src/main/java/org/apache/datasketches/characterization/uniquecount/BaseBoundsAccuracyProfile.java - datasketches-characterization - Git at Google

 /*
  * 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.datasketches.characterization.uniquecount;

 import static org.apache.datasketches.GaussianRanks.GAUSSIANS_3SD;
 import static org.apache.datasketches.common.Util.milliSecToString;
 import static org.apache.datasketches.common.Util.pwr2SeriesNext;

 import java.io.PrintWriter;

 import org.apache.datasketches.Job;
 import org.apache.datasketches.JobProfile;
 import org.apache.datasketches.Properties;
 import org.apache.datasketches.characterization.BoundsAccuracyStats;
 import org.apache.datasketches.quantiles.DoublesSketch;

 /**
  * @author Lee Rhodes
  */
 public abstract class BaseBoundsAccuracyProfile implements JobProfile {
   Job job;
   PrintWriter pw;
   public Properties prop;
   public long vIn = 0;
   int lgMinT;
   int lgMaxT;
   int tPPO;
   int lgMinU;
   int lgMaxU;
   int uPPO;
   int lgQK;
   public int lgK;
   boolean interData;
   boolean postPMFs;
   public boolean getSize = false;
   public BoundsAccuracyStats[] qArr;

   //JobProfile
   @Override
   public void start(final Job job) {
     this.job = job;
     pw = job.getPrintWriter();
     prop = job.getProperties();
     lgMinT = Integer.parseInt(prop.mustGet("Trials_lgMinT"));
     lgMaxT = Integer.parseInt(prop.mustGet("Trials_lgMaxT"));
     tPPO = Integer.parseInt(prop.mustGet("Trials_TPPO"));
     lgMinU = Integer.parseInt(prop.mustGet("Trials_lgMinU"));
     lgMaxU = Integer.parseInt(prop.mustGet("Trials_lgMaxU"));
     interData = Boolean.parseBoolean(prop.mustGet("Trials_interData"));
     postPMFs = Boolean.parseBoolean(prop.mustGet("Trials_postPMFs"));
     uPPO = Integer.parseInt(prop.mustGet("Trials_UPPO"));
     lgQK = Integer.parseInt(prop.mustGet("Trials_lgQK"));
     qArr = BoundsAccuracyStats.buildAccuracyStatsArray(lgMinU, lgMaxU, uPPO, lgQK);
     lgK = Integer.parseInt(prop.mustGet("LgK"));
     configure();
     doTrials();
     shutdown();
     cleanup();
   }

   @Override
   public void shutdown() {}

   @Override
   public void cleanup() {}

   //end JobProfile

   public abstract void configure();

   /**
    * An accuracy trial is one pass through all uniques, pausing to store the estimate into a
    * quantiles sketch at each point along the unique axis.
    */
   public abstract void doTrial();

   /**
    * Manages multiple trials for measuring accuracy.
    *
    * <p>An accuracy trial is run along the count axis (X-axis) first. The "points" along the X-axis
    * where accuracy data is collected is controlled by the data loaded into the CountAccuracyStats
    * array. A single trial consists of a single sketch being updated with the Trials_lgMaxU unique
    * values, stopping at the configured x-axis points along the way where the accuracy is recorded
    * into the corresponding stats array. Each stats array retains the distribution of
    * the accuracies measured for all the trials at that x-axis point.
    *
    * <p>Because accuracy trials take a long time, this profile will output intermediate
    * accuracy results starting after Trials_lgMinT trials and then again at trial intervals
    * determined by Trials_TPPO until Trials_lgMaxT.  This allows you to stop the testing at
    * any intermediate trials point if you feel you have sufficient trials for the accuracy you
    * need.
    */
   private void doTrials() {
     final int minT = 1 << lgMinT;
     final int maxT = 1 << lgMaxT;
     final int maxU = 1 << lgMaxU;

