OR-Tools  9.1
lp_data.h
Go to the documentation of this file.
1 // Copyright 2010-2021 Google LLC
2 // Licensed under the Apache License, Version 2.0 (the "License");
3 // you may not use this file except in compliance with the License.
4 // You may obtain a copy of the License at
5 //
6 // http://www.apache.org/licenses/LICENSE-2.0
7 //
8 // Unless required by applicable law or agreed to in writing, software
9 // distributed under the License is distributed on an "AS IS" BASIS,
10 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
11 // See the License for the specific language governing permissions and
12 // limitations under the License.
13 
14 //
15 // Storage classes for Linear Programs.
16 //
17 // LinearProgram stores the complete data for a Linear Program:
18 // - objective coefficients and offset,
19 // - cost coefficients,
20 // - coefficient matrix,
21 // - bounds for each variable,
22 // - bounds for each constraint.
23 
24 #ifndef OR_TOOLS_LP_DATA_LP_DATA_H_
25 #define OR_TOOLS_LP_DATA_LP_DATA_H_
26 
27 #include <algorithm> // for max
28 #include <cstdint>
29 #include <map>
30 #include <string> // for string
31 #include <vector> // for vector
32 
33 #include "absl/container/flat_hash_map.h"
34 #include "absl/container/flat_hash_set.h"
35 #include "ortools/base/hash.h"
36 #include "ortools/base/int_type.h"
37 #include "ortools/base/logging.h" // for CHECK*
38 #include "ortools/base/macros.h" // for DISALLOW_COPY_AND_ASSIGN, NULL
41 #include "ortools/lp_data/sparse.h"
42 #include "ortools/util/fp_utils.h"
43 
44 namespace operations_research {
45 namespace glop {
46 
47 class SparseMatrixScaler;
48 
49 // The LinearProgram class is used to store a linear problem in a form
50 // accepted by LPSolver.
51 //
52 // In addition to the simple setter functions used to create such problems, the
53 // class also contains a few more advanced modification functions used primarily
54 // by preprocessors. A client shouldn't need to use them directly.
56  public:
57  enum class VariableType {
58  // The variable can take any value between and including its lower and upper
59  // bound.
60  CONTINUOUS,
61  // The variable must only take integer values.
62  INTEGER,
63  // The variable is implied integer variable i.e. it was continuous variable
64  // in the LP and was detected to take only integer values.
66  };
67 
68  LinearProgram();
69 
70  // Clears, i.e. reset the object to its initial value.
71  void Clear();
72 
73  // Name setter and getter.
74  void SetName(const std::string& name) { name_ = name; }
75  const std::string& name() const { return name_; }
76 
77  // Creates a new variable and returns its index.
78  // By default, the column bounds will be [0, infinity).
79  ColIndex CreateNewVariable();
80 
81  // Creates a new slack variable and returns its index. Do not use this method
82  // to create non-slack variables.
83  ColIndex CreateNewSlackVariable(bool is_integer_slack_variable,
86  const std::string& name);
87 
88  // Creates a new constraint and returns its index.
89  // By default, the constraint bounds will be [0, 0].
90  RowIndex CreateNewConstraint();
91 
92  // Same as CreateNewVariable() or CreateNewConstraint() but also assign an
93  // immutable id to the variable or constraint so they can be retrieved later.
94  // By default, the name is also set to this id, but it can be changed later
95  // without changing the id.
96  //
97  // Note that these ids are NOT copied over by the Populate*() functions.
98  //
99  // TODO(user): Move these and the two corresponding hash_table into a new
100  // LinearProgramBuilder class to simplify the code of some functions like
101  // DeleteColumns() here and make the behavior on copy clear? or simply remove
102  // them as it is almost as easy to maintain a hash_table on the client side.
103  ColIndex FindOrCreateVariable(const std::string& variable_id);
104  RowIndex FindOrCreateConstraint(const std::string& constraint_id);
105 
106  // Functions to set the name of a variable or constraint. Note that you
107  // won't be able to find those named variables/constraints with
108  // FindOrCreate{Variable|Constraint}().
109  // TODO(user): Add PopulateIdsFromNames() so names added via
110  // Set{Variable|Constraint}Name() can be found.
111  void SetVariableName(ColIndex col, absl::string_view name);
112  void SetConstraintName(RowIndex row, absl::string_view name);
113 
114  // Set the type of the variable.
115  void SetVariableType(ColIndex col, VariableType type);
116 
117  // Returns whether the variable at column col is constrained to be integer.
118  bool IsVariableInteger(ColIndex col) const;
119 
120  // Returns whether the variable at column col must take binary values or not.
