dragon.matrix

Class AbstractGiantSparseMatrix

java.lang.Object

dragon.matrix.AbstractMatrix

dragon.matrix.AbstractSparseMatrix

dragon.matrix.AbstractGiantSparseMatrix

All Implemented Interfaces:

Matrix, SparseMatrix, Serializable

Direct Known Subclasses:

DoubleGiantSparseMatrix, IntGiantSparseMatrix

public abstract class AbstractGiantSparseMatrix
extends AbstractSparseMatrix

The abstract sparse matrix for handling extreme large sparse matrirwhich will write matrix data to disk whenever it's over fulsh interval by default 1000,000 and superior to AbstractSuperSparseMatrix, however it's lack of some basic matrix operation functions such as getNonZeroColumnInRow privided by AbstractSuperSparseMatrix because it focuses on storing and loading matrix to disk efficiently

Copyright: Copyright (c) 2005

Company: IST, Drexel University

Version:

1.0

Author:

Davis Zhou

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected AbstractFlatSparseMatrix	cacheMatrix
protected int	flushInterval
protected String	indexFilename
protected SparseMatrixFactory	matrixFactory
protected String	matrixFilename
protected RandomAccessFile	rafIndex
protected RandomAccessFile	rafMatrix
protected int	totalCell

Fields inherited from class dragon.matrix.AbstractSparseMatrix

isFinalized, mergeMode, miniMode

Fields inherited from class dragon.matrix.AbstractMatrix

cellDataLength, columnBase, columns, rowBase, rows, transposeMatrix

Constructor Summary

Constructors

Constructor and Description
AbstractGiantSparseMatrix(String indexFilename, String matrixFilename, int cellDataLength)
AbstractGiantSparseMatrix(String indexFilename, String matrixFilename, int cellDataLength, boolean mergeMode, boolean miniMode)

Method Summary

Modifier and Type	Method and Description
boolean	add(Cell cell) Add a cell to the sparse matrix.
void	close() Close the matrix and release all resources.
protected abstract AbstractFlatSparseMatrix	createFlatSparseMatrix(boolean mergeMode, boolean miniMode)
protected abstract Row	createRow(int row, int columns, byte[] data)
boolean	finalizeData(boolean sorting) Any sparse matrix has two possible modes, writing mode and reading mode.
void	flush() Many implementations of sparse matrix cache cells in memory.
Cell	getCell(int row, int col) If the specified cell doesn't exist, this method will return null.
String	getIndexFilename()
String	getMatrixFilename()
Cell	getNonZeroCellInRow(int row, int index) This method returns the index-th non-zero cell in the given column.
int	getNonZeroColumnInRow(int row, int index) This method returns the column index of the index-th non-zero cell in the given row.
int[]	getNonZeroColumnsInRow(int row) If the given row has n non-zero cells, this method returns a n-length integer array.
int	getNonZeroNum()
int	getNonZeroNumInRow(int row)
protected Row	getRow(int index)
void	setFlushInterval(int interval)

Methods inherited from class dragon.matrix.AbstractSparseMatrix

cosine, finalizeData, genCooccurrenceMatrix, genCooccurrenceMatrix, genCooccurrenceMatrix, genTranslationMatrix, getCooccurrenceCount, getDouble, getInt, getNonZeroCellInColumn, getNonZeroDoubleScoreInColumn, getNonZeroDoubleScoreInRow, getNonZeroDoubleScoresInColumn, getNonZeroDoubleScoresInRow, getNonZeroIntScoreInColumn, getNonZeroIntScoreInRow, getNonZeroIntScoresInColumn, getNonZeroIntScoresInRow, getNonZeroNumInColumn, getNonZeroRowInColumn, getNonZeroRowsInColumn, isFinalized, transpose

Methods inherited from class dragon.matrix.AbstractMatrix

columns, getBaseColumn, getBaseRow, getCellDataLength, getTranspose, rows, setTranspose

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface dragon.matrix.SparseMatrix

createCell, createCell, createSparseMatrix

Methods inherited from interface dragon.matrix.Matrix

columns, getBaseColumn, getBaseRow, getCellDataLength, getTranspose, rows, setTranspose

