Package org.apache.solr.search.function.distance

Class VectorDistanceFunction

java.lang.Object

org.apache.lucene.queries.function.ValueSource

org.apache.solr.search.function.distance.VectorDistanceFunction

Direct Known Subclasses:

SquaredEuclideanFunction

public class VectorDistanceFunction
extends org.apache.lucene.queries.function.ValueSource

Calculate the p-norm for a Vector. See http://en.wikipedia.org/wiki/Lp_space

Common cases:

• 0 = Sparseness calculation
• 1 = Manhattan distance
• 2 = Euclidean distance
• Integer.MAX_VALUE = infinite norm

See Also:

for the special case

Field Summary

Fields

Modifier and Type	Field	Description
protected float	oneOverPower
protected float	power
protected org.apache.lucene.queries.function.valuesource.MultiValueSource	source1
protected org.apache.lucene.queries.function.valuesource.MultiValueSource	source2

Constructor Summary

Constructors

Constructor	Description
VectorDistanceFunction(float power, org.apache.lucene.queries.function.valuesource.MultiValueSource source1, org.apache.lucene.queries.function.valuesource.MultiValueSource source2)

Method Summary

Modifier and Type	Method	Description
void	createWeight(Map context, org.apache.lucene.search.IndexSearcher searcher)
String	description()
protected double	distance(int doc, org.apache.lucene.queries.function.FunctionValues dv1, org.apache.lucene.queries.function.FunctionValues dv2)	Calculate the distance
static double	distSquaredCartesian(double[] vec1, double[] vec2)	The square of the cartesian Distance.
boolean	equals(Object o)
org.apache.lucene.queries.function.FunctionValues	getValues(Map context, org.apache.lucene.index.LeafReaderContext readerContext)
int	hashCode()
protected String	name()
static double	vectorDistance(double[] vec1, double[] vec2, double power)	Calculate the p-norm (i.e.
static double	vectorDistance(double[] vec1, double[] vec2, double power, double oneOverPower)	Calculate the p-norm (i.e.

Methods inherited from class org.apache.lucene.queries.function.ValueSource

asDoubleValuesSource, asLongValuesSource, fromDoubleValuesSource, getSortField, newContext, toString

Methods inherited from class java.lang.Object

clone, finalize, getClass, notify, notifyAll, wait, wait, wait

Field Detail

source1

protected org.apache.lucene.queries.function.valuesource.MultiValueSource source1

source2

protected org.apache.lucene.queries.function.valuesource.MultiValueSource source2

power

protected float power

oneOverPower

protected float oneOverPower

Constructor Detail

VectorDistanceFunction

public VectorDistanceFunction(float power,
                              org.apache.lucene.queries.function.valuesource.MultiValueSource source1,
                              org.apache.lucene.queries.function.valuesource.MultiValueSource source2)

Method Detail

name

protected String name()

distance

protected double distance(int doc,
                          org.apache.lucene.queries.function.FunctionValues dv1,
                          org.apache.lucene.queries.function.FunctionValues dv2)
                   throws IOException

Calculate the distance

Parameters:

doc - The current doc

dv1 - The values from the first MultiValueSource

dv2 - The values from the second MultiValueSource

Returns:

The distance

Throws:

IOException

vectorDistance

public static double vectorDistance(double[] vec1,
                                    double[] vec2,
                                    double power)

Calculate the p-norm (i.e. length) between two vectors.

See Lp space

Parameters:

vec1 - The first vector

vec2 - The second vector

power - The power (2 for cartesian distance, 1 for manhattan, etc.)

Returns:

The length.

See Also:

vectorDistance(double[], double[], double, double)

vectorDistance

public static double vectorDistance(double[] vec1,
                                    double[] vec2,
                                    double power,
                                    double oneOverPower)

Calculate the p-norm (i.e. length) between two vectors.

Parameters:

vec1 - The first vector

vec2 - The second vector

power - The power (2 for cartesian distance, 1 for manhattan, etc.)

oneOverPower - If you've pre-calculated oneOverPower and cached it, use this method to save one division operation over vectorDistance(double[], double[], double).

Returns:

The length.

distSquaredCartesian

public static double distSquaredCartesian(double[] vec1,
                                          double[] vec2)

The square of the cartesian Distance. Not really a distance, but useful if all that matters is comparing the result to another one.

Parameters:

vec1 - The first point

vec2 - The second point

Returns:

The squared cartesian distance

getValues

public org.apache.lucene.queries.function.FunctionValues getValues(Map context,
                                                                   org.apache.lucene.index.LeafReaderContext readerContext)
                                                            throws IOException

Specified by:

getValues in class org.apache.lucene.queries.function.ValueSource

Throws:

IOException

createWeight

public void createWeight(Map context,
                         org.apache.lucene.search.IndexSearcher searcher)
                  throws IOException

Overrides:

createWeight in class org.apache.lucene.queries.function.ValueSource

Throws:

IOException

equals

public boolean equals(Object o)

Specified by:

equals in class org.apache.lucene.queries.function.ValueSource

hashCode

public int hashCode()

Specified by:

hashCode in class org.apache.lucene.queries.function.ValueSource

description

public String description()

Specified by:

description in class org.apache.lucene.queries.function.ValueSource