIP_CALL( SCIPreleaseRow(scip, &row) );
2430  SCIP_CALL( SCIPbacktrackProbing(scip, 0) );
2431 
2432  /* reset objective function */
2433  SCIP_CALL( SCIPchgVarObjProbing(scip, x, 0.0) );
2434 
2435  return SCIP_OKAY;
2436 }
2437 
2438 /* applies obbt for finding valid inequalities for bilinear terms; function works as follows:
2439  *
2440  * 1. start probing mode
2441  * 2. add objective cutoff (if possible)
2442  * 3. set objective function to zero
2443  * 4. set feasibility, optimality, and relative bound improvement tolerances of SCIP
2444  * 5. main loop
2445  * 6. restore old tolerances
2446  *
2447  */
2448 static
2450  SCIP* scip, /**< SCIP data structure */
2451  SCIP_PROPDATA* propdata, /**< data of the obbt propagator */
2452  SCIP_Longint itlimit, /**< LP iteration limit (-1: no limit) */
2453  SCIP_RESULT* result /**< result pointer */
2454  )
2455 {
2456  SCIP_VAR** vars;
2457  SCIP_Real oldfeastol;
2458  SCIP_Bool lperror;
2459  SCIP_Longint nolditerations;
2460  SCIP_Longint nleftiterations;
2461  SCIP_CONSHDLR* conshdlr;
2462  SCIP_NLHDLR* bilinearnlhdlr;
2463  int nvars;
2464  int i;
2465 
2466  assert(scip != NULL);
2467  assert(propdata != NULL);
2468  assert(itlimit == -1 || itlimit >= 0);
2469  assert(result != NULL);
2470 
2471  if( propdata->nbilinbounds <= 0 || SCIPgetDepth(scip) != 0 || propdata->lastbilinidx >= propdata->nbilinbounds )
2472  return SCIP_OKAY;
2473 
2474  SCIPdebugMsg(scip, "call applyObbtBilinear starting from %d\n", propdata->lastbilinidx);
2475 
2476  /* find nonlinear handler for bilinear terms */
2477  conshdlr = SCIPfindConshdlr(scip, "nonlinear");
2478  bilinearnlhdlr = conshdlr != NULL ? SCIPfindNlhdlrNonlinear(conshdlr, "bilinear") : NULL;
2479 
2480  /* no nonlinear handler available -> skip */
2481  if( bilinearnlhdlr == NULL )
2482  return SCIP_OKAY;
2483 
2484  vars = SCIPgetVars(scip);
2485  nvars = SCIPgetNVars(scip);
2486 
2487  nolditerations = SCIPgetNLPIterations(scip);
2488  nleftiterations = getIterationsLeft(scip, nolditerations, itlimit);
2489  SCIPdebugMsg(scip, "iteration limit: %lld\n", nleftiterations);
2490 
2491  /* 1. start probing */
2492  SCIP_CALL( SCIPstartProbing(scip) );
2493 
2494  /* 2. add objective cutoff */
2495  SCIP_CALL( addObjCutoff(scip, propdata) );
2496 
2497  /* 3. set objective function to zero */
2498  for( i = 0; i < nvars; ++i )
2499  {
2500  SCIP_CALL( SCIPchgVarObjProbing(scip, vars[i], 0.0) );
2501  }
2502 
2503  /* 4. tighten LP feasibility tolerance to be at most feastol/10.0 */
2504  oldfeastol = SCIPchgRelaxfeastol(scip, SCIPfeastol(scip) / 10.0);
2505 
2506  /* we need to solve the probing LP before creating new probing nodes in solveBilinearLP() */
2507  SCIP_CALL( SCIPsolveProbingLP(scip, (int)nleftiterations, &lperror, NULL) );
2508 
2509  if( lperror )
2510  goto TERMINATE;
2511 
2512  /* 5. main loop */
2513  for( i = propdata->lastbilinidx; i < propdata->nbilinbounds
2514  && (nleftiterations > 0 || nleftiterations == -1)
2515  && (propdata->itlimitbilin < 0 || propdata->itlimitbilin > propdata->itusedbilin )
2516  && !SCIPisStopped(scip); ++i )
2517  {
2518  CORNER corners[4] = {LEFTBOTTOM, LEFTTOP, RIGHTTOP, RIGHTBOTTOM};
2519  BILINBOUND* bilinbound;
2520  int k;
2521 
2522  bilinbound = propdata->bilinbounds[i];
2523  assert(bilinbound != NULL);
2524 
2525  SCIPdebugMsg(scip, "process %d: %s %s done=%u filtered=%d nunderest=%d noverest=%d\n", i,
2526  SCIPvarGetName(bilinboundGetX(bilinbound)), SCIPvarGetName(bilinboundGetY(bilinbound)), bilinbound->done,
2527  bilinbound->filtered, bilinboundGetLocksNeg(bilinbound), bilinboundGetLocksPos(bilinbound));
2528 
2529  /* we already solved LPs for this bilinear term */
2530  if( bilinbound->done || bilinbound->filtered == (int)FILTERED )
2531  continue;
2532 
2533  /* iterate through all corners
2534  *
2535  * 0: (xs,ys)=(ubx,lby) (xt,yt)=(lbx,uby) -> underestimate
2536  * 1: (xs,ys)=(ubx,uby) (xt,yt)=(lbx,lby) -> overestimate
2537  * 2: (xs,ys)=(lbx,uby) (xt,yt)=(ubx,lby) -> underestimate
2538  * 3: (xs,ys)=(lbx,lby) (xt,yt)=(ubx,uby) -> overestimate
2539  */
2540  for( k = 0; k < 4; ++k )
2541  {
2542  CORNER corner = corners[k];
2543  SCIP_VAR* x = bilinboundGetX(bilinbound);
2544  SCIP_VAR* y = bilinboundGetY(bilinbound);
2545  SCIP_Real xcoef;
2546  SCIP_Real ycoef;
2547  SCIP_Real constant;
2548  SCIP_Real xs = SCIP_INVALID;
2549  SCIP_Real ys = SCIP_INVALID;
2550  SCIP_Real xt = SCIP_INVALID;
2551  SCIP_Real yt = SCIP_INVALID;
2552  int nnonzduals = 0;
2553 
2554  /* skip corners that lead to an under- or overestimate that is not needed */
2555  if( ((corner == LEFTTOP || corner == RIGHTBOTTOM) && bilinboundGetLocksPos(bilinbound) == 0)
2556  || ((corner == LEFTBOTTOM || corner == RIGHTTOP) && bilinboundGetLocksNeg(bilinbound) == 0) )
2557  continue;
2558 
2559  /* check whether corner has been filtered already */
2560  if( (bilinbound->filtered & corner) != 0 ) /*lint !e641*/
2561  continue;
2562 
2563  /* get corners (xs,ys) and (xt,yt) */
2564  getCorners(x, y, corner, &xs, &ys, &xt, &yt);
2565 
2566  /* skip target corner points with too large values */
2567  if( SCIPisHugeValue(scip, REALABS(xt)) || SCIPisHugeValue(scip, REALABS(yt)) )
2568  continue;
2569 
2570  /* compute inequality */
2571  propdata->itusedbilin -= SCIPgetNLPIterations(scip);
2572  SCIP_CALL( solveBilinearLP(scip, x, y, xs, ys, xt, yt, &xcoef, &ycoef, &constant, -1L,
2573  propdata->createlincons ? &nnonzduals : NULL) ); /*lint !e826*/
2574  propdata->itusedbilin += SCIPgetNLPIterations(scip);
2575 
2576  /* update number of LP iterations */
2577  nleftiterations = getIterationsLeft(scip, nolditerations, itlimit);
2578  SCIPdebugMsg(scip, "LP iterations left: %lld\n", nleftiterations);
2579 
2580  /* add inequality to quadratic constraint handler if it separates (xt,yt) */
2581  if( !SCIPisHugeValue(scip, xcoef) && !SCIPisFeasZero(scip, xcoef)
2582  && REALABS(ycoef) < 1e+3 && REALABS(ycoef) > 1e-3 /* TODO add a parameter for this */
2583  && SCIPisFeasGT(scip, (xcoef*xt - ycoef*yt - constant) / sqrt(SQR(xcoef) + SQR(ycoef) + SQR(constant)), 1e-2) )
2584  {
2585  SCIP_Bool success;
2586 
2587  /* add inequality to the associated product expression */
2588  SCIP_CALL( SCIPaddIneqBilinear(scip, bilinearnlhdlr, bilinbound->expr, xcoef, ycoef,
2589  constant, &success) );
2590 
2591  /* check whether the inequality has been accepted */
2592  if( success )
2593  {
2594  *result = SCIP_REDUCEDDOM;
2595  SCIPdebugMsg(scip, " found %g x <= %g y + %g with violation %g\n", xcoef, ycoef, constant,
2596  (xcoef*xt - ycoef*yt - constant) / sqrt(SQR(xcoef) + SQR(ycoef) + SQR(constant)));
2597 
2598  /* create a linear constraint that is only used for propagation */
2599  if( propdata->createlincons && nnonzduals > 1 )
2600  {
2601  SCIP_CONS* cons;
2602  char name[SCIP_MAXSTRLEN];
2603  SCIP_VAR* linvars[2] = {x, y};
2604  SCIP_Real linvals[2] = {xcoef, -ycoef};
2605  SCIP_Real rhs = constant;
2606 
2607  (void) SCIPsnprintf(name, SCIP_MAXSTRLEN, "bilincons_%s_%s", SCIPvarGetName(x), SCIPvarGetName(y));
2608  SCIP_CALL( SCIPcreateConsLinear(scip, &cons, name, 2, linvars, linvals, -SCIPinfinity(scip), rhs,
2610 
2611  SCIP_CALL( SCIPaddCons(scip, cons) );
2612  SCIP_CALL( SCIPreleaseCons(scip, &cons) );
2613  }
2614  }
2615  }
2616  }
2617 
2618  /* mark the bound as processed */
2619  bilinbound->done = TRUE;
2620  }
2621 
2622  /* remember last unprocessed bilinear term */