121  bool IsVariableBinary(ColIndex col) const;
122 
123  // Defines lower and upper bounds for the variable at col. Note that the
124  // bounds may be set to +/- infinity. The variable must have been created
125  // before or this will crash in non-debug mode.
128 
129  // Defines lower and upper bounds for the constraint at row. Note that the
130  // bounds may be set to +/- infinity. If the constraint wasn't created before,
131  // all the rows from the current GetNumberOfRows() to the given row will be
132  // created with a range [0,0].
135 
136  // Defines the coefficient for col / row.
137  void SetCoefficient(RowIndex row, ColIndex col, Fractional value);
138 
139  // Defines the objective coefficient of column col.
140  // It is set to 0.0 by default.
142 
143  // Define the objective offset (0.0 by default) and scaling factor (positive
144  // and equal to 1.0 by default). This is mainly used for displaying purpose
145  // and the real objective is factor * (objective + offset).
148 
149  // Defines the optimization direction. When maximize is true (resp. false),
150  // the objective is maximized (resp. minimized). The default is false.
151  void SetMaximizationProblem(bool maximize);
152 
153  // Calls CleanUp() on each columns.
154  // That is, removes duplicates, zeros, and orders the coefficients by row.
155  void CleanUp();
156 
157  // Returns true if all the columns are ordered by rows and contain no
158  // duplicates or zero entries (i.e. if IsCleanedUp() is true for all columns).
159  bool IsCleanedUp() const;
160 
161  // Functions that return the name of a variable or constraint. If the name is
162  // empty, they return a special name that depends on the index.
163  std::string GetVariableName(ColIndex col) const;
164  std::string GetConstraintName(RowIndex row) const;
165 
166  // Returns the type of variable.
167  VariableType GetVariableType(ColIndex col) const;
168 
169  // Returns true (resp. false) when the problem is a maximization
170  // (resp. minimization) problem.
171  bool IsMaximizationProblem() const { return maximize_; }
172 
173  // Returns the underlying SparseMatrix or its transpose (which may need to be
174  // computed).
175  const SparseMatrix& GetSparseMatrix() const { return matrix_; }
176  const SparseMatrix& GetTransposeSparseMatrix() const;
177 
178  // Some transformations are better done on the transpose representation. These
179  // two functions are here for that. Note that calling the first function and
180  // modifying the matrix does not change the result of any function in this
181  // class until UseTransposeMatrixAsReference() is called. This is because the
182  // transpose matrix is only used by GetTransposeSparseMatrix() and this
183  // function will recompute the whole transpose from the matrix. In particular,
184  // do not call GetTransposeSparseMatrix() while you modify the matrix returned
185  // by GetMutableTransposeSparseMatrix() otherwise all your changes will be
186  // lost.
187  //
188  // IMPORTANT: The matrix dimension cannot change. Otherwise this will cause
189  // problems. This is checked in debug mode when calling
190  // UseTransposeMatrixAsReference().
193 
194  // Release the memory used by the transpose matrix.
195  void ClearTransposeMatrix();
196 
197  // Gets the underlying SparseColumn with the given index.
198  // This is the same as GetSparseMatrix().column(col);
199  const SparseColumn& GetSparseColumn(ColIndex col) const;
200 
201  // Gets a pointer to the underlying SparseColumn with the given index.
203 
204  // Returns the number of variables.
205  ColIndex num_variables() const { return matrix_.num_cols(); }
206 
207  // Returns the number of constraints.
208  RowIndex num_constraints() const { return matrix_.num_rows(); }
209 
210  // Returns the number of entries in the linear program matrix.
211  EntryIndex num_entries() const { return matrix_.num_entries(); }
212 
213  // Return the lower bounds (resp. upper bounds) of constraints as a column
214  // vector. Note that the bound values may be +/- infinity.
216  return constraint_lower_bounds_;
217  }
219  return constraint_upper_bounds_;
220  }
221 
222  // Returns the objective coefficients (or cost) of variables as a row vector.
224  return objective_coefficients_;
225  }
226 
227  // Return the lower bounds (resp. upper bounds) of variables as a row vector.
228  // Note that the bound values may be +/- infinity.
230  return variable_lower_bounds_;
231  }
233  return variable_upper_bounds_;
234  }
235 
236  // Returns a row vector of VariableType representing types of variables.
238  return variable_types_;
239  }
240 
241  // Returns a list (technically a vector) of the ColIndices of the integer
242  // variables. This vector is lazily computed.
243  const std::vector<ColIndex>& IntegerVariablesList() const;
244 
245  // Returns a list (technically a vector) of the ColIndices of the binary
246  // integer variables. This vector is lazily computed.
