Package org.apache.sysds.runtime.matrix.data

Interface LibMatrixCountDistinct

public interface LibMatrixCountDistinct

This class contains various methods for counting the number of distinct values inside a MatrixBlock

Field Summary

Fields

Modifier and Type	Field	Description
static org.apache.commons.logging.Log	LOG
static int	minimumSize	The minimum number NonZero of cells in the input before using approximate techniques for counting number of distinct values.

Method Summary

Static Methods

Modifier and Type	Method	Description
static MatrixBlock	countDistinctValuesFromSketch(CountDistinctOperator op, CorrMatrixBlock corrBlkIn)
static CorrMatrixBlock	createSketch(CountDistinctOperator op, MatrixBlock blkIn)
static MatrixBlock	estimateDistinctValues(MatrixBlock in, CountDistinctOperator op)	Public method to count the number of distinct values inside a matrix.
static CorrMatrixBlock	unionSketch(CountDistinctOperator op, CorrMatrixBlock corrBlkIn0, CorrMatrixBlock corrBlkIn1)

Field Detail

LOG

static final org.apache.commons.logging.Log LOG

minimumSize

static final int minimumSize

The minimum number NonZero of cells in the input before using approximate techniques for counting number of distinct values.

See Also:

Constant Field Values

Method Detail

estimateDistinctValues

static MatrixBlock estimateDistinctValues(MatrixBlock in,
                                          CountDistinctOperator op)

Public method to count the number of distinct values inside a matrix. Depending on which CountDistinctOperator selected it either gets the absolute number or a estimated value. TODO: If the MatrixBlock type is CompressedMatrix, simply read the values from the ColGroups.

Parameters:

in - the input matrix to count number distinct values in

op - the selected operator to use

Returns:

A matrix block containing the absolute distinct count for the entire input or along given row/col axis

countDistinctValuesFromSketch

static MatrixBlock countDistinctValuesFromSketch(CountDistinctOperator op,
                                                 CorrMatrixBlock corrBlkIn)

createSketch

static CorrMatrixBlock createSketch(CountDistinctOperator op,
                                    MatrixBlock blkIn)

unionSketch

static CorrMatrixBlock unionSketch(CountDistinctOperator op,
                                   CorrMatrixBlock corrBlkIn0,
                                   CorrMatrixBlock corrBlkIn1)