org.knime.base.node.preproc.pmml.missingval

Class MissingCellHandler

java.lang.Object

org.knime.base.node.preproc.pmml.missingval.MissingCellHandler

Direct Known Subclasses:

DefaultMissingCellHandler, DoubleMovingAverageMissingCellHandler, FixedDoubleValueMissingCellHandler

public abstract class MissingCellHandler
extends Object

The base class for all missing cell handlers. It is created by a MissingCellHandlerFactory and used for a single column.

Author:

Alexander Fillbrunn

Field Summary

Fields

Modifier and Type	Field and Description
static String	CUSTOM_HANDLER_EXTENSION_NAME Name of the pmml extension that holds the name of a custom missing value handler.

Constructor Summary

Constructors

Constructor and Description
MissingCellHandler(DataColumnSpec col) Constructor for the missing cell handler.

Method Summary

Modifier and Type	Method and Description
protected org.dmg.pmml.DerivedFieldDocument.DerivedField	createExtensionDerivedField(org.dmg.pmml.DATATYPE.Enum dataType, String factoryID) Creates a derived field that contains an extension which contains the name of the factory to use for the replacement.
protected org.dmg.pmml.DerivedFieldDocument.DerivedField	createValueReplacingDerivedField(org.dmg.pmml.DATATYPE.Enum dataType, String value) Helper method for creating a derived field that replaces a field's value with a fixed value.
static MissingCellHandler	fromPMMLExtension(DataColumnSpec column, org.dmg.pmml.ExtensionDocument.Extension ext) Creates a missing cell handler from an extension that is inside of a PMML derived field.
abstract DataCell	getCell(RowKey key, DataColumnWindow window) Calculates the replacement value for a cell.
DataColumnSpec	getColumnSpec()
abstract int	getNextCellsWindowSize()
DataType	getOutputDataType()
protected org.dmg.pmml.DATATYPE.Enum	getPMMLDataTypeForColumn() Maps the columns data type to a PMML data type.
abstract org.dmg.pmml.DerivedFieldDocument.DerivedField	getPMMLDerivedField() Creates a derived field for the documentation of the operation in PMML.
abstract int	getPreviousCellsWindowSize()
abstract Statistic	getStatistic() Returns a Knime statistic to be evaluated before missing values are replaced.
String	getWarningMessage()
abstract void	loadSettingsFrom(NodeSettingsRO settings) Loads user settings.
void	nonMissingValueSeen(RowKey key, DataColumnWindow window) Is called when the iterator over the table passes over a row that has a non-missing cell in this handler's column.
void	rowRemoved(RowKey key) Is called when the current value is missing but the current row is skipped because a previous column had a handler that requested to remove the row.
abstract void	saveSettingsTo(NodeSettingsWO settings) Saves user settings.

Methods inherited from class java.lang.Object

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

Field Detail

CUSTOM_HANDLER_EXTENSION_NAME

public static final String CUSTOM_HANDLER_EXTENSION_NAME

Name of the pmml extension that holds the name of a custom missing value handler.

See Also:

Constant Field Values

Constructor Detail

MissingCellHandler

public MissingCellHandler(DataColumnSpec col)

Constructor for the missing cell handler.

Parameters:

col - the column this handler is used for

Method Detail

getColumnSpec

public DataColumnSpec getColumnSpec()

Returns:

the name of the column this handler is used for.

getWarningMessage

public String getWarningMessage()

Returns:

Any warning messages that occured during execution.

loadSettingsFrom

public abstract void loadSettingsFrom(NodeSettingsRO settings)
                               throws InvalidSettingsException

Loads user settings.

Parameters:

settings - the settings

Throws:

InvalidSettingsException - when the settings cannot be loaded

saveSettingsTo

public abstract void saveSettingsTo(NodeSettingsWO settings)

Saves user settings.

Parameters:

settings - the settings

getStatistic

public abstract Statistic getStatistic()

Returns a Knime statistic to be evaluated before missing values are replaced. Can return null to indicate that no statistics are needed.

Returns:

a Knime statistic object

getCell

public abstract DataCell getCell(RowKey key,
                                 DataColumnWindow window)

Calculates the replacement value for a cell. If the returned value is null, the row this cell in is removed from the output table.

Parameters:

key - the row key of the row where the cell should be replaced.

window - the window over the previous and next cells as defined by getPreviousCellsWindowSize and getNextCellsWindowSize.

Returns:

a data cell with the replacement value or null if the row should be removed completely

nonMissingValueSeen

public void nonMissingValueSeen(RowKey key,
                                DataColumnWindow window)

Is called when the iterator over the table passes over a row that has a non-missing cell in this handler's column.

Parameters:

key - the key of the row

window - the sliding window over the column

rowRemoved

public void rowRemoved(RowKey key)

Is called when the current value is missing but the current row is skipped because a previous column had a handler that requested to remove the row. When a handler requests to remove a row after this handler's column has been treated (getCell has been called already) then this method is not called. This method should be used to advance iterators for statistics etc.

Parameters:

key - the key of the row that is removed

getPMMLDerivedField

public abstract org.dmg.pmml.DerivedFieldDocument.DerivedField getPMMLDerivedField()

Creates a derived field for the documentation of the operation in PMML.

Returns:

the derived field

getPreviousCellsWindowSize

public abstract int getPreviousCellsWindowSize()

Returns:

the number of cells to be remembered as lookbehind.

getNextCellsWindowSize

public abstract int getNextCellsWindowSize()

Returns:

the number of cells to be read ahead as lookahead.

getOutputDataType

public DataType getOutputDataType()

Returns:

The type of the column with the replaced values. If this value is null, the handler returns the data type of the input column.

createValueReplacingDerivedField

protected org.dmg.pmml.DerivedFieldDocument.DerivedField createValueReplacingDerivedField(org.dmg.pmml.DATATYPE.Enum dataType,
                                                                                          String value)

Helper method for creating a derived field that replaces a field's value with a fixed value.

Parameters:

dataType - the data type of the field.

value - the replacement value for the field

Returns:

the derived field

createExtensionDerivedField

protected org.dmg.pmml.DerivedFieldDocument.DerivedField createExtensionDerivedField(org.dmg.pmml.DATATYPE.Enum dataType,
                                                                                     String factoryID)

Creates a derived field that contains an extension which contains the name of the factory to use for the replacement. The result may be adjusted to contain necessary information for the handler.

Parameters:

dataType - the data type of the derived field

factoryID - the id of the factory

Returns:

the derived field

getPMMLDataTypeForColumn

protected org.dmg.pmml.DATATYPE.Enum getPMMLDataTypeForColumn()

Maps the columns data type to a PMML data type.

Returns:

The PMML data type for this handler's column

fromPMMLExtension

public static MissingCellHandler fromPMMLExtension(DataColumnSpec column,
                                                   org.dmg.pmml.ExtensionDocument.Extension ext)
                                            throws InvalidSettingsException

Creates a missing cell handler from an extension that is inside of a PMML derived field.

Parameters:

column - the column this handler is used for

ext - the extension containing the necessary information

Returns:

a missing cell handler that was initialized from the extension

Throws:

InvalidSettingsException - if the the factory from the extension is not applicable for the column