247  const std::vector<ColIndex>& BinaryVariablesList() const;
248 
249  // Returns a list (technically a vector) of the ColIndices of the non-binary
250  // integer variables. This vector is lazily computed.
251  const std::vector<ColIndex>& NonBinaryVariablesList() const;
252 
253  // Returns the objective coefficient (or cost) of the given variable for the
254  // minimization version of the problem. That is, this is the same as
255  // GetObjectiveCoefficient() for a minimization problem and the opposite for a
256  // maximization problem.
258 
259  // Returns the objective offset and scaling factor.
260  Fractional objective_offset() const { return objective_offset_; }
262  return objective_scaling_factor_;
263  }
264 
265  // Checks if each variable respects its bounds, nothing else.
266  bool SolutionIsWithinVariableBounds(const DenseRow& solution,
267  Fractional absolute_tolerance) const;
268 
269  // Tests if the solution is LP-feasible within the given tolerance,
270  // i.e., satisfies all linear constraints within the absolute tolerance level.
271  // The solution does not need to satisfy the integer constraints.
272  bool SolutionIsLPFeasible(const DenseRow& solution,
273  Fractional absolute_tolerance) const;
274 
275  // Tests if the solution is integer within the given tolerance, i.e., all
276  // integer variables have integer values within the absolute tolerance level.
277  // The solution does not need to satisfy the linear constraints.
278  bool SolutionIsInteger(const DenseRow& solution,
279  Fractional absolute_tolerance) const;
280 
281  // Tests if the solution is both LP-feasible and integer within the tolerance.
282  bool SolutionIsMIPFeasible(const DenseRow& solution,
283  Fractional absolute_tolerance) const;
284 
285  // Fills the value of the slack from the other variable values.
286  // This requires that the slack have been added.
287  void ComputeSlackVariableValues(DenseRow* solution) const;
288 
289  // Functions to translate the sum(solution * objective_coefficients()) to
290  // the real objective of the problem and back. Note that these can also
291  // be used to translate bounds of the objective in the same way.
294 
295  // A short string with the problem dimension.
296  std::string GetDimensionString() const;
297 
298  // A short line with some stats on the problem coefficients.
299  std::string GetObjectiveStatsString() const;
300  std::string GetBoundsStatsString() const;
301 
302  // Returns a stringified LinearProgram. We use the LP file format used by
303  // lp_solve (see http://lpsolve.sourceforge.net/5.1/index.htm).
304  std::string Dump() const;
305 
306  // Returns a string that contains the provided solution of the LP in the
307  // format var1 = X, var2 = Y, var3 = Z, ...
308  std::string DumpSolution(const DenseRow& variable_values) const;
309 
310  // Returns a comma-separated string of integers containing (in that order)
311  // num_constraints_, num_variables_in_file_, num_entries_,
312  // num_objective_non_zeros_, num_rhs_non_zeros_, num_less_than_constraints_,
313  // num_greater_than_constraints_, num_equal_constraints_,
314  // num_range_constraints_, num_non_negative_variables_, num_boxed_variables_,
315  // num_free_variables_, num_fixed_variables_, num_other_variables_
316  // Very useful for reporting in the way used in journal articles.
317  std::string GetProblemStats() const;
318 
319  // Returns a string containing the same information as with GetProblemStats(),
320  // but in a much more human-readable form, for example:
321  // Number of rows : 27
322  // Number of variables in file : 32
323  // Number of entries (non-zeros) : 83
324  // Number of entries in the objective : 5
325  // Number of entries in the right-hand side : 7
326  // Number of <= constraints : 19
327  // Number of >= constraints : 0
328  // Number of = constraints : 8
329  // Number of range constraints : 0
330  // Number of non-negative variables : 32
331  // Number of boxed variables : 0
332  // Number of free variables : 0
333  // Number of fixed variables : 0
334  // Number of other variables : 0
335  std::string GetPrettyProblemStats() const;
336 
337  // Returns a comma-separated string of numbers containing (in that order)
338  // fill rate, max number of entries (length) in a row, average row length,
339  // standard deviation of row length, max column length, average column length,
340  // standard deviation of column length
341  // Useful for profiling algorithms.
342  //
343  // TODO(user): Theses are statistics about the underlying matrix and should be
344  // moved to SparseMatrix.
