CoinPackedMatrix Class Reference

Sparse Matrix Base Class. More...

#include <CoinPackedMatrix.hpp>

Inheritance diagram for CoinPackedMatrix:

Public Member Functions
Query members

double	getExtraGap () const
	Return the current setting of the extra gap.
double	getExtraMajor () const
	Return the current setting of the extra major.
void	reserve (const int newMaxMajorDim, const CoinBigIndex newMaxSize, bool create=false)
	Reserve sufficient space for appending major-ordered vectors.
void	clear ()
	Clear the data, but do not free any arrays.
bool	isColOrdered () const
	Whether the packed matrix is column major ordered or not.
CoinBigIndex	getNumElements () const
	Number of entries in the packed matrix.
int	getNumCols () const
	Number of columns.
int	getNumRows () const
	Number of rows.
const double *	getElements () const
	A vector containing the elements in the packed matrix.
const int *	getIndices () const
	A vector containing the minor indices of the elements in the packed matrix.
int	getSizeVectorStarts () const
	The size of the `vectorStarts` array.
int	getSizeVectorLengths () const
	The size of the `vectorLengths` array.
const CoinBigIndex *	getVectorStarts () const
	The positions where the major-dimension vectors start in elements and indices.
const int *	getVectorLengths () const
	The lengths of the major-dimension vectors.
CoinBigIndex	getVectorFirst (const int i) const
	The position of the first element in the i'th major-dimension vector.
CoinBigIndex	getVectorLast (const int i) const
	The position of the last element (well, one entry past the last) in the i'th major-dimension vector.
int	getVectorSize (const int i) const
	The length of i'th vector.
const CoinShallowPackedVector	getVector (int i) const
	Return the i'th vector in matrix.
int *	getMajorIndices () const
	Returns an array containing major indices.
Modifying members.

void	setDimensions (int numrows, int numcols)
	Set the dimansions of the matrix.
void	setExtraGap (const double newGap)
	Set the extra gap to be allocated to the specified value.
void	setExtraMajor (const double newMajor)
	Set the extra major to be allocated to the specified value.
void	appendCol (const CoinPackedVectorBase &vec)
	Append a column to the end of the matrix.
void	appendCol (const int vecsize, const int vecind, const double vecelem)
	Append a column to the end of the matrix.
void	appendCols (const int numcols, const CoinPackedVectorBase const cols)
	Append a set of columns to the end of the matrix.
int	appendCols (const int numcols, const CoinBigIndex columnStarts, const int row, const double *element, int numberRows=-1)
	Append a set of columns to the end of the matrix.
void	appendRow (const CoinPackedVectorBase &vec)
	Append a row to the end of the matrix.
void	appendRow (const int vecsize, const int vecind, const double vecelem)
	Append a row to the end of the matrix.
void	appendRows (const int numrows, const CoinPackedVectorBase const rows)
	Append a set of rows to the end of the matrix.
int	appendRows (const int numrows, const CoinBigIndex rowStarts, const int column, const double *element, int numberColumns=-1)
	Append a set of rows to the end of the matrix.
void	rightAppendPackedMatrix (const CoinPackedMatrix &matrix)
	Append the argument to the "right" of the current matrix.
void	bottomAppendPackedMatrix (const CoinPackedMatrix &matrix)
	Append the argument to the "bottom" of the current matrix.
void	deleteCols (const int numDel, const int *indDel)
	Delete the columns whose indices are listed in `indDel`.
void	deleteRows (const int numDel, const int *indDel)
	Delete the rows whose indices are listed in `indDel`.
void	replaceVector (const int index, const int numReplace, const double *newElements)
	Replace the elements of a vector.
void	modifyCoefficient (int row, int column, double newElement, bool keepZero=false)
	Modify one element of packed matrix.
double	getCoefficient (int row, int column) const
	Return one element of packed matrix.
int	compress (double threshold)
	Eliminate all elements in matrix whose absolute value is less than threshold.
int	eliminateDuplicates (double threshold)
	Eliminate all duplicate AND small elements in matrix The column starts are not affected.
void	orderMatrix ()
	Sort all columns so indices are increasing.in each column.
int	cleanMatrix (double threshold=1.0e-20)
	Really clean up matrix.
Methods that reorganize the whole matrix

void	removeGaps (double removeValue=-1.0)
	Remove the gaps from the matrix if there were any Can also remove small elements fabs() <= removeValue.
void	submatrixOf (const CoinPackedMatrix &matrix, const int numMajor, const int *indMajor)
	Extract a submatrix from matrix.
void	submatrixOfWithDuplicates (const CoinPackedMatrix &matrix, const int numMajor, const int *indMajor)
	Extract a submatrix from matrix.
void	copyOf (const CoinPackedMatrix &rhs)
	Copy method.
void	copyOf (const bool colordered, const int minor, const int major, const CoinBigIndex numels, const double elem, const int ind, const CoinBigIndex start, const int len, const double extraMajor=0.0, const double extraGap=0.0)
	Copy the arguments to the matrix.
void	copyReuseArrays (const CoinPackedMatrix &rhs)
	Copy method.
void	reverseOrderedCopyOf (const CoinPackedMatrix &rhs)
	Reverse copy method.
void	assignMatrix (const bool colordered, const int minor, const int major, const CoinBigIndex numels, double &elem, int &ind, CoinBigIndex &start, int &len, const int maxmajor=-1, const CoinBigIndex maxsize=-1)
	Assign the arguments to the matrix.
CoinPackedMatrix &	operator= (const CoinPackedMatrix &rhs)
	Assignment operator.
void	reverseOrdering ()
	Reverse the ordering of the packed matrix.
void	transpose ()
	Transpose the matrix.
void	swap (CoinPackedMatrix &matrix)
	Swap the content of the two packed matrix.
Matrix times vector methods