2623  propdata->lastbilinidx = i;
2624 
2625  TERMINATE:
2626  /* end probing */
2627  SCIP_CALL( SCIPendProbing(scip) );
2628 
2629  /* release cutoff row if there is one */
2630  if( propdata->cutoffrow != NULL )
2631  {
2632  assert(!SCIProwIsInLP(propdata->cutoffrow));
2633  SCIP_CALL( SCIPreleaseRow(scip, &(propdata->cutoffrow)) );
2634  }
2635 
2636  /* 6. restore old tolerance */
2637  (void) SCIPchgRelaxfeastol(scip, oldfeastol);
2638 
2639  return SCIP_OKAY;
2640 }
2641 
2642 /** computes the score of a bound */
2643 static
2644 unsigned int getScore(
2645  SCIP* scip, /**< SCIP data structure */
2646  BOUND* bound, /**< pointer of bound */
2647  int nlcount, /**< number of nonlinear constraints containing the bounds variable */
2648  int nindcount, /**< number of indicator constraints containing the bounds variable */
2649  int maxnlcount, /**< maximal number of nonlinear and indicator constraints a variable appears in */
2650  SCIP_Real smallub /**< variables with upper bound smaller than this value are counted in half iff part of indicator constraints */
2651  )
2652 {
2653  SCIP_Real counter;
2654  unsigned int score; /* score to be computed */
2655 
2656  assert(scip != NULL);
2657  assert(bound != NULL);
2658  assert(nlcount >= 0);
2659  assert(nindcount >= 0);
2660  assert(maxnlcount >= nlcount + nindcount);
2661 
2662  counter = nlcount;
2663  if( nindcount > 0 )
2664  {
2665  /* variables with small upper bound are counted in half
2666  * since the probability is high that the corresponding indicator constraint is already reformulated as bigM constraint
2667  */
2668  if( SCIPvarGetUbLocal(bound->var) <= smallub )
2669  counter += 0.5 * nindcount;
2670  else
2671  counter += nindcount;
2672  }
2673 
2674  /* score = ( nlcount * ( BASE - 1 ) / maxnlcount ) * BASE^2 + vartype * BASE + boundtype */
2675  score = (unsigned int) ( counter > 0 ? (OBBT_SCOREBASE * counter * ( OBBT_SCOREBASE - 1 )) / maxnlcount : 0 ); /*lint !e414*/
2676  switch( SCIPvarGetType(bound->var) )
2677  {
2678  case SCIP_VARTYPE_INTEGER:
2679  score += 1;
2680  break;
2681  case SCIP_VARTYPE_IMPLINT:
2682  score += 2;
2683  break;
2685  score += 3;
2686  break;
2687  case SCIP_VARTYPE_BINARY:
2688  score += 4;
2689  break;
2690  default:
2691  break;
2692  }
2693 
2694  score *= OBBT_SCOREBASE;
2695  if( bound->boundtype == SCIP_BOUNDTYPE_UPPER )
2696  score += 1;
2697 
2698  return score;
2699 }
2700 
2701 /** count how often each variable is used in a nonconvex term */
2702 static
2704  SCIP* scip, /**< SCIP data structure */
2705  unsigned int* nccounts /**< store the number each variable appears in a
2706  * non-convex term */
2707  )
2708 {
2709  SCIP_CONSHDLR* conshdlr;
2710  SCIP_HASHMAP* var2expr;
2711  int nvars;
2712  int i;
2713 
2714  assert(scip != NULL);
2715  assert(nccounts != NULL);
2717  nvars = SCIPgetNVars(scip);
2718 
2719  /* initialize nccounts to zero */
2720  BMSclearMemoryArray(nccounts, nvars);
2721 
2722  /* get nonlinear constraint handler */
2723  conshdlr = SCIPfindConshdlr(scip, "nonlinear");
2724  if( conshdlr == NULL || SCIPconshdlrGetNConss(conshdlr) == 0 )
2725  return SCIP_OKAY;
2726 
2727  var2expr = SCIPgetVarExprHashmapNonlinear(conshdlr);
2728  assert(var2expr != NULL);
2729 
2730  for( i = 0; i < SCIPgetNVars(scip); ++i )
2731  {
2732  SCIP_VAR* var;
2733 
2734  var = SCIPgetVars(scip)[i];
2735  assert(var != NULL);
2736 
2737  if( SCIPhashmapExists(var2expr, (void*) var) )
2738  {
2739  SCIP_EXPR* expr = (SCIP_EXPR*)SCIPhashmapGetImage(var2expr, (void*) var);
2740  assert(expr != NULL);
2741  assert(SCIPisExprVar(scip, expr));
2742 
2744  }
2745  }
2746 
2747 #ifdef SCIP_DEBUG
2748  for( i = 0; i < SCIPgetNVars(scip); ++i)
2749  {
2750  SCIP_VAR* var = SCIPgetVars(scip)[i];
2751  assert(var != NULL);
2752  SCIPdebugMsg(scip, "nccounts[%s] = %u\n", SCIPvarGetName(var), nccounts[SCIPvarGetProbindex(var)]);
2753  }
2754 #endif
2755 
2756  return SCIP_OKAY;
2757 }
2758 
2759 /** computes for each variable the number of indicator constraints in which the variable appears */
2760 static
2762  SCIP* scip, /**< SCIP data structure */
2763  int* nindcount /**< array that stores in how many indicator conss each variable appears */
2764  )
2765 {
2766  SCIP_CONSHDLR* conshdlr;
2767  SCIP_CONS** indconss;
2768  int nvars;
2769  int nindconss;
2770 
2771  assert(scip != NULL);
2772  assert(nindcount != NULL);
2773 
2774  nvars = SCIPgetNVars(scip);
2775 
2776  /* initialize nindcount to zero */
2777  BMSclearMemoryArray(nindcount, nvars);
2778 
2779  /* get indicator constraint handler */
2780  conshdlr = SCIPfindConshdlr(scip, "indicator");
2781  if( conshdlr == NULL || SCIPconshdlrGetNConss(conshdlr) == 0 )
2782  return SCIP_OKAY;
2783 
2784  nindconss = SCIPconshdlrGetNConss(conshdlr);
2785  indconss = SCIPconshdlrGetConss(conshdlr);
2786 
2787  for( int i = 0; i < nindconss; ++i )
2788  {
2789  SCIP_VAR** consvars;
2790  SCIP_VAR* slackvar;
2791  SCIP_CONS* lincons;
2792  int nconsvars;
2793 
2794  lincons = SCIPgetLinearConsIndicator(indconss[i]);
2795  assert(lincons!=NULL);
2796 
2797  nconsvars = SCIPgetNVarsLinear(scip, lincons);
2798  consvars = SCIPgetVarsLinear(scip, lincons);
2799  assert(consvars != NULL);
2800  slackvar = SCIPgetSlackVarIndicator(indconss[i]);
2801 
2802  for( int v = 0; v < nconsvars; ++v )
2803  {
2804  /* we should skip the slackvariable */
2805  if( consvars[v] == slackvar )
2806  continue;
2807 
2808  /* consider only active variables */
2809  if( SCIPvarGetStatus(consvars[v]) == SCIP_VARSTATUS_COLUMN )
2810  {
2811  nindcount[SCIPvarGetProbindex(consvars[v])] += 1;
2812  }
2813  }
2814  }
2815 
2816  return SCIP_OKAY;
2817 }
2818 
2819 /** determines whether a variable is interesting */
2820 static
2822  SCIP* scip, /**< SCIP data structure */
2823  SCIP_VAR* var, /**< variable to check */
2824  int nlcount, /**< number of nonlinear constraints containing the variable
2825  * or number of non-convex terms containing the variable
2826  * (depends on propdata->onlynonconvexvars) */
2827  int nindcount /**< number of indicator constraints containing the variable
2828  * or 0 (depends on propdata->indicators) */
2829  )
2830 {
2831  assert(SCIPgetDepth(scip) == 0);
2832 
2833  return !SCIPvarIsBinary(var) && SCIPvarGetStatus(var) == SCIP_VARSTATUS_COLUMN && (nlcount > 0 || nindcount > 0)
2834  && !varIsFixedLocal(scip, var);
2835 }
2836 
2837 /** initializes interesting bounds */
2838 static
2840  SCIP* scip, /**< SCIP data structure */
2841  SCIP_PROPDATA* propdata /**< data of the obbt propagator */
2842  )
2843 {
2844  SCIP_CONSHDLR* conshdlr;
2845  SCIP_VAR** vars; /* array of the problems variables */
2846  int* nlcount; /* array that stores in how many nonlinearities each variable appears */
2847  int* nindcount; /* array that stores in how many indicator constraints each variable appears */
2848  unsigned int* nccount; /* array that stores in how many nonconvexities each variable appears */
2849  SCIP_Real maxcouplingvalue;
2850  SCIP_Real sepacouplingvalue;
2851  SCIP_Real smallub;
2853  int bdidx; /* bound index inside propdata->bounds */
2854  int maxnlcount; /* maximal number of nonlinear and indicator constraints a variable appears in */
2855  int nvars; /* number of the problems variables */
2856  int i;
2857 
2858  assert(scip != NULL);
2859  assert(propdata != NULL);
2860  assert(SCIPisNLPConstructed(scip) || propdata->indicators);
