SampleStatistic.java

// ========================================================================
// Copyright (c) 2009-2009 Mort Bay Consulting Pty. Ltd.
// ------------------------------------------------------------------------
// All rights reserved. This program and the accompanying materials
// are made available under the terms of the Eclipse Public License v1.0
// and Apache License v2.0 which accompanies this distribution.
// The Eclipse Public License is available at 
// http://www.eclipse.org/legal/epl-v10.html
// The Apache License v2.0 is available at
// http://www.opensource.org/licenses/apache2.0.php
// You may elect to redistribute this code under either of these licenses. 
// ========================================================================

package org.eclipse.jetty.util.statistic;

import java.util.concurrent.atomic.AtomicLong;


/* ------------------------------------------------------------ */
/**
 * SampledStatistics
 * <p>
 * Provides max, total, mean, count, variance, and standard 
 * deviation of continuous sequence of samples.
 * <p>
 * Calculates estimates of mean, variance, and standard deviation
 * characteristics of a sample using a non synchronized
 * approximation of the on-line algorithm presented
 * in Donald Knuth's Art of Computer Programming, Volume 2,
 * Seminumerical Algorithms, 3rd edition, page 232,
 * Boston: Addison-Wesley. that cites a 1962 paper by B.P. Welford
 * that can be found by following the link http://www.jstor.org/pss/1266577
 * <p>
 * This algorithm is also described in Wikipedia at
 * http://en.wikipedia.org/w/index.php?title=Algorithms_for_calculating_variance§ion=4#On-line_algorithm
 */
public class SampleStatistic
{
    protected final AtomicLong _max = new AtomicLong();
    protected final AtomicLong _total = new AtomicLong();
    protected final AtomicLong _count = new AtomicLong();
    protected final AtomicLong _totalVariance100 = new AtomicLong();

    public void reset()
    {
        _max.set(0);
        _total.set(0);
        _count.set(0);
        _totalVariance100.set(0);
    }

    public void set(final long sample)
    {
        long total = _total.addAndGet(sample);
        long count = _count.incrementAndGet();
        
        if (count>1)
        {
            long mean10 = total*10/count;
            long delta10 = sample*10 - mean10;
            _totalVariance100.addAndGet(delta10*delta10);
        }        
        
        long oldMax = _max.get();
        while (sample > oldMax)
        {
            if (_max.compareAndSet(oldMax, sample))
                break;
            oldMax = _max.get();
        }
        
    }

    /* ------------------------------------------------------------ */
    /**
     * @return the max value
     */
    public long getMax()
    {
        return _max.get();
    }

    /* ------------------------------------------------------------ */
    public long getTotal()
    {
        return _total.get();
    }

    /* ------------------------------------------------------------ */
    public long getCount()
    {
        return _count.get();
    }

    /* ------------------------------------------------------------ */
    public double getMean()
    {
        return (double)_total.get()/_count.get();
    }

    /* ------------------------------------------------------------ */
    public double getVariance()
    {
        final long variance100 = _totalVariance100.get();
        final long count = _count.get();
        
        return count>1?((double)variance100)/100.0/(count-1):0.0;
    }

    /* ------------------------------------------------------------ */
    public double getStdDev()
    {
        return Math.sqrt(getVariance());
    }

}