Package elki.datasource.parser

Class NumberVectorLabelParser<V extends NumberVector>

java.lang.Object

elki.datasource.parser.AbstractStreamingParser

elki.datasource.parser.NumberVectorLabelParser<V>

Type Parameters:

V - the type of NumberVector used

All Implemented Interfaces:

BundleStreamSource, Parser, StreamingParser

Direct Known Subclasses:

BitVectorLabelParser, CategorialDataAsNumberVectorParser, SparseNumberVectorLabelParser, TermFrequencyParser

public class NumberVectorLabelParser<V extends NumberVector>
extends AbstractStreamingParser

Parser for a simple CSV type of format, with columns separated by the given pattern (default: whitespace).

Several labels may be given per point. A label must not be parseable as double. Lines starting with "#" will be ignored.

An index can be specified to identify an entry to be treated as class label. This index counts all entries (numeric and labels as well) starting with 0.

Since:

0.1

Author:

Arthur Zimek, Erich Schubert

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static class	NumberVectorLabelParser.Par<V extends NumberVector>	Parameterization class.

Nested classes/interfaces inherited from interface elki.datasource.bundle.BundleStreamSource

BundleStreamSource.Event

Field Summary

Fields

Modifier and Type	Field	Description
protected DoubleArray	attributes	Double array storing the numerical attributes during parsing.
protected java.util.List<java.lang.String>	columnnames	Column names.
protected LabelList	curlbl	Current labels.
protected V	curvec	Current vector.
protected NumberVector.Factory<V>	factory	Vector factory class.
protected boolean	haslabels	Whether or not the data set has labels.
private long[]	labelIndices	Keeps the indices of the attributes to be treated as a string label.
(package private) java.util.ArrayList<java.lang.String>	labels	(Reused) store for labels.
private static Logging	LOG	Logging class.
protected int	maxdim	Dimensionality reported.
protected BundleMeta	meta	Metadata.
protected int	mindim	Dimensionality reported.
(package private) BundleStreamSource.Event	nextevent	Event to report next.
(package private) it.unimi.dsi.fastutil.objects.ObjectOpenHashSet<java.lang.String>	unique	For String unification.
(package private) boolean	warnedDim	Emit a dimensionality change warning once.
(package private) boolean	warnedPrecision	Emit a double-precision limit warning once.

Fields inherited from class elki.datasource.parser.AbstractStreamingParser

reader, tokenizer

Constructor Summary

Constructors

Constructor	Description
NumberVectorLabelParser(NumberVector.Factory<V> factory)	Constructor with defaults.
NumberVectorLabelParser(CSVReaderFormat format, long[] labelIndices, NumberVector.Factory<V> factory)	Constructor.
NumberVectorLabelParser(java.util.regex.Pattern colSep, java.lang.String quoteChars, java.util.regex.Pattern comment, long[] labelIndices, NumberVector.Factory<V> factory)	Constructor.

Method Summary

Modifier and Type	Method	Description
protected void	buildMeta()	Update the meta element.
void	cleanup()	Perform cleanup operations after parsing.
protected V	createVector()	Creates a database object of type V.
java.lang.Object	data(int rnum)	Access a particular object and representation.
protected Logging	getLogger()	Get the logger for this class.
BundleMeta	getMeta()	Get the current meta data.
(package private) SimpleTypeInformation<V>	getTypeInformation(int mindim, int maxdim)	Get a prototype object for the given dimensionality.
void	initStream(java.io.InputStream in)	Init the streaming parser for the given input stream.
protected boolean	isLabelColumn(int col)	Test if the current column is marked as label column.
BundleStreamSource.Event	nextEvent()	Get the next event
protected boolean	parseLineInternal()	Internal method for parsing a single line.

