xiangfu0 commented on code in PR #20233: URL: https://github.com/apache/pinot/pull/20233#discussion_r4545099120
########## pinot-core/src/main/java/org/apache/pinot/core/query/aggregation/function/CorrelationAggregationFunction.java: ########## @@ -0,0 +1,256 @@ +/** + * 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 java.util.ArrayList; +import java.util.List; +import java.util.Map; +import javax.annotation.Nullable; +import org.apache.pinot.common.CustomObject; +import org.apache.pinot.common.request.context.ExpressionContext; +import org.apache.pinot.common.utils.DataSchema.ColumnDataType; +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.core.query.aggregation.utils.StatisticalAggregationFunctionUtils; +import org.apache.pinot.segment.local.customobject.CorrelationTuple; +import org.apache.pinot.segment.spi.AggregationFunctionType; + + +/// Aggregation function which returns the population Pearson correlation coefficient of 2 expressions. +/// CORR(exp1, exp2) = covarPop(exp1, exp2) / (stdDevPop(exp1) \* stdDevPop(exp2)) +/// +/// Pearson's correlation between two random variables X and Y is defined as +/// corr(X,Y) = cov(X,Y) / (stdDev(X) \* stdDev(Y)). +/// Using the raw sums kept in the tuple (count, sum(X), sum(Y), sum(X\*Y), sum(X^2), sum(Y^2)) this becomes +/// corr(X,Y) = (count \* sum(X\*Y) - sum(X) \* sum(Y)) +/// / sqrt((count \* sum(X^2) - sum(X)^2) \* (count \* sum(Y^2) - sum(Y)^2)). +/// @see <a href="https://en.wikipedia.org/wiki/Pearson_correlation_coefficient";>Pearson correlation coefficient</a> +/// +/// The coefficient does not depend on whether the population or sample forms of covariance and standard deviation are +/// used, because the sample sizes cancel out, so only the population form is computed here. The result is undefined +/// when fewer than one row contributes or when either column has zero variance (a constant column); in those cases the +/// function returns NULL when null handling is enabled and a sentinel otherwise. +public class CorrelationAggregationFunction extends BaseAggregationFunction<CorrelationTuple, Double> { + private static final int CORRELATION_TUPLE_TYPE = 33; + + private static final double DEFAULT_FINAL_RESULT = Double.NEGATIVE_INFINITY; + protected final ExpressionContext _expression1; + protected final ExpressionContext _expression2; + + public CorrelationAggregationFunction(List<ExpressionContext> arguments, boolean nullHandlingEnabled) { + super(nullHandlingEnabled); + _expression1 = arguments.get(0); + _expression2 = arguments.get(1); + } + + @Override + public AggregationFunctionType getType() { + return AggregationFunctionType.CORR; + } + + @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 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression1); + double[] values2 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression2); + + CorrelationTuple tuple = new CorrelationTuple(0.0, 0.0, 0.0, 0.0, 0.0, 0L); + forEachNotNull(length, blockValSetMap.get(_expression1), blockValSetMap.get(_expression2), (from, to) -> { + double sumX = 0.0; + double sumY = 0.0; + double sumXY = 0.0; + double sumX2 = 0.0; + double sumY2 = 0.0; + for (int i = from; i < to; i++) { + double x = values1[i]; + double y = values2[i]; + sumX += x; + sumY += y; + sumXY += x * y; + sumX2 += x * x; + sumY2 += y * y; + } + tuple.apply(sumX, sumY, sumXY, sumX2, sumY2, to - from); + }); + + // Leaving the holder untouched is how "nothing was aggregated" reaches extractFinalResult + if (_nullHandlingEnabled && tuple.getCount() == 0L) { + return; + } + setAggregationResult(aggregationResultHolder, tuple.getSumX(), tuple.getSumY(), tuple.getSumXY(), tuple.getSumX2(), + tuple.getSumY2(), tuple.getCount()); + } + + protected void setAggregationResult(AggregationResultHolder aggregationResultHolder, double sumX, double sumY, + double sumXY, double sumX2, double sumY2, long count) { + CorrelationTuple correlationTuple = aggregationResultHolder.getResult(); + if (correlationTuple == null) { + aggregationResultHolder.setValue(new