/**
* 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.hadoop.hive.ql.udf.generic;
import java.util.ArrayList;
import java.util.HashSet;
import java.util.List;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.apache.hadoop.hive.common.type.HiveDecimal;
import org.apache.hadoop.hive.ql.exec.Description;
import org.apache.hadoop.hive.ql.exec.PTFPartition;
import org.apache.hadoop.hive.ql.exec.UDFArgumentTypeException;
import org.apache.hadoop.hive.ql.metadata.HiveException;
import org.apache.hadoop.hive.ql.parse.SemanticException;
import org.apache.hadoop.hive.ql.plan.ptf.PTFExpressionDef;
import org.apache.hadoop.hive.ql.plan.ptf.WindowFrameDef;
import org.apache.hadoop.hive.ql.udf.generic.GenericUDAFEvaluator.AbstractAggregationBuffer;
import org.apache.hadoop.hive.ql.udf.generic.GenericUDAFEvaluator.AggregationType;
import org.apache.hadoop.hive.ql.udf.ptf.BasePartitionEvaluator;
import org.apache.hadoop.hive.ql.util.JavaDataModel;
import org.apache.hadoop.hive.serde2.io.DoubleWritable;
import org.apache.hadoop.hive.serde2.io.HiveDecimalWritable;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorFactory;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorUtils;
import org.apache.hadoop.hive.serde2.objectinspector.PrimitiveObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.StructField;
import org.apache.hadoop.hive.serde2.objectinspector.StructObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorUtils.ObjectInspectorCopyOption;
import org.apache.hadoop.hive.serde2.objectinspector.ObjectInspectorUtils.ObjectInspectorObject;
import org.apache.hadoop.hive.serde2.objectinspector.primitive.DoubleObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.primitive.HiveDecimalObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.primitive.LongObjectInspector;
import org.apache.hadoop.hive.serde2.objectinspector.primitive.PrimitiveObjectInspectorFactory;
import org.apache.hadoop.hive.serde2.objectinspector.primitive.PrimitiveObjectInspectorUtils;
import org.apache.hadoop.hive.serde2.typeinfo.DecimalTypeInfo;
import org.apache.hadoop.hive.serde2.typeinfo.PrimitiveTypeInfo;
import org.apache.hadoop.hive.serde2.typeinfo.TypeInfo;
import org.apache.hadoop.hive.serde2.typeinfo.TypeInfoFactory;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.util.StringUtils;
/**
* GenericUDAFAverage.
*
*/
@Description(name = "avg", value = "_FUNC_(x) - Returns the mean of a set of numbers")
public class GenericUDAFAverage extends AbstractGenericUDAFResolver {
static final Logger LOG = LoggerFactory.getLogger(GenericUDAFAverage.class.getName());
@Override
public GenericUDAFEvaluator getEvaluator(TypeInfo[] parameters)
throws SemanticException {
if (parameters.length != 1) {
throw new UDFArgumentTypeException(parameters.length - 1,
"Exactly one argument is expected.");
}
if (parameters[0].getCategory() != ObjectInspector.Category.PRIMITIVE) {
throw new UDFArgumentTypeException(0,
"Only primitive type arguments are accepted but "
+ parameters[0].getTypeName() + " is passed.");
}
switch (((PrimitiveTypeInfo) parameters[0]).getPrimitiveCategory()) {
case BYTE:
case SHORT:
case INT:
case LONG:
case FLOAT:
case DOUBLE:
case STRING:
case VARCHAR:
case CHAR:
case TIMESTAMP:
return new GenericUDAFAverageEvaluatorDouble();
case DECIMAL:
return new GenericUDAFAverageEvaluatorDecimal();
case BOOLEAN:
case DATE:
default:
throw new UDFArgumentTypeException(0,
"Only numeric or string type arguments are accepted but "
+ parameters[0].getTypeName() + " is passed.");
}
}
@Override
public GenericUDAFEvaluator getEvaluator(GenericUDAFParameterInfo paramInfo)
throws SemanticException {
if (paramInfo.isAllColumns()) {
throw new SemanticException(
"The specified syntax for UDAF invocation is invalid.");
}
AbstractGenericUDAFAverageEvaluator eval =
(AbstractGenericUDAFAverageEvaluator) getEvaluator(paramInfo.getParameters());
eval.avgDistinct = paramInfo.isDistinct();
eval.isWindowing = paramInfo.isWindowing();
return eval;
}
public static class GenericUDAFAverageEvaluatorDouble extends AbstractGenericUDAFAverageEvaluator<Double> {
@Override
public void doReset(AverageAggregationBuffer<Double> aggregation) throws HiveException {
aggregation.count = 0;
aggregation.sum = new Double(0);
aggregation.uniqueObjects = new HashSet<ObjectInspectorObject>();
}
@Override
protected ObjectInspector getSumFieldJavaObjectInspector() {
return PrimitiveObjectInspectorFactory.javaDoubleObjectInspector;
}
@Override
protected ObjectInspector getSumFieldWritableObjectInspector() {
return PrimitiveObjectInspectorFactory.writableDoubleObjectInspector;
}
@Override
protected void doIterate(AverageAggregationBuffer<Double> aggregation,
PrimitiveObjectInspector oi, Object parameter) {
double value = PrimitiveObjectInspectorUtils.getDouble(parameter, oi);
aggregation.count++;
aggregation.sum += value;
}
@Override
protected void doