Package org.apache.sysds.runtime.transform.encode

Interface Encoder

All Superinterfaces:

Externalizable, Serializable

All Known Implementing Classes:

ColumnEncoder, ColumnEncoderBin, ColumnEncoderComposite, ColumnEncoderDummycode, ColumnEncoderFeatureHash, ColumnEncoderPassThrough, ColumnEncoderRecode, ColumnEncoderUDF, ColumnEncoderWordEmbedding, MultiColumnEncoder

public interface Encoder
extends Externalizable

Method Summary

Modifier and Type	Method	Description
void	allocateMetaData(FrameBlock meta)	Pre-allocate a FrameBlock for metadata collection.
MatrixBlock	apply(CacheBlock<?> in, MatrixBlock out, int outputCol)	Apply the generated metadata to the FrameBlock and saved the result in out.
void	build(CacheBlock<?> in)	Build the transform meta data for the given block input.
void	buildPartial(FrameBlock in)	Partial build of internal data structures (e.g., in distributed spark operations).
FrameBlock	getMetaData(FrameBlock out)	Construct a frame block out of the transform meta data.
void	initMetaData(FrameBlock meta)	Sets up the required meta data for a subsequent call to apply.
void	prepareBuildPartial()	Allocates internal data structures for partial build.
void	updateIndexRanges(long[] beginDims, long[] endDims, int offset)	Update index-ranges to after encoding.

Methods inherited from interface java.io.Externalizable

readExternal, writeExternal

Method Detail

build

void build(CacheBlock<?> in)

Build the transform meta data for the given block input. This call modifies and keeps meta data as encoder state.

Parameters:

in - input frame block

apply

MatrixBlock apply(CacheBlock<?> in,
                  MatrixBlock out,
                  int outputCol)

Apply the generated metadata to the FrameBlock and saved the result in out.

Parameters:

in - input frame block

out - output matrix block

outputCol - is a offset in the output matrix. column in FrameBlock + outputCol = column in out

Returns:

output matrix block

allocateMetaData

void allocateMetaData(FrameBlock meta)

Pre-allocate a FrameBlock for metadata collection.

Parameters:

meta - frame block

getMetaData

FrameBlock getMetaData(FrameBlock out)

Construct a frame block out of the transform meta data.

Parameters:

out - output frame block

Returns:

output frame block?

initMetaData

void initMetaData(FrameBlock meta)

Sets up the required meta data for a subsequent call to apply.

Parameters:

meta - frame block

prepareBuildPartial

void prepareBuildPartial()

Allocates internal data structures for partial build.

buildPartial

void buildPartial(FrameBlock in)

Partial build of internal data structures (e.g., in distributed spark operations).

Parameters:

in - input frame block

updateIndexRanges

void updateIndexRanges(long[] beginDims,
                       long[] endDims,
                       int offset)

Update index-ranges to after encoding. Note that only Dummycoding changes the ranges.

Parameters:

beginDims - begin dimensions of range

endDims - end dimensions of range

offset - is applied to begin and endDims