345  std::string GetNonZeroStats() const;
346 
347  // Returns a string containing the same information as with GetNonZeroStats(),
348  // but in a much more human-readable form, for example:
349  // Fill rate : 9.61%
350  // Entries in row (Max / average / std, dev.) : 9 / 3.07 / 1.94
351  // Entries in column (Max / average / std, dev.): 4 / 2.59 / 0.96
352  std::string GetPrettyNonZeroStats() const;
353 
354  // Adds slack variables to the problem for all rows which don't have slack
355  // variables. The new slack variables have bounds set to opposite of the
356  // bounds of the corresponding constraint, and changes all constraints to
357  // equality constraints with both bounds set to 0.0. If a constraint uses only
358  // integer variables and all their coefficients are integer, it will mark the
359  // slack variable as integer too.
360  //
361  // It is an error to call CreateNewVariable() or CreateNewConstraint() on a
362  // linear program on which this method was called.
363  //
364  // Note that many of the slack variables may not be useful at all, but in
365  // order not to recompute the matrix from one Solve() to the next, we always
366  // include all of them for a given lp matrix.
367  //
368  // TODO(user): investigate the impact on the running time. It seems low
369  // because we almost never iterate on fixed variables.
370  void AddSlackVariablesWhereNecessary(bool detect_integer_constraints);
371 
372  // Returns the index of the first slack variable in the linear program.
373  // Returns kInvalidCol if slack variables were not injected into the problem
374  // yet.
375  ColIndex GetFirstSlackVariable() const;
376 
377  // Returns the index of the slack variable corresponding to the given
378  // constraint. Returns kInvalidCol if slack variables were not injected into
379  // the problem yet.
380  ColIndex GetSlackVariable(RowIndex row) const;
381 
382  // Populates the calling object with the dual of the LinearProgram passed as
383  // parameter.
384  // For the general form that we solve,
385  // min c.x
386  // s.t. A_1 x = b_1
387  // A_2 x <= b_2
388  // A_3 x >= b_3
389  // l <= x <= u
390  // With x: n-column of unknowns
391  // l,u: n-columns of bound coefficients
392  // c: n-row of cost coefficients
393  // A_i: m_i×n-matrix of coefficients
394  // b_i: m_i-column of right-hand side coefficients
395  //
396  // The dual is
397  //
398  // max b_1.y_1 + b_2.y_2 + b_3.y_3 + l.v + u.w
399  // s.t. y_1 A_1 + y_2 A_2 + y_3 A_3 + v + w = c
400  // y_1 free, y_2 <= 0, y_3 >= 0, v >= 0, w <= 0
401  // With:
402  // y_i: m_i-row of unknowns
403  // v,w: n-rows of unknowns
404  //
405  // If range constraints are present, each of the corresponding row is
406  // duplicated (with one becoming lower bounded and the other upper bounded).
407  // For such ranged row in the primal, duplicated_rows[row] is set to the
408  // column index in the dual of the corresponding column duplicate. For
409  // non-ranged row, duplicated_rows[row] is set to kInvalidCol.
410  //
411  // IMPORTANT: The linear_program argument must not have any free constraints.
412  //
413  // IMPORTANT: This function always interprets the argument in its minimization
414  // form. So the dual solution of the dual needs to be negated if we want to
415  // compute the solution of a maximization problem given as an argument.
416  //
417  // TODO(user): Do not interpret as a minimization problem?
418  void PopulateFromDual(const LinearProgram& dual,
419  RowToColMapping* duplicated_rows);
420 
421  // Populates the calling object with the given LinearProgram.
422  void PopulateFromLinearProgram(const LinearProgram& linear_program);
423 
424  // Populates the calling object with the given LinearProgram while permuting
425  // variables and constraints. This is useful mainly for testing to generate
426  // a model with the same optimal objective value.
428  const LinearProgram& lp, const RowPermutation& row_permutation,
429  const ColumnPermutation& col_permutation);
430 
431  // Populates the calling object with the variables of the given LinearProgram.
432  // The function preserves the bounds, the integrality, the names of the
433  // variables and their objective coefficients. No constraints are copied (the
434  // matrix in the destination has 0 rows).
435  void PopulateFromLinearProgramVariables(const LinearProgram& linear_program);
436 
437  // Adds constraints to the linear program. The constraints are specified using
438  // a sparse matrix of the coefficients, and vectors that represent the
439  // left-hand side and the right-hand side of the constraints, i.e.
440  // left_hand_sides <= coefficients * variables <= right_hand_sides.
441  // The sizes of the columns and the names must be the same as the number of
442  // rows of the sparse matrix; the number of columns of the matrix must be
443  // equal to the number of variables of the linear program.
445  const DenseColumn& left_hand_sides,
446  const DenseColumn& right_hand_sides,
448 
449  // Calls the AddConstraints method. After adding the constraints it adds slack
450  // variables to the constraints.
