Sparse Matrix Base Class. More...

#include <CoinPackedMatrix.hpp>

Public Member Functions
Query members
double	getExtraGap () const
	Return the current setting of the extra gap. More...

double	getExtraMajor () const
	Return the current setting of the extra major. More...

void	reserve (const int newMaxMajorDim, const CoinBigIndex newMaxSize, bool create=false)
	Reserve sufficient space for appending major-ordered vectors. More...

void	clear ()
	Clear the data, but do not free any arrays. More...

bool	isColOrdered () const
	Whether the packed matrix is column major ordered or not. More...

bool	hasGaps () const
	Whether the packed matrix has gaps or not. More...

CoinBigIndex	getNumElements () const
	Number of entries in the packed matrix. More...

int	getNumCols () const
	Number of columns. More...

int	getNumRows () const
	Number of rows. More...

const double *	getElements () const
	A vector containing the elements in the packed matrix. More...

const int *	getIndices () const
	A vector containing the minor indices of the elements in the packed matrix. More...

int	getSizeVectorStarts () const
	The size of the `vectorStarts` array. More...

int	getSizeVectorLengths () const
	The size of the `vectorLengths` array. More...

const CoinBigIndex *	getVectorStarts () const
	The positions where the major-dimension vectors start in elements and indices. More...

const int *	getVectorLengths () const
	The lengths of the major-dimension vectors. More...

CoinBigIndex	getVectorFirst (const int i) const
	The position of the first element in the i'th major-dimension vector. More...

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. More...

int	getVectorSize (const int i) const
	The length of i'th vector. More...

const CoinShallowPackedVector	getVector (int i) const
	Return the i'th vector in matrix. More...

int *	getMajorIndices () const
	Returns an array containing major indices. More...

Modifying members
void	setDimensions (int numrows, int numcols)
	Set the dimensions of the matrix. More...

void	setExtraGap (const double newGap)
	Set the extra gap to be allocated to the specified value. More...

void	setExtraMajor (const double newMajor)
	Set the extra major to be allocated to the specified value. More...

void	appendCol (const CoinPackedVectorBase &vec)
	Append a column to the end of the matrix. More...

void	appendCol (const int vecsize, const int vecind, const double vecelem)
	Append a column to the end of the matrix. More...

void	appendCols (const int numcols, const CoinPackedVectorBase const cols)
	Append a set of columns to the end of the matrix. More...

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. More...

void	appendRow (const CoinPackedVectorBase &vec)
	Append a row to the end of the matrix. More...

void	appendRow (const int vecsize, const int vecind, const double vecelem)
	Append a row to the end of the matrix. More...

void	appendRows (const int numrows, const CoinPackedVectorBase const rows)
	Append a set of rows to the end of the matrix. More...

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. More...

void	rightAppendPackedMatrix (const CoinPackedMatrix &matrix)
	Append the argument to the "right" of the current matrix. More...

void	bottomAppendPackedMatrix (const CoinPackedMatrix &matrix)
	Append the argument to the "bottom" of the current matrix. More...

void	deleteCols (const int numDel, const int *indDel)
	Delete the columns whose indices are listed in `indDel`. More...

void	deleteRows (const int numDel, const int *indDel)
	Delete the rows whose indices are listed in `indDel`. More...

void	replaceVector (const int index, const int numReplace, const double *newElements)
	Replace the elements of a vector. More...

void	modifyCoefficient (int row, int column, double newElement, bool keepZero=false)
	Modify one element of packed matrix. More...

double	getCoefficient (int row, int column) const
	Return one element of packed matrix. More...

CoinBigIndex	compress (double threshold)
	Eliminate all elements in matrix whose absolute value is less than threshold. More...

CoinBigIndex	eliminateDuplicates (double threshold)
	Eliminate all duplicate AND small elements in matrix The column starts are not affected. More...

void	orderMatrix ()
	Sort all columns so indices are increasing.in each column. More...

CoinBigIndex	cleanMatrix (double threshold=1.0e-20)
	Really clean up matrix. More...

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. More...

void	submatrixOf (const CoinPackedMatrix &matrix, const int numMajor, const int *indMajor)
	Extract a submatrix from matrix. More...

void	submatrixOfWithDuplicates (const CoinPackedMatrix &matrix, const int numMajor, const int *indMajor)
	Extract a submatrix from matrix. More...

void	copyOf (const CoinPackedMatrix &rhs)
	Copy method. More...

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. More...

void	copyReuseArrays (const CoinPackedMatrix &rhs)
	Copy method. More...

void	reverseOrderedCopyOf (const CoinPackedMatrix &rhs)
	Make a reverse-ordered copy. More...

