This is an automated email from the ASF dual-hosted git repository.
aherbert pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/commons-statistics.git
commit 0f68f194cc92558e15ca46c933f38760bf565825
Author: Alex Herbert <aherb...@apache.org>
AuthorDate: Fri Mar 14 13:23:09 2025 +0000
STATISTICS-89: Update descriptive statistics array creation benchmark
Support all the current double statistic implementations. Support
creation of statistics using part of the array range.
---
.../descriptive/StatisticCreationPerformance.java | 369 +++++++++++++++++++--
1 file changed, 344 insertions(+), 25 deletions(-)
diff --git
a/commons-statistics-examples/examples-jmh/src/main/java/org/apache/commons/statistics/examples/jmh/descriptive/StatisticCreationPerformance.java
b/commons-statistics-examples/examples-jmh/src/main/java/org/apache/commons/statistics/examples/jmh/descriptive/StatisticCreationPerformance.java
index fbe88a6..f293d38 100644
---
a/commons-statistics-examples/examples-jmh/src/main/java/org/apache/commons/statistics/examples/jmh/descriptive/StatisticCreationPerformance.java
+++
b/commons-statistics-examples/examples-jmh/src/main/java/org/apache/commons/statistics/examples/jmh/descriptive/StatisticCreationPerformance.java
@@ -19,9 +19,24 @@ package
org.apache.commons.statistics.examples.jmh.descriptive;
import java.util.Arrays;
import java.util.concurrent.TimeUnit;
-import java.util.function.DoubleConsumer;
import java.util.function.DoubleSupplier;
+import java.util.function.Function;
+import java.util.function.Supplier;
import org.apache.commons.rng.simple.RandomSource;
+import org.apache.commons.statistics.descriptive.DoubleStatistic;
+import org.apache.commons.statistics.descriptive.GeometricMean;
+import org.apache.commons.statistics.descriptive.Kurtosis;
+import org.apache.commons.statistics.descriptive.Max;
+import org.apache.commons.statistics.descriptive.Mean;
+import org.apache.commons.statistics.descriptive.Min;
+import org.apache.commons.statistics.descriptive.Product;
+import org.apache.commons.statistics.descriptive.Skewness;
+import org.apache.commons.statistics.descriptive.StandardDeviation;
+import org.apache.commons.statistics.descriptive.Statistic;
+import org.apache.commons.statistics.descriptive.Sum;
+import org.apache.commons.statistics.descriptive.SumOfLogs;
+import org.apache.commons.statistics.descriptive.SumOfSquares;
+import org.apache.commons.statistics.descriptive.Variance;
import org.openjdk.jmh.annotations.Benchmark;
import org.openjdk.jmh.annotations.BenchmarkMode;
import org.openjdk.jmh.annotations.Fork;
@@ -64,44 +79,261 @@ public class StatisticCreationPerformance {
return data;
}
+ /**
+ * @return the start inclusive of the sub-range.
+ */
+ public int from() {
+ // Approximately 1/4
+ return data.length >> 2;
+ }
+
+ /**
+ * @return the end exclusive of the sub-range.
+ */
+ public int to() {
+ // Approximately 3/4
+ return (data.length >> 1) + (data.length >> 2);
+ }
+
/**
* Create the data.
*/
@Setup(Level.Iteration)
public void setup() {
- // Data will be randomized per iteration
- data =
RandomSource.XO_RO_SHI_RO_128_PP.create().doubles(length).toArray();
+ // Data will be randomized per iteration.
+ // Ideally the product should not underflow/overflow.
+ // A product of 1 would have a sum of logs of 0.
+ // Create a uniform sum of logs around zero and transform:
+ // log x in [-0.5, 0.5) => x in [0.607, 1.649)
+ data = RandomSource.XO_RO_SHI_RO_128_PP.create().doubles(length)
+ .map(x -> Math.exp(x - 0.5)).toArray();
}
}
/**
- * A sum of {@code double} data.
+ * Source of a {@code Statistic}.
*/
- static class SimpleSum implements DoubleConsumer, DoubleSupplier {
- /** The sum. */
- private double sum;
+ @State(Scope.Benchmark)
+ public static class StatisticSource {
+ /** The statistic to create. */
+ @Param()
+ private Statistic statistic;
+
+ /** Statistic factory. */
+ private Supplier<DoubleStatistic> supplier;
+
+ /** Statistic factory using input data. */
+ private Function<double[], DoubleStatistic> factory;
+
+ /**
+ * @return a statistic instance
+ */
+ public DoubleStatistic create() {
+ return supplier.get();
+ }
- @Override
- public void accept(double value) {
- sum += value;
+ /**
+ * @param x Values.