452  const SparseMatrix& coefficients, const DenseColumn& left_hand_sides,
453  const DenseColumn& right_hand_sides,
455  bool detect_integer_constraints_for_slack);
456 
457  // Swaps the content of this LinearProgram with the one passed as argument.
458  // Works in O(1).
459  void Swap(LinearProgram* linear_program);
460 
461  // Removes the given column indices from the LinearProgram.
462  // This needs to allocate O(num_variables) memory to update variable_table_.
463  void DeleteColumns(const DenseBooleanRow& columns_to_delete);
464 
465  // Removes slack variables from the linear program. The method restores the
466  // bounds on constraints from the bounds of the slack variables, resets the
467  // index of the first slack variable, and removes the relevant columns from
468  // the matrix.
469  void DeleteSlackVariables();
470 
471  // Scales the problem using the given scaler.
472  void Scale(SparseMatrixScaler* scaler);
473 
474  // While Scale() makes sure the coefficients inside the linear program matrix
475  // are in [-1, 1], the objective coefficients, variable bounds and constraint
476  // bounds can still take large values (originally or due to the matrix
477  // scaling).
478  //
479  // It makes a lot of sense to also scale them given that internally we use
480  // absolute tolerances, and that it is nice to have the same behavior if users
481  // scale their problems. For instance one could change the unit of ALL the
482  // variables from Bytes to MBytes if they denote memory quantities. Or express
483  // a cost in dollars instead of thousands of dollars.
484  //
485  // Here, we are quite prudent and just make sure that the range of the
486  // non-zeros magnitudes contains one. So for instance if all non-zeros costs
487  // are in [1e4, 1e6], we will divide them by 1e4 so that the new range is
488  // [1, 1e2].
489  //
490  // TODO(user): Another more aggressive idea is to set the median/mean/geomean
491  // of the magnitudes to one. Investigate if this leads to better results. It
492  // does look more robust.
493  //
494  // Both functions update objective_scaling_factor()/objective_offset() and
495  // return the scaling coefficient so that:
496  // - For ScaleObjective(), the old coefficients can be retrieved by
497  // multiplying the new ones by the returned factor.
498  // - For ScaleBounds(), the old variable and constraint bounds can be
499  // retrieved by multiplying the new ones by the returned factor.
502 
503  // Removes the given row indices from the LinearProgram.
504  // This needs to allocate O(num_variables) memory.
505  void DeleteRows(const DenseBooleanColumn& rows_to_delete);
506 
507  // Does basic checking on the linear program:
508  // - returns false if some coefficient are NaNs.
509  // - returns false if some coefficient other than the bounds are +/- infinity.
510  // Note that these conditions are also guarded by DCHECK on each of the
511  // SetXXX() function above.
512  bool IsValid() const;
513 
514  // Updates the bounds of the variables to the intersection of their original
515  // bounds and the bounds specified by variable_lower_bounds and
516  // variable_upper_bounds. If the new bounds of all variables are non-empty,
517  // returns true; otherwise, returns false.
521 
522  // Returns true if the linear program is in equation form Ax = 0 and all slack
523  // variables have been added. This is also called "computational form" in some
524  // of the literature.
525  bool IsInEquationForm() const;
526 
527  // Returns true if all integer variables in the linear program have strictly
528  // integer bounds.
529  bool BoundsOfIntegerVariablesAreInteger(Fractional tolerance) const;
530 
531  // Returns true if all integer constraints in the linear program have strictly
532  // integer bounds.
533  bool BoundsOfIntegerConstraintsAreInteger(Fractional tolerance) const;
534 
535  // Advanced usage. Bypass the costly call to CleanUp() when we known that the
536  // change we made kept the matrix columns "clean" (see the comment of
537  // CleanUp()). This is unsafe but can save a big chunk of the running time
538  // when one does a small amount of incremental changes to the problem (like
539  // adding a new row with no duplicates or zero entries).
541  DCHECK(matrix_.IsCleanedUp());
542  columns_are_known_to_be_clean_ = true;
543  }
544 
545  // If true, checks bound validity in debug mode.
546  void SetDcheckBounds(bool dcheck_bounds) { dcheck_bounds_ = dcheck_bounds; }
547 
548  // In our presolve, the calls and the extra test inside SetConstraintBounds()
549  // can be visible when a lot of substitutions are performed.
551  return &constraint_lower_bounds_;
552  }
554  return &constraint_upper_bounds_;
555  }
556 
557  private:
558  // A helper function that updates the vectors integer_variables_list_,
559  // binary_variables_list_, and non_binary_variables_list_.
