walterddr commented on code in PR #9236: URL: https://github.com/apache/pinot/pull/9236#discussion_r954314178
########## pinot-core/src/main/java/org/apache/pinot/core/query/aggregation/function/CovarianceAggregationFunction.java: ########## @@ -0,0 +1,237 @@ +/** + * 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.pinot.core.query.aggregation.function; + +import com.google.common.base.Preconditions; +import java.util.ArrayList; +import java.util.List; +import java.util.Map; +import org.apache.pinot.common.request.context.ExpressionContext; +import org.apache.pinot.common.utils.DataSchema; +import org.apache.pinot.core.common.BlockValSet; +import org.apache.pinot.core.query.aggregation.AggregationResultHolder; +import org.apache.pinot.core.query.aggregation.ObjectAggregationResultHolder; +import org.apache.pinot.core.query.aggregation.groupby.GroupByResultHolder; +import org.apache.pinot.core.query.aggregation.groupby.ObjectGroupByResultHolder; +import org.apache.pinot.segment.local.customobject.CovarianceTuple; +import org.apache.pinot.segment.spi.AggregationFunctionType; + + +/** + * Aggregation function which returns the population covariance of 2 expressions. + * COVAR_POP(exp1, exp2) = mean(exp1 * exp2) - mean(exp1) * mean(exp2) + * COVAR_SAMP(exp1, exp2) = (sum(exp1 * exp2) - sum(exp1) * sum(exp2)) / (count - 1) + * + * Population covariance between two random variables X and Y is defined as either + * covarPop(X,Y) = E[(X - E[X]) * (Y - E[Y])] or + * covarPop(X,Y) = E[X*Y] - E[X] * E[Y], + * here E[X] represents mean of X + * @see <a href="https://en.wikipedia.org/wiki/Covariance";>Covariance</a> + * The calculations here are based on the second definition shown above. + * Sample covariance = covarPop(X, Y) * besselCorrection + * @see <a href="https://en.wikipedia.org/wiki/Bessel%27s_correction";>Bessel's correction</a> + */ +public class CovarianceAggregationFunction implements AggregationFunction<CovarianceTuple, Double> { + private static final double DEFAULT_FINAL_RESULT = Double.NEGATIVE_INFINITY; + protected final ExpressionContext _expression1; + protected final ExpressionContext _expression2; + protected final boolean _isSample; + + public CovarianceAggregationFunction(List<ExpressionContext> arguments, boolean isSample) { + _expression1 = arguments.get(0); + _expression2 = arguments.get(1); + _isSample = isSample; + } + + @Override + public AggregationFunctionType getType() { + if (_isSample) { + return AggregationFunctionType.COVARSAMP; + } + return AggregationFunctionType.COVARPOP; + } + + @Override + public String getColumnName() { + return getType().getName() + "_" + _expression1 + "_" + _expression2; + } + + @Override + public String getResultColumnName() { + return getType().getName().toLowerCase() + "(" + _expression1 + "," + _expression2 + ")"; + } + + @Override + public List<ExpressionContext> getInputExpressions() { + ArrayList<ExpressionContext> inputExpressions = new ArrayList<>(); + inputExpressions.add(_expression1); + inputExpressions.add(_expression2); + return inputExpressions; + } + + @Override + public AggregationResultHolder createAggregationResultHolder() { + return new ObjectAggregationResultHolder(); + } + + @Override + public GroupByResultHolder createGroupByResultHolder(int initialCapacity, int maxCapacity) { + return new ObjectGroupByResultHolder(initialCapacity, maxCapacity); + } + + @Override + public void aggregate(int length, AggregationResultHolder aggregationResultHolder, + Map<ExpressionContext, BlockValSet> blockValSetMap) { + double[] values1 = getValSet(blockValSetMap, _expression1); + double[] values2 = getValSet(blockValSetMap, _expression2); + + double sumX = 0.0; + double sumY = 0.0; + double sumXY = 0.0; + + for (int i = 0; i < length; i++) { + sumX += values1[i]; + sumY += values2[i]; + sumXY += values1[i] * values2[i]; + } + setAggregationResult(aggregationResultHolder, sumX, sumY, sumXY, length); + } + + protected void setAggregationResult(AggregationResultHolder aggregationResultHolder, double sumX, double sumY, + double sumXY, long count) { + CovarianceTuple covarianceTuple = aggregationResultHolder.getResult(); + if (covarianceTuple == null) { + aggregationResultHolder.setValue(new CovarianceTuple(sumX, sumY, sumXY, count)); + } else { + covarianceTuple.apply(sumX, sumY, sumXY, count); + } + } + + protected void setGroupByResult(int groupKey, GroupByResultHolder groupByResultHolder, double sumX, double sumY, + double sumXY, long count) { + CovarianceTuple covarianceTuple = groupByResultHolder.getResult(groupKey); + if (covarianceTuple == null) { + groupByResultHolder.setValueForKey(groupKey, new CovarianceTuple(sumX, sumY, sumXY, count)); + } else { + covarianceTuple.apply(sumX, sumY, sumXY, count); + } + } + + private double[] getValSet(Map<ExpressionContext, BlockValSet> blockValSetMap, ExpressionContext expression) { + BlockValSet blockValSet = blockValSetMap.get(expression); + //TODO: Add MV support for covariance + Preconditions.checkState(blockValSet.isSingleValue(), + "Covariance function currently only supports single-valued column"); + switch (blockValSet.getValueType().getStoredType()) { + case INT: + case LONG: + case FLOAT: + case DOUBLE: + return blockValSet.getDoubleValuesSV(); + default: + throw new IllegalStateException( + "Cannot compute covariance for non-numeric type: " + blockValSet.getValueType()); + } + } + + @Override + public void aggregateGroupBySV(int length, int[] groupKeyArray, GroupByResultHolder groupByResultHolder, + Map<ExpressionContext, BlockValSet> blockValSetMap) { + double[] values1 = getValSet(blockValSetMap, _expression1); + double[] values2 = getValSet(blockValSetMap, _expression2); + for (int i = 0; i < length; i++) { + setGroupByResult(groupKeyArray[i], groupByResultHolder, values1[i], values2[i], values1[i] * values2[i], 1L); + } + } + + @Override + public void aggregateGroupByMV(int length, int[][] groupKeysArray, GroupByResultHolder groupByResultHolder, + Map<ExpressionContext, BlockValSet> blockValSetMap) { + double[] values1 = getValSet(blockValSetMap, _expression1); + double[] values2 = getValSet(blockValSetMap, _expression2); + for (int i = 0; i < length; i++) { + for (int groupKey : groupKeysArray[i]) { + setGroupByResult(groupKey, groupByResultHolder, values1[i], values2[i], values1[i] * values2[i], 1L); + } + } + } + + @Override + public CovarianceTuple extractAggregationResult(AggregationResultHolder aggregationResultHolder) { + CovarianceTuple covarianceTuple = aggregationResultHolder.getResult(); + if (covarianceTuple == null) { + return new CovarianceTuple(0.0, 0.0, 0.0, 0L); + } else { + return covarianceTuple; + } + } + + @Override + public CovarianceTuple extractGroupByResult(GroupByResultHolder groupByResultHolder, int groupKey) { + return groupByResultHolder.getResult(groupKey); + } + + @Override + public CovarianceTuple merge(CovarianceTuple intermediateResult1, CovarianceTuple intermediateResult2) { + intermediateResult1.apply(intermediateResult2); + return intermediateResult1; + } + + @Override + public DataSchema.ColumnDataType getIntermediateResultColumnType() { + return DataSchema.ColumnDataType.OBJECT; + } + + @Override + public DataSchema.ColumnDataType getFinalResultColumnType() { + return DataSchema.ColumnDataType.DOUBLE; + } + + @Override + public Double extractFinalResult(CovarianceTuple covarianceTuple) { + long count = covarianceTuple.getCount(); + if (count == 0L) { + return DEFAULT_FINAL_RESULT; + } else { + double sumX = covarianceTuple.getSumX(); + double sumY = covarianceTuple.getSumY(); + double sumXY = covarianceTuple.getSumXY(); + double popCov = (sumXY / count) - (sumX / count) * (sumY / count); + double besselCorrection = count / (count - 1); + if (_isSample) { + return popCov * besselCorrection; + } Review Comment: ```suggestion if (_isSample) { double besselCorrection = count / (count - 1); return popCov * besselCorrection; } ``` ########## pinot-segment-local/src/main/java/org/apache/pinot/segment/local/customobject/CovarianceTuple.java: ########## @@ -0,0 +1,121 @@ +/** + * 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.pinot.segment.local.customobject; + +import java.nio.ByteBuffer; +import javax.annotation.Nonnull; + + +/** + * Intermediate state used by CovarianceAggregationFunction which helps calculate + * population covariance and sample covariance + */ +public class CovarianceTuple implements Comparable<CovarianceTuple> { + + private double _sumX; + private double _sumY; + private double _sumXY; + private long _count; + + public CovarianceTuple(double sumX, double sumY, double sumXY, long count) { + _sumX = sumX; + _sumY = sumY; + _sumXY = sumXY; + _count = count; + } + + public void apply(double sumX, double sumY, double sumXY, long count) { + _sumX += sumX; + _sumY += sumY; + _sumXY += sumXY; + _count += count; + } + + public void apply(@Nonnull CovarianceTuple covarianceTuple) { + _sumX += covarianceTuple._sumX; + _sumY += covarianceTuple._sumY; + _sumXY += covarianceTuple._sumXY; + _count += covarianceTuple._count; + } + + public double getSumX() { + return _sumX; + } + + public double getSumY() { + return _sumY; + } + + public double getSumXY() { + return _sumXY; + } + + public long getCount() { + return _count; + } + + @Nonnull + public byte[] toBytes() { + ByteBuffer byteBuffer = ByteBuffer.allocate(Double.BYTES + Double.BYTES + Double.BYTES + Long.BYTES); + byteBuffer.putDouble(_sumX); + byteBuffer.putDouble(_sumY); + byteBuffer.putDouble(_sumXY); + byteBuffer.putLong(_count); + return byteBuffer.array(); + } + + @Nonnull + public static CovarianceTuple fromBytes(byte[] bytes) { + return fromByteBuffer(ByteBuffer.wrap(bytes)); + } + + @Nonnull + public static CovarianceTuple fromByteBuffer(ByteBuffer byteBuffer) { + return new CovarianceTuple(byteBuffer.getDouble(), byteBuffer.getDouble(), byteBuffer.getDouble(), + byteBuffer.getLong()); + } + + @Override + public int compareTo(@Nonnull CovarianceTuple covarianceTuple) { Review Comment: compareTo should return the diff value, not just 0 and 1/-1. although I am not sure how to convert the double into int, is there any typical comparison written in other SQL system? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@pinot.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@pinot.apache.org For additional commands, e-mail: commits-h...@pinot.apache.org