+ * @return a statistic instance
+ */
+ public DoubleStatistic create(double[] x) {
+ return factory.apply(x);
}
- @Override
- public double getAsDouble() {
- return sum;
+ /**
+ * Create the factory functions.
+ */
+ @Setup(Level.Trial)
+ public void setup() {
+ switch (statistic) {
+ case GEOMETRIC_MEAN:
+ supplier = GeometricMean::create;
+ factory = GeometricMean::of;
+ break;
+ case KURTOSIS:
+ supplier = Kurtosis::create;
+ factory = Kurtosis::of;
+ break;
+ case MAX:
+ supplier = Max::create;
+ factory = Max::of;
+ break;
+ case MEAN:
+ supplier = Mean::create;
+ factory = Mean::of;
+ break;
+ case MIN:
+ supplier = Min::create;
+ factory = Min::of;
+ break;
+ case PRODUCT:
+ supplier = Product::create;
+ factory = Product::of;
+ break;
+ case SKEWNESS:
+ supplier = Skewness::create;
+ factory = Skewness::of;
+ break;
+ case STANDARD_DEVIATION:
+ supplier = StandardDeviation::create;
+ factory = StandardDeviation::of;
+ break;
+ case SUM:
+ supplier = Sum::create;
+ factory = Sum::of;
+ break;
+ case SUM_OF_LOGS:
+ supplier = SumOfLogs::create;
+ factory = SumOfLogs::of;
+ break;
+ case SUM_OF_SQUARES:
+ supplier = SumOfSquares::create;
+ factory = SumOfSquares::of;
+ break;
+ case VARIANCE:
+ supplier = Variance::create;
+ factory = Variance::of;
+ break;
+ default:
+ throw new IllegalStateException("Unsupported statistic: " +
statistic);
+ }
}
}
+ /**
+ * Source of a {@code Statistic} created using a custom implementation.
+ * This contains alternative version of creating statistics from an array
+ * for benchmarking performance.
+ */
+ @State(Scope.Benchmark)
+ public static class CustomStatisticSource {
+ /** The statistic to create. */
+ @Param({"min", "product"})
+ private String statistic;
+
+ /** Statistic factory using input data. */
+ private Function<double[], DoubleSupplier> factory;
+
+ /**
+ * @param x Values.
+ * @return a statistic instance
+ */
+ public DoubleSupplier create(double[] x) {
+ return factory.apply(x);
+ }
+
+ /**
+ * Create the factory functions.
+ */
+ @Setup(Level.Trial)
+ public void setup() {
+ if ("min".equals(statistic)) {
+ factory = CMin::of;
+ } else if ("product".equals(statistic)) {
+ factory = CProduct::of;
+ } else {
+ throw new IllegalStateException("Unsupported custom statistic:
" + statistic);
+ }
+ }
+
+ /** Compute the minimum. */
+ static final class CMin implements DoubleSupplier {
+ /** Current statistic. */
+ private double s;
+
+ /**
+ * Create an instance.
+ * @param s Statistic value.
+ */
+ private CMin(double s) {
+ this.s = s;
+ }
+
+ /**
+ * @param values Values.
+ * @return instance.
+ */
+ static CMin of(double... values) {
+ double s = Double.POSITIVE_INFINITY;
+ for (final double x : values) {
+ s = Math.min(s, x);
+ }
+ return new CMin(s);
+ }
+
+ @Override
+ public double getAsDouble() {
+ return s;
+ }
+ }
+
+ /** Compute the product. */
+ static final class CProduct implements DoubleSupplier {
+ /** Current statistic. */
+ private double s;
+
+ /**
+ * Create an instance.
+ * @param s Statistic value.
+ */
+ private CProduct(double s) {
+ this.s = s;
+ }
+
+ /**
+ * @param values Values.
+ * @return instance.
+ */
+ static CProduct of(double... values) {
+ double s = 1;
+ for (final double x : values) {
+ s *= x;
+ }
+ return new CProduct(s);
+ }
+
+ @Override
+ public double getAsDouble() {
+ return s;
+ }
+ }
+ }
+
+ /**
+ * Create the statistic using an array.
+ *
+ * @param dataSource Source of the data.
+ * @param statisticSource Source of the statistic.
+ * @return the statistic
+ */
+ @Benchmark
+ public double array(DataSource dataSource, StatisticSource
statisticSource) {
+ return statisticSource.create(dataSource.getData()).getAsDouble();
+ }
+
+ /**
+ * Create the statistic using an array.
+ *
+ * @param dataSource Source of the data.
+ * @param statisticSource Source of the statistic.
+ * @return the statistic
+ */
+ @Benchmark
+ public double customArray(DataSource dataSource, CustomStatisticSource
statisticSource) {
+ return statisticSource.create(dataSource.getData()).getAsDouble();
+ }
+
/**
* Create the statistic using a for loop.
*
- * @param source Source of the data.