void	times (const double x, double y) const
	Return `A * x` in `y`.
void	times (const CoinPackedVectorBase &x, double *y) const
	Return `A * x` in `y`.
void	transposeTimes (const double x, double y) const
	Return `x * A` in `y`.
void	transposeTimes (const CoinPackedVectorBase &x, double *y) const
	Return `x * A` in `y`.
Queries

int *	countOrthoLength () const
	Count the number of entries in every minor-dimension vector and return an array containing these lengths.
void	countOrthoLength (int *counts) const
	Count the number of entries in every minor-dimension vector and fill in an array containing these lengths.
int	getMajorDim () const
	Major dimension.
int	getMinorDim () const
	Minor dimension.
int	getMaxMajorDim () const
	Current maximum for major dimension.
void	dumpMatrix (const char *fname=NULL) const
	Dump the matrix on stdout.
void	printMatrixElement (const int row_val, const int col_val) const
	Print a single matrix element.
Append vectors. <br>
When libosi is compiled with a COIN_DEBUG defined then these methods throws an exception if the major (minor) vector contains an index that's larger than the minor (major) dimension (-1). Otherwise the methods assume that every index fits into the matrix.
void	appendMajorVector (const CoinPackedVectorBase &vec)
	Append a major-dimension vector to the end of the matrix.
void	appendMajorVector (const int vecsize, const int vecind, const double vecelem)
	Append a major-dimension vector to the end of the matrix.
void	appendMajorVectors (const int numvecs, const CoinPackedVectorBase const vecs)
	Append several major-dimensonvectors to the end of the matrix.
void	appendMinorVector (const CoinPackedVectorBase &vec)
	Append a minor-dimension vector to the end of the matrix.
void	appendMinorVector (const int vecsize, const int vecind, const double vecelem)
	Append a minor-dimension vector to the end of the matrix.
void	appendMinorVectors (const int numvecs, const CoinPackedVectorBase const vecs)
	Append several minor-dimensonvectors to the end of the matrix.
Append matrices.
We'll document these methods assuming that the current matrix is column major ordered (Hence in the `...SameOrdered()` methods the argument is column ordered, in the `OrthoOrdered()` methods the argument is row ordered. )
void	majorAppendSameOrdered (const CoinPackedMatrix &matrix)
	Append the columns of the argument to the right end of this matrix.
void	minorAppendSameOrdered (const CoinPackedMatrix &matrix)
	Append the columns of the argument to the bottom end of this matrix.
void	majorAppendOrthoOrdered (const CoinPackedMatrix &matrix)
	Append the rows of the argument to the right end of this matrix.
void	minorAppendOrthoOrdered (const CoinPackedMatrix &matrix)
	Append the rows of the argument to the bottom end of this matrix.
Delete vectors

void	deleteMajorVectors (const int numDel, const int *indDel)
	Delete the major-dimension vectors whose indices are listed in `indDel`.
void	deleteMinorVectors (const int numDel, const int *indDel)
	Delete the minor-dimension vectors whose indices are listed in `indDel`.
Various dot products.

void	timesMajor (const double x, double y) const
	Return `A * x` (multiplied from the "right" direction) in `y`.
void	timesMajor (const CoinPackedVectorBase &x, double *y) const
	Return `A * x` (multiplied from the "right" direction) in `y`.
void	timesMinor (const double x, double y) const
	Return `A * x` (multiplied from the "right" direction) in `y`.
void	timesMinor (const CoinPackedVectorBase &x, double *y) const
	Return `A * x` (multiplied from the "right" direction) in `y`.
Logical Operations.

template<class FloatEqual >
bool	isEquivalent (const CoinPackedMatrix &rhs, const FloatEqual &eq) const
	Equivalence.
bool	isEquivalent2 (const CoinPackedMatrix &rhs) const
	Equivalence.
bool	isEquivalent (const CoinPackedMatrix &rhs) const
	The default equivalence test is that the entries are relatively equal.
non const methods.
This is to be used with great care when doing column generation etc
double *	getMutableElements () const
	A vector containing the elements in the packed matrix.
int *	getMutableIndices () const
	A vector containing the minor indices of the elements in the packed matrix.
CoinBigIndex *	getMutableVectorStarts () const
	The positions where the major-dimension vectors start in elements and indices.
int *	getMutableVectorLengths () const
	The lengths of the major-dimension vectors.
void	setNumElements (CoinBigIndex value)
	Change size after modifying - be careful.
void	nullElementArray ()
	NULLify element array - used when space is very tight.
void	nullStartArray ()
	NULLify start array - used when space is very tight.
void	nullLengthArray ()
	NULLify length array - used when space is very tight.
void	nullIndexArray ()
	NULLify index array - used when space is very tight.
Constructors, destructors and major modifying methods