560  void UpdateAllIntegerVariableLists() const;
561 
562  // A helper function to format problem statistics. Used by GetProblemStats()
563  // and GetPrettyProblemStats().
564  std::string ProblemStatFormatter(const absl::string_view format) const;
565 
566  // A helper function to format non-zero statistics. Used by GetNonZeroStats()
567  // and GetPrettyNonZeroStats().
568  std::string NonZeroStatFormatter(const absl::string_view format) const;
569 
570  // Resizes all row vectors to include index 'row'.
571  void ResizeRowsIfNeeded(RowIndex row);
572 
573  // Populates the definitions of variables, name and objective in the calling
574  // linear program with the data from the given linear program. The method does
575  // not touch the data structures for storing constraints.
576  void PopulateNameObjectiveAndVariablesFromLinearProgram(
577  const LinearProgram& linear_program);
578 
579  // Stores the linear program coefficients.
580  SparseMatrix matrix_;
581 
582  // The transpose of matrix_. This will be lazily recomputed by
583  // GetTransposeSparseMatrix() if transpose_matrix_is_consistent_ is false.
584  mutable SparseMatrix transpose_matrix_;
585 
586  // Constraint related quantities.
587  DenseColumn constraint_lower_bounds_;
588  DenseColumn constraint_upper_bounds_;
589  StrictITIVector<RowIndex, std::string> constraint_names_;
590 
591  // Variable related quantities.
592  DenseRow objective_coefficients_;
593  DenseRow variable_lower_bounds_;
594  DenseRow variable_upper_bounds_;
597 
598  // The vector of the indices of variables constrained to be integer.
599  // Note(user): the set of indices in integer_variables_list_ is the union
600  // of the set of indices in binary_variables_list_ and of the set of indices
601  // in non_binary_variables_list_ below.
602  mutable std::vector<ColIndex> integer_variables_list_;
603 
604  // The vector of the indices of variables constrained to be binary.
605  mutable std::vector<ColIndex> binary_variables_list_;
606 
607  // The vector of the indices of variables constrained to be integer, but not
608  // binary.
609  mutable std::vector<ColIndex> non_binary_variables_list_;
610 
611  // Map used to find the index of a variable based on its id.
612  absl::flat_hash_map<std::string, ColIndex> variable_table_;
613 
614  // Map used to find the index of a constraint based on its id.
615  absl::flat_hash_map<std::string, RowIndex> constraint_table_;
616 
617  // Offset of the objective, i.e. value of the objective when all variables
618  // are set to zero.
619  Fractional objective_offset_;
620  Fractional objective_scaling_factor_;
621 
622  // Boolean true (resp. false) when the problem is a maximization
623  // (resp. minimization) problem.
624  bool maximize_;
625 
626  // Boolean to speed-up multiple calls to IsCleanedUp() or
627  // CleanUp(). Mutable so IsCleanedUp() can be const.
628  mutable bool columns_are_known_to_be_clean_;
629 
630  // Whether transpose_matrix_ is guaranteed to be the transpose of matrix_.
631  mutable bool transpose_matrix_is_consistent_;
632 
633  // Whether integer_variables_list_ is consistent with the current
634  // LinearProgram.
635  mutable bool integer_variables_list_is_consistent_;
636 
637  // The name of the LinearProgram.
638  std::string name_;
639 
640  // The index of the first slack variable added to the linear program by
641  // LinearProgram::AddSlackVariablesForAllRows().
642  ColIndex first_slack_variable_;
643 
644  // If true, checks bounds in debug mode.
645  bool dcheck_bounds_ = true;
646 
647  friend void Scale(LinearProgram* lp, SparseMatrixScaler* scaler,
648  GlopParameters::ScalingAlgorithm scaling_method);
649 
650  DISALLOW_COPY_AND_ASSIGN(LinearProgram);
651 };
652 
653 // --------------------------------------------------------
654 // ProblemSolution
655 // --------------------------------------------------------
656 // Contains the solution of a LinearProgram as returned by a preprocessor.
658  ProblemSolution(RowIndex num_rows, ColIndex num_cols)
660  primal_values(num_cols, 0.0),
661  dual_values(num_rows, 0.0),
664  // The solution status.
666 
667  // The actual primal/dual solution values. This is what most clients will
668  // need, and this is enough for LPSolver to easily check the optimality.
671 
672  // The status of the variables and constraints which is difficult to
673  // reconstruct from the solution values alone. Some remarks:
674  // - From this information alone, by factorizing the basis, it is easy to
675  // reconstruct the primal and dual values.