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. More...

CoinPackedMatrix &	operator= (const CoinPackedMatrix &rhs)
	Assignment operator. More...

void	reverseOrdering ()
	Reverse the ordering of the packed matrix. More...

void	transpose ()
	Transpose the matrix. More...

void	swap (CoinPackedMatrix &matrix)
	Swap the content of two packed matrices. More...

Matrix times vector methods
void	times (const double x, double y) const
	Return `A * x` in `y`. More...

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

void	transposeTimes (const double x, double y) const
	Return `x * A` in `y`. More...

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

Queries
int *	countOrthoLength () const
	Count the number of entries in every minor-dimension vector and return an array containing these lengths. More...

void	countOrthoLength (int *counts) const
	Count the number of entries in every minor-dimension vector and fill in an array containing these lengths. More...

int	getMajorDim () const
	Major dimension. More...

void	setMajorDim (int value)
	Set major dimension. More...

int	getMinorDim () const
	Minor dimension. More...

void	setMinorDim (int value)
	Set minor dimension. More...

int	getMaxMajorDim () const
	Current maximum for major dimension. More...

void	dumpMatrix (const char *fname=NULL) const
	Dump the matrix on stdout. More...

void	printMatrixElement (const int row_val, const int col_val) const
	Print a single matrix element. More...

Append vectors
When compiled with COIN_DEBUG defined these methods throw an exception if the major (minor) vector contains an index that's invalid for the minor (major) dimension. 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. More...

void	appendMajorVector (const int vecsize, const int vecind, const double vecelem)
	Append a major-dimension vector to the end of the matrix. More...

void	appendMajorVectors (const int numvecs, const CoinPackedVectorBase const vecs)
	Append several major-dimensonvectors to the end of the matrix. More...

void	appendMinorVector (const CoinPackedVectorBase &vec)
	Append a minor-dimension vector to the end of the matrix. More...

void	appendMinorVector (const int vecsize, const int vecind, const double vecelem)
	Append a minor-dimension vector to the end of the matrix. More...

void	appendMinorVectors (const int numvecs, const CoinPackedVectorBase const vecs)
	Append several minor-dimension vectors to the end of the matrix. More...

void	appendMinorFast (const int number, const CoinBigIndex starts, const int index, const double *element)
	Append a set of rows (columns) to the end of a column (row) ordered matrix. More...

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. More...

void	minorAppendSameOrdered (const CoinPackedMatrix &matrix)
	Append the columns of the argument to the bottom end of this matrix. More...

void	majorAppendOrthoOrdered (const CoinPackedMatrix &matrix)
	Append the rows of the argument to the right end of this matrix. More...

void	minorAppendOrthoOrdered (const CoinPackedMatrix &matrix)
	Append the rows of the argument to the bottom end of this matrix. More...

Delete vectors
void	deleteMajorVectors (const int numDel, const int *indDel)
	Delete the major-dimension vectors whose indices are listed in `indDel`. More...

void	deleteMinorVectors (const int numDel, const int *indDel)
	Delete the minor-dimension vectors whose indices are listed in `indDel`. More...

Various dot products.
void	timesMajor (const double x, double y) const
	Return `A * x` (multiplied from the "right" direction) in `y`. More...

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

void	timesMinor (const double x, double y) const
	Return `A * x` (multiplied from the "right" direction) in `y`. More...

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

Logical Operations.
template<class FloatEqual >
bool	isEquivalent (const CoinPackedMatrix &rhs, const FloatEqual &eq) const
	Test for equivalence. More...

bool	isEquivalent2 (const CoinPackedMatrix &rhs) const
	Test for equivalence and report differences. More...

bool	isEquivalent (const CoinPackedMatrix &rhs) const
	Test for equivalence. More...

Non-const methods
These are to be used with great care when doing column generation, etc.
double *	getMutableElements () const
	A vector containing the elements in the packed matrix. More...

int *	getMutableIndices () const
	A vector containing the minor indices of the elements in the packed matrix. More...

CoinBigIndex *	getMutableVectorStarts () const
	The positions where the major-dimension vectors start in element_ and index_. More...

int *	getMutableVectorLengths () const
	The lengths of the major-dimension vectors. More...

void	setNumElements (CoinBigIndex value)
	Change the size of the bulk store after modifying - be careful. More...

void	nullElementArray ()
	NULLify element array. More...

void	nullStartArray ()
	NULLify start array. More...

void	nullLengthArray ()
	NULLify length array. More...

void	nullIndexArray ()
	NULLify index array. More...