2861 
2862  SCIPdebugMsg(scip, "initialize bounds\n");
2863 
2864  /* get variable data */
2865  SCIP_CALL( SCIPgetVarsData(scip, &vars, &nvars, NULL, NULL, NULL, NULL) );
2866 
2867  SCIP_CALL( SCIPallocBufferArray(scip, &nlcount, nvars) );
2868  SCIP_CALL( SCIPallocBufferArray(scip, &nccount, nvars) );
2869  SCIP_CALL( SCIPallocBufferArray(scip, &nindcount, nvars) );
2870 
2871  /* count nonlinearities */
2872  if( SCIPisNLPConstructed(scip) )
2873  {
2874  assert(SCIPgetNNLPVars(scip) == nvars);
2875  SCIP_CALL( SCIPgetNLPVarsNonlinearity(scip, nlcount) );
2876  SCIP_CALL( getNLPVarsNonConvexity(scip, nccount) );
2877  }
2878  else
2879  {
2880  /* initialize to zero */
2881  BMSclearMemoryArray(nlcount, nvars);
2882  BMSclearMemoryArray(nccount, nvars);
2883  }
2884 
2885  /* count indicators */
2886  if( propdata->indicators )
2887  {
2888  SCIP_CALL( getNVarsIndicators(scip, nindcount) );
2889  }
2890  else
2891  {
2892  /* initialize to zero */
2893  BMSclearMemoryArray(nindcount, nvars);
2894  }
2895 
2896  maxnlcount = 0;
2897  for( i = 0; i < nvars; i++ )
2898  {
2899  if( maxnlcount < (nlcount[i] + nindcount[i]) )
2900  maxnlcount = nlcount[i] + nindcount[i];
2901  }
2902 
2903  /* allocate interesting bounds array */
2904  propdata->boundssize = 2 * nvars;
2905  SCIP_CALL( SCIPallocBlockMemoryArray(scip, &(propdata->bounds), 2 * nvars) );
2906 
2907  SCIP_CALL( SCIPgetRealParam(scip, "constraints/indicator/maxcouplingvalue", &maxcouplingvalue) );
2908  SCIP_CALL( SCIPgetRealParam(scip, "constraints/indicator/sepacouplingvalue", &sepacouplingvalue) );
2909 
2910  smallub = MIN(maxcouplingvalue, sepacouplingvalue);
2911 
2912  /* get all interesting variables and their bounds */
2913  bdidx = 0;
2914  for( i = 0; i < nvars; i++ )
2915  {
2916  if( varIsInteresting(scip, vars[i], (propdata->onlynonconvexvars ? (int)nccount[i] : nlcount[i]), (propdata->indicators ? nindcount[i] : 0))
2917  && indicatorVarIsInteresting(scip, vars[i], (propdata->onlynonconvexvars ? (int)nccount[i] : nlcount[i]), (propdata->indicators ? nindcount[i] : 0), propdata->indicatorthreshold) )
2918  {
2919  BOUND** bdaddress;
2920 
2921  /* create lower bound */
2922  bdaddress = &(propdata->bounds[bdidx]);
2923  SCIP_CALL( SCIPallocBlockMemory(scip, bdaddress) );
2924  propdata->bounds[bdidx]->boundtype = SCIP_BOUNDTYPE_LOWER;
2925  propdata->bounds[bdidx]->var = vars[i];
2926  propdata->bounds[bdidx]->found = FALSE;
2927  propdata->bounds[bdidx]->filtered = FALSE;
2928  propdata->bounds[bdidx]->newval = 0.0;
2929  propdata->bounds[bdidx]->score = getScore(scip, propdata->bounds[bdidx], nlcount[i], nindcount[i], maxnlcount, smallub);
2930  propdata->bounds[bdidx]->done = FALSE;
2931  propdata->bounds[bdidx]->nonconvex = (nccount[i] > 0);
2932  propdata->bounds[bdidx]->indicator = (nindcount[i] > 0);
2933  propdata->bounds[bdidx]->index = bdidx;
2934  bdidx++;
2935 
2936  /* create upper bound */
2937  bdaddress = &(propdata->bounds[bdidx]);
2938  SCIP_CALL( SCIPallocBlockMemory(scip, bdaddress) );
2939  propdata->bounds[bdidx]->boundtype = SCIP_BOUNDTYPE_UPPER;
2940  propdata->bounds[bdidx]->var = vars[i];
2941  propdata->bounds[bdidx]->found = FALSE;
2942  propdata->bounds[bdidx]->filtered = FALSE;
2943  propdata->bounds[bdidx]->newval = 0.0;
2944  propdata->bounds[bdidx]->score = getScore(scip, propdata->bounds[bdidx], nlcount[i], nindcount[i], maxnlcount, smallub);
2945  propdata->bounds[bdidx]->done = FALSE;
2946  propdata->bounds[bdidx]->nonconvex = (nccount[i] > 0);
2947  propdata->bounds[bdidx]->indicator = (nindcount[i] > 0);
2948  propdata->bounds[bdidx]->index = bdidx;
2949  bdidx++;
2950  }
2951  }
2952 
2953  /* set number of interesting bounds */
2954  propdata->nbounds = bdidx;
2955 
2956  conshdlr = SCIPfindConshdlr(scip, "nonlinear");
2957 
2958  /* get all product expressions from nonlinear constraint handler */
2959  if( propdata->nbounds > 0 && conshdlr != NULL && propdata->createbilinineqs )
2960  {
2961  SCIP_NLHDLR* bilinnlhdlr;
2962  SCIP_EXPR** exprs;
2963  int nexprs;
2964 
2965  /* find nonlinear handler for bilinear terms */
2966  bilinnlhdlr = SCIPfindNlhdlrNonlinear(conshdlr, "bilinear");
2967  assert(bilinnlhdlr != NULL);
2968 
2969  /* collect all bilinear product in all nonlinear constraints */
2970  exprs = SCIPgetExprsBilinear(bilinnlhdlr);
2971  nexprs = SCIPgetNExprsBilinear(bilinnlhdlr);
2972 
2973  if( nexprs > 0 )
2974  {
2975  SCIP_CALL( SCIPallocBlockMemoryArray(scip, &propdata->bilinbounds, nexprs) );
2976  propdata->bilinboundssize = nexprs;
2977  propdata->nbilinbounds = 0;
2978 
2979  /* store candidates as bilinear bounds */
2980  for( i = 0; i < nexprs; ++i )
2981  {
2982  BILINBOUND* bilinbound;
2983  SCIP_VAR* x;
2984  SCIP_VAR* y;
2985 
2986  assert(exprs[i] != NULL);
2987  assert(SCIPexprGetNChildren(exprs[i]) == 2);
2988 
2991  assert(x != NULL);
2992  assert(y != NULL);
2993  assert(x != y);
2994 
2995  /* skip almost fixed variables */
2996  if( !varIsInteresting(scip, x, 1, 0) || !varIsInteresting(scip, y, 1, 0) )
2997  continue;
2998 
2999  /* create bilinear bound */
3000  SCIP_CALL( SCIPallocBlockMemory(scip, &propdata->bilinbounds[propdata->nbilinbounds]) ); /*lint !e866*/
3001  bilinbound = propdata->bilinbounds[propdata->nbilinbounds];
3002  BMSclearMemory(bilinbound);
3003 
3004  /* store and capture expression */
3005  bilinbound->expr = exprs[i];
3006  SCIPcaptureExpr(bilinbound->expr);
3007 
3008  /* compute a descent score */
3009  bilinbound->score = bilinboundGetScore(scip, propdata->randnumgen, bilinbound);
3010 
3011  /* increase the number of bilinear bounds */
3012  ++(propdata->nbilinbounds);
3013 
3014  SCIPdebugMsg(scip, "added bilinear bound for %s %s\n", SCIPvarGetName(x), SCIPvarGetName(y));
3015  }
3016  }
3017 
3018  /* sort bounds according to decreasing score */
3019  if( propdata->nbilinbounds > 1 )
3020  {
3021  SCIPsortDownPtr((void**) propdata->bilinbounds, compBilinboundsScore, propdata->nbilinbounds);
3022  }
3023  }
3024 
3025  /* free memory for buffering nonlinearities */
3026  assert(nlcount != NULL);
3027  assert(nccount != NULL);
3028  assert(nindcount != NULL);
3029  SCIPfreeBufferArray(scip, &nindcount);
3030  SCIPfreeBufferArray(scip, &nccount);
3031  SCIPfreeBufferArray(scip, &nlcount);
3032 
3033  /* propdata->bounds array if empty */
3034  if( propdata->nbounds <= 0 )
3035  {
3036  assert(propdata->nbounds == 0);
3037  assert(propdata->boundssize >= 0 );
3038  SCIPfreeBlockMemoryArray(scip, &(propdata->bounds), propdata->boundssize);
3039  }
3040 
3041  SCIPdebugMsg(scip, "problem has %d/%d interesting bounds\n", propdata->nbounds, 2 * nvars);
3042 
3043  if( propdata->nbounds > 0 )
3044  {
3045  /* sort bounds according to decreasing score; although this initial order will be overruled by the distance
3046  * criterion later, gives a more well-defined starting situation for OBBT and might help to reduce solver
3047  * variability
3048  */
3049  SCIPsortDownPtr((void**) propdata->bounds, compBoundsScore, propdata->nbounds);
3050  }
3051 
3052  return SCIP_OKAY;
3053 }
3054 
3055 /*
3056  * Callback methods of propagator
3057  */
3058 
3059 /** copy method for propagator plugins (called when SCIP copies plugins)
3060  *
3061  * @note The UG framework assumes that all default plug-ins of SCIP implement a copy callback. We check
3062  * SCIPgetSubscipDepth() in PROPEXEC to prevent the propagator to run in a sub-SCIP.