	CoinPackedMatrix ()
	Default Constructor creates an empty column ordered packed matrix.
	CoinPackedMatrix (const bool colordered, const double extraMajor, const double extraGap)
	A constructor where the ordering and the gaps are specified.
	CoinPackedMatrix (const bool colordered, const int minor, const int major, const CoinBigIndex numels, const double elem, const int ind, const CoinBigIndex start, const int len, const double extraMajor, const double extraGap)
	Default Constructor creates an empty column ordered packed matrix.
	CoinPackedMatrix (const bool colordered, const int minor, const int major, const CoinBigIndex numels, const double elem, const int ind, const CoinBigIndex start, const int len)
	Default Constructor creates an empty column ordered packed matrix.
	CoinPackedMatrix (const bool colordered, const int rowIndices, const int colIndices, const double *elements, CoinBigIndex numels)
	Create packed matrix from triples.
	CoinPackedMatrix (const CoinPackedMatrix &m)
	Copy constructor.
	CoinPackedMatrix (const CoinPackedMatrix &m, int extraForMajor, int extraElements, bool reverseOrdering=false)
	Copy constructor - fine tuning - allowing extra space and/or reverse ordering.
	CoinPackedMatrix (const CoinPackedMatrix &wholeModel, int numberRows, const int whichRows, int numberColumns, const int whichColumns)
	Subset constructor (without gaps).
virtual	~CoinPackedMatrix ()
	Destructor.
Protected Member Functions
void	gutsOfDestructor ()
void	gutsOfCopyOf (const bool colordered, const int minor, const int major, const CoinBigIndex numels, const double elem, const int ind, const CoinBigIndex start, const int len, const double extraMajor=0.0, const double extraGap=0.0)
void	gutsOfCopyOfNoGaps (const bool colordered, const int minor, const int major, const double elem, const int ind, const CoinBigIndex *start)
	When no gaps we can do faster.
void	gutsOfOpEqual (const bool colordered, const int minor, const int major, const CoinBigIndex numels, const double elem, const int ind, const CoinBigIndex start, const int len)
void	resizeForAddingMajorVectors (const int numVec, const int *lengthVec)
void	resizeForAddingMinorVectors (const int *addedEntries)
int	appendMajor (const int number, const CoinBigIndex starts, const int index, const double *element, int numberOther=-1)
	Append a set of rows/columns to the end of the matrix.
int	appendMinor (const int number, const CoinBigIndex starts, const int index, const double *element, int numberOther=-1)
	Append a set of rows/columns to the end of the matrix.
Protected Attributes
Data members
The data members are protected to allow access for derived classes.
bool	colOrdered_
	A flag indicating whether the matrix is column or row major ordered.
double	extraGap_
	This much times more space should be allocated for each major-dimension vector (with respect to the number of entries in the vector) when the matrix is resized.
double	extraMajor_
	his much times more space should be allocated for major-dimension vectors when the matrix is resized.
double *	element_
	List of nonzero element values.
int *	index_
	List of nonzero element minor-dimension indices.
CoinBigIndex *	start_
	Starting positions of major-dimension vectors.
int *	length_
	Lengths of major-dimension vectors.
int	majorDim_
	number of vectors in matrix
int	minorDim_
	size of other dimension
CoinBigIndex	size_
	the number of nonzero entries
int	maxMajorDim_
	max space allocated for major-dimension
CoinBigIndex	maxSize_
	max space allocated for entries
Private Member Functions
CoinBigIndex	getLastStart () const
Friends
void	CoinPackedMatrixUnitTest ()
	A function that tests the methods in the CoinPackedMatrix class.

Detailed Description

Sparse Matrix Base Class.

This class is used for storing a matrix by rows or columns.

The sparse represention can be completely compact or it can have "extra" space. The extra space can be added at the end of rows and/or columns. Incorporating extra space into the sparse matrix representation can improve performance in cases where new data needs to be inserted into the packed matrix.

For example if the matrix:

     3  1  0   -2   -1  0  0   -1                 
     0  2  1.1  0    0  0  0    0                       
     0  0  1    0    0  1  0    0         
     0  0  0    2.8  0  0 -1.2  0   
   5.6  0  0    0    1  0  0    1.9

  was stored by rows (with no extra space) in 
  CoinPackedMatrix r then: 
    r.getElements() returns a vector containing: 
      3 1 -2 -1 -1 2 1.1 1 1 2.8 -1.2 5.6 1 1.9 
    r.getIndices() returns a vector containing: 
      0 1  3  4  7 1 2   2 5 3    6   0   4 7 
    r.getVectorStarts() returns a vector containing: 
      0 5 7 9 11 14 
    r.getNumElements() returns 14. 
    r.getMajorDim() returns 5. 
    r.getVectorSize(0) returns 5. 
    r.getVectorSize(1) returns 2. 
    r.getVectorSize(2) returns 2. 
    r.getVectorSize(3) returns 2. 
    r.getVectorSize(4) returns 3. 
 
  If stored by columns (with no extra space) then: 
    c.getElements() returns a vector containing: 
      3 5.6 1 2 1.1 1 -2 2.8 -1 1 1 -1.2 -1 1.9 
    c.getIndices() returns a vector containing: 
      0  4  0 1 1   2  0 3    0 4 2  3    0 4 
    c.getVectorStarts() returns a vector containing: 
      0 2 4 6 8 10 11 12 14 
    c.getNumElements() returns 14. 
    c.getMajorDim() returns 8.

Definition at line 61 of file CoinPackedMatrix.hpp.

