Package org.terracotta.statistics.derived.histogram

Class ExponentialHistogram

java.lang.Object

org.terracotta.statistics.derived.histogram.ExponentialHistogram

public class ExponentialHistogram
extends Object

An implementation of the Exponential Histogram sketch as outlined by Datar et al.

This class is *not thread-safe*, safe consumption in a multi-threaded environment will require some form of external locking.

See Also:

Maintaining Stream Statistics over Sliding Windows

Constructor Summary

Constructors

Constructor	Description
ExponentialHistogram(double epsilon, long window)	Creates an exponential histogram maintaining a count over window to within @{epsilon} fractional accuracy.

Method Summary

Modifier and Type	Method	Description
long	count()	Returns the approximate current count.
double	epsilon()	Return the fractional accuracy of this exponential histogram
long	expire(long time)	Expire old events.
void	insert(long time)	Insert a single event at time
void	insert(long time, long count)	Bulk insert count events at time.
void	merge(ExponentialHistogram b)	Merge the supplied ExponentialHistogram in to this one.
ExponentialHistogram	split(double fraction)	Split an exponential histogram off this one.
String	toString()

Methods inherited from class java.lang.Object

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

Constructor Detail

ExponentialHistogram

public ExponentialHistogram(double epsilon,
                            long window)

Creates an exponential histogram maintaining a count over window to within @{epsilon} fractional accuracy.

Parameters:

epsilon - fractional accuracy

window - sliding window size

Method Detail

merge

public void merge(ExponentialHistogram b)

Merge the supplied ExponentialHistogram in to this one.

Parameters:

b - histogram to merge

Throws:

IllegalArgumentException - if the two merge-thresholds are not equals

insert

public void insert(long time,
                   long count)
            throws IllegalArgumentException

Bulk insert count events at time.

Parameters:

time - event time

count - event count

Throws:

IllegalArgumentException - if count is negative

insert

public void insert(long time)

Insert a single event at time

Parameters:

time - event timestamp

expire

public long expire(long time)

Expire old events.

Parameters:

time - current timestamp

Returns:

the count following expiry

count

public long count()

Returns the approximate current count.

Returns:

the approximate count

split

public ExponentialHistogram split(double fraction)

Split an exponential histogram off this one.

The returned histogram will contain {code fraction} of the events in this one.

Parameters:

fraction - splitting fraction

Returns:

the new histogram

toString

public String toString()

Overrides:

toString in class Object

epsilon

public double epsilon()

Return the fractional accuracy of this exponential histogram

Returns:

fractional accuracy