3063  */
3064 static
3065 SCIP_DECL_PROPCOPY(propCopyObbt)
3066 { /*lint --e{715}*/
3067  assert(scip != NULL);
3068  assert(prop != NULL);
3069  assert(strcmp(SCIPpropGetName(prop), PROP_NAME) == 0);
3070 
3071  /* call inclusion method of constraint handler */
3072  SCIP_CALL( SCIPincludePropObbt(scip) );
3073 
3074  return SCIP_OKAY;
3075 }
3076 
3077 /** solving process initialization method of propagator (called when branch and bound process is about to begin) */
3078 static
3079 SCIP_DECL_PROPINITSOL(propInitsolObbt)
3080 { /*lint --e{715}*/
3081  SCIP_PROPDATA* propdata;
3082 
3083  assert(scip != NULL);
3084  assert(prop != NULL);
3085  assert(strcmp(SCIPpropGetName(prop), PROP_NAME) == 0);
3086 
3087  /* get propagator data */
3088  propdata = SCIPpropGetData(prop);
3089  assert(propdata != NULL);
3090 
3091  propdata->bounds = NULL;
3092  propdata->nbounds = -1;
3093  propdata->boundssize = 0;
3094  propdata->cutoffrow = NULL;
3095  propdata->lastnode = -1;
3096 
3097  /* if genvbounds propagator is not available, we cannot create genvbounds */
3098  propdata->genvboundprop = propdata->creategenvbounds ? SCIPfindProp(scip, GENVBOUND_PROP_NAME) : NULL;
3099 
3100  SCIPdebugMsg(scip, "creating genvbounds: %s\n", propdata->genvboundprop != NULL ? "true" : "false");
3101 
3102  /* create random number generator */
3103  SCIP_CALL( SCIPcreateRandom(scip, &propdata->randnumgen, DEFAULT_RANDSEED, TRUE) );
3104 
3105  return SCIP_OKAY;
3106 }
3107 
3108 /** execution method of propagator */
3109 static
3110 SCIP_DECL_PROPEXEC(propExecObbt)
3111 { /*lint --e{715}*/
3112  SCIP_PROPDATA* propdata;
3113  SCIP_Longint itlimit;
3114 
3115  assert(scip != NULL);
3116  assert(prop != NULL);
3117  assert(strcmp(SCIPpropGetName(prop), PROP_NAME) == 0);
3118 
3119  *result = SCIP_DIDNOTRUN;
3120 
3121  /* do not run in: presolving, repropagation, probing mode, if no objective propagation is allowed */
3123  return SCIP_OKAY;
3124 
3125  /* do not run propagator in a sub-SCIP */
3126  if( SCIPgetSubscipDepth(scip) > 0 )
3127  return SCIP_OKAY;
3128 
3129  /* get propagator data */
3130  propdata = SCIPpropGetData(prop);
3131  assert(propdata != NULL);
3132 
3133  /* only run for nonlinear problems, i.e., if NLP is constructed
3134  * or if indicator constraints exists and should be considered
3135  */
3136  if( !SCIPisNLPConstructed(scip)
3137  && (!propdata->indicators || SCIPfindConshdlr(scip, "indicator") == NULL || SCIPconshdlrGetNConss(SCIPfindConshdlr(scip, "indicator")) == 0) )
3138  {
3139  SCIPdebugMsg(scip, "NLP not constructed and no indicator constraints available, skipping obbt\n");
3140  return SCIP_OKAY;
3141  }
3142 
3143  /* only run if LP all columns are in the LP, i.e., the LP is a relaxation; e.g., do not run if pricers are active
3144  * since pricing is not performed in probing mode
3145  */
3146  if( !SCIPallColsInLP(scip) )
3147  {
3148  SCIPdebugMsg(scip, "not all columns in LP, skipping obbt\n");
3149  return SCIP_OKAY;
3150  }
3151 
3152  /* ensure that bounds are initialized */
3153  if( propdata->nbounds == -1 )
3154  {
3155  /* bounds must be initialized at root node */
3156  if( SCIPgetDepth(scip) == 0 )
3157  {
3158  SCIP_CALL( initBounds(scip, propdata) );
3159  }
3160  else
3161  {
3162  assert(!SCIPinProbing(scip));
3163  return SCIP_OKAY;
3164  }
3165  }
3166  assert(propdata->nbounds >= 0);
3167 
3168  /* do not run if there are no interesting bounds */
3169  /**@todo disable */
3170  if( propdata->nbounds <= 0 )
3171  {
3172  SCIPdebugMsg(scip, "there are no interesting bounds\n");
3173  return SCIP_OKAY;
3174  }
3175 
3176  /* only run once in a node != root */
3177  if( SCIPgetDepth(scip) > 0 && SCIPnodeGetNumber(SCIPgetCurrentNode(scip)) == propdata->lastnode )
3178  {
3179  return SCIP_OKAY;
3180  }
3181 
3182  SCIPdebugMsg(scip, "applying obbt for problem <%s> at depth %d\n", SCIPgetProbName(scip), SCIPgetDepth(scip));
3183 
3184  /* without an optimal LP solution we don't want to run; this may be because propagators with higher priority have
3185  * already found reductions or numerical troubles occured during LP solving
3186  */
3188  {
3189  SCIPdebugMsg(scip, "aborting since no optimal LP solution is at hand\n");
3190  return SCIP_OKAY;
3191  }
3192 
3193  /* compute iteration limit */
3194  if( propdata->itlimitfactor > 0.0 )
3195  itlimit = (SCIP_Longint) MAX(propdata->itlimitfactor * SCIPgetNRootLPIterations(scip),
3196  propdata->minitlimit); /*lint !e666*/
3197  else
3198  itlimit = -1;
3199 
3200  /* apply obbt */
3201  SCIP_CALL( applyObbt(scip, propdata, itlimit, result) );
3202  assert(*result != SCIP_DIDNOTRUN);
3203 
3204  /* compute globally inequalities for bilinear terms */
3205  if( propdata->createbilinineqs )
3206  {
3207  /* set LP iteration limit */
3208  if( propdata->itlimitbilin == 0L )
3209  {
3210  /* no iteration limit if itlimit < 0 or itlimitfactorbilin < 0 */
3211  propdata->itlimitbilin = (itlimit < 0 || propdata->itlimitfactorbilin < 0)
3212  ? -1L : (SCIP_Longint)(itlimit * propdata->itlimitfactorbilin);
3213  }
3214 
3215  SCIP_CALL( applyObbtBilinear(scip, propdata, itlimit, result) );
3216  }
3217 
3218  /* set current node as last node */
3219  propdata->lastnode = SCIPnodeGetNumber(SCIPgetCurrentNode(scip));
3220 
3221  return SCIP_OKAY;
3222 }
3223 
3224 /** propagation conflict resolving method of propagator */
3225 static
3226 SCIP_DECL_PROPRESPROP(propRespropObbt)
3227 { /*lint --e{715}*/
3228  *result = SCIP_DIDNOTFIND;
3229 
3230  return SCIP_OKAY;
3231 }
3232 
3233 /** solving process deinitialization method of propagator (called before branch and bound process data is freed) */
3234 static
3235 SCIP_DECL_PROPEXITSOL(propExitsolObbt)
3236 { /*lint --e{715}*/
3237  SCIP_PROPDATA* propdata;
3238  int i;
3240  assert(scip != NULL);
3241  assert(prop != NULL);
3242  assert(strcmp(SCIPpropGetName(prop), PROP_NAME) == 0);
3243 
3244  /* get propagator data */
3245  propdata = SCIPpropGetData(prop);
3246  assert(propdata != NULL);
3247 
3248  /* free random number generator */
3249  SCIPfreeRandom(scip, &propdata->randnumgen);
3250  propdata->randnumgen = NULL;
3251 
3252  /* note that because we reset filtered flags to false at each call to obbt, the same bound may be filtered multiple
3253  * times
3254  */
3255  SCIPstatisticMessage("DIVE-LP: %" SCIP_LONGINT_FORMAT " NFILTERED: %d NTRIVIALFILTERED: %d NSOLVED: %d "
3256  "FILTER-LP: %" SCIP_LONGINT_FORMAT " NGENVB(dive): %d NGENVB(aggr.): %d NGENVB(triv.) %d\n",
3257  propdata->nprobingiterations, propdata->nfiltered, propdata->ntrivialfiltered, propdata->nsolvedbounds,
3258  propdata->nfilterlpiters, propdata->ngenvboundsprobing, propdata->ngenvboundsaggrfil, propdata->ngenvboundstrivfil);
3259 
3260  /* free bilinear bounds */
3261  if( propdata->bilinboundssize > 0 )
3262  {
3263  for( i = propdata->nbilinbounds - 1; i >= 0; --i )
3264  {
3265  assert(propdata->bilinbounds[i] != NULL);
3266  assert(propdata->bilinbounds[i]->expr != NULL);
3267 
3268  /* release expression */
3269  SCIP_CALL( SCIPreleaseExpr(scip, &propdata->bilinbounds[i]->expr) );
3270 
3271  SCIPfreeBlockMemory(scip, &propdata->bilinbounds[i]); /*lint !e866*/
3272  }
3273  SCIPfreeBlockMemoryArray(scip, &propdata->bilinbounds, propdata->bilinboundssize);
3274  propdata->bilinboundssize = 0;
3275  propdata->nbilinbounds = 0;
3276  }
3277 
3278  /* free memory allocated for the bounds */
3279  if( propdata->nbounds > 0 )
3280  {
3281  /* free bounds */
3282  for( i = propdata->nbounds - 1; i >= 0; i-- )
3283  {
3284  SCIPfreeBlockMemory(scip, &(propdata->bounds[i])); /*lint !e866*/
3285  }
3286  SCIPfreeBlockMemoryArray(scip, &(propdata->bounds), propdata->boundssize);
3287  }
3288 
3289  propdata->nbounds = -1;
3290  propdata->itlimitbilin = 0;
3291  propdata->itusedbilin = 0;
3292 
3293  return SCIP_OKAY;
3294 }
3295 
3296 /** destructor of propagator to free user data (called when SCIP is exiting) */
3297 static
3298 SCIP_DECL_PROPFREE(propFreeObbt)
3299 { /*lint --e{715}*/
3300  SCIP_PROPDATA* propdata;
3301 
3302  assert(strcmp(SCIPpropGetName(prop), PROP_NAME) == 0);
