All Downloads are FREE. Search and download functionalities are using the official Maven repository.

com.aliyun.openservices.shade.io.prometheus.client.CKMSQuantiles Maven / Gradle / Ivy

There is a newer version: 1.9.4.Final
Show newest version
package com.aliyun.openservices.shade.io.prometheus.client;

// The original implementation was copied from
// https://raw.githubusercontent.com/Netflix/ocelli/master/ocelli-core/src/main/java/netflix/ocelli/stats/CKMSQuantiles.java
// Revision d0357b8bf5c17a173ce94d6b26823775b3f999f6 from Jan 21, 2015.
// However, it has been heavily refactored in the meantime.

/*
 Copyright 2012 Andrew Wang ([email protected])

 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
 You may obtain a copy of the License at

 http://www.apache.org/licenses/LICENSE-2.0

 Unless required by applicable law or agreed to in writing, software
 distributed under the License is distributed on an "AS IS" BASIS,
 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License.
 */

import java.util.Arrays;
import java.util.Iterator;
import java.util.LinkedList;
import java.util.ListIterator;

/**
 * Algorithm solving the "Targeted Quantile Problem" as described in
 * "Effective Computation of Biased Quantiles over Data Streams"
 * by Cormode, Korn, Muthukrishnan, and Srivastava.
 *
 */
final class CKMSQuantiles {

    final Quantile[] quantiles;

    /**
     * Total number of observations (not including those that are still in the buffer).
     */
    int n = 0;

    /**
     * List of sampled observations, ordered by Sample.value.
     */
    final LinkedList samples = new LinkedList();

    /**
     * Compress is called every compressInterval inserts.
     * Note that the buffer is flushed whenever get() is called, so we
     * cannot just wait until the buffer is full before we call compress.
     */
    private final int compressInterval = 128;
    private int insertsSinceLastCompress = 0;

    /**
     * Note that the buffer size could as well be less than the compressInterval.
     * However, the buffer size should not be greater than the compressInterval,
     * because the compressInterval is not respected in flush(), so if you want
     * to compress more often than calling flush() that won't work.
     */
    private final double[] buffer = new double[compressInterval];
    private int bufferPos = 0;

    public CKMSQuantiles(Quantile... quantiles) {
        if (quantiles.length == 0) {
            throw new IllegalArgumentException("quantiles cannot be empty");
        }
        this.quantiles = quantiles;
    }

    /**
     * Add an observed value
     */
    public void insert(double value) {
        buffer[bufferPos++] = value;

        if (bufferPos == buffer.length) {
            flush();
        }

        if (++insertsSinceLastCompress == compressInterval) {
            compress();
            insertsSinceLastCompress = 0;
        }
    }

    private void flush() {
        Arrays.sort(buffer, 0, bufferPos);
        insertBatch(buffer, bufferPos);
        bufferPos = 0;
    }

    /**
     * Inserts the elements from index 0 to index toIndex from the sortedBuffer.
     */
    void insertBatch(double[] sortedBuffer, int toIndex) {
        if (toIndex == 0) {
            return;
        }
        ListIterator iterator = samples.listIterator();
        int i = 0; // position in buffer
        int r = 0; // sum of g's left of the current sample
        while (iterator.hasNext() && i < toIndex) {
            Sample item = iterator.next();
            while (i < toIndex) {
                if (sortedBuffer[i] > item.value) {
                    break;
                }
                insertBefore(iterator, sortedBuffer[i], r);
                r++; // new item with g=1 was inserted before, so increment r
                i++;
                n++;
            }
            r += item.g;
        }
        while (i < toIndex) {
            samples.add(new Sample(sortedBuffer[i], 0));
            i++;
            n++;
        }
    }

    private void insertBefore(ListIterator iterator, double value, int r) {
        if (!iterator.hasPrevious()) {
            samples.addFirst(new Sample(value, 0));
        } else {
            iterator.previous();
            iterator.add(new Sample(value, f(r) - 1));
            iterator.next();
        }
    }