Constructors and destructors
	CoinPackedMatrix ()
	Default Constructor creates an empty column ordered packed matrix. More...

	CoinPackedMatrix (const bool colordered, const double extraMajor, const double extraGap)
	A constructor where the ordering and the gaps are specified. More...

	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)

	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)

	CoinPackedMatrix (const bool colordered, const int rowIndices, const int colIndices, const double *elements, CoinBigIndex numels)
	Create packed matrix from triples. More...

	CoinPackedMatrix (const CoinPackedMatrix &m)
	Copy constructor. More...

	CoinPackedMatrix (const CoinPackedMatrix &m, int extraForMajor, int extraElements, bool reverseOrdering=false)
	Copy constructor with fine tuning. More...

	CoinPackedMatrix (const CoinPackedMatrix &wholeModel, int numberRows, const int whichRows, int numberColumns, const int whichColumns)
	Subset constructor (without gaps). More...

virtual	~CoinPackedMatrix ()
	Destructor. More...

Debug Utilities
int	verifyMtx (int verbosity=1, bool zeroesAreError=false) const
	Scan the matrix for anomalies. More...

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. More...

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 a row (colum) ordered matrix. More...

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 a column (row) ordered matrix. More...

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. More...

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. More...

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

double *	element_
	List of nonzero element values. More...

int *	index_
	List of nonzero element minor-dimension indices. More...

CoinBigIndex *	start_
	Starting positions of major-dimension vectors. More...

int *	length_
	Lengths of major-dimension vectors. More...

int	majorDim_
	number of vectors in matrix More...

int	minorDim_
	size of other dimension More...

CoinBigIndex	size_
	the number of nonzero entries More...

int	maxMajorDim_
	max space allocated for major-dimension More...

CoinBigIndex	maxSize_
	max space allocated for entries More...

Private Member Functions
CoinBigIndex	getLastStart () const

Friends
void	CoinPackedMatrixUnitTest ()
	Test the methods in the CoinPackedMatrix class. More...

Detailed Description

Sparse Matrix Base Class.

This class is intended to represent sparse matrices using row-major or column-major ordering. The representation is very efficient for adding, deleting, or retrieving major-dimension vectors. Adding a minor-dimension vector is less efficient, but can be helped by providing "extra" space as described in the next paragraph. Deleting a minor-dimension vector requires inspecting all coefficients in the matrix. Retrieving a minor-dimension vector would incur the same cost and is not supported (except in the sense that you can write a loop to retrieve all coefficients one at a time). Consider physically transposing the matrix, or keeping a second copy with the other major-vector ordering.

The sparse represention can be completely compact or it can have "extra" space available at the end of each major vector. Incorporating extra space into the sparse matrix representation can improve performance in cases where new data needs to be inserted into the packed matrix against the major-vector orientation (e.g, inserting a row into a matrix stored in column-major order).

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.

Compiling this class with CLP_NO_VECTOR defined will excise all methods which use CoinPackedVectorBase, CoinPackedVector, or CoinShallowPackedVector as parameters or return types.

Compiling this class with COIN_FAST_CODE defined removes index range checks.

Definition at line 79 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
	)

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
	)

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 with fine tuning.

This constructor allows for the specification of an exact amount of extra space and/or reverse ordering.

extraForMajor is the exact number of spare major vector slots after any possible reverse ordering. If extraForMajor < 0, all gaps and small elements will be removed from the copy, otherwise gaps and small elements are preserved.

extraElements is the exact number of spare element entries.

The usual multipliers, extraMajor_ and extraGap_, are 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 87 of file CoinPackedMatrix.hpp.

double CoinPackedMatrix::getExtraMajor ( ) const

inline

Return the current setting of the extra major.

Definition at line 89 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.

bool CoinPackedMatrix::isColOrdered ( ) const

inline

Whether the packed matrix is column major ordered or not.

Definition at line 99 of file CoinPackedMatrix.hpp.

bool CoinPackedMatrix::hasGaps ( ) const

inline

Whether the packed matrix has gaps or not.

Definition at line 102 of file CoinPackedMatrix.hpp.

CoinBigIndex CoinPackedMatrix::getNumElements ( ) const

inline

Number of entries in the packed matrix.

Definition at line 105 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getNumCols ( ) const

inline

Number of columns.

Definition at line 108 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getNumRows ( ) const

inline

Number of rows.

Definition at line 114 of file CoinPackedMatrix.hpp.

const double* CoinPackedMatrix::getElements ( ) const

inline

A vector containing the elements in the packed matrix.