     //This will generate a table of data up for each intermediate Trials point
     int lastT = 0;
     while (lastT < maxT) {
       final int nextT = lastT == 0 ? minT : (int)pwr2SeriesNext(tPPO, lastT);
       final int delta = nextT - lastT;
       for (int i = 0; i < delta; i++) {
         doTrial();
       }
       lastT = nextT;
       final StringBuilder sb = new StringBuilder();
       if (nextT < maxT) { // intermediate
         if (interData) {
           job.println(getHeader());
           process(qArr, lastT, sb);
           job.println(sb.toString());
         }
       } else { //done
         job.println(getHeader());
         process(qArr, lastT, sb);
         job.println(sb.toString());
       }

       job.println(prop.extractKvPairs());
       job.println("Cum Trials             : " + lastT);
       job.println("Cum Updates            : " + vIn);
       final long currentTime_mS = System.currentTimeMillis();
       final long cumTime_mS = currentTime_mS - job.getStartTime();
       job.println("Cum Time               : " + milliSecToString(cumTime_mS));
       final double timePerTrial_mS = cumTime_mS * 1.0 / lastT;
       final double avgUpdateTime_ns = timePerTrial_mS * 1e6 / maxU;
       job.println("Time Per Trial, mSec   : " + timePerTrial_mS);
       job.println("Avg Update Time, nSec  : " + avgUpdateTime_ns);
       job.println("Date Time              : "
           + job.getReadableDateString(currentTime_mS));

       final long timeToComplete_mS = (long)(timePerTrial_mS * (maxT - lastT));
       job.println("Est Time to Complete   : " + milliSecToString(timeToComplete_mS));
       job.println("Est Time at Completion : "
           + job.getReadableDateString(timeToComplete_mS + currentTime_mS));
       job.println("");
       if (postPMFs) {
         for (int i = 0; i < qArr.length; i++) {
           job.println(outputPMF(qArr[i]));
         }
       }
       job.flush();
     }
   }

   private static void process(final BoundsAccuracyStats[] qArr,
       final int cumTrials, final StringBuilder sb) {

     final int points = qArr.length;
     sb.setLength(0);
     for (int pt = 0; pt < points; pt++) {
       final BoundsAccuracyStats q = qArr[pt];
       final double uniques = q.trueValue;
       final double relLb3 = q.sumLB3 / cumTrials / uniques - 1.0;
       final double relLb2 = q.sumLB2 / cumTrials / uniques - 1.0;
       final double relLb1 = q.sumLB1 / cumTrials / uniques - 1.0;
       final double relUb1 = q.sumUB1 / cumTrials / uniques - 1.0;
       final double relUb2 = q.sumUB2 / cumTrials / uniques - 1.0;
       final double relUb3 = q.sumUB3 / cumTrials / uniques - 1.0;

       //OUTPUT
       sb.append(uniques).append(TAB);
       //TRIALS
       sb.append(cumTrials).append(TAB);

       //Quantiles
       final double[] quants = q.qskEst.getQuantiles(GAUSSIANS_3SD);
       final int len = quants.length;
       for (int i = 0; i < len; i++) {
         sb.append(quants[i] / uniques - 1.0).append(TAB);
       }
       //Bound averages
       sb.append(relLb3).append(TAB);
       sb.append(relLb2).append(TAB);
       sb.append(relLb1).append(TAB);
       sb.append(relUb1).append(TAB);
       sb.append(relUb2).append(TAB);
       sb.append(relUb3).append(TAB);
       sb.append(LS);
     }
   }

   private static String getHeader() {
     final StringBuilder sb = new StringBuilder();
     sb.append("InU").append(TAB);
     //Trials
     sb.append("Trials").append(TAB);
     //Quantiles
     sb.append("Min").append(TAB);
     sb.append("Q(.00135)").append(TAB);
     sb.append("Q(.02275)").append(TAB);
     sb.append("Q(.15866)").append(TAB);
     sb.append("Q(.5)").append(TAB);
     sb.append("Q(.84134)").append(TAB);
     sb.append("Q(.97725)").append(TAB);
     sb.append("Q(.99865)").append(TAB);
     sb.append("Max").append(TAB);
     sb.append("avgLB3").append(TAB);
     sb.append("avgLB2").append(TAB);
     sb.append("avgLB1").append(TAB);
     sb.append("avgUB1").append(TAB);
     sb.append("avgUB2").append(TAB);
     sb.append("avgUB3").append(TAB);
     return sb.toString();
   }

