hadoop-tools/hadoop-rumen/src/test/java/org/apache/hadoop/tools/rumen/TestPiecewiseLinearInterpolation.java

/**
 * Licensed to the Apache Software Foundation (ASF) under one
 * or more contributor license agreements.  See the NOTICE file
 * distributed with this work for additional information
 * regarding copyright ownership.  The ASF licenses this file
 * to you 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.
 */

package org.apache.hadoop.tools.rumen;

import java.util.ArrayList;

import org.junit.Test;
import static org.junit.Assert.*;

public class TestPiecewiseLinearInterpolation {

  static private double maximumRelativeError = 0.002D;

  static private LoggedSingleRelativeRanking makeRR(double ranking, long datum) {
    LoggedSingleRelativeRanking result = new LoggedSingleRelativeRanking();

    result.setDatum(datum);
    result.setRelativeRanking(ranking);

    return result;
  }

  @Test
  public void testOneRun() {
    LoggedDiscreteCDF input = new LoggedDiscreteCDF();

    input.setMinimum(100000L);
    input.setMaximum(1100000L);

    ArrayList<LoggedSingleRelativeRanking> rankings = new ArrayList<LoggedSingleRelativeRanking>();

    rankings.add(makeRR(0.1, 200000L));
    rankings.add(makeRR(0.5, 800000L));
    rankings.add(makeRR(0.9, 1000000L));

    input.setRankings(rankings);
    input.setNumberValues(3);

    CDFRandomGenerator gen = new CDFPiecewiseLinearRandomGenerator(input);
    Histogram values = new Histogram();

    for (int i = 0; i < 1000000; ++i) {
      long value = gen.randomValue();
      values.enter(value);
    }

    /*
     * Now we build a percentiles CDF, and compute the sum of the squares of the
     * actual percentiles vrs. the predicted percentiles
     */
    int[] percentiles = new int[99];

    for (int i = 0; i < 99; ++i) {
      percentiles[i] = i + 1;
    }

    long[] result = values.getCDF(100, percentiles);
    long sumErrorSquares = 0L;

    for (int i = 0; i < 10; ++i) {
      long error = result[i] - (10000L * i + 100000L);
      System.out.println("element " + i + ", got " + result[i] + ", expected "
          + (10000L * i + 100000L) + ", error = " + error);
      sumErrorSquares += error * error;
    }

    for (int i = 10; i < 50; ++i) {
      long error = result[i] - (15000L * i + 50000L);
      System.out.println("element " + i + ", got " + result[i] + ", expected "
          + (15000L * i + 50000L) + ", error = " + error);
      sumErrorSquares += error * error;
    }

    for (int i = 50; i < 90; ++i) {
      long error = result[i] - (5000L * i + 550000L);
      System.out.println("element " + i + ", got " + result[i] + ", expected "
          + (5000L * i + 550000L) + ", error = " + error);
      sumErrorSquares += error * error;
    }

    for (int i = 90; i <= 100; ++i) {
      long error = result[i] - (10000L * i + 100000L);
      System.out.println("element " + i + ", got " + result[i] + ", expected "
          + (10000L * i + 100000L) + ", error = " + error);
      sumErrorSquares += error * error;
    }

    // normalize the error
    double realSumErrorSquares = (double) sumErrorSquares;

    double normalizedError = realSumErrorSquares / 100
        / rankings.get(1).getDatum() / rankings.get(1).getDatum();
    double RMSNormalizedError = Math.sqrt(normalizedError);

    System.out.println("sumErrorSquares = " + sumErrorSquares);

    System.out.println("normalizedError: " + normalizedError
        + ", RMSNormalizedError: " + RMSNormalizedError);

    System.out.println("Cumulative error is " + RMSNormalizedError);

    assertTrue("The RMS relative error per bucket, " + RMSNormalizedError
        + ", exceeds our tolerance of " + maximumRelativeError,
        RMSNormalizedError <= maximumRelativeError);

  }
}