3303 
3304  /* free propagator data */
3305  propdata = SCIPpropGetData(prop);
3306  assert(propdata != NULL);
3307 
3308  SCIPfreeBlockMemory(scip, &propdata);
3309 
3310  SCIPpropSetData(prop, NULL);
3312  return SCIP_OKAY;
3313 }
3314 
3315 /*
3316  * propagator specific interface methods
3317  */
3318 
3319 /** creates the obbt propagator and includes it in SCIP */
3321  SCIP* scip /**< SCIP data structure */
3322  )
3323 {
3324  SCIP_PROPDATA* propdata;
3325  SCIP_PROP* prop;
3326 
3327  /* create obbt propagator data */
3328  SCIP_CALL( SCIPallocBlockMemory(scip, &propdata) );
3329  BMSclearMemory(propdata);
3330 
3331  /* initialize statistic variables */
3332  propdata->nprobingiterations = 0;
3333  propdata->nfiltered = 0;
3334  propdata->ntrivialfiltered = 0;
3335  propdata->nsolvedbounds = 0;
3336  propdata->ngenvboundsprobing = 0;
3337  propdata->ngenvboundsaggrfil = 0;
3338  propdata->ngenvboundstrivfil = 0;
3339  propdata->nfilterlpiters = 0;
3340  propdata->lastidx = -1;
3341  propdata->propagatecounter = 0;
3342  propdata->npropagatedomreds = 0;
3343 
3344  /* include propagator */
3346  propExecObbt, propdata) );
3347 
3348  SCIP_CALL( SCIPsetPropCopy(scip, prop, propCopyObbt) );
3349  SCIP_CALL( SCIPsetPropFree(scip, prop, propFreeObbt) );
3350  SCIP_CALL( SCIPsetPropExitsol(scip, prop, propExitsolObbt) );
3351  SCIP_CALL( SCIPsetPropInitsol(scip, prop, propInitsolObbt) );
3352  SCIP_CALL( SCIPsetPropResprop(scip, prop, propRespropObbt) );
3353 
3354  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/creategenvbounds",
3355  "should obbt try to provide genvbounds if possible?",
3356  &propdata->creategenvbounds, TRUE, DEFAULT_CREATE_GENVBOUNDS, NULL, NULL) );
3357 
3358  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/normalize",
3359  "should coefficients in filtering be normalized w.r.t. the domains sizes?",
3360  &propdata->normalize, TRUE, DEFAULT_FILTERING_NORM, NULL, NULL) );
3361 
3362  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/applyfilterrounds",
3363  "try to filter bounds in so-called filter rounds by solving auxiliary LPs?",
3364  &propdata->applyfilterrounds, TRUE, DEFAULT_APPLY_FILTERROUNDS, NULL, NULL) );
3365 
3366  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/applytrivialfilter",
3367  "try to filter bounds with the LP solution after each solve?",
3368  &propdata->applytrivialfilter, TRUE, DEFAULT_APPLY_TRIVIALFITLERING, NULL, NULL) );
3369 
3370  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/genvbdsduringfilter",
3371  "should we try to generate genvbounds during trivial and aggressive filtering?",
3372  &propdata->genvbdsduringfilter, TRUE, DEFAULT_GENVBDSDURINGFILTER, NULL, NULL) );
3373 
3374  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/genvbdsduringsepa",
3375  "try to create genvbounds during separation process?",
3376  &propdata->genvbdsduringsepa, TRUE, DEFAULT_GENVBDSDURINGSEPA, NULL, NULL) );
3377 
3378  SCIP_CALL( SCIPaddIntParam(scip, "propagating/" PROP_NAME "/minfilter",
3379  "minimal number of filtered bounds to apply another filter round",
3380  &propdata->nminfilter, TRUE, DEFAULT_FILTERING_MIN, 1, INT_MAX, NULL, NULL) );
3381 
3382  SCIP_CALL( SCIPaddRealParam(scip, "propagating/" PROP_NAME "/itlimitfactor",
3383  "multiple of root node LP iterations used as total LP iteration limit for obbt (<= 0: no limit )",
3384  &propdata->itlimitfactor, FALSE, DEFAULT_ITLIMITFACTOR, SCIP_REAL_MIN, SCIP_REAL_MAX, NULL, NULL) );
3385 
3386  SCIP_CALL( SCIPaddRealParam(scip, "propagating/" PROP_NAME "/itlimitfactorbilin",
3387  "multiple of OBBT LP limit used as total LP iteration limit for solving bilinear inequality LPs (< 0 for no limit)",
3388  &propdata->itlimitfactorbilin, FALSE, DEFAULT_ITLIMITFAC_BILININEQS, SCIP_REAL_MIN, SCIP_REAL_MAX, NULL, NULL) );
3389 
3390  SCIP_CALL( SCIPaddRealParam(scip, "propagating/" PROP_NAME "/minnonconvexity",
3391  "minimum absolute value of nonconvex eigenvalues for a bilinear term",
3392  &propdata->minnonconvexity, FALSE, DEFAULT_MINNONCONVEXITY, 0.0, SCIP_REAL_MAX, NULL, NULL) );
3393 
3394  SCIP_CALL( SCIPaddLongintParam(scip, "propagating/" PROP_NAME "/minitlimit",
3395  "minimum LP iteration limit",
3396  &propdata->minitlimit, FALSE, DEFAULT_MINITLIMIT, 0L, SCIP_LONGINT_MAX, NULL, NULL) );
3397 
3398  SCIP_CALL( SCIPaddRealParam(scip, "propagating/" PROP_NAME "/dualfeastol",
3399  "feasibility tolerance for reduced costs used in obbt; this value is used if SCIP's dual feastol is greater",
3400  &propdata->dualfeastol, FALSE, DEFAULT_DUALFEASTOL, 0.0, SCIP_REAL_MAX, NULL, NULL) );
3401 
3402  SCIP_CALL( SCIPaddRealParam(scip, "propagating/" PROP_NAME "/conditionlimit",
3403  "maximum condition limit used in LP solver (-1.0: no limit)",
3404  &propdata->conditionlimit, FALSE, DEFAULT_CONDITIONLIMIT, -1.0, SCIP_REAL_MAX, NULL, NULL) );
3405 
3406  SCIP_CALL( SCIPaddRealParam(scip, "propagating/" PROP_NAME "/boundstreps",
3407  "minimal relative improve for strengthening bounds",
3408  &propdata->boundstreps, FALSE, DEFAULT_BOUNDSTREPS, 0.0, 1.0, NULL, NULL) );
3409 
3410  SCIP_CALL( SCIPaddRealParam(scip, "propagating/" PROP_NAME "/indicatorthreshold",
3411  "threshold whether upper bounds of vars of indicator conss are considered or tightened",
3412  &propdata->indicatorthreshold, TRUE, DEFAULT_INDICATORTHRESHOLD, 0.0, SCIP_REAL_MAX, NULL, NULL) );
3413 
3414  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/onlynonconvexvars",
3415  "only apply obbt on non-convex variables",
3416  &propdata->onlynonconvexvars, TRUE, DEFAULT_ONLYNONCONVEXVARS, NULL, NULL) );
3417 
3418  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/indicators",
3419  "apply obbt on variables of indicator constraints? (independent of convexity)",
3420  &propdata->indicators, TRUE, DEFAULT_INDICATORS, NULL, NULL) );
3421 
3422  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/tightintboundsprobing",
3423  "should integral bounds be tightened during the probing mode?",
3424  &propdata->tightintboundsprobing, TRUE, DEFAULT_TIGHTINTBOUNDSPROBING, NULL, NULL) );
3425 
3426  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/tightcontboundsprobing",
3427  "should continuous bounds be tightened during the probing mode?",
3428  &propdata->tightcontboundsprobing, TRUE, DEFAULT_TIGHTCONTBOUNDSPROBING, NULL, NULL) );
3429 
3430  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/createbilinineqs",
3431  "solve auxiliary LPs in order to find valid inequalities for bilinear terms?",
3432  &propdata->createbilinineqs, TRUE, DEFAULT_CREATE_BILININEQS, NULL, NULL) );
3433 
3434  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/createlincons",
3435  "create linear constraints from inequalities for bilinear terms?",
3436  &propdata->createlincons, TRUE, DEFAULT_CREATE_LINCONS, NULL, NULL) );
3437 
3438  SCIP_CALL( SCIPaddIntParam(scip, "propagating/" PROP_NAME "/orderingalgo",
3439  "select the type of ordering algorithm which should be used (0: no special ordering, 1: greedy, 2: greedy reverse)",
3440  &propdata->orderingalgo, TRUE, DEFAULT_ORDERINGALGO, 0, 2, NULL, NULL) );
3441 
3442  SCIP_CALL( SCIPaddBoolParam(scip, "propagating/" PROP_NAME "/separatesol",
3443  "should the obbt LP solution be separated?",
3444  &propdata->separatesol, TRUE, DEFAULT_SEPARATESOL, NULL, NULL) );
3445 
3446  SCIP_CALL( SCIPaddIntParam(scip, "propagating/" PROP_NAME "/sepaminiter",
3447  "minimum number of iteration spend to separate an obbt LP solution",
3448  &propdata->sepaminiter, TRUE, DEFAULT_SEPAMINITER, 0, INT_MAX, NULL, NULL) );
3449 
3450  SCIP_CALL( SCIPaddIntParam(scip, "propagating/" PROP_NAME "/sepamaxiter",
3451  "maximum number of iteration spend to separate an obbt LP solution",
3452  &propdata->sepamaxiter, TRUE, DEFAULT_SEPAMAXITER, 0, INT_MAX, NULL, NULL) );
3453 
3454  SCIP_CALL( SCIPaddIntParam(scip, "propagating/" PROP_NAME "/propagatefreq",
3455  "trigger a propagation round after that many bound tightenings (0: no propagation)",
3456  &propdata->propagatefreq, TRUE, DEFAULT_PROPAGATEFREQ, 0, INT_MAX, NULL, NULL) );
