XPRSpresolverow

Purpose

Presolves a row formulated in terms of the original variables such that it can be added to a presolved matrix.

Topic areas

Synopsis

 int XPRS_CC XPRSpresolverow(XPRSprob prob, char rowtype, int norigcoefs, const int origcolind[], const double origrowcoef[], double origrhs, int maxcoefs, int * p_ncoefs, int colind[], double rowcoef[], double * p_rhs, int * p_status);

Arguments

prob

The current problem.

rowtype

The type of the row:

L	indicates a ≤ row;
G	indicates a ≥ row.
E	indicates a ≥ row.

norigcoefs

Number of elements in the origcolind and origrowcoef arrays.

origcolind

Integer array of length norigcoefs containing the column indices of the row to presolve.

origrowcoef

Double array of length norigcoefs containing the non-zero coefficients of the row to presolve.

origrhs

The right-hand side constant of the row to presolve.

maxcoefs

Maximum number of elements to return in the colind and rowcoef arrays. The function will not write more elements than this to colind or rowcoef, even if the presolved row is longer.

p_ncoefs

Pointer to the integer where the number of non-zero elements in the presolved row will be returned (this may be bigger than maxcoefs).

colind

Integer array of length maxcoefs which will be filled with the column indices of the presolved row. May be NULL if not required. Allocate this to COLS elements and pass COLS as maxcoef> to make sure it is large enough to hold the presolved row.

rowcoef

Double array of length maxcoefs which will be filled with the coefficients of the presolved row. May be NULL if not required. Allocate this to COLS elements and pass COLS as maxcoef> to make sure it is large enough to hold the presolved row.

p_rhs

Pointer to the double where the presolved right-hand side will be returned.

p_status

Status of the presolved row:

-5	Failed to presolve the row due to presolve operations making the row nonlinear;
-4	Failed to presolve the equality row due to presolve operations requiring relaxation of the row;
-3	Failed to presolve the row due to presolve dual reductions;
-2	Failed to presolve the row due to presolve duplicate column reductions;
-1	Failed to presolve the row due to an error. Check the Optimizer error code for the cause;
0	The row was successfully presolved;
1	The row was presolved, but may be relaxed.

May be NULL if not required.

Related controls

Integer

PRESOLVE	Turns presolve on or off.
PRESOLVEOPS	Selects the presolve operations.

Example

Suppose we want to add the row 2x ₁ + x ₂ ≤ 1 to our presolved matrix. This could be done in the following way:

int mindo[] = { 1, 2 };
int origrowcoef[] = { 2.0, 1 };
char rowtype = 'L';
double origrhs = 1;
int ncols, ncoefs, status, mtype, mstart[2], *mindp;
double rhs, *rowcoef;
...
XPRSgetintattrib(prob, XPRS_COLS, &ncols);
mindp = (int*) malloc(ncols*sizeof(int));
rowcoef = (double*) malloc(ncols*sizeof(double));
XPRSpresolverow(prob, rowtype, 2, mindo, origrowcoef, origrhs, ncols,
                &ncoefs, mindp, rowcoef, &rhs, &status);
if (status >= 0) {
  mtype = 0;
  mstart[0] = 0; mstart[1] = ncoefs;
  XPRSaddcuts(prob, 1, &mtype, &rowtype, &rhs, mstart, mindp,
              rowcoef);
}

See also examples els_usercuts.c, tsp.c.

Further information

1. There are certain presolve operations that can prevent a row from being presolved exactly. If the row contains a coefficient for a column that was eliminated due to duplicate column reductions or singleton column reductions, the row might have to be relaxed to remain valid for the presolved problem. The relaxation will be done automatically by the XPRSpresolverow function, but a return status of +1 will be returned. If it is not possible to relax the row, a status of -2 will be returned instead. Likewise, it is possible that certain dual reductions prevents the row from being presolved. In such a case a status of -3 will be returned instead. Also transforming a convex quadratic objective transfer row to a convex quadratic objective and nonlinear eliminations will lead to a row not being presolved and return a status of -5 instead.

2. If XPRSpresolverow will be used for presolving hard constraints, such as e.g. branching bounds or delayed constraints, then dual reductions should be disabled, by setting MIPDUALREDUCTIONS to 0.

3. If the user knows in advance which columns will have non-zero coefficients in rows that will be presolved, it is possible to protect these individual columns through the XPRSloadsecurevecs function. This way the Optimizer is left free to apply all possible reductions to the remaining columns.

Contents

Index

Glossary

Search Results

XPRSpresolverow