Mwolters/pred enhancements (#2159)

* adding new binding for single array int types

* new binding for var length int arrays

* new binding for var length int arrays cleanup

* add embedding to convert int array to string

* add script for hdf5 generation

* list version of var length int embedding

* updating target byte code for groovy scripts

* adding pytorch version

* removing datagen files to put them in tools repo

* make IntArrayToString threadsafe
This commit is contained in:
Mark Wolters 2025-02-08 07:05:49 -04:00 committed by GitHub
parent cacd39231f
commit d617796f99
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
10 changed files with 257 additions and 4 deletions

View File

@ -36,10 +36,13 @@ public abstract class AbstractHdfFileToVectorType {
long[] sliceOffset = new long[dims.length];
sliceOffset[0] = (l % dims[0]);
int[] sliceDimensions = new int[dims.length];
// We always want to read a single vector
// We always want to read a single value
sliceDimensions[0] = 1;
// Number of elements in the vector
sliceDimensions[1] = dims[1];
return dataset.getData(sliceOffset, sliceDimensions);
if (dims.length > 1) {
sliceDimensions[1] = dims[1];
return dataset.getData(sliceOffset, sliceDimensions);
} else {
return dataset.getData();
}
}
}

View File

@ -0,0 +1,62 @@
/*
* Copyright (c) nosqlbench
*
* 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.
*/
package io.nosqlbench.virtdata.library.hdf5.from_long.to_array;
import io.nosqlbench.virtdata.api.annotations.Categories;
import io.nosqlbench.virtdata.api.annotations.Category;
import io.nosqlbench.virtdata.api.annotations.ThreadSafeMapper;
import io.nosqlbench.virtdata.library.hdf5.from_long.AbstractHdfFileToVectorType;
import java.util.function.LongFunction;
/**
* This function reads a vector dataset from an HDF5 file. The dataset itself is not
* read into memory, only the metadata (the "dataset" Java Object). The lambda function
* reads a single vector from the dataset, based on the long input value. As currently
* written this class will only work for datasets with 2 dimensions where the 1st dimension
* specifies the number of vectors and the 2nd dimension specifies the number of elements in
* each vector. Only datatypes short, int, and float are supported at this time.
* <p>
* This implementation is specific to returning an array of ints from a dataset that contains
* variable length arrays of ints.
*/
@ThreadSafeMapper
@Categories(Category.experimental)
public class HdfFileToVarLengthIntArray extends AbstractHdfFileToVectorType implements LongFunction<int[]> {
public HdfFileToVarLengthIntArray(String filename, String datasetName) {
super(filename, datasetName);
}
@Override
public int[] apply(long l) {
Object data = getDataFrom(l);
return extractIds(data, l);
}
/**
* @param o The dataset from hdf5 file. In this case because the number of ints in each array is variable length it
* is returned by hdf5 as a one dimensional array of Objects
* @param l The current cycle
* @return The int array at the offset of l
*/
private int[] extractIds(Object o, long l) {
Object[] objects = (Object[]) o;
Object o1 = objects[(int) l % objects.length];
return (int[]) o1;
}
}

View File

@ -0,0 +1,53 @@
/*
* Copyright (c) nosqlbench
*
* 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.
*/
package io.nosqlbench.virtdata.library.hdf5.from_long.to_int;
import io.nosqlbench.virtdata.api.annotations.Categories;
import io.nosqlbench.virtdata.api.annotations.Category;
import io.nosqlbench.virtdata.api.annotations.ThreadSafeMapper;
import io.nosqlbench.virtdata.library.hdf5.from_long.AbstractHdfFileToVectorType;
import io.nosqlbench.virtdata.library.hdf5.helpers.EmbeddingGenerator;
import io.nosqlbench.virtdata.library.hdf5.helpers.EmbeddingGeneratorFactory;
import java.util.function.LongFunction;
/**
* This function reads a vector dataset from an HDF5 file. The dataset itself is not
* read into memory, only the metadata (the "dataset" Java Object). The lambda function
* reads a single vector from the dataset, based on the long input value. As currently
* written this class will only work for datasets with 2 dimensions where the 1st dimension
* specifies the number of vectors and the 2nd dimension specifies the number of elements in
* each vector. Only datatypes short, int, and float are supported at this time.
* <p>
* This implementation is specific to returning a single int
*/
@ThreadSafeMapper
@Categories(Category.experimental)
public class HdfFileToInt extends AbstractHdfFileToVectorType implements LongFunction<Integer> {
private final EmbeddingGenerator embeddingGenerator;
public HdfFileToInt(String filename, String datasetName) {
super(filename, datasetName);
embeddingGenerator = EmbeddingGeneratorFactory.getGenerator(dataset.getJavaType().getSimpleName().toLowerCase());
}
@Override
public Integer apply(long l) {
Object data = getDataFrom(l);
return embeddingGenerator.generateIntFrom(data, l);
}
}