Methods inherited from class elki.datasource.parser.AbstractStreamingParser

asMultipleObjectsBundle, assignDBID, hasDBIDs, parse

Methods inherited from class java.lang.Object

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

Field Detail

LOG

private static final Logging LOG

Logging class.

labelIndices

private long[] labelIndices

Keeps the indices of the attributes to be treated as a string label.

factory

protected NumberVector.Factory<V extends NumberVector> factory

Vector factory class.

mindim

protected int mindim

Dimensionality reported.

maxdim

protected int maxdim

Dimensionality reported.

meta

protected BundleMeta meta

Metadata.

columnnames

protected java.util.List<java.lang.String> columnnames

Column names.

haslabels

protected boolean haslabels

Whether or not the data set has labels.

curvec

protected V extends NumberVector curvec

Current vector.

curlbl

protected LabelList curlbl

Current labels.

attributes

protected DoubleArray attributes

Double array storing the numerical attributes during parsing.

labels

final java.util.ArrayList<java.lang.String> labels

(Reused) store for labels.

unique

it.unimi.dsi.fastutil.objects.ObjectOpenHashSet<java.lang.String> unique

For String unification.

nextevent

BundleStreamSource.Event nextevent

Event to report next.

warnedPrecision

boolean warnedPrecision

Emit a double-precision limit warning once.

warnedDim

boolean warnedDim

Emit a dimensionality change warning once.

Constructor Detail

NumberVectorLabelParser

public NumberVectorLabelParser(CSVReaderFormat format,
                               long[] labelIndices,
                               NumberVector.Factory<V> factory)

Constructor.

Parameters:

format - Input format

labelIndices - Column indexes that are not numeric.

factory - Vector factory

NumberVectorLabelParser

public NumberVectorLabelParser(NumberVector.Factory<V> factory)

Constructor with defaults.

Parameters:

factory - Vector factory

NumberVectorLabelParser

public NumberVectorLabelParser(java.util.regex.Pattern colSep,
                               java.lang.String quoteChars,
                               java.util.regex.Pattern comment,
                               long[] labelIndices,
                               NumberVector.Factory<V> factory)

Constructor.

Parameters:

colSep - Column separator

quoteChars - Quote character

comment - Comment pattern

labelIndices - Column indexes that are not numeric.

factory - Vector factory

Method Detail

isLabelColumn

protected boolean isLabelColumn(int col)

Test if the current column is marked as label column.

Parameters:

col - Column number

Returns:

true when a label column.

initStream

public void initStream(java.io.InputStream in)

Description copied from interface: StreamingParser

Init the streaming parser for the given input stream.

Specified by:

initStream in interface StreamingParser

Overrides:

initStream in class AbstractStreamingParser

Parameters:

in - the stream to parse objects from

getMeta

public BundleMeta getMeta()

Description copied from interface: BundleStreamSource

Get the current meta data.

Returns:

Metadata

nextEvent

public BundleStreamSource.Event nextEvent()

Description copied from interface: BundleStreamSource

Get the next event

Returns:

Event type

cleanup

public void cleanup()

Description copied from interface: Parser

Perform cleanup operations after parsing.

Specified by:

cleanup in interface Parser

Overrides:

cleanup in class AbstractStreamingParser

buildMeta

protected void buildMeta()

Update the meta element.

data

public java.lang.Object data(int rnum)

Description copied from interface: BundleStreamSource

Access a particular object and representation.

Parameters:

rnum - Representation number

Returns:

Contained data

parseLineInternal

protected boolean parseLineInternal()

Internal method for parsing a single line. Used by both line based parsing as well as block parsing. This saves the building of meta data for each line.

Returns:

true when a valid line was read, false on a label row.

createVector

protected V createVector()

Creates a database object of type V.

Returns:

a vector of type V containing the given attribute values

getTypeInformation

SimpleTypeInformation<V> getTypeInformation(int mindim,
                                            int maxdim)

Get a prototype object for the given dimensionality.

Parameters:

mindim - Minimum dimensionality

maxdim - Maximum dimensionality

Returns:

Prototype object

getLogger

protected Logging getLogger()

Description copied from class: AbstractStreamingParser

Get the logger for this class.

Specified by:

getLogger in class AbstractStreamingParser

Returns:

Logger.