Constructor & Destructor Documentation

CoinPackedMatrix::CoinPackedMatrix ( )

Default Constructor creates an empty column ordered packed matrix.

CoinPackedMatrix::CoinPackedMatrix	(	const bool	colordered,
		const double	extraMajor,
		const double	extraGap
	)

A constructor where the ordering and the gaps are specified.

CoinPackedMatrix::CoinPackedMatrix	(	const bool	colordered,
		const int	minor,
		const int	major,
		const CoinBigIndex	numels,
		const double *	elem,
		const int *	ind,
		const CoinBigIndex *	start,
		const int *	len,
		const double	extraMajor,
		const double	extraGap
	)

Default Constructor creates an empty column ordered packed matrix.

CoinPackedMatrix::CoinPackedMatrix	(	const bool	colordered,
		const int	minor,
		const int	major,
		const CoinBigIndex	numels,
		const double *	elem,
		const int *	ind,
		const CoinBigIndex *	start,
		const int *	len
	)

Default Constructor creates an empty column ordered packed matrix.

CoinPackedMatrix::CoinPackedMatrix	(	const bool	colordered,
		const int *	rowIndices,
		const int *	colIndices,
		const double *	elements,
		CoinBigIndex	numels
	)

Create packed matrix from triples.

If colordered is true then the created matrix will be column ordered. Duplicate matrix elements are allowed. The created matrix will have the sum of the duplicates.
For example if:
rowIndices[0]=2; colIndices[0]=5; elements[0]=2.0
rowIndices[1]=2; colIndices[1]=5; elements[1]=0.5
then the created matrix will contain a value of 2.5 in row 2 and column 5.
The matrix is created without gaps.

CoinPackedMatrix::CoinPackedMatrix ( const CoinPackedMatrix & m )

Copy constructor.

CoinPackedMatrix::CoinPackedMatrix	(	const CoinPackedMatrix &	m,
		int	extraForMajor,
		int	extraElements,
		bool	reverseOrdering = `false`
	)

Copy constructor - fine tuning - allowing extra space and/or reverse ordering.

extraForMajor is exact extra after any possible reverse ordering. extraMajor_ and extraGap_ set to zero.

CoinPackedMatrix::CoinPackedMatrix	(	const CoinPackedMatrix &	wholeModel,
		int	numberRows,
		const int *	whichRows,
		int	numberColumns,
		const int *	whichColumns
	)

Subset constructor (without gaps).

Duplicates are allowed and order is as given

virtual CoinPackedMatrix::~CoinPackedMatrix ( ) [virtual]

Destructor.

Member Function Documentation

double CoinPackedMatrix::getExtraGap ( ) const [inline]

Return the current setting of the extra gap.

Definition at line 71 of file CoinPackedMatrix.hpp.

double CoinPackedMatrix::getExtraMajor ( ) const [inline]

Return the current setting of the extra major.

Definition at line 73 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::reserve	(	const int	newMaxMajorDim,
		const CoinBigIndex	newMaxSize,
		bool	create = `false`
	)

Reserve sufficient space for appending major-ordered vectors.

If create is true, empty columns are created (for column generation)

void CoinPackedMatrix::clear ( )

Clear the data, but do not free any arrays.

Reimplemented in BCP_lp_relax.

bool CoinPackedMatrix::isColOrdered ( ) const [inline]

Whether the packed matrix is column major ordered or not.

Definition at line 83 of file CoinPackedMatrix.hpp.

CoinBigIndex CoinPackedMatrix::getNumElements ( ) const [inline]

Number of entries in the packed matrix.

Definition at line 85 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getNumCols ( ) const [inline]

Number of columns.

Definition at line 87 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getNumRows ( ) const [inline]

Number of rows.

Definition at line 89 of file CoinPackedMatrix.hpp.

const double* CoinPackedMatrix::getElements ( ) const [inline]

A vector containing the elements in the packed matrix.

Note that there might be gaps in this list, entries that do not belong to any major-dimension vector. To get the actual elements one should look at this vector together with vectorStarts and vectorLengths.

Definition at line 95 of file CoinPackedMatrix.hpp.

const int* CoinPackedMatrix::getIndices ( ) const [inline]

A vector containing the minor indices of the elements in the packed matrix.

Note that there might be gaps in this list, entries that do not belong to any major-dimension vector. To get the actual elements one should look at this vector together with vectorStarts and vectorLengths.

Definition at line 101 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getSizeVectorStarts ( ) const [inline]

The size of the vectorStarts array.

Definition at line 104 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getSizeVectorLengths ( ) const [inline]

The size of the vectorLengths array.

Definition at line 106 of file CoinPackedMatrix.hpp.

const CoinBigIndex* CoinPackedMatrix::getVectorStarts ( ) const [inline]

The positions where the major-dimension vectors start in elements and indices.

Definition at line 109 of file CoinPackedMatrix.hpp.

const int* CoinPackedMatrix::getVectorLengths ( ) const [inline]

The lengths of the major-dimension vectors.

Definition at line 111 of file CoinPackedMatrix.hpp.

CoinBigIndex CoinPackedMatrix::getVectorFirst ( const int i ) const [inline]

The position of the first element in the i'th major-dimension vector.

Definition at line 116 of file CoinPackedMatrix.hpp.

CoinBigIndex CoinPackedMatrix::getVectorLast ( const int i ) const [inline]

The position of the last element (well, one entry past the last) in the i'th major-dimension vector.