    /**
     * Get the estimated value at the specified quantile.
     */
    public double get(double q) {
        flush();

        if (samples.size() == 0) {
            return Double.NaN;
        }

        if (q == 0.0) {
            return samples.getFirst().value;
        }

        if (q == 1.0) {
            return samples.getLast().value;
        }

        int r = 0; // sum of g's left of the current sample
        int desiredRank = (int) Math.ceil(q * n);

        ListIterator iterator = samples.listIterator();
        while (iterator.hasNext()) {
            Sample sample = iterator.next();
            if (r + sample.g + sample.delta > desiredRank + f(desiredRank) / 2) {
                iterator.previous(); // roll back the item.next() above
                if (iterator.hasPrevious()) {
                    Sample result = iterator.previous();
                    return result.value;
                } else {
                    return sample.value;
                }
            }
            r += sample.g;
        }
        return samples.getLast().value;
    }

    /**
     * Error function, as in definition 5 of the paper.
     */
    int f(int r) {
        int minResult = Integer.MAX_VALUE;
        for (Quantile q : quantiles) {
            if (q.quantile == 0 || q.quantile == 1) {
                continue;
            }
            int result;
            // We had a numerical error here with the following example:
            // quantile = 0.95, epsilon = 0.01, (n-r) = 30.
            // The expected result of (2*0.01*30)/(1-0.95) is 12. The actual result is 11.99999999999999.
            // To avoid running into these types of error we add 0.00000000001 before rounding down.
            if (r >= q.quantile * n) {
                result = (int) (q.v * r + 0.00000000001);
            } else {
                result = (int) (q.u * (n - r) + 0.00000000001);
            }
            if (result < minResult) {
                minResult = result;
            }
        }
        return Math.max(minResult, 1);
    }

    /**
     * Merge pairs of consecutive samples if this doesn't violate the error function.
     */
    void compress() {
        if (samples.size() < 3) {
            return;
        }
        Iterator descendingIterator = samples.descendingIterator();
        int r = n; // n is equal to the sum of the g's of all samples

        Sample right;
        Sample left = descendingIterator.next();
        r -= left.g;

        while (descendingIterator.hasNext()) {
            right = left;
            left = descendingIterator.next();
            r = r - left.g;
            if (left == samples.getFirst()) {
                // The min sample must never be merged.
                break;
            }
            if (left.g + right.g + right.delta < f(r)) {
                right.g += left.g;
                descendingIterator.remove();
                left = right;
            }
        }
    }

    static class Sample {

        /**
         * Observed value.
         */
        final double value;

        /**
         * Difference between the lowest possible rank of this sample and its predecessor.
         * This always starts with 1, but will be updated when compress() merges Samples.
         */
        int g = 1;

        /**
         * Difference between the greatest possible rank of this sample and the lowest possible rank of this sample.
         */
        final int delta;

        Sample(double value, int delta) {
            this.value = value;
            this.delta = delta;
        }

        @Override
        public String toString() {
            return String.format("Sample{val=%.3f, g=%d, delta=%d}", value, g, delta);
        }
    }

    static class Quantile {

        /**
         * Quantile. Must be between 0 and 1.
         */
        final double quantile;

        /**
         * Allowed error. Must be between 0 and 1.
         */
        final double epsilon;

        /**
         * Helper used in the error function f(), see definition 5 in the paper.
         */
        final double u;

        /**
         * Helper used in the error function f(), see definition 5 in the paper.
         */
        final double v;

        Quantile(double quantile, double epsilon) {
            if (quantile < 0.0 || quantile > 1.0) throw new IllegalArgumentException("Quantile must be between 0 and 1");
            if (epsilon < 0.0 || epsilon > 1.0) throw new IllegalArgumentException("Epsilon must be between 0 and 1");

            this.quantile = quantile;
            this.epsilon = epsilon;
            u = 2.0 * epsilon / (1.0 - quantile); // if quantile == 1 this will be Double.NaN
            v = 2.0 * epsilon / quantile; // if quantile == 0 this will be Double.NaN
        }

        @Override
        public String toString() {
            return String.format("Quantile{q=%.3f, epsilon=%.3f}", quantile, epsilon);
        }
    }
}




© 2015 - 2025 Weber Informatics LLC | Privacy Policy