3457 
3458  return SCIP_OKAY;
3459 }
enum SCIP_Result SCIP_RESULT
Definition: type_result.h:61
void SCIPfreeRandom(SCIP *scip, SCIP_RANDNUMGEN **randnumgen)
#define SCIPfreeBlockMemoryArray(scip, ptr, num)
Definition: scip_mem.h:110
#define DEFAULT_APPLY_FILTERROUNDS
Definition: prop_obbt.c:96
#define DEFAULT_GENVBDSDURINGFILTER
Definition: prop_obbt.c:100
enum SCIP_BoundType SCIP_BOUNDTYPE
Definition: type_lp.h:59
SCIP_ROW ** SCIPgetLPRows(SCIP *scip)
Definition: scip_lp.c:605
SCIP_Bool SCIPisFeasZero(SCIP *scip, SCIP_Real val)
#define PROP_NAME
Definition: prop_obbt.c:83
#define PROP_FREQ
Definition: prop_obbt.c:87
SCIP_Bool SCIPinRepropagation(SCIP *scip)
Definition: scip_tree.c:146
SCIP_Longint SCIPgetNRootLPIterations(SCIP *scip)
#define NULL
Definition: def.h:267
SCIP_Real SCIPfeastol(SCIP *scip)
unsigned int indicator
Definition: prop_obbt.c:162
#define SCIPallocBlockMemoryArray(scip, ptr, num)
Definition: scip_mem.h:93
SCIP_Bool SCIPisNLPConstructed(SCIP *scip)
Definition: scip_nlp.c:110
SCIP_RETCODE SCIPtightenVarLb(SCIP *scip, SCIP_VAR *var, SCIP_Real newbound, SCIP_Bool force, SCIP_Bool *infeasible, SCIP_Bool *tightened)
Definition: scip_var.c:5205
SCIP_RETCODE SCIPcacheRowExtensions(SCIP *scip, SCIP_ROW *row)
Definition: scip_lp.c:1635
unsigned int done
Definition: prop_obbt.c:160
#define DEFAULT_CREATE_LINCONS
Definition: prop_obbt.c:142
SCIP_Bool SCIPisFeasEQ(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
public methods for SCIP parameter handling
SCIP_NODE * SCIPgetCurrentNode(SCIP *scip)
Definition: scip_tree.c:91
static SCIP_DECL_SORTPTRCOMP(compBoundsScore)
Definition: prop_obbt.c:1608
SCIP_STAGE SCIPgetStage(SCIP *scip)
Definition: scip_general.c:380
static SCIP_RETCODE filterExistingLP(SCIP *scip, SCIP_PROPDATA *propdata, int *nfiltered, BOUND *currbound)
Definition: prop_obbt.c:977
public methods for branch and bound tree
SCIP_RETCODE SCIPbacktrackProbing(SCIP *scip, int probingdepth)
Definition: scip_probing.c:225
#define DEFAULT_MINITLIMIT
Definition: prop_obbt.c:112
SCIP_Longint SCIPgetNLPIterations(SCIP *scip)
static SCIP_RETCODE sortBounds(SCIP *scip, SCIP_PROPDATA *propdata)
Definition: prop_obbt.c:1654
SCIP_RETCODE SCIPcreateEmptyRowUnspec(SCIP *scip, SCIP_ROW **row, const char *name, SCIP_Real lhs, SCIP_Real rhs, SCIP_Bool local, SCIP_Bool modifiable, SCIP_Bool removable)
Definition: scip_lp.c:1482
static void exchangeBounds(SCIP_PROPDATA *propdata, int i)
Definition: prop_obbt.c:756
public methods for memory management
SCIP_CONSHDLR * SCIPfindConshdlr(SCIP *scip, const char *name)
Definition: scip_cons.c:941
SCIP_Real SCIPgetCutoffbound(SCIP *scip)
int SCIPexprGetNChildren(SCIP_EXPR *expr)
Definition: expr.c:3854
SCIP_RETCODE SCIPflushRowExtensions(SCIP *scip, SCIP_ROW *row)
Definition: scip_lp.c:1658
SCIP_Bool SCIPisUbBetter(SCIP *scip, SCIP_Real newub, SCIP_Real oldlb, SCIP_Real oldub)
enum SCIP_BaseStat SCIP_BASESTAT
Definition: type_lpi.h:96
SCIP_RETCODE SCIPapplyCutsProbing(SCIP *scip, SCIP_Bool *cutoff)
Definition: scip_probing.c:948
#define DEFAULT_BOUNDSTREPS
Definition: prop_obbt.c:105
SCIP_PROP * SCIPfindProp(SCIP *scip, const char *name)
Definition: scip_prop.c:329
SCIP_Real SCIPvarGetLbGlobal(SCIP_VAR *var)
Definition: var.c:18079
SCIP_RETCODE SCIPgetRealParam(SCIP *scip, const char *name, SCIP_Real *value)
Definition: scip_param.c:307
#define SCIP_MAXSTRLEN
Definition: def.h:288
#define DEFAULT_SEPAMAXITER
Definition: prop_obbt.c:136
SCIP_BASESTAT SCIPcolGetBasisStatus(SCIP_COL *col)
Definition: lp.c:17031
#define DEFAULT_MINNONCONVEXITY
Definition: prop_obbt.c:144
static SCIP_DECL_PROPEXITSOL(propExitsolObbt)
Definition: prop_obbt.c:3248
static long bound
SCIP_RETCODE SCIPaddVarToRow(SCIP *scip, SCIP_ROW *row, SCIP_VAR *var, SCIP_Real val)
Definition: scip_lp.c:1701
#define DEFAULT_ITLIMITFACTOR
Definition: prop_obbt.c:109
SCIP_Real SCIPgetVarRedcost(SCIP *scip, SCIP_VAR *var)
Definition: scip_var.c:1866
#define SQR(x)
Definition: def.h:214
SCIP_Bool SCIPisPositive(SCIP *scip, SCIP_Real val)
SCIP_Real SCIPvarGetLbLocal(SCIP_VAR *var)
Definition: var.c:18135
int filtered
Definition: prop_obbt.c:182
static SCIP_RETCODE filterRound(SCIP *scip, SCIP_PROPDATA *propdata, int itlimit, int *nfiltered, SCIP_Real *objcoefs, int *objcoefsinds, int nobjcoefs)
Definition: prop_obbt.c:1150
SCIP_RETCODE SCIPincludePropObbt(SCIP *scip)
Definition: prop_obbt.c:3333
SCIP_Bool SCIPvarIsBinary(SCIP_VAR *var)
Definition: var.c:17600
constraint handler for indicator constraints
SCIP_Real SCIPdualfeastol(SCIP *scip)
SCIP_Bool SCIPisFeasGE(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
SCIP_RETCODE SCIPgetVarsData(SCIP *scip, SCIP_VAR ***vars, int *nvars, int *nbinvars, int *nintvars, int *nimplvars, int *ncontvars)
Definition: scip_prob.c:1866
static int nextBound(SCIP *scip, SCIP_PROPDATA *propdata, SCIP_Bool convexphase)
Definition: prop_obbt.c:1686
SCIP_CONS ** SCIPconshdlrGetConss(SCIP_CONSHDLR *conshdlr)
Definition: cons.c:4595
#define FALSE
Definition: def.h:94
#define DEFAULT_SEPAMINITER
Definition: prop_obbt.c:135
SCIP_RETCODE SCIPaddLongintParam(SCIP *scip, const char *name, const char *desc, SCIP_Longint *valueptr, SCIP_Bool isadvanced, SCIP_Longint defaultvalue, SCIP_Longint minvalue, SCIP_Longint maxvalue, SCIP_DECL_PARAMCHGD((*paramchgd)), SCIP_PARAMDATA *paramdata)
Definition: scip_param.c:111
SCIP_Real SCIPrelDiff(SCIP_Real val1, SCIP_Real val2)
Definition: misc.c:11184
int SCIPgetSubscipDepth(SCIP *scip)
Definition: scip_copy.c:2605
SCIP_Real SCIPinfinity(SCIP *scip)
int SCIPsnprintf(char *t, int len, const char *s,...)
Definition: misc.c:10877
SCIP_Bool SCIPisNegative(SCIP *scip, SCIP_Real val)
#define DEFAULT_APPLY_TRIVIALFITLERING
Definition: prop_obbt.c:99
#define TRUE
Definition: def.h:93
#define SCIPdebug(x)
Definition: pub_message.h:93
enum SCIP_Retcode SCIP_RETCODE
Definition: type_retcode.h:63
#define SCIPstatisticMessage
Definition: pub_message.h:123
SCIP_NLHDLR * SCIPfindNlhdlrNonlinear(SCIP_CONSHDLR *conshdlr, const char *name)
static SCIP_Real getFilterCoef(SCIP *scip, SCIP_PROPDATA *propdata, SCIP_VAR *var, SCIP_BOUNDTYPE boundtype)
Definition: prop_obbt.c:496
int SCIPvarGetProbindex(SCIP_VAR *var)
Definition: var.c:17769
int index
Definition: prop_obbt.c:163
public methods for problem variables
#define DEFAULT_CREATE_BILININEQS
Definition: prop_obbt.c:141
SCIP_RETCODE SCIPtightenVarUb(SCIP *scip, SCIP_VAR *var, SCIP_Real newbound, SCIP_Bool force, SCIP_Bool *infeasible, SCIP_Bool *tightened)
Definition: scip_var.c:5322
#define SCIPfreeBlockMemory(scip, ptr)
Definition: scip_mem.h:108
#define DEFAULT_ITLIMITFAC_BILININEQS
Definition: prop_obbt.c:143
#define SCIPdebugMessage
Definition: pub_message.h:96
SCIP_RETCODE SCIPchgVarLbProbing(SCIP *scip, SCIP_VAR *var, SCIP_Real newbound)
Definition: scip_probing.c:301
SCIP_RETCODE SCIPgetNLPVarsNonlinearity(SCIP *scip, int *nlcount)
Definition: scip_nlp.c:223
static int bilinboundGetLocksPos(BILINBOUND *bilinbound)
Definition: prop_obbt.c:894
SCIP_EXPR * expr
Definition: prop_obbt.c:181
void * SCIPhashmapGetImage(SCIP_HASHMAP *hashmap, void *origin)
Definition: misc.c:3261
SCIP_Bool SCIPisEQ(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
#define GENVBOUND_PROP_NAME
Definition: prop_obbt.c:128
#define SCIP_LONGINT_MAX
Definition: def.h:159
static SCIP_RETCODE getNVarsIndicators(SCIP *scip, int *nindcount)
Definition: prop_obbt.c:2774
#define SCIPfreeBufferArray(scip, ptr)
Definition: scip_mem.h:136
enum SCIP_LPSolStat SCIP_LPSOLSTAT
Definition: type_lp.h:51
void SCIPcaptureExpr(SCIP_EXPR *expr)
Definition: scip_expr.c:1409
static void getCorner(SCIP_VAR *x, SCIP_VAR *y, CORNER corner, SCIP_Real *px, SCIP_Real *py)
Definition: prop_obbt.c:779
#define SCIPallocBlockMemory(scip, ptr)
Definition: scip_mem.h:89
public methods for SCIP variables
SCIP_RETCODE SCIPsetRealParam(SCIP *scip, const char *name, SCIP_Real value)
Definition: scip_param.c:603
Corner
Definition: prop_obbt.c:168
SCIP_Real SCIProwGetDualsol(SCIP_ROW *row)
Definition: lp.c:17312
void SCIPwarningMessage(SCIP *scip, const char *formatstr,...)