CorrelationTuple(sumX, sumY, sumXY, sumX2, sumY2, count)); + } else { + correlationTuple.apply(sumX, sumY, sumXY, sumX2, sumY2, count); + } + } + + protected void setGroupByResult(int groupKey, GroupByResultHolder groupByResultHolder, double sumX, double sumY, + double sumXY, double sumX2, double sumY2, long count) { + CorrelationTuple correlationTuple = groupByResultHolder.getResult(groupKey); + if (correlationTuple == null) { + groupByResultHolder.setValueForKey(groupKey, new CorrelationTuple(sumX, sumY, sumXY, sumX2, sumY2, count)); + } else { + correlationTuple.apply(sumX, sumY, sumXY, sumX2, sumY2, count); + } + } + + @Override + public void aggregateGroupBySV(int length, int[] groupKeyArray, GroupByResultHolder groupByResultHolder, + Map<ExpressionContext, BlockValSet> blockValSetMap) { + double[] values1 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression1); + double[] values2 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression2); + forEachNotNull(length, blockValSetMap.get(_expression1), blockValSetMap.get(_expression2), (from, to) -> { + for (int i = from; i < to; i++) { + double x = values1[i]; + double y = values2[i]; + setGroupByResult(groupKeyArray[i], groupByResultHolder, x, y, x * y, x * x, y * y, 1L); + } + }); + } + + @Override + public void aggregateGroupByMV(int length, int[][] groupKeysArray, GroupByResultHolder groupByResultHolder, + Map<ExpressionContext, BlockValSet> blockValSetMap) { + double[] values1 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression1); + double[] values2 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression2); + forEachNotNull(length, blockValSetMap.get(_expression1), blockValSetMap.get(_expression2), (from, to) -> { + for (int i = from; i < to; i++) { + double x = values1[i]; + double y = values2[i]; + for (int groupKey : groupKeysArray[i]) { + setGroupByResult(groupKey, groupByResultHolder, x, y, x * y, x * x, y * y, 1L); + } + } + }); + } + + @Nullable + @Override + public CorrelationTuple extractAggregationResult(AggregationResultHolder aggregationResultHolder) { + return aggregationResultHolder.getResult(); + } + + @Nullable + @Override + public CorrelationTuple extractGroupByResult(GroupByResultHolder groupByResultHolder, int groupKey) { + return groupByResultHolder.getResult(groupKey); + } + + @Override + public CorrelationTuple merge(CorrelationTuple intermediateResult1, CorrelationTuple intermediateResult2) { + intermediateResult1.apply(intermediateResult2); + return intermediateResult1; + } + + @Override + public ColumnDataType getIntermediateResultColumnType() { + return ColumnDataType.OBJECT; + } + + @Override + public SerializedResult serializeIntermediateResult(CorrelationTuple correlationTuple) { + return new SerializedBytes(CORRELATION_TUPLE_TYPE, correlationTuple.toBytes()); + } + + @Override + public CorrelationTuple deserializeIntermediateResult(CustomObject customObject) { + return CorrelationTuple.fromByteBuffer(customObject.getBuffer()); + } + + @Override + public ColumnDataType getFinalResultColumnType() { + return ColumnDataType.DOUBLE; + } + + @Nullable + @Override + public Double extractFinalResult(@Nullable CorrelationTuple correlationTuple) { + // A null intermediate result means nothing was aggregated, and so does a zero count, which is what a + // deserialized peer can still carry. With null handling enabled the correlation of nothing is NULL; with it + // disabled it is what an untouched tuple renders to, which is the sentinel below. + long count = correlationTuple != null ? correlationTuple.getCount() : 0L; + if (count == 0L) { + return _nullHandlingEnabled ? null : DEFAULT_FINAL_RESULT; + } + double sumX = correlationTuple.getSumX(); + double sumY = correlationTuple.getSumY(); + double sumXY = correlationTuple.getSumXY(); + double sumX2 = correlationTuple.getSumX2(); + double sumY2 = correlationTuple.getSumY2(); + // count * variance of each column, up to the shared 1 / count^2 factor that cancels in the ratio + double varX = count * sumX2 - sumX * sumX; Review Comment: Blocking: `count * sumX2 - sumX * sumX` cancels catastrophically when values are large relative to their spread. With two independent columns of mean 1e9 and sd 10 (n=1,000), this returns 0.603 where the true value is 0.048, and NULL / `-Infinity` at n=100k. Could the tuple keep Welford/Chan co-moments `(n, meanX, meanY, m2X, m2Y, cXY)`, the way `VarianceTuple` does, so that `corr = cXY / sqrt(m2X * m2Y)`? --- AI-generated review via IntentLab · Model: Claude Opus 5.5 (claude-opus-5-5) ########## pinot-core/src/main/java/org/apache/pinot/core/query/aggregation/function/CorrelationAggregationFunction.java: ########## @@ -0,0 +1,256 @@ +/** + * 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 java.util.ArrayList; +import java.util.List; +import java.util.Map; +import javax.annotation.Nullable; +import org.apache.pinot.common.CustomObject; +import org.apache.pinot.common.request.context.ExpressionContext; +import org.apache.pinot.common.utils.DataSchema.ColumnDataType; +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.core.query.aggregation.utils.StatisticalAggregationFunctionUtils; +import org.apache.pinot.segment.local.customobject.CorrelationTuple; +import org.apache.pinot.segment.spi.AggregationFunctionType; + + +/// Aggregation function which returns the population Pearson correlation coefficient of 2 expressions. +/// CORR(exp1, exp2) = covarPop(exp1, exp2) / (stdDevPop(exp1) \* stdDevPop(exp2)) +/// +/// Pearson's correlation between two random variables X and Y is defined as +/// corr(X,Y) = cov(X,Y) / (stdDev(X) \* stdDev(Y)). +/// Using the raw sums kept in the tuple (count, sum(X), sum(Y), sum(X\*Y), sum(X^2), sum(Y^2)) this becomes +/// corr(X,Y) = (count \* sum(X\*Y) - sum(X) \* sum(Y)) +/// / sqrt((count \* sum(X^2) - sum(X)^2) \* (count \* sum(Y^2) - sum(Y)^2)). +/// @see <a href="https://en.wikipedia.org/wiki/Pearson_correlation_coefficient";>Pearson correlation coefficient</a> +/// +/// The coefficient does not depend on whether the population or sample forms of covariance and standard deviation are +/// used, because the sample sizes cancel out, so only the population form is computed here. The result is undefined +/// when fewer than one row contributes or when either column has zero variance (a constant column); in those cases the +/// function returns NULL when null handling is enabled and a sentinel otherwise. +public class CorrelationAggregationFunction extends BaseAggregationFunction<CorrelationTuple, Double> { + private static final int CORRELATION_TUPLE_TYPE = 33; + + private static final double DEFAULT_FINAL_RESULT = Double.NEGATIVE_INFINITY; + protected final ExpressionContext _expression1; + protected final ExpressionContext _expression2; + + public CorrelationAggregationFunction(List<ExpressionContext> arguments, boolean nullHandlingEnabled) { + super(nullHandlingEnabled); + _expression1 = arguments.get(0); + _expression2 = arguments.get(1); + } + + @Override + public AggregationFunctionType getType() { + return AggregationFunctionType.CORR; + } + + @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 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression1); + double[] values2 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression2); + + CorrelationTuple tuple = new CorrelationTuple(0.0, 0.0, 0.0, 0.0, 0.0, 0L); + forEachNotNull(length, blockValSetMap.get(_expression1), blockValSetMap.get(_expression2), (from, to) -> { + double sumX = 0.0; + double sumY = 0.0; + double sumXY = 0.0; + double sumX2 = 0.0; + double sumY2 = 0.0; + for (int i = from; i < to; i++) { + double x = values1[i]; + double y = values2[i]; + sumX += x; + sumY += y; + sumXY += x * y; + sumX2 += x * x; + sumY2 += y * y; + } + tuple.apply(sumX, sumY, sumXY, sumX2, sumY2, to - from); + }); + + // Leaving the holder untouched is how "nothing was aggregated" reaches extractFinalResult + if (_nullHandlingEnabled && tuple.getCount() == 