Field Detail

indexFilename

protected String indexFilename

matrixFilename

protected String matrixFilename

totalCell

protected int totalCell

rafIndex

protected RandomAccessFile rafIndex

rafMatrix

protected RandomAccessFile rafMatrix

cacheMatrix

protected AbstractFlatSparseMatrix cacheMatrix

flushInterval

protected int flushInterval

matrixFactory

protected SparseMatrixFactory matrixFactory

Constructor Detail

AbstractGiantSparseMatrix

public AbstractGiantSparseMatrix(String indexFilename,
                                 String matrixFilename,
                                 int cellDataLength)

AbstractGiantSparseMatrix

public AbstractGiantSparseMatrix(String indexFilename,
                                 String matrixFilename,
                                 int cellDataLength,
                                 boolean mergeMode,
                                 boolean miniMode)

Method Detail

createRow

protected abstract Row createRow(int row,
                                 int columns,
                                 byte[] data)

createFlatSparseMatrix

protected abstract AbstractFlatSparseMatrix createFlatSparseMatrix(boolean mergeMode,
                                                                   boolean miniMode)

setFlushInterval

public void setFlushInterval(int interval)

close

public void close()

Description copied from interface: Matrix

Close the matrix and release all resources.

add

public boolean add(Cell cell)

Description copied from interface: SparseMatrix

Add a cell to the sparse matrix. If the cell already exists in the sparse matrix, the behavior of the addition is subject to the implementations. It may simply add the score to the existing cell. It may return false. It may cause errors even if the method returns true. Thus, one should be cautious when adding possibly existing cells to the sparse matrix.

Parameters:

cell - the input cell

Returns:

true if successfully added.

flush

public void flush()

Description copied from interface: SparseMatrix

Many implementations of sparse matrix cache cells in memory. This method gives the chance to dump cells in the cache to the disk file.

finalizeData

public boolean finalizeData(boolean sorting)

Description copied from interface: SparseMatrix

Any sparse matrix has two possible modes, writing mode and reading mode. After calling this method, writing is not allowed any more. Instead, one call read out data from the sparse matrix. In most cases, one should set the sorting option to true. In one case that all cells are added to the sparse matrix in their natural order (from top to bottom, from left to right), one can set the sorting option to false for saving time.

Parameters:

sorting - if need to sort the added cells before finalization.

Returns:

true if finalizing successfully

getMatrixFilename

public String getMatrixFilename()

getIndexFilename

public String getIndexFilename()

getNonZeroNum

public int getNonZeroNum()

Returns:

the number of non-zero cells

getNonZeroNumInRow

public int getNonZeroNumInRow(int row)

Parameters:

row - the index of the row

Returns:

the number of non-zero cells in the given row.

getNonZeroColumnInRow

public int getNonZeroColumnInRow(int row,
                                 int index)

Description copied from interface: SparseMatrix

This method returns the column index of the index-th non-zero cell in the given row. The index starts from zero.

Parameters:

row - the index of the row

index - the index-th non-zero cell

Returns:

the column index of the specified cell

getNonZeroColumnsInRow

public int[] getNonZeroColumnsInRow(int row)

Description copied from interface: SparseMatrix

If the given row has n non-zero cells, this method returns a n-length integer array. Each element of the array stores the column index of the corresponding non-zero cell.

Parameters:

row - the index of row

Returns:

the array of non-zero columns in the given row

getCell

public Cell getCell(int row,
                    int col)

Description copied from interface: SparseMatrix

If the specified cell doesn't exist, this method will return null.

Parameters:

row - the index of row in the matrix

col - the index of column in the matrix

Returns:

the cell

getNonZeroCellInRow

public Cell getNonZeroCellInRow(int row,
                                int index)

Description copied from interface: SparseMatrix

This method returns the index-th non-zero cell in the given column. The index starts from zero.

Parameters:

row - the index of the row

index - the index-th non-zero cell

Returns:

the cell

getRow

protected Row getRow(int index)