   /**
    * Outputs the Probability Mass Function given the AccuracyStats.
    * @param q the given AccuracyStats
    */
   private static String outputPMF(final BoundsAccuracyStats q) {
     final DoublesSketch qSk = q.qskEst;
     final double[] splitPoints = qSk.getQuantiles(GAUSSIANS_3SD); //1:1
     final double[] reducedSp = reduceSplitPoints(splitPoints);
     final double[] pmfArr = qSk.getPMF(reducedSp); //pmfArr is one larger
     final long trials = qSk.getN();
     final StringBuilder sb = new StringBuilder();

     //output Histogram
     final String hdr = String.format("%10s%4s%12s", "Trials", "    ", "Est");
     final String fmt = "%10d%4s%12.2f";
     sb.append("Histogram At " + q.trueValue).append(LS);
     sb.append(hdr).append(LS);
     for (int i = 0; i < reducedSp.length; i++) {
       final int hits = (int)(pmfArr[i + 1] * trials);
       final double est = reducedSp[i];
       final String line = String.format(fmt, hits, " >= ", est);
       sb.append(line).append(LS);
     }
     return sb.toString();
   }

   /**
    *
    * @param splitPoints the given splitPoints
    * @return the reduced array of splitPoints
    */
   private static double[] reduceSplitPoints(final double[] splitPoints) {
     int num = 1;
     double lastV = splitPoints[0];
     for (int i = 0; i < splitPoints.length; i++) {
       final double v = splitPoints[i];
       if (v <= lastV) { continue; }
       num++;
       lastV = v;
     }
     lastV = splitPoints[0];
     int idx = 0;
     final double[] sp = new double[num];
     sp[0] = lastV;
     for (int i = 0; i < splitPoints.length; i++) {
       final double v = splitPoints[i];
       if (v <= lastV) { continue; }
       sp[++idx] = v;
       lastV = v;
     }
     return sp;
   }

 }
	/*
	* 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.datasketches.characterization.uniquecount;

	import static org.apache.datasketches.GaussianRanks.GAUSSIANS_3SD;
	import static org.apache.datasketches.common.Util.milliSecToString;
	import static org.apache.datasketches.common.Util.pwr2SeriesNext;

	import java.io.PrintWriter;

	import org.apache.datasketches.Job;
	import org.apache.datasketches.JobProfile;
	import org.apache.datasketches.Properties;
	import org.apache.datasketches.characterization.BoundsAccuracyStats;
	import org.apache.datasketches.quantiles.DoublesSketch;

	/**
	* @author Lee Rhodes
	*/
	public abstract class BaseBoundsAccuracyProfile implements JobProfile {
	Job job;
	PrintWriter pw;
	public Properties prop;
	public long vIn = 0;
	int lgMinT;
	int lgMaxT;
	int tPPO;
	int lgMinU;
	int lgMaxU;
	int uPPO;
	int lgQK;
	public int lgK;
	boolean interData;
	boolean postPMFs;
	public boolean getSize = false;
	public BoundsAccuracyStats[] qArr;

	//JobProfile
	@Override
	public void start(final Job job) {
	this.job = job;
	pw = job.getPrintWriter();
	prop = job.getProperties();
	lgMinT = Integer.parseInt(prop.mustGet("Trials_lgMinT"));
	lgMaxT = Integer.parseInt(prop.mustGet("Trials_lgMaxT"));
	tPPO = Integer.parseInt(prop.mustGet("Trials_TPPO"));
	lgMinU = Integer.parseInt(prop.mustGet("Trials_lgMinU"));
	lgMaxU = Integer.parseInt(prop.mustGet("Trials_lgMaxU"));
	interData = Boolean.parseBoolean(prop.mustGet("Trials_interData"));
	postPMFs = Boolean.parseBoolean(prop.mustGet("Trials_postPMFs"));
	uPPO = Integer.parseInt(prop.mustGet("Trials_UPPO"));
	lgQK = Integer.parseInt(prop.mustGet("Trials_lgQK"));
	qArr = BoundsAccuracyStats.buildAccuracyStatsArray(lgMinU, lgMaxU, uPPO, lgQK);
	lgK = Integer.parseInt(prop.mustGet("LgK"));
	configure();
	doTrials();
	shutdown();
	cleanup();
	}