0L) { + return; + } + setAggregationResult(aggregationResultHolder, tuple.getSumX(), tuple.getSumY(), tuple.getSumXY(), tuple.getSumX2(), + tuple.getSumY2(), tuple.getCount()); + } + + protected void setAggregationResult(AggregationResultHolder aggregationResultHolder, double sumX, double sumY, + double sumXY, double sumX2, double sumY2, long count) { + CorrelationTuple correlationTuple = aggregationResultHolder.getResult(); + if (correlationTuple == null) { + aggregationResultHolder.setValue(new CorrelationTuple(sumX, sumY, sumXY, sumX2, sumY2, count)); + } else { + correlationTuple.apply(sumX, sumY, sumXY, sumX2, sumY2, count); + } + } + + protected void setGroupByResult(int groupKey, GroupByResultHolder groupByResultHolder, double sumX, double sumY, + double sumXY, double sumX2, double sumY2, long count) { + CorrelationTuple correlationTuple = groupByResultHolder.getResult(groupKey); + if (correlationTuple == null) { + groupByResultHolder.setValueForKey(groupKey, new CorrelationTuple(sumX, sumY, sumXY, sumX2, sumY2, count)); + } else { + correlationTuple.apply(sumX, sumY, sumXY, sumX2, sumY2, count); + } + } + + @Override + public void aggregateGroupBySV(int length, int[] groupKeyArray, GroupByResultHolder groupByResultHolder, + Map<ExpressionContext, BlockValSet> blockValSetMap) { + double[] values1 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression1); + double[] values2 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression2); + forEachNotNull(length, blockValSetMap.get(_expression1), blockValSetMap.get(_expression2), (from, to) -> { + for (int i = from; i < to; i++) { + double x = values1[i]; + double y = values2[i]; + setGroupByResult(groupKeyArray[i], groupByResultHolder, x, y, x * y, x * x, y * y, 1L); + } + }); + } + + @Override + public void aggregateGroupByMV(int length, int[][] groupKeysArray, GroupByResultHolder groupByResultHolder, + Map<ExpressionContext, BlockValSet> blockValSetMap) { + double[] values1 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression1); + double[] values2 = StatisticalAggregationFunctionUtils.getValSet(blockValSetMap, _expression2); + forEachNotNull(length, blockValSetMap.get(_expression1), blockValSetMap.get(_expression2), (from, to) -> { + for (int i = from; i < to; i++) { + double x = values1[i]; + double y = values2[i]; + for (int groupKey : groupKeysArray[i]) { + setGroupByResult(groupKey, groupByResultHolder, x, y, x * y, x * x, y * y, 1L); + } + } + }); + } + + @Nullable + @Override + public CorrelationTuple extractAggregationResult(AggregationResultHolder aggregationResultHolder) { + return aggregationResultHolder.getResult(); + } + + @Nullable + @Override + public CorrelationTuple extractGroupByResult(GroupByResultHolder groupByResultHolder, int groupKey) { + return groupByResultHolder.getResult(groupKey); + } + + @Override + public CorrelationTuple merge(CorrelationTuple intermediateResult1, CorrelationTuple intermediateResult2) { + intermediateResult1.apply(intermediateResult2); + return intermediateResult1; + } + + @Override + public ColumnDataType getIntermediateResultColumnType() { + return ColumnDataType.OBJECT; + } + + @Override + public SerializedResult serializeIntermediateResult(CorrelationTuple correlationTuple) { + return new SerializedBytes(CORRELATION_TUPLE_TYPE, correlationTuple.toBytes()); + } + + @Override + public CorrelationTuple deserializeIntermediateResult(CustomObject customObject) { + return CorrelationTuple.fromByteBuffer(customObject.getBuffer()); + } + + @Override + public ColumnDataType getFinalResultColumnType() { + return ColumnDataType.DOUBLE; + } + + @Nullable + @Override + public Double extractFinalResult(@Nullable CorrelationTuple correlationTuple) { + // A null intermediate result means nothing was aggregated, and so does a zero count, which is what a + // deserialized peer can still carry. With null handling enabled the correlation of nothing is NULL; with it + // disabled it is what an untouched tuple renders to, which is the sentinel below. + long count = correlationTuple != null ? correlationTuple.getCount() : 0L; + if (count == 0L) { + return _nullHandlingEnabled ? null : DEFAULT_FINAL_RESULT; + } + double sumX = correlationTuple.getSumX(); + double sumY = correlationTuple.getSumY(); + double sumXY = correlationTuple.getSumXY(); + double sumX2 = correlationTuple.getSumX2(); + double sumY2 = correlationTuple.getSumY2(); + // count * variance of each column, up to the shared 1 / count^2 factor that cancels in the ratio + double varX = count * sumX2 - sumX * sumX; + double varY = count * sumY2 - sumY * sumY; + // Correlation is undefined when either column never varies; floating point can also drive a non-negative variance + // slightly below zero, which would turn the square root into NaN, so both are treated as undefined here + if (varX <= 0.0 || varY <= 0.0) { + return _nullHandlingEnabled ? null : DEFAULT_FINAL_RESULT; + } + double cov = count * sumXY - sumX * sumY; + double correlation = cov / Math.sqrt(varX * varY); + // Clamp rounding error that can push the ratio a hair outside the mathematical [-1, 1] range + return Math.max(-1.0, Math.min(1.0, correlation)); Review Comment: This clamp (together with the `<= 0` guard above) turns large cancellation errors into plausible-looking results. A dataset with true r = 0.9963 shifted by +1e9 comes back as exactly `1.0`. Once the accumulation is numerically stable, the clamp is only needed for ulp-level noise. --- AI-generated review via IntentLab · Model: Claude Opus 5.5 (claude-opus-5-5) ########## pinot-core/src/main/java/org/apache/pinot/core/query/aggregation/function/CorrelationAggregationFunction.java: ########## @@ -0,0 +1,256 @@ +/** + * 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 java.util.ArrayList; +import java.util.List; +import java.util.Map; +import javax.annotation.Nullable; +import org.apache.pinot.common.CustomObject; +import org.apache.pinot.common.request.context.ExpressionContext; +import org.apache.pinot.common.utils.DataSchema.ColumnDataType; +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.core.query.aggregation.utils.StatisticalAggregationFunctionUtils; +import org.apache.pinot.segment.local.customobject.CorrelationTuple; +import org.apache.pinot.segment.spi.AggregationFunctionType; + + +/// Aggregation function which returns the population Pearson correlation coefficient of 2 expressions. +/// CORR(exp1, exp2) = covarPop(exp1, exp2) / (stdDevPop(exp1) \* stdDevPop(exp2)) +/// +/// Pearson's correlation between two random variables X and Y is defined as +/// corr(X,Y) = cov(X,Y) / (stdDev(X) \* stdDev(Y)). +/// Using the raw sums kept in the tuple (count, sum(X), sum(Y), sum(X\*Y), sum(X^2), sum(Y^2)) this becomes +/// corr(X,Y) = (count \* sum(X\*Y) - sum(X) \* sum(Y)) +/// / sqrt((count \* sum(X^2) - sum(X)^2) \* (count \* sum(Y^2) - sum(Y)^2)). +/// @see <a href="https://en.wikipedia.org/wiki/Pearson_correlation_coefficient";>Pearson correlation coefficient</a> +/// +/// The coefficient does not depend on whether the population or sample forms of covariance and standard deviation are +/// used, because the sample sizes cancel out, so only the population form is computed here. The result is undefined +/// when fewer than one row contributes or when either column has zero variance (a constant column); in those cases the +/// function returns NULL when null handling is enabled and a sentinel otherwise. +public class CorrelationAggregationFunction extends BaseAggregationFunction<CorrelationTuple, Double> { + private static final int CORRELATION_TUPLE_TYPE = 33; Review Comment: `33` is already `VARIANCE_TUPLE_TYPE`. `ObjectSerDeUtils` says new intermediate types take a value above every existing one and never reuse one, so this should be `58`. No runtime break today (dispatch is per function), but the value is frozen once released. Please also add a CORR row to `IntermediateResultByteBufferSerDeTest`. --- AI-generated review via IntentLab · Model: Claude Opus 5.5 (claude-opus-5-5) -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