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

Definition at line 126 of file CoinPackedMatrix.hpp.

const int* CoinPackedMatrix::getIndices ( ) const

inline

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

Returns index_. 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 (start_) and vectorLengths (length_).

Definition at line 136 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getSizeVectorStarts ( ) const

inline

The size of the vectorStarts array.

See start_.

Definition at line 142 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getSizeVectorLengths ( ) const

inline

The size of the vectorLengths array.

See length_.

Definition at line 151 of file CoinPackedMatrix.hpp.

const CoinBigIndex* CoinPackedMatrix::getVectorStarts ( ) const

inline

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

See start_.

Definition at line 158 of file CoinPackedMatrix.hpp.

const int* CoinPackedMatrix::getVectorLengths ( ) const

inline

The lengths of the major-dimension vectors.

See length_.

Definition at line 164 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 168 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 178 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::getVectorSize ( const int i ) const

inline

The length of i'th vector.

Definition at line 187 of file CoinPackedMatrix.hpp.

const CoinShallowPackedVector CoinPackedMatrix::getVector ( int i ) const

inline

Return the i'th vector in matrix.

Definition at line 197 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 dimensions of the matrix.

The method name is deceptive; the effect is to append empty columns and/or rows to the matrix to reach the specified dimensions. A negative number for either dimension means that that dimension doesn't change. An exception will be thrown if the specified dimensions are smaller than the current dimensions.

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 compiled with COIN_DEBUG defined this method throws an exception if the column vector specifies a nonexistent row index. 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 compiled with COIN_DEBUG defined this method throws an exception if the column vector specifies a nonexistent row index. 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 compiled with COIN_DEBUG defined this method throws an exception if any of the column vectors specify a nonexistent row index. 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 the number of errors (nonexistent or duplicate row index). No error checking is performed if numberRows < 0.

void CoinPackedMatrix::appendRow ( const CoinPackedVectorBase & vec )

Append a row to the end of the matrix.

When compiled with COIN_DEBUG defined this method throws an exception if the row vector specifies a nonexistent column index. 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 compiled with COIN_DEBUG defined this method throws an exception if the row vector specifies a nonexistent column index. 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 compiled with COIN_DEBUG defined this method throws an exception if any of the row vectors specify a nonexistent column index. 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 the number of errors (nonexistent or duplicate column index). No error checking is performed 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

CoinBigIndex 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

CoinBigIndex 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.

CoinBigIndex 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 )

Make a reverse-ordered copy.

This method makes an exact replica of the argument with the major vector orientation changed from row (column) to column (row). The extra space parameters are also copied and reversed. (Cf. reverseOrdering, which does the same thing in place.)

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.

void CoinPackedMatrix::reverseOrdering ( )

Reverse the ordering of the packed matrix.

Change the major vector orientation of the matrix data structures from row (column) to column (row). (Cf. reverseOrderedCopyOf, which does the same thing but produces a new matrix.)

void CoinPackedMatrix::transpose ( )

Transpose the matrix.

Note: If you start with a column-ordered matrix and invoke transpose, you will have a row-ordered transposed matrix. To change the major vector orientation (e.g., to transform a column-ordered matrix to a column-ordered transposed matrix), invoke transpose() followed by reverseOrdering().

void CoinPackedMatrix::swap ( CoinPackedMatrix & matrix )

Swap the content of two packed matrices.

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 504 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::setMajorDim ( int value )

inline

Set major dimension.

For row ordered matrix this would be the number of rows. Use with great care.

Definition at line 507 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 510 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::setMinorDim ( int value )

inline

Set minor dimension.

For row ordered matrix this would be the number of columns. Use with great care.

Definition at line 513 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 517 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-dimension vectors to the end of the matrix.

void CoinPackedMatrix::appendMinorFast	(	const int	number,
		const CoinBigIndex *	starts,
		const int *	index,
		const double *	element
	)

Append a set of rows (columns) to the end of a column (row) ordered matrix.

This case is when we know there are no gaps and majorDim_ will not change.

Todo:: This method really belongs in the group of protected methods with appendMinor; there are no safeties here even with COIN_DEBUG. Apparently this method was needed in ClpPackedMatrix and giving it proper visibility was too much trouble. Should be moved.

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

Test for equivalence.

Two matrices are equivalent if they are both row- or column-ordered, they have the same dimensions, and each (major) vector is equivalent. The operator used to test for equality can be specified using the FloatEqual template parameter.

Definition at line 663 of file CoinPackedMatrix.hpp.