	@Override
	public void shutdown() {}

	@Override
	public void cleanup() {}

	//end JobProfile

	public abstract void configure();

	/**
	* An accuracy trial is one pass through all uniques, pausing to store the estimate into a
	* quantiles sketch at each point along the unique axis.
	*/
	public abstract void doTrial();

	/**
	* Manages multiple trials for measuring accuracy.
	*
	* <p>An accuracy trial is run along the count axis (X-axis) first. The "points" along the X-axis
	* where accuracy data is collected is controlled by the data loaded into the CountAccuracyStats
	* array. A single trial consists of a single sketch being updated with the Trials_lgMaxU unique
	* values, stopping at the configured x-axis points along the way where the accuracy is recorded
	* into the corresponding stats array. Each stats array retains the distribution of
	* the accuracies measured for all the trials at that x-axis point.
	*
	* <p>Because accuracy trials take a long time, this profile will output intermediate
	* accuracy results starting after Trials_lgMinT trials and then again at trial intervals
	* determined by Trials_TPPO until Trials_lgMaxT. This allows you to stop the testing at
	* any intermediate trials point if you feel you have sufficient trials for the accuracy you
	* need.
	*/
	private void doTrials() {
	final int minT = 1 << lgMinT;
	final int maxT = 1 << lgMaxT;
	final int maxU = 1 << lgMaxU;

	//This will generate a table of data up for each intermediate Trials point
	int lastT = 0;
	while (lastT < maxT) {
	final int nextT = lastT == 0 ? minT : (int)pwr2SeriesNext(tPPO, lastT);
	final int delta = nextT - lastT;
	for (int i = 0; i < delta; i++) {
	doTrial();
	}
	lastT = nextT;
	final StringBuilder sb = new StringBuilder();
	if (nextT < maxT) { // intermediate
	if (interData) {
	job.println(getHeader());
	process(qArr, lastT, sb);
	job.println(sb.toString());
	}
	} else { //done
	job.println(getHeader());
	process(qArr, lastT, sb);
	job.println(sb.toString());
	}

	job.println(prop.extractKvPairs());
	job.println("Cum Trials : " + lastT);
	job.println("Cum Updates : " + vIn);
	final long currentTime_mS = System.currentTimeMillis();
	final long cumTime_mS = currentTime_mS - job.getStartTime();
	job.println("Cum Time : " + milliSecToString(cumTime_mS));
	final double timePerTrial_mS = cumTime_mS * 1.0 / lastT;
	final double avgUpdateTime_ns = timePerTrial_mS * 1e6 / maxU;
	job.println("Time Per Trial, mSec : " + timePerTrial_mS);
	job.println("Avg Update Time, nSec : " + avgUpdateTime_ns);
	job.println("Date Time : "
	+ job.getReadableDateString(currentTime_mS));

	final long timeToComplete_mS = (long)(timePerTrial_mS * (maxT - lastT));
	job.println("Est Time to Complete : " + milliSecToString(timeToComplete_mS));
	job.println("Est Time at Completion : "
	+ job.getReadableDateString(timeToComplete_mS + currentTime_mS));
	job.println("");
	if (postPMFs) {
	for (int i = 0; i < qArr.length; i++) {
	job.println(outputPMF(qArr[i]));
	}
	}
	job.flush();
	}
	}

	private static void process(final BoundsAccuracyStats[] qArr,
	final int cumTrials, final StringBuilder sb) {

	final int points = qArr.length;
	sb.setLength(0);
	for (int pt = 0; pt < points; pt++) {
	final BoundsAccuracyStats q = qArr[pt];
	final double uniques = q.trueValue;
	final double relLb3 = q.sumLB3 / cumTrials / uniques - 1.0;
	final double relLb2 = q.sumLB2 / cumTrials / uniques - 1.0;
	final double relLb1 = q.sumLB1 / cumTrials / uniques - 1.0;
	final double relUb1 = q.sumUB1 / cumTrials / uniques - 1.0;
	final double relUb2 = q.sumUB2 / cumTrials / uniques - 1.0;
	final double relUb3 = q.sumUB3 / cumTrials / uniques - 1.0;