+ * @param dataSource Source of the data.
+ * @param statisticSource Source of the statistic.
* @return the statistic
*/
@Benchmark
- public double forLoop(DataSource source) {
- final double[] data = source.getData();
- final SimpleSum s = new SimpleSum();
+ public double forLoop(DataSource dataSource, StatisticSource
statisticSource) {
+ final double[] data = dataSource.getData();
+ final DoubleStatistic s = statisticSource.create();
for (int i = 0; i < data.length; i++) {
s.accept(data[i]);
}
@@ -111,13 +343,14 @@ public class StatisticCreationPerformance {
/**
* Create the statistic using a for-each loop.
*
- * @param source Source of the data.
+ * @param dataSource Source of the data.
+ * @param statisticSource Source of the statistic.
* @return the statistic
*/
@Benchmark
- public double forEachLoop(DataSource source) {
- final double[] data = source.getData();
- final SimpleSum s = new SimpleSum();
+ public double forEachLoop(DataSource dataSource, StatisticSource
statisticSource) {
+ final double[] data = dataSource.getData();
+ final DoubleStatistic s = statisticSource.create();
for (final double x : data) {
s.accept(x);
}
@@ -127,14 +360,100 @@ public class StatisticCreationPerformance {
/**
* Create the statistic using a stream.
*
- * @param source Source of the data.
+ * @param dataSource Source of the data.
+ * @param statisticSource Source of the statistic.
* @return the statistic
*/
@Benchmark
- public double streamForEach(DataSource source) {
- final double[] data = source.getData();
- final SimpleSum s = new SimpleSum();
+ public double streamForEach(DataSource dataSource, StatisticSource
statisticSource) {
+ final double[] data = dataSource.getData();
+ final DoubleStatistic s = statisticSource.create();
Arrays.stream(data).forEach(s::accept);
return s.getAsDouble();
}
+
+ /**
+ * Create the statistic using an array.
+ *
+ * @param dataSource Source of the data.
+ * @param statisticSource Source of the statistic.
+ * @return the statistic
+ */
+ @Benchmark
+ public double arrayRange(DataSource dataSource, StatisticSource
statisticSource) {
+ final int from = dataSource.from();
+ final int to = dataSource.to();
+ final double[] data = Arrays.copyOfRange(dataSource.getData(), from,
to);
+ return statisticSource.create(data).getAsDouble();
+ }
+
+ /**
+ * Create the statistic using an array.
+ *
+ * @param dataSource Source of the data.
+ * @param statisticSource Source of the statistic.
+ * @return the statistic
+ */
+ @Benchmark
+ public double customArrayRange(DataSource dataSource,
CustomStatisticSource statisticSource) {
+ final int from = dataSource.from();
+ final int to = dataSource.to();
+ final double[] data = Arrays.copyOfRange(dataSource.getData(), from,
to);
+ return statisticSource.create(data).getAsDouble();
+ }
+
+ /**
+ * Create the statistic using a for loop on a range of the data.
+ *
+ * @param dataSource Source of the data.
+ * @param statisticSource Source of the statistic.
+ * @return the statistic
+ */
+ @Benchmark
+ public double forLoopRange(DataSource dataSource, StatisticSource
statisticSource) {
+ final int from = dataSource.from();
+ final int to = dataSource.to();
+ final double[] data = dataSource.getData();
+ final DoubleStatistic s = statisticSource.create();
+ for (int i = from; i < to; i++) {
+ s.accept(data[i]);
+ }
+ return s.getAsDouble();
+ }
+
+ /**
+ * Create the statistic using a for-each loop on a range of the data.
+ *
+ * @param dataSource Source of the data.
+ * @param statisticSource Source of the statistic.
+ * @return the statistic
+ */
+ @Benchmark
+ public double forEachLoopRange(DataSource dataSource, StatisticSource
statisticSource) {
+ final int from = dataSource.from();
+ final int to = dataSource.to();
+ final double[] data = Arrays.copyOfRange(dataSource.getData(), from,
to);
+ final DoubleStatistic s = statisticSource.create();
+ for (final double x : data) {
+ s.accept(x);
+ }
+ return s.getAsDouble();
+ }
+
+ /**
+ * Create the statistic using a stream on a range of the data.
+ *
+ * @param dataSource Source of the data.
+ * @param statisticSource Source of the statistic.
+ * @return the statistic
+ */
+ @Benchmark
+ public double streamForEachRange(DataSource dataSource, StatisticSource
statisticSource) {
+ final int from = dataSource.from();
+ final int to = dataSource.to();
+ final double[] data = dataSource.getData();
+ final DoubleStatistic s = statisticSource.create();
+ Arrays.stream(data, from, to).forEach(s::accept);
+ return s.getAsDouble();
+ }
}