View File

@ -0,0 +1,67 @@
/*
* Copyright (c) nosqlbench
*
* 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.
*/
package io.nosqlbench.virtdata.library.hdf5.from_long.to_list;
import io.nosqlbench.virtdata.api.annotations.Categories;
import io.nosqlbench.virtdata.api.annotations.Category;
import io.nosqlbench.virtdata.api.annotations.ThreadSafeMapper;
import io.nosqlbench.virtdata.library.hdf5.from_long.AbstractHdfFileToVectorType;
import java.util.List;
import java.util.function.LongFunction;
/**
* This function reads a vector dataset from an HDF5 file. The dataset itself is not
* read into memory, only the metadata (the "dataset" Java Object). The lambda function
* reads a single vector from the dataset, based on the long input value. As currently
* written this class will only work for datasets with 2 dimensions where the 1st dimension
* specifies the number of vectors and the 2nd dimension specifies the number of elements in
* each vector. Only datatypes short, int, and float are supported at this time.
* <p>
* This implementation is specific to returning a List of type Integer from a dataset that contains
* variable length arrays of ints.
*/
@ThreadSafeMapper
@Categories(Category.experimental)
public class HdfFileToVarLengthIntList extends AbstractHdfFileToVectorType implements LongFunction<List<Integer>> {
public HdfFileToVarLengthIntList(String filename, String datasetName) {
super(filename, datasetName);
}
@Override
public List<Integer> apply(long l) {
Object data = getDataFrom(l);
return extractIds(data, l);
}
/**
* @param o The dataset from hdf5 file. In this case because the number of ints in each array is variable length it
* is returned by hdf5 as a one dimensional array of Objects
* @param l The current cycle
* @return A List of the integers at the offset of l
*/
private List<Integer> extractIds(Object o, long l) {
Object[] objects = (Object[]) o;
int[] ints = (int[]) objects[(int) l % objects.length];
Integer[] integers = new Integer[ints.length];
for (int i = 0; i < ints.length; i++) {
integers[i] = ints[i];
}
return List.of(integers);
}
}

View File

@ -0,0 +1,42 @@
/*
* Copyright (c) nosqlbench
*
* 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.
*/
package io.nosqlbench.virtdata.library.hdf5.from_long.to_string;
import io.nosqlbench.virtdata.api.annotations.Categories;
import io.nosqlbench.virtdata.api.annotations.Category;
import io.nosqlbench.virtdata.api.annotations.ThreadSafeMapper;
import java.util.function.Function;
@ThreadSafeMapper
@Categories(Category.experimental)
public class IntArrayToString implements Function<int[],String> {
private final ThreadLocal<StringBuilder> threadSb = ThreadLocal.withInitial(StringBuilder::new);
@Override
public String apply(int[] ints) {
StringBuilder sb = threadSb.get();
sb.setLength(0);
for (int i : ints) {
sb.append(i);
if (i != ints[ints.length - 1]) {
sb.append(",");
}
}
return sb.toString();
}
}

View File

@ -81,4 +81,10 @@ public class DoubleEmbeddingGenerator implements EmbeddingGenerator {
return vector2;
}
@Override
public int generateIntFrom(Object o, long l) {
double[] source = (double[]) o;
return (int) source[(int) (l % source.length)];
}
}

View File

@ -30,4 +30,6 @@ public interface EmbeddingGenerator {
List<Integer> generateIntListEmbeddingFrom(Object data, int[] dims);
int[] generateIntArrayEmbeddingFrom(Object data, int[] dims);
int generateIntFrom(Object data, long l);
}

View File

@ -76,4 +76,10 @@ public class FloatEmbeddingGenerator implements EmbeddingGenerator {
return vector2;
}
@Override
public int generateIntFrom(Object o, long l) {
float[] source = (float[]) o;
return (int) source[(int) (l % source.length)];
}
}

View File

@ -74,4 +74,10 @@ public class IntEmbeddingGenerator implements EmbeddingGenerator {
public int[] generateIntArrayEmbeddingFrom(Object o, int[] dims) {
return ((int[][]) o)[0];
}
@Override
public int generateIntFrom(Object o, long l) {
int[] source = (int[]) o;
return source[(int) (l % source.length)];
}
}

View File

@ -75,4 +75,10 @@ public class LongEmbeddingGenerator implements EmbeddingGenerator {
}
return vector2;
}
@Override
public int generateIntFrom(Object o, long l) {
long[] source = (long[]) o;
return Math.toIntExact(source[(int) (l % source.length)]);
}
}