Definition at line 123 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getVectorSize ( const int i ) const [inline]

The length of i'th vector.

Definition at line 129 of file CoinPackedMatrix.hpp.

const CoinShallowPackedVector CoinPackedMatrix::getVector ( int i ) const [inline]

Return the i'th vector in matrix.

Definition at line 136 of file CoinPackedMatrix.hpp.

int* CoinPackedMatrix::getMajorIndices ( ) const

Returns an array containing major indices.

The array is getNumElements long and if getVectorStarts() is 0,2,5 then the array would start 0,0,1,1,1,2... This method is provided to go back from a packed format to a triple format. It returns NULL if there are gaps in matrix so user should use removeGaps() if there are any gaps. It does this as this array has to match getElements() and getIndices() and because it makes no sense otherwise. The returned array is allocated with new int[], free it with delete[].

void CoinPackedMatrix::setDimensions	(	int	numrows,
		int	numcols
	)

Set the dimansions of the matrix.

In effect, append new empty columns/rows to the matrix. A negative number for either dimension means that that dimension doesn't change. Otherwise the new dimensions MUST be at least as large as the current ones otherwise an exception is thrown.

void CoinPackedMatrix::setExtraGap ( const double newGap )

Set the extra gap to be allocated to the specified value.

void CoinPackedMatrix::setExtraMajor ( const double newMajor )

Set the extra major to be allocated to the specified value.

void CoinPackedMatrix::appendCol ( const CoinPackedVectorBase & vec )

Append a column to the end of the matrix.

When libosi is compiled with a COIN_DEBUG defined then this method throws an exception if the new column contains an index that's larger than the number of rows (-1). Otherwise the method assumes that every index fits into the matrix.

void CoinPackedMatrix::appendCol	(	const int	vecsize,
		const int *	vecind,
		const double *	vecelem
	)

Append a column to the end of the matrix.

When libosi is compiled with a COIN_DEBUG defined then this method throws an exception if the new column contains an index that's larger than the number of rows (-1). Otherwise the method assumes that every index fits into the matrix.

void CoinPackedMatrix::appendCols	(	const int	numcols,
		const CoinPackedVectorBase const	cols
	)

Append a set of columns to the end of the matrix.

When libosi is compiled with a COIN_DEBUG defined then this method throws an exception if any of the new columns contain an index that's larger than the number of rows (-1). Otherwise the method assumes that every index fits into the matrix.

int CoinPackedMatrix::appendCols	(	const int	numcols,
		const CoinBigIndex *	columnStarts,
		const int *	row,
		const double *	element,
		int	numberRows = `-1`
	)

Append a set of columns to the end of the matrix.

Returns number of errors i.e. if any of the new columns contain an index that's larger than the number of rows-1 (if numberRows>0) or duplicates (if numberRows>0).

void CoinPackedMatrix::appendRow ( const CoinPackedVectorBase & vec )

Append a row to the end of the matrix.

When libosi is compiled with a COIN_DEBUG defined then this method throws an exception if the new row contains an index that's larger than the number of columns (-1). Otherwise the method assumes that every index fits into the matrix.

void CoinPackedMatrix::appendRow	(	const int	vecsize,
		const int *	vecind,
		const double *	vecelem
	)

Append a row to the end of the matrix.

When libosi is compiled with a COIN_DEBUG defined then this method throws an exception if the new row contains an index that's larger than the number of columns (-1). Otherwise the method assumes that every index fits into the matrix.

void CoinPackedMatrix::appendRows	(	const int	numrows,
		const CoinPackedVectorBase const	rows
	)

Append a set of rows to the end of the matrix.

When libosi is compiled with a COIN_DEBUG defined then this method throws an exception if any of the new rows contain an index that's larger than the number of columns (-1). Otherwise the method assumes that every index fits into the matrix.

int CoinPackedMatrix::appendRows	(	const int	numrows,
		const CoinBigIndex *	rowStarts,
		const int *	column,
		const double *	element,
		int	numberColumns = `-1`
	)

Append a set of rows to the end of the matrix.

Returns number of errors i.e. if any of the new rows contain an index that's larger than the number of columns-1 (if numberColumns>0) or duplicates (if numberColumns>0).

void CoinPackedMatrix::rightAppendPackedMatrix ( const CoinPackedMatrix & matrix )

Append the argument to the "right" of the current matrix.

Imagine this as adding new columns (don't worry about how the matrices are ordered, that is taken care of). An exception is thrown if the number of rows is different in the matrices.

void CoinPackedMatrix::bottomAppendPackedMatrix ( const CoinPackedMatrix & matrix )

Append the argument to the "bottom" of the current matrix.

Imagine this as adding new rows (don't worry about how the matrices are ordered, that is taken care of). An exception is thrown if the number of columns is different in the matrices.

void CoinPackedMatrix::deleteCols	(	const int	numDel,
		const int *	indDel
	)

Delete the columns whose indices are listed in indDel.

void CoinPackedMatrix::deleteRows	(	const int	numDel,
		const int *	indDel
	)

Delete the rows whose indices are listed in indDel.

void CoinPackedMatrix::replaceVector	(	const int	index,
		const int	numReplace,
		const double *	newElements
	)

Replace the elements of a vector.