676  // - The main difficulty to construct this from the solution values is to
677  // reconstruct the optimal basis if some basic variables are exactly at
678  // one of their bounds (and their reduced costs are close to zero).
679  // - The non-basic information (VariableStatus::FIXED_VALUE,
680  // VariableStatus::AT_LOWER_BOUND, VariableStatus::AT_UPPER_BOUND,
681  // VariableStatus::FREE) is easy to construct for variables (because
682  // they are at their exact bounds). They can be guessed for constraints
683  // (here a small precision error is unavoidable). However, it is useful to
684  // carry this exact information during post-solve.
687 
688  std::string DebugString() const;
689 };
690 
691 // Helper function to check the bounds of the SetVariableBounds() and
692 // SetConstraintBounds() functions.
694  if (std::isnan(lower_bound)) return false;
695  if (std::isnan(upper_bound)) return false;
696  if (lower_bound == kInfinity && upper_bound == kInfinity) return false;
697  if (lower_bound == -kInfinity && upper_bound == -kInfinity) return false;
698  if (lower_bound > upper_bound) return false;
699  return true;
700 }
701 
702 } // namespace glop
703 } // namespace operations_research
704 
705 #endif // OR_TOOLS_LP_DATA_LP_DATA_H_
bool IsVariableBinary(ColIndex col) const
Definition: lp_data.cc:300
std::string GetPrettyNonZeroStats() const
Definition: lp_data.cc:690
std::string GetObjectiveStatsString() const
Definition: lp_data.cc:452
const StrictITIVector< ColIndex, VariableType > variable_types() const
Definition: lp_data.h:237
const std::vector< ColIndex > & NonBinaryVariablesList() const
Definition: lp_data.cc:290
bool SolutionIsLPFeasible(const DenseRow &solution, Fractional absolute_tolerance) const
Definition: lp_data.cc:497
ProblemSolution(RowIndex num_rows, ColIndex num_cols)
Definition: lp_data.h:658
const SparseMatrix & GetTransposeSparseMatrix() const
Definition: lp_data.cc:376
void SetObjectiveCoefficient(ColIndex col, Fractional value)
Definition: lp_data.cc:326
ColIndex FindOrCreateVariable(const std::string &variable_id)
Definition: lp_data.cc:205
bool IsVariableInteger(ColIndex col) const
Definition: lp_data.cc:295
void Swap(LinearProgram *linear_program)
Definition: lp_data.cc:1031
Fractional ApplyObjectiveScalingAndOffset(Fractional value) const
Definition: lp_data.cc:550
VariableType GetVariableType(ColIndex col) const
Definition: lp_data.cc:372
void DeleteColumns(const DenseBooleanRow &columns_to_delete)
Definition: lp_data.cc:1065
bool SolutionIsInteger(const DenseRow &solution, Fractional absolute_tolerance) const
Definition: lp_data.cc:517
std::string GetPrettyProblemStats() const
Definition: lp_data.cc:664
void SetVariableType(ColIndex col, VariableType type)
Definition: lp_data.cc:236
void SetDcheckBounds(bool dcheck_bounds)
Definition: lp_data.h:546
void SetVariableName(ColIndex col, absl::string_view name)
Definition: lp_data.cc:232
ColIndex col
Definition: markowitz.cc:183
const std::vector< ColIndex > & BinaryVariablesList() const
Definition: lp_data.cc:285
RowIndex FindOrCreateConstraint(const std::string &constraint_id)
Definition: lp_data.cc:218
void SetConstraintBounds(RowIndex row, Fractional lower_bound, Fractional upper_bound)
Definition: lp_data.cc:309
void SetObjectiveOffset(Fractional objective_offset)
Definition: lp_data.cc:331
RowIndex row
Definition: markowitz.cc:182
SparseMatrix * GetMutableTransposeSparseMatrix()
Definition: lp_data.cc:386
bool AreBoundsValid(Fractional lower_bound, Fractional upper_bound)
Definition: lp_data.h:693
const DenseRow & objective_coefficients() const
Definition: lp_data.h:223
const DenseColumn & constraint_upper_bounds() const
Definition: lp_data.h:218
void ComputeSlackVariableValues(DenseRow *solution) const
Definition: lp_data.cc:535
Fractional objective_scaling_factor() const
Definition: lp_data.h:261
double upper_bound
std::string GetDimensionString() const
Definition: lp_data.cc:425
const DenseColumn & constraint_lower_bounds() const
Definition: lp_data.h:215
