Fixed bug in bucketizing function from histogram

Author: ArnauPrat

src/main/java/ldbc/snb/datagen/generator/distribution/AltmannDistribution.java

@@ -12,7 +12,7 @@
  */
 public class AltmannDistribution extends BucketedDistribution {
 
-    private int NUM_BUCKETS_ = 100;
+    private int NUM_BUCKETS_ = 10000;
     private int POPULATION_ = 10000;
     private double normalization_factor_ = 0.0;
     private double GAMMA_ = 0.4577;
@@ -21,14 +21,9 @@ public class AltmannDistribution extends BucketedDistribution {
     private ArrayList<Bucket> buckets_ = null;
 
     @Override
-    public ArrayList<Bucket> getBuckets() {
-        return buckets_;
-    }
-
-    @Override
-    public void initialize(Configuration conf) {
-        throw new UnsupportedOperationException("Distribution not implemented");
-        /*POPULATION_ = DatagenParams.numPersons;
+        public ArrayList<Bucket> getBuckets() {
+        //throw new UnsupportedOperationException("Distribution not implemented");
+        POPULATION_ = DatagenParams.numPersons;
         for( int i = 1; i <= POPULATION_; ++i ) {
             normalization_factor_+= Math.pow(i,-GAMMA_)*Math.exp(-DELTA_*i);
         }
@@ -40,9 +35,10 @@ public void initialize(Configuration conf) {
         double scale_factor = DatagenParams.numPersons / POPULATION_;
         buckets_ = Bucket.bucketizeHistogram(histogram, NUM_BUCKETS_);
         for( Bucket e : buckets_) {
+            System.out.println((e.min()*scale_factor)+" "+(e.max()*scale_factor));
             e.max(e.max()*scale_factor);
             e.min(e.min()*scale_factor);
         }
-        */
+        return buckets_;
     }
 }

src/main/java/ldbc/snb/datagen/generator/distribution/WeibullDistribution.java

@@ -20,11 +20,6 @@ public class WeibullDistribution extends BucketedDistribution {
 
     @Override
     public ArrayList<Bucket> getBuckets() {
-        return buckets_;
-    }
-
-    @Override
-    public void initialize(Configuration conf) {
         weibull_ = new org.apache.commons.math3.distribution.WeibullDistribution(LAMBDA_,K_);
 
         ArrayList<Double> histogram = new ArrayList<Double>();
@@ -34,5 +29,6 @@ public void initialize(Configuration conf) {
         }
 
         buckets_ = Bucket.bucketizeHistogram(histogram,100);
+        return buckets_;
     }
 }

src/main/java/ldbc/snb/datagen/generator/distribution/ZetaDistribution.java renamed to src/main/java/ldbc/snb/datagen/generator/distribution/ZipfDistribution.java

@@ -1,23 +1,20 @@
 package ldbc.snb.datagen.generator.distribution;
 
 import ldbc.snb.datagen.generator.DatagenParams;
-import ldbc.snb.datagen.generator.distribution.utils.Bucket;
-import ldbc.snb.datagen.generator.distribution.utils.BucketedDistribution;
-import org.apache.commons.math3.distribution.ZipfDistribution;
 import org.apache.hadoop.conf.Configuration;
 
 import java.util.ArrayList;
 
 /**
  * Created by aprat on 5/03/15.
  */
-public class ZetaDistribution implements DegreeDistribution {
+public class ZipfDistribution implements DegreeDistribution {
 
-    private ZipfDistribution zipf_;
+    private org.apache.commons.math3.distribution.ZipfDistribution zipf_;
     private double ALPHA_ = 1.7;
 
     public void initialize(Configuration conf) {
-        zipf_ = new ZipfDistribution(DatagenParams.numPersons, ALPHA_);
+        zipf_ = new org.apache.commons.math3.distribution.ZipfDistribution(10000, ALPHA_);
     }
 
     public void reset (long seed) {

src/main/java/ldbc/snb/datagen/generator/distribution/utils/Bucket.java

@@ -14,13 +14,13 @@ public static ArrayList<Bucket> bucketizeHistogram(ArrayList<Double> histogram,
         for( Double d : histogram )  {
            population+=d;
         }
-        int percentile_size = population.intValue() / num_buckets;
+        double percentile = 1 / (double)num_buckets;
         int current_histogram_index = 0;
         int current_histogram_counter = histogram.get(current_histogram_index).intValue();
         for( int i = 0; i < num_buckets; ++i ) {
-            double min = population.intValue();
+            double min = population;
             double max = 0;
-            for( int j = 0; j < percentile_size; ++j ) {
+            for( int j = 0; j/(double)population < percentile; ++j ) {
                 min = min > (current_histogram_index+1) ? (current_histogram_index+1) : min;
                 max = max < (current_histogram_index+1) ? (current_histogram_index+1) : max;
                 if(--current_histogram_counter <= 0) {

src/main/java/ldbc/snb/datagen/generator/distribution/utils/BucketedDistribution.java

@@ -37,8 +37,11 @@ public void reset(long seed) {
     public long nextDegree() {
         int idx = randomPercentile_.nextInt(buckets_.size());
         int minRange = (int)(buckets_.get(idx).min());
-        int maxRange = (int)(Math.floor(buckets_.get(idx).max()));
-        if( maxRange < minRange ) maxRange = minRange;
-        return randomDegree_.get(idx).nextInt( maxRange - minRange  + 1) + minRange;
+        int maxRange = (int)(buckets_.get(idx).max());
+        if( maxRange < minRange ) {
+            maxRange = minRange;
+        }
+        long ret= randomDegree_.get(idx).nextInt( maxRange - minRange  + 1) + minRange;
+        return ret;
     }
 }