Definition: scip_message.c:120
#define SCIPdebugMsg
Definition: scip_message.h:78
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:83
SCIP_VAR ** x
Definition: circlepacking.c:63
SCIP_Real SCIPepsilon(SCIP *scip)
SCIP_HASHMAP * SCIPgetVarExprHashmapNonlinear(SCIP_CONSHDLR *conshdlr)
bilinear nonlinear handler
public methods for numerical tolerances
static SCIP_RETCODE tightenBoundProbing(SCIP *scip, BOUND *bound, SCIP_Real newval, SCIP_Bool *tightened)
Definition: prop_obbt.c:1547
#define DEFAULT_ORDERINGALGO
Definition: prop_obbt.c:124
public methods for querying solving statistics
const char * SCIPgetProbName(SCIP *scip)
Definition: scip_prob.c:1067
SCIP_Bool SCIProwIsInLP(SCIP_ROW *row)
Definition: lp.c:17523
SCIP_Bool SCIPhashmapExists(SCIP_HASHMAP *hashmap, void *origin)
Definition: misc.c:3423
public methods for the branch-and-bound tree
SCIP_VAR * SCIPgetSlackVarIndicator(SCIP_CONS *cons)
SCIP_Bool SCIPisLbBetter(SCIP *scip, SCIP_Real newlb, SCIP_Real oldlb, SCIP_Real oldub)
SCIP_Longint SCIPnodeGetNumber(SCIP_NODE *node)
Definition: tree.c:7490
static SCIP_RETCODE createGenVBound(SCIP *scip, SCIP_PROPDATA *propdata, BOUND *bound, SCIP_Bool *found)
Definition: prop_obbt.c:570
SCIP_EXPR ** SCIPexprGetChildren(SCIP_EXPR *expr)
Definition: expr.c:3864
SCIP_Real SCIPvarGetUbGlobal(SCIP_VAR *var)
Definition: var.c:18089
#define DEFAULT_DUALFEASTOL
Definition: prop_obbt.c:101
public methods for managing constraints
#define DEFAULT_FILTERING_NORM
Definition: prop_obbt.c:93
int SCIPgetNNLPVars(SCIP *scip)
Definition: scip_nlp.c:201
static SCIP_Bool indicatorVarIsInteresting(SCIP *scip, SCIP_VAR *var, int nlcount, int nindcount, SCIP_Real threshold)
Definition: prop_obbt.c:954
static SCIP_RETCODE applyBoundChgs(SCIP *scip, SCIP_PROPDATA *propdata, SCIP_RESULT *result)
Definition: prop_obbt.c:1472
static SCIP_Bool includeVarGenVBound(SCIP *scip, SCIP_VAR *var)
Definition: prop_obbt.c:439
SCIP_RETCODE SCIPaddCons(SCIP *scip, SCIP_CONS *cons)
Definition: scip_prob.c:2770
SCIP_Bool SCIPisLT(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
SCIP_RETCODE SCIPpropagateProbing(SCIP *scip, int maxproprounds, SCIP_Bool *cutoff, SCIP_Longint *ndomredsfound)
Definition: scip_probing.c:580
static SCIP_DECL_PROPRESPROP(propRespropObbt)
Definition: prop_obbt.c:3239
void SCIPsortDownPtr(void **ptrarray, SCIP_DECL_SORTPTRCOMP((*ptrcomp)), int len)
#define PROP_DESC
Definition: prop_obbt.c:84
SCIP_RETCODE SCIPgenVBoundAdd(SCIP *scip, SCIP_PROP *genvboundprop, SCIP_VAR **vars, SCIP_VAR *var, SCIP_Real *coefs, int ncoefs, SCIP_Real coefcutoffbound, SCIP_Real constant, SCIP_BOUNDTYPE boundtype)
#define SCIPfreeBufferArrayNull(scip, ptr)
Definition: scip_mem.h:137
#define MAX3(x, y, z)
Definition: def.h:247
SCIP_RETCODE SCIPendProbing(SCIP *scip)
Definition: scip_probing.c:260
const char * SCIPvarGetName(SCIP_VAR *var)
Definition: var.c:17420
#define DEFAULT_GENVBDSDURINGSEPA
Definition: prop_obbt.c:137
SCIP_RETCODE SCIPseparateSol(SCIP *scip, SCIP_SOL *sol, SCIP_Bool pretendroot, SCIP_Bool allowlocal, SCIP_Bool onlydelayed, SCIP_Bool *delayed, SCIP_Bool *cutoff)
Definition: scip_cut.c:735
#define REALABS(x)
Definition: def.h:197
SCIP_RETCODE SCIPchgDualfeastol(SCIP *scip, SCIP_Real dualfeastol)
SCIP_RETCODE SCIPgetIntParam(SCIP *scip, const char *name, int *value)
Definition: scip_param.c:269
SCIP_Real SCIPvarGetLPSol(SCIP_VAR *var)
Definition: var.c:18453
int SCIPgetNLPRows(SCIP *scip)
Definition: scip_lp.c:626
public methods for problem copies
SCIP_Real SCIPgetVarObjProbing(SCIP *scip, SCIP_VAR *var)
Definition: scip_probing.c:388
#define SCIP_CALL(x)
Definition: def.h:380
unsigned int done
Definition: prop_obbt.c:183
static SCIP_Real bilinboundGetScore(SCIP *scip, SCIP_RANDNUMGEN *randnumgen, BILINBOUND *bilinbound)
Definition: prop_obbt.c:905
#define DEFAULT_PROPAGATEFREQ
Definition: prop_obbt.c:138
SCIP_Bool SCIPisFeasGT(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
SCIP_RETCODE SCIPsolveProbingLP(SCIP *scip, int itlim, SCIP_Bool *lperror, SCIP_Bool *cutoff)
Definition: scip_probing.c:820
#define OBBT_SCOREBASE
Definition: prop_obbt.c:127
SCIP_Bool SCIPisFeasLE(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
unsigned int score
Definition: prop_obbt.c:157
static int getIterationsLeft(SCIP *scip, SCIP_Longint nolditerations, SCIP_Longint itlimit)
Definition: prop_obbt.c:466
int SCIPconshdlrGetNConss(SCIP_CONSHDLR *conshdlr)
Definition: cons.c:4638
public methods for constraint handler plugins and constraints
public methods for NLP management
SCIP_Bool SCIPisHugeValue(SCIP *scip, SCIP_Real val)
static unsigned int getScore(SCIP *scip, BOUND *bound, int nlcount, int nindcount, int maxnlcount, SCIP_Real smallub)
Definition: prop_obbt.c:2657
SCIP_RETCODE SCIPcreateRandom(SCIP *scip, SCIP_RANDNUMGEN **randnumgen, unsigned int initialseed, SCIP_Bool useglobalseed)
static SCIP_RETCODE solveBilinearLP(SCIP *scip, SCIP_VAR *x, SCIP_VAR *y, SCIP_Real xs, SCIP_Real ys, SCIP_Real xt, SCIP_Real yt, SCIP_Real *xcoef, SCIP_Real *ycoef, SCIP_Real *constant, SCIP_Longint iterlim, int *nnonzduals)
Definition: prop_obbt.c:2314
int SCIPgetExprNLocksNegNonlinear(SCIP_EXPR *expr)
#define SCIPallocBufferArray(scip, ptr, num)
Definition: scip_mem.h:124
public data structures and miscellaneous methods
#define SCIP_Bool
Definition: def.h:91
SCIP_LPSOLSTAT SCIPgetLPSolstat(SCIP *scip)
Definition: scip_lp.c:168
SCIP_Real newval
Definition: prop_obbt.c:155
int SCIPgetDepth(SCIP *scip)
Definition: scip_tree.c:670
SCIP_RETCODE SCIPreleaseExpr(SCIP *scip, SCIP_EXPR **expr)
Definition: scip_expr.c:1417
constraint handler for nonlinear constraints specified by algebraic expressions
optimization-based bound tightening propagator
#define MIN(x, y)
Definition: def.h:243
SCIP_RETCODE SCIPsetIntParam(SCIP *scip, const char *name, int value)
Definition: scip_param.c:487
public methods for LP management
public methods for cuts and aggregation rows
#define DEFAULT_CREATE_GENVBOUNDS
Definition: prop_obbt.c:92
SCIP_Real SCIPvarGetObj(SCIP_VAR *var)
Definition: var.c:17927
SCIP_RETCODE SCIPsetPropCopy(SCIP *scip, SCIP_PROP *prop, SCIP_DECL_PROPCOPY((*propcopy)))
Definition: scip_prop.c:151
static SCIP_DECL_PROPFREE(propFreeObbt)
Definition: prop_obbt.c:3311
SCIP_Bool SCIPallowWeakDualReds(SCIP *scip)
Definition: scip_var.c:8658
SCIP_BOUNDTYPE boundtype
Definition: prop_obbt.c:156
SCIP_COL * SCIPvarGetCol(SCIP_VAR *var)
Definition: var.c:17790
Constraint handler for linear constraints in their most general form, .