The indices remain the same. At most the number specified will be replaced. The index is between 0 and major dimension of matrix

void CoinPackedMatrix::modifyCoefficient	(	int	row,
		int	column,
		double	newElement,
		bool	keepZero = `false`
	)

Modify one element of packed matrix.

An element may be added. This works for either ordering If the new element is zero it will be deleted unless keepZero true

double CoinPackedMatrix::getCoefficient	(	int	row,
		int	column
	)			const

Return one element of packed matrix.

This works for either ordering If it is not present will return 0.0

int CoinPackedMatrix::compress ( double threshold )

Eliminate all elements in matrix whose absolute value is less than threshold.

The column starts are not affected. Returns number of elements eliminated. Elements eliminated are at end of each vector

int CoinPackedMatrix::eliminateDuplicates ( double threshold )

Eliminate all duplicate AND small elements in matrix The column starts are not affected.

Returns number of elements eliminated.

void CoinPackedMatrix::orderMatrix ( )

Sort all columns so indices are increasing.in each column.

int CoinPackedMatrix::cleanMatrix ( double threshold = 1.0e-20 )

Really clean up matrix.

a) eliminate all duplicate AND small elements in matrix b) remove all gaps and set extraGap_ and extraMajor_ to 0.0 c) reallocate arrays and make max lengths equal to lengths d) orders elements returns number of elements eliminated

void CoinPackedMatrix::removeGaps ( double removeValue = -1.0 )

Remove the gaps from the matrix if there were any Can also remove small elements fabs() <= removeValue.

void CoinPackedMatrix::submatrixOf	(	const CoinPackedMatrix &	matrix,
		const int	numMajor,
		const int *	indMajor
	)

Extract a submatrix from matrix.

Those major-dimension vectors of the matrix comprise the submatrix whose indices are given in the arguments. Does not allow duplicates.

void CoinPackedMatrix::submatrixOfWithDuplicates	(	const CoinPackedMatrix &	matrix,
		const int	numMajor,
		const int *	indMajor
	)

Extract a submatrix from matrix.

Those major-dimension vectors of the matrix comprise the submatrix whose indices are given in the arguments. Allows duplicates and keeps order.

void CoinPackedMatrix::copyOf ( const CoinPackedMatrix & rhs )

Copy method.

This method makes an exact replica of the argument, including the extra space parameters.

void CoinPackedMatrix::copyOf	(	const bool	colordered,
		const int	minor,
		const int	major,
		const CoinBigIndex	numels,
		const double *	elem,
		const int *	ind,
		const CoinBigIndex *	start,
		const int *	len,
		const double	extraMajor = `0.0`,
		const double	extraGap = `0.0`
	)

Copy the arguments to the matrix.

If len is a NULL pointer then the matrix is assumed to have no gaps in it and len will be created accordingly.

void CoinPackedMatrix::copyReuseArrays ( const CoinPackedMatrix & rhs )

Copy method.

This method makes an exact replica of the argument, including the extra space parameters. If there is room it will re-use arrays

void CoinPackedMatrix::reverseOrderedCopyOf ( const CoinPackedMatrix & rhs )

Reverse copy method.

This method makes an exact replica of the argument, but the major ordering reversed. The extra space parameters are copied and reversed, too.

void CoinPackedMatrix::assignMatrix	(	const bool	colordered,
		const int	minor,
		const int	major,
		const CoinBigIndex	numels,
		double *&	elem,
		int *&	ind,
		CoinBigIndex *&	start,
		int *&	len,
		const int	maxmajor = `-1`,
		const CoinBigIndex	maxsize = `-1`
	)

Assign the arguments to the matrix.

If len is a NULL pointer then the matrix is assumed to have no gaps in it and len will be created accordingly.
NOTE 1: After this method returns the pointers passed to the method will be NULL pointers!
NOTE 2: When the matrix is eventually destructed the arrays will be deleted by delete[]. Hence one should use this method ONLY if all array swere allocated by new[]!

CoinPackedMatrix& CoinPackedMatrix::operator= ( const CoinPackedMatrix & rhs )

Assignment operator.

This copies out the data, but uses the current matrix's extra space parameters.

Reimplemented in BCP_lp_relax.

void CoinPackedMatrix::reverseOrdering ( )

Reverse the ordering of the packed matrix.

void CoinPackedMatrix::transpose ( )

Transpose the matrix.

NOTE: All this routine does is to flip the ordering! Of course, then the matrix describes the transposed matrix. To get the matrix physically transposed (e.g., for a column ordered matrix to get the transpose in column ordered format) one has to invoke this method AND reverseOrdering().

void CoinPackedMatrix::swap ( CoinPackedMatrix & matrix )

Swap the content of the two packed matrix.

void CoinPackedMatrix::times	(	const double *	x,
		double *	y
	)			const

Return A * x in y.

Precondition:: x must be of size numColumns(); y must be of size numRows()

void CoinPackedMatrix::times	(	const CoinPackedVectorBase &	x,
		double *	y
	)			const

Return A * x in y.

Same as the previous method, just x is given in the form of a packed vector.

void CoinPackedMatrix::transposeTimes	(	const double *	x,
		double *	y
	)			const

Return x * A in y.

Precondition:: x must be of size numRows(); y must be of size numColumns()

void CoinPackedMatrix::transposeTimes	(	const CoinPackedVectorBase &	x,
		double *	y
	)			const

Return x * A in y.