	//OUTPUT
	sb.append(uniques).append(TAB);
	//TRIALS
	sb.append(cumTrials).append(TAB);

	//Quantiles
	final double[] quants = q.qskEst.getQuantiles(GAUSSIANS_3SD);
	final int len = quants.length;
	for (int i = 0; i < len; i++) {
	sb.append(quants[i] / uniques - 1.0).append(TAB);
	}
	//Bound averages
	sb.append(relLb3).append(TAB);
	sb.append(relLb2).append(TAB);
	sb.append(relLb1).append(TAB);
	sb.append(relUb1).append(TAB);
	sb.append(relUb2).append(TAB);
	sb.append(relUb3).append(TAB);
	sb.append(LS);
	}
	}

	private static String getHeader() {
	final StringBuilder sb = new StringBuilder();
	sb.append("InU").append(TAB);
	//Trials
	sb.append("Trials").append(TAB);
	//Quantiles
	sb.append("Min").append(TAB);
	sb.append("Q(.00135)").append(TAB);
	sb.append("Q(.02275)").append(TAB);
	sb.append("Q(.15866)").append(TAB);
	sb.append("Q(.5)").append(TAB);
	sb.append("Q(.84134)").append(TAB);
	sb.append("Q(.97725)").append(TAB);
	sb.append("Q(.99865)").append(TAB);
	sb.append("Max").append(TAB);
	sb.append("avgLB3").append(TAB);
	sb.append("avgLB2").append(TAB);
	sb.append("avgLB1").append(TAB);
	sb.append("avgUB1").append(TAB);
	sb.append("avgUB2").append(TAB);
	sb.append("avgUB3").append(TAB);
	return sb.toString();
	}

	/**
	* Outputs the Probability Mass Function given the AccuracyStats.
	* @param q the given AccuracyStats
	*/
	private static String outputPMF(final BoundsAccuracyStats q) {
	final DoublesSketch qSk = q.qskEst;
	final double[] splitPoints = qSk.getQuantiles(GAUSSIANS_3SD); //1:1
	final double[] reducedSp = reduceSplitPoints(splitPoints);
	final double[] pmfArr = qSk.getPMF(reducedSp); //pmfArr is one larger
	final long trials = qSk.getN();
	final StringBuilder sb = new StringBuilder();

	//output Histogram
	final String hdr = String.format("%10s%4s%12s", "Trials", " ", "Est");
	final String fmt = "%10d%4s%12.2f";
	sb.append("Histogram At " + q.trueValue).append(LS);
	sb.append(hdr).append(LS);
	for (int i = 0; i < reducedSp.length; i++) {
	final int hits = (int)(pmfArr[i + 1] * trials);
	final double est = reducedSp[i];
	final String line = String.format(fmt, hits, " >= ", est);
	sb.append(line).append(LS);
	}
	return sb.toString();
	}

	/**
	*
	* @param splitPoints the given splitPoints
	* @return the reduced array of splitPoints
	*/
	private static double[] reduceSplitPoints(final double[] splitPoints) {
	int num = 1;
	double lastV = splitPoints[0];
	for (int i = 0; i < splitPoints.length; i++) {
	final double v = splitPoints[i];
	if (v <= lastV) { continue; }
	num++;
	lastV = v;
	}
	lastV = splitPoints[0];
	int idx = 0;
	final double[] sp = new double[num];
	sp[0] = lastV;
	for (int i = 0; i < splitPoints.length; i++) {
	final double v = splitPoints[i];
	if (v <= lastV) { continue; }
	sp[++idx] = v;
	lastV = v;
	}
	return sp;
	}

	}