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

Test for equivalence and report differences.

Equivalence is defined as for isEquivalent. In addition, this method will print differences to std::cerr. Intended for use in unit tests and for debugging.

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

Test for equivalence.

The test for element equality is the default CoinRelFltEq operator.

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 start_ and length_.

Definition at line 713 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 start_ and length_.

Definition at line 719 of file CoinPackedMatrix.hpp.

CoinBigIndex* CoinPackedMatrix::getMutableVectorStarts ( ) const

inline

The positions where the major-dimension vectors start in element_ and index_.

Definition at line 723 of file CoinPackedMatrix.hpp.

int* CoinPackedMatrix::getMutableVectorLengths ( ) const

inline

The lengths of the major-dimension vectors.

Definition at line 725 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::setNumElements ( CoinBigIndex value )

inline

Change the size of the bulk store after modifying - be careful.

Definition at line 727 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::nullElementArray ( )

inline

NULLify element array.

Used when space is very tight. Does not free the space!

Definition at line 735 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::nullStartArray ( )

inline

NULLify start array.

Used when space is very tight. Does not free the space!

Definition at line 741 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::nullLengthArray ( )

inline

NULLify length array.

Used when space is very tight. Does not free the space!

Definition at line 747 of file CoinPackedMatrix.hpp.

void CoinPackedMatrix::nullIndexArray ( )

inline

NULLify index array.

Used when space is very tight. Does not free the space!

Definition at line 753 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::verifyMtx	(	int	verbosity = `1`,
		bool	zeroesAreError = `false`
	)		const

Scan the matrix for anomalies.

Returns the number of anomalies. Scans the structure for gaps, obviously bogus indices and coefficients, and inconsistencies. Gaps are not an error unless hasGaps() says the matrix should be gap-free. Zeroes are not an error unless zeroesAreError is set to true.

Values for verbosity are:

0: No messages, just the return value
1: Messages about errors
2: If there are no errors, a message indicating the matrix was checked is printed (positive confirmation).
3: Adds a bit more information about the matrix.
4: Prints warnings about zeroes even if they're not considered errors.

Obviously bogus coefficients are coefficients that are NaN or have absolute value greater than 1e50. Zeros have absolute value less than 1e-50.

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 a row (colum) ordered matrix.

If numberOther > 0 the method will check if any of the new rows (columns) contain duplicate indices or invalid indices and return the number of errors. A valid minor index must satisfy

0 <= k < numberOther

If numberOther < 0 no checking is performed.

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 a column (row) ordered matrix.

If numberOther > 0 the method will check if any of the new rows (columns) contain duplicate indices or indices outside the current range for the major dimension and return the number of violations. If numberOther <= 0 the major dimension will be expanded as necessary and there are no checks for duplicate indices.

CoinBigIndex CoinPackedMatrix::getLastStart ( ) const

inlineprivate

Definition at line 896 of file CoinPackedMatrix.hpp.

Friends And Related Function Documentation

void CoinPackedMatrixUnitTest ( )

friend

Test 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 907 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 912 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 916 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 920 of file CoinPackedMatrix.hpp.

int* CoinPackedMatrix::index_

protected

List of nonzero element minor-dimension indices.

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

Definition at line 923 of file CoinPackedMatrix.hpp.

CoinBigIndex* CoinPackedMatrix::start_

protected

Starting positions of major-dimension vectors.

Definition at line 925 of file CoinPackedMatrix.hpp.

int* CoinPackedMatrix::length_

protected

Lengths of major-dimension vectors.

Definition at line 927 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::majorDim_

protected

number of vectors in matrix

Definition at line 930 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::minorDim_

protected

size of other dimension

Definition at line 932 of file CoinPackedMatrix.hpp.

CoinBigIndex CoinPackedMatrix::size_

protected

the number of nonzero entries

Definition at line 934 of file CoinPackedMatrix.hpp.

int CoinPackedMatrix::maxMajorDim_

protected

max space allocated for major-dimension

Definition at line 937 of file CoinPackedMatrix.hpp.

CoinBigIndex CoinPackedMatrix::maxSize_

protected

max space allocated for entries

Definition at line 939 of file CoinPackedMatrix.hpp.

The documentation for this class was generated from the following file:

/tmp/DyLP-1.10.4/CoinUtils/src/CoinPackedMatrix.hpp

Public Member Functions

Protected Member Functions

Protected Attributes

Private Member Functions

Friends

Detailed Description

Constructor & Destructor Documentation

Member Function Documentation

Friends And Related Function Documentation

Member Data Documentation