Same as the previous method, just x is given in the form of a packed vector.

int* CoinPackedMatrix::countOrthoLength ( ) const

Count the number of entries in every minor-dimension vector and return an array containing these lengths.

The returned array is allocated with new int[], free it with delete[].

void CoinPackedMatrix::countOrthoLength ( int * counts ) const

Count the number of entries in every minor-dimension vector and fill in an array containing these lengths.

int CoinPackedMatrix::getMajorDim ( ) const [inline]

Major dimension.

For row ordered matrix this would be the number of rows.

Definition at line 410 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getMinorDim ( ) const [inline]

Minor dimension.

For row ordered matrix this would be the number of columns.

Definition at line 413 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getMaxMajorDim ( ) const [inline]

Current maximum for major dimension.

For row ordered matrix this many rows can be added without reallocating the vector related to the major dimension (start_ and length_).

Definition at line 417 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::dumpMatrix ( const char * fname = NULL ) const

Dump the matrix on stdout.

When in dire straits this method can help.

void CoinPackedMatrix::printMatrixElement	(	const int	row_val,
		const int	col_val
	)			const

Print a single matrix element.

void CoinPackedMatrix::appendMajorVector ( const CoinPackedVectorBase & vec )

Append a major-dimension vector to the end of the matrix.

void CoinPackedMatrix::appendMajorVector	(	const int	vecsize,
		const int *	vecind,
		const double *	vecelem
	)

Append a major-dimension vector to the end of the matrix.

void CoinPackedMatrix::appendMajorVectors	(	const int	numvecs,
		const CoinPackedVectorBase const	vecs
	)

Append several major-dimensonvectors to the end of the matrix.

void CoinPackedMatrix::appendMinorVector ( const CoinPackedVectorBase & vec )

Append a minor-dimension vector to the end of the matrix.

void CoinPackedMatrix::appendMinorVector	(	const int	vecsize,
		const int *	vecind,
		const double *	vecelem
	)

Append a minor-dimension vector to the end of the matrix.

void CoinPackedMatrix::appendMinorVectors	(	const int	numvecs,
		const CoinPackedVectorBase const	vecs
	)

Append several minor-dimensonvectors to the end of the matrix.

void CoinPackedMatrix::majorAppendSameOrdered ( const CoinPackedMatrix & matrix )

Append the columns of the argument to the right end of this matrix.

Precondition:: minorDim_ == matrix.minorDim_
This method throws an exception if the minor dimensions are not the same.

void CoinPackedMatrix::minorAppendSameOrdered ( const CoinPackedMatrix & matrix )

Append the columns of the argument to the bottom end of this matrix.

Precondition:: majorDim_ == matrix.majorDim_
This method throws an exception if the major dimensions are not the same.

void CoinPackedMatrix::majorAppendOrthoOrdered ( const CoinPackedMatrix & matrix )

Append the rows of the argument to the right end of this matrix.

Precondition:: minorDim_ == matrix.majorDim_
This method throws an exception if the minor dimension of the current matrix is not the same as the major dimension of the argument matrix.

void CoinPackedMatrix::minorAppendOrthoOrdered ( const CoinPackedMatrix & matrix )

Append the rows of the argument to the bottom end of this matrix.

Precondition:: majorDim_ == matrix.minorDim_
This method throws an exception if the major dimension of the current matrix is not the same as the minor dimension of the argument matrix.

void CoinPackedMatrix::deleteMajorVectors	(	const int	numDel,
		const int *	indDel
	)

Delete the major-dimension vectors whose indices are listed in indDel.

void CoinPackedMatrix::deleteMinorVectors	(	const int	numDel,
		const int *	indDel
	)

Delete the minor-dimension vectors whose indices are listed in indDel.

void CoinPackedMatrix::timesMajor	(	const double *	x,
		double *	y
	)			const

Return A * x (multiplied from the "right" direction) in y.

Precondition:: x must be of size majorDim(); y must be of size minorDim()

void CoinPackedMatrix::timesMajor	(	const CoinPackedVectorBase &	x,
		double *	y
	)			const

Return A * x (multiplied from the "right" direction) in y.

Same as the previous method, just x is given in the form of a packed vector.

void CoinPackedMatrix::timesMinor	(	const double *	x,
		double *	y
	)			const

Return A * x (multiplied from the "right" direction) in y.

Precondition:: x must be of size minorDim(); y must be of size majorDim()

void CoinPackedMatrix::timesMinor	(	const CoinPackedVectorBase &	x,
		double *	y
	)			const

Return A * x (multiplied from the "right" direction) in y.

Same as the previous method, just x is given in the form of a packed vector.

template<class FloatEqual >

bool CoinPackedMatrix::isEquivalent	(	const CoinPackedMatrix &	rhs,
		const FloatEqual &	eq
	)			const `[inline]`

Equivalence.

Two matrices are equivalent if they are both by rows or both by columns, they have the same dimensions, and each vector is equivalent. In this method the FloatEqual function operator can be specified.

Definition at line 541 of file CoinPackedMatrix.hpp.

bool CoinPackedMatrix::isEquivalent2 ( const CoinPackedMatrix & rhs ) const

Equivalence.

Two matrices are equivalent if they are both by rows or both by columns, they have the same dimensions, and each vector is equivalent. In this method the FloatEqual function operator can be specified.

bool CoinPackedMatrix::isEquivalent ( const CoinPackedMatrix & rhs ) const [inline]

The default equivalence test is that the entries are relatively equal.

Definition at line 569 of file CoinPackedMatrix.hpp.

double* CoinPackedMatrix::getMutableElements ( ) const [inline]

A vector containing the elements in the packed matrix.

Note that there might be gaps in this list, entries that do not belong to any major-dimension vector. To get the actual elements one should look at this vector together with vectorStarts and vectorLengths.

Definition at line 583 of file CoinPackedMatrix.hpp.

int* CoinPackedMatrix::getMutableIndices ( ) const [inline]

A vector containing the minor indices of the elements in the packed matrix.

Note that there might be gaps in this list, entries that do not belong to any major-dimension vector. To get the actual elements one should look at this vector together with vectorStarts and vectorLengths.

Definition at line 589 of file CoinPackedMatrix.hpp.

CoinBigIndex* CoinPackedMatrix::getMutableVectorStarts ( ) const [inline]

The positions where the major-dimension vectors start in elements and indices.

Definition at line 593 of file CoinPackedMatrix.hpp.

int* CoinPackedMatrix::getMutableVectorLengths ( ) const [inline]

The lengths of the major-dimension vectors.

Definition at line 595 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::setNumElements ( CoinBigIndex value ) [inline]

Change size after modifying - be careful.

Definition at line 597 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::nullElementArray ( ) [inline]

NULLify element array - used when space is very tight.

Definition at line 600 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::nullStartArray ( ) [inline]

NULLify start array - used when space is very tight.

Definition at line 602 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::nullLengthArray ( ) [inline]

NULLify length array - used when space is very tight.

Definition at line 604 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::nullIndexArray ( ) [inline]

NULLify index array - used when space is very tight.

Definition at line 606 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::gutsOfDestructor ( ) [protected]

void CoinPackedMatrix::gutsOfCopyOf	(	const bool	colordered,
		const int	minor,
		const int	major,
		const CoinBigIndex	numels,
		const double *	elem,
		const int *	ind,
		const CoinBigIndex *	start,
		const int *	len,
		const double	extraMajor = `0.0`,
		const double	extraGap = `0.0`
	)			`[protected]`

void CoinPackedMatrix::gutsOfCopyOfNoGaps	(	const bool	colordered,
		const int	minor,
		const int	major,
		const double *	elem,
		const int *	ind,
		const CoinBigIndex *	start
	)			`[protected]`

When no gaps we can do faster.

void CoinPackedMatrix::gutsOfOpEqual	(	const bool	colordered,
		const int	minor,
		const int	major,
		const CoinBigIndex	numels,
		const double *	elem,
		const int *	ind,
		const CoinBigIndex *	start,
		const int *	len
	)			`[protected]`

void CoinPackedMatrix::resizeForAddingMajorVectors	(	const int	numVec,
		const int *	lengthVec
	)			`[protected]`

void CoinPackedMatrix::resizeForAddingMinorVectors ( const int * addedEntries ) [protected]

int CoinPackedMatrix::appendMajor	(	const int	number,
		const CoinBigIndex *	starts,
		const int *	index,
		const double *	element,
		int	numberOther = `-1`
	)			`[protected]`

Append a set of rows/columns to the end of the matrix.

Returns number of errors i.e. if any of the new rows/columns contain an index that's larger than the number of columns-1/rows-1 (if numberOther>0) or duplicates This version is easy one i.e. adding columns to column ordered

int CoinPackedMatrix::appendMinor	(	const int	number,
		const CoinBigIndex *	starts,
		const int *	index,
		const double *	element,
		int	numberOther = `-1`
	)			`[protected]`

Append a set of rows/columns to the end of the matrix.

Returns number of errors i.e. if any of the new rows/columns contain an index that's larger than the number of columns-1/rows-1 (if numberOther>0) or duplicates This version is harder one i.e. adding columns to row ordered

CoinBigIndex CoinPackedMatrix::getLastStart ( ) const [inline, private]

Definition at line 699 of file CoinPackedMatrix.hpp.

Friends And Related Function Documentation

void CoinPackedMatrixUnitTest ( ) [friend]

A function that tests the methods in the CoinPackedMatrix class.

The only reason for it not to be a member method is that this way it doesn't have to be compiled into the library. And that's a gain, because the library should be compiled with optimization on, but this method should be compiled with debugging.

Member Data Documentation

bool CoinPackedMatrix::colOrdered_ [protected]

A flag indicating whether the matrix is column or row major ordered.

Definition at line 709 of file CoinPackedMatrix.hpp.

double CoinPackedMatrix::extraGap_ [protected]

This much times more space should be allocated for each major-dimension vector (with respect to the number of entries in the vector) when the matrix is resized.

The purpose of these gaps is to allow fast insertion of new minor-dimension vectors.

Definition at line 714 of file CoinPackedMatrix.hpp.

double CoinPackedMatrix::extraMajor_ [protected]

his much times more space should be allocated for major-dimension vectors when the matrix is resized.

The purpose of these gaps is to allow fast addition of new major-dimension vectors.

Definition at line 718 of file CoinPackedMatrix.hpp.

double* CoinPackedMatrix::element_ [protected]

List of nonzero element values.

The entries in the gaps between major-dimension vectors are undefined.

Definition at line 722 of file CoinPackedMatrix.hpp.

int* CoinPackedMatrix::index_ [protected]

List of nonzero element minor-dimension indices.