bool SolutionIsWithinVariableBounds(const DenseRow &solution, Fractional absolute_tolerance) const
Definition: lp_data.cc:481
void PopulateFromLinearProgramVariables(const LinearProgram &linear_program)
Definition: lp_data.cc:935
bool BoundsOfIntegerConstraintsAreInteger(Fractional tolerance) const
Definition: lp_data.cc:1500
const SparseMatrix & GetSparseMatrix() const
Definition: lp_data.h:175
Fractional GetObjectiveCoefficientForMinimizationVersion(ColIndex col) const
Definition: lp_data.cc:419
void PopulateFromPermutedLinearProgram(const LinearProgram &lp, const RowPermutation &row_permutation, const ColumnPermutation &col_permutation)
Definition: lp_data.cc:883
void AddConstraints(const SparseMatrix &coefficients, const DenseColumn &left_hand_sides, const DenseColumn &right_hand_sides, const StrictITIVector< RowIndex, std::string > &names)
Definition: lp_data.cc:972
double lower_bound
DenseColumn * mutable_constraint_lower_bounds()
Definition: lp_data.h:550
const double kInfinity
Definition: lp_types.h:84
ColIndex CreateNewSlackVariable(bool is_integer_slack_variable, Fractional lower_bound, Fractional upper_bound, const std::string &name)
Definition: lp_data.cc:176
bool SolutionIsMIPFeasible(const DenseRow &solution, Fractional absolute_tolerance) const
Definition: lp_data.cc:529
SparseColumn * GetMutableSparseColumn(ColIndex col)
Definition: lp_data.cc:413
void AddConstraintsWithSlackVariables(const SparseMatrix &coefficients, const DenseColumn &left_hand_sides, const DenseColumn &right_hand_sides, const StrictITIVector< RowIndex, std::string > &names, bool detect_integer_constraints_for_slack)
Definition: lp_data.cc:997
ColIndex GetSlackVariable(RowIndex row) const
Definition: lp_data.cc:755
const std::string & name() const
Definition: lp_data.h:75
void SetName(const std::string &name)
Definition: lp_data.h:74
ConstraintStatusColumn constraint_statuses
Definition: lp_data.h:686
#define DCHECK(condition)
Definition: base/logging.h:885
const DenseRow & variable_upper_bounds() const
Definition: lp_data.h:232
const DenseRow & variable_lower_bounds() const
Definition: lp_data.h:229
void SetVariableBounds(ColIndex col, Fractional lower_bound, Fractional upper_bound)
Definition: lp_data.cc:249
Fractional RemoveObjectiveScalingAndOffset(Fractional value) const
Definition: lp_data.cc:555
void DeleteRows(const DenseBooleanColumn &rows_to_delete)
Definition: lp_data.cc:1258
bool UpdateVariableBoundsToIntersection(const DenseRow &variable_lower_bounds, const DenseRow &variable_upper_bounds)
Definition: lp_data.cc:1006
void SetMaximizationProblem(bool maximize)
Definition: lp_data.cc:343
DenseColumn * mutable_constraint_upper_bounds()
Definition: lp_data.h:553
Collection of objects used to extend the Constraint Solver library.
std::string GetConstraintName(RowIndex row) const
Definition: lp_data.cc:366
absl::Span< const double > coefficients
std::string DumpSolution(const DenseRow &variable_values) const
Definition: lp_data.cc:647
void SetObjectiveScalingFactor(Fractional objective_scaling_factor)
Definition: lp_data.cc:336
std::string GetVariableName(ColIndex col) const
Definition: lp_data.cc:360
Fractional ScaleObjective(GlopParameters::CostScalingAlgorithm method)
Definition: lp_data.cc:1188
std::string GetBoundsStatsString() const
Definition: lp_data.cc:465
void SetCoefficient(RowIndex row, ColIndex col, Fractional value)
Definition: lp_data.cc:317
void Scale(SparseMatrixScaler *scaler)
bool BoundsOfIntegerVariablesAreInteger(Fractional tolerance) const
Definition: lp_data.cc:1484
void SetConstraintName(RowIndex row, absl::string_view name)
Definition: lp_data.cc:245
int64_t value
const std::vector< ColIndex > & IntegerVariablesList() const
Definition: lp_data.cc:280
void PopulateFromLinearProgram(const LinearProgram &linear_program)
Definition: lp_data.cc:862
const SparseColumn & GetSparseColumn(ColIndex col) const
Definition: lp_data.cc:409
void PopulateFromDual(const LinearProgram &dual, RowToColMapping *duplicated_rows)
Definition: lp_data.cc:764
void AddSlackVariablesWhereNecessary(bool detect_integer_constraints)
Definition: lp_data.cc:697