static SCIP_Bool varIsInteresting(SCIP *scip, SCIP_VAR *var, int nlcount, int nindcount)
Definition: prop_obbt.c:2834
SCIP_Bool SCIPisInfinity(SCIP *scip, SCIP_Real val)
static SCIP_RETCODE applySeparation(SCIP *scip, SCIP_PROPDATA *propdata, BOUND *currbound, SCIP_Longint *nleftiterations, SCIP_Bool *success)
Definition: prop_obbt.c:1739
#define BMSclearMemory(ptr)
Definition: memory.h:129
static SCIP_RETCODE setObjProbing(SCIP *scip, SCIP_PROPDATA *propdata, BOUND *bound, SCIP_Real coef)
Definition: prop_obbt.c:392
SCIP_RETCODE SCIPsetPropExitsol(SCIP *scip, SCIP_PROP *prop, SCIP_DECL_PROPEXITSOL((*propexitsol)))
Definition: scip_prop.c:231
static SCIP_DECL_PROPCOPY(propCopyObbt)
Definition: prop_obbt.c:3078
SCIP_Real SCIPrandomGetReal(SCIP_RANDNUMGEN *randnumgen, SCIP_Real minrandval, SCIP_Real maxrandval)
Definition: misc.c:10130
SCIP_Bool SCIPinProbing(SCIP *scip)
Definition: scip_probing.c:97
public methods for the LP relaxation, rows and columns
const char * SCIPpropGetName(SCIP_PROP *prop)
Definition: prop.c:941
SCIP_EXPR ** SCIPgetExprsBilinear(SCIP_NLHDLR *nlhdlr)
int SCIPgetNVars(SCIP *scip)
Definition: scip_prob.c:1992
#define SCIP_REAL_MAX
Definition: def.h:174
unsigned int found
Definition: prop_obbt.c:159
public methods for nonlinear relaxation
SCIP_Real * r
Definition: circlepacking.c:59
#define SCIP_REAL_MIN
Definition: def.h:175
methods for sorting joint arrays of various types
SCIP_RETCODE SCIPaddIneqBilinear(SCIP *scip, SCIP_NLHDLR *nlhdlr, SCIP_EXPR *expr, SCIP_Real xcoef, SCIP_Real ycoef, SCIP_Real constant, SCIP_Bool *success)
SCIP_RETCODE SCIPcreateConsLinear(SCIP *scip, SCIP_CONS **cons, const char *name, int nvars, SCIP_VAR **vars, SCIP_Real *vals, SCIP_Real lhs, SCIP_Real rhs, SCIP_Bool initial, SCIP_Bool separate, SCIP_Bool enforce, SCIP_Bool check, SCIP_Bool propagate, SCIP_Bool local, SCIP_Bool modifiable, SCIP_Bool dynamic, SCIP_Bool removable, SCIP_Bool stickingatnode)
#define SCIP_LONGINT_FORMAT
Definition: def.h:165
static SCIP_RETCODE solveLP(SCIP *scip, int itlimit, SCIP_Bool *error, SCIP_Bool *optimal)
Definition: prop_obbt.c:260
int SCIPgetExprNLocksPosNonlinear(SCIP_EXPR *expr)
SCIP_RETCODE SCIPreleaseRow(SCIP *scip, SCIP_ROW **row)
Definition: scip_lp.c:1562
general public methods
#define MAX(x, y)
Definition: def.h:239
#define PROP_TIMING
Definition: prop_obbt.c:85
SCIP_RETCODE SCIPsetPropResprop(SCIP *scip, SCIP_PROP *prop, SCIP_DECL_PROPRESPROP((*propresprop)))
Definition: scip_prop.c:312
SCIP_Real SCIPgetLPObjval(SCIP *scip)
Definition: scip_lp.c:247
SCIP_Bool SCIPisGT(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
SCIP_VAR ** SCIPgetVarsLinear(SCIP *scip, SCIP_CONS *cons)
public methods for random numbers
#define DEFAULT_RANDSEED
Definition: prop_obbt.c:145
public methods for the probing mode
SCIP_RETCODE SCIPreleaseCons(SCIP *scip, SCIP_CONS **cons)
Definition: scip_cons.c:1174
#define DEFAULT_SEPARATESOL
Definition: prop_obbt.c:130
int SCIPgetNCuts(SCIP *scip)
Definition: scip_cut.c:787
SCIP_RETCODE SCIPaddRowProbing(SCIP *scip, SCIP_ROW *row)
Definition: scip_probing.c:908
public methods for message output
static SCIP_VAR * bilinboundGetX(BILINBOUND *bilinbound)
Definition: prop_obbt.c:859
static SCIP_VAR * bilinboundGetY(BILINBOUND *bilinbound)
Definition: prop_obbt.c:871
static SCIP_RETCODE findNewBounds(SCIP *scip, SCIP_PROPDATA *propdata, SCIP_Longint *nleftiterations, SCIP_Bool convexphase)
Definition: prop_obbt.c:1821
#define DEFAULT_ONLYNONCONVEXVARS
Definition: prop_obbt.c:113
SCIP_Bool SCIPisExprVar(SCIP *scip, SCIP_EXPR *expr)
Definition: scip_expr.c:1431
SCIP_VAR ** SCIPgetVars(SCIP *scip)
Definition: scip_prob.c:1947
SCIP_RETCODE SCIPsetPropInitsol(SCIP *scip, SCIP_PROP *prop, SCIP_DECL_PROPINITSOL((*propinitsol)))
Definition: scip_prop.c:215
#define DEFAULT_INDICATORTHRESHOLD
Definition: prop_obbt.c:115
SCIP_VARSTATUS SCIPvarGetStatus(SCIP_VAR *var)
Definition: var.c:17539
int SCIPgetNExprsBilinear(SCIP_NLHDLR *nlhdlr)
#define SCIP_Real
Definition: def.h:173
struct SCIP_PropData SCIP_PROPDATA
Definition: type_prop.h:52
SCIP_Bool SCIPisStopped(SCIP *scip)
Definition: scip_general.c:718
SCIP_VAR ** y
Definition: circlepacking.c:64
public methods for message handling
SCIP_Real score
Definition: prop_obbt.c:184
#define SCIP_INVALID
Definition: def.h:193
static SCIP_Real evalBound(SCIP *scip, BOUND *bound)
Definition: prop_obbt.c:1670
SCIP_VAR * var
Definition: prop_obbt.c:154
SCIP_RETCODE SCIPsetPropFree(SCIP *scip, SCIP_PROP *prop, SCIP_DECL_PROPFREE((*propfree)))
Definition: scip_prop.c:167
SCIP_PROPDATA * SCIPpropGetData(SCIP_PROP *prop)
Definition: prop.c:789
void SCIPpropSetData(SCIP_PROP *prop, SCIP_PROPDATA *propdata)
Definition: prop.c:799
static int bilinboundGetLocksNeg(BILINBOUND *bilinbound)
Definition: prop_obbt.c:883
static SCIP_RETCODE filterBounds(SCIP *scip, SCIP_PROPDATA *propdata, SCIP_Longint itlimit)
Definition: prop_obbt.c:1310
#define SCIP_Longint
Definition: def.h:158
SCIP_Bool SCIPallColsInLP(SCIP *scip)
Definition: scip_lp.c:649
public methods for propagator plugins
#define DEFAULT_INDICATORS
Definition: prop_obbt.c:114
SCIP_Real SCIPchgRelaxfeastol(SCIP *scip, SCIP_Real relaxfeastol)
#define DEFAULT_TIGHTCONTBOUNDSPROBING
Definition: prop_obbt.c:121
static SCIP_RETCODE addObjCutoff(SCIP *scip, SCIP_PROPDATA *propdata)
Definition: prop_obbt.c:332
static SCIP_Bool varIsFixedLocal(SCIP *scip, SCIP_VAR *var)
Definition: prop_obbt.c:382
static SCIP_DECL_PROPINITSOL(propInitsolObbt)
Definition: prop_obbt.c:3092
static SCIP_RETCODE getNLPVarsNonConvexity(SCIP *scip, unsigned int *nccounts)
Definition: prop_obbt.c:2716
SCIP_VARTYPE SCIPvarGetType(SCIP_VAR *var)
Definition: var.c:17585
SCIP_Bool SCIPisZero(SCIP *scip, SCIP_Real val)
unsigned int filtered
Definition: prop_obbt.c:158
SCIP_Bool SCIPisLE(SCIP *scip, SCIP_Real val1, SCIP_Real val2)
#define PROP_DELAY
Definition: prop_obbt.c:88
#define PROP_PRIORITY
Definition: prop_obbt.c:86
#define DEFAULT_CONDITIONLIMIT
Definition: prop_obbt.c:104
SCIP_Real SCIPvarGetUbLocal(SCIP_VAR *var)
Definition: var.c:18145
SCIP_RETCODE SCIPnewProbingNode(SCIP *scip)
Definition: scip_probing.c:165
unsigned int SCIPgetExprNSepaUsesActivityNonlinear(SCIP_EXPR *expr)
unsigned int nonconvex
Definition: prop_obbt.c:161
#define DEFAULT_TIGHTINTBOUNDSPROBING
Definition: prop_obbt.c:118
SCIP_RETCODE SCIPstartProbing(SCIP *scip)
Definition: scip_probing.c:119
SCIP_VAR * SCIPgetExprAuxVarNonlinear(SCIP_EXPR *expr)
#define BMSclearMemoryArray(ptr, num)
Definition: memory.h:130
static SCIP_RETCODE initBounds(SCIP *scip, SCIP_PROPDATA *propdata)
Definition: prop_obbt.c:2852
SCIP_Real SCIPceil(SCIP *scip, SCIP_Real val)
static SCIP_DECL_PROPEXEC(propExecObbt)
Definition: prop_obbt.c:3123
SCIP_CONS * SCIPgetLinearConsIndicator(SCIP_CONS *cons)
enum Corner CORNER
Definition: prop_obbt.c:176
#define SCIPABORT()
Definition: def.h:352
public methods for global and local (sub)problems
static SCIP_RETCODE applyObbt(SCIP *scip, SCIP_PROPDATA *propdata, SCIP_Longint itlimit, SCIP_RESULT *result)
Definition: prop_obbt.c:2060
SCIP_Bool SCIPvarIsIntegral(SCIP_VAR *var)
Definition: var.c:17611
SCIP_RETCODE SCIPchgVarUbProbing(SCIP *scip, SCIP_VAR *var, SCIP_Real newbound)
Definition: scip_probing.c:345
int SCIPgetNVarsLinear(SCIP *scip, SCIP_CONS *cons)
#define DEFAULT_FILTERING_MIN
Definition: prop_obbt.c:106
SCIP_RETCODE SCIPaddRealParam(SCIP *scip, const char *name, const char *desc, SCIP_Real *valueptr, SCIP_Bool isadvanced, SCIP_Real defaultvalue, SCIP_Real minvalue, SCIP_Real maxvalue, SCIP_DECL_PARAMCHGD((*paramchgd)), SCIP_PARAMDATA *paramdata)
Definition: scip_param.c:139
SCIP_Real SCIPfloor(SCIP *scip, SCIP_Real val)
#define EPSZ(x, eps)
Definition: def.h:203
static SCIP_RETCODE applyObbtBilinear(SCIP *scip, SCIP_PROPDATA *propdata, SCIP_Longint itlimit, SCIP_RESULT *result)
Definition: prop_obbt.c:2462
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:57
static void getCorners(SCIP_VAR *x, SCIP_VAR *y, CORNER corner, SCIP_Real *xs, SCIP_Real *ys, SCIP_Real *xt, SCIP_Real *yt)
Definition: prop_obbt.c:817
public methods for propagators
generalized variable bounds propagator
SCIP_RETCODE SCIPincludePropBasic(SCIP *scip, SCIP_PROP **propptr, const char *name, const char *desc, int priority, int freq, SCIP_Bool delay, SCIP_PROPTIMING timingmask, SCIP_DECL_PROPEXEC((*propexec)), SCIP_PROPDATA *propdata)
Definition: scip_prop.c:114
SCIP_RETCODE SCIPchgVarObjProbing(SCIP *scip, SCIP_VAR *var, SCIP_Real newobj)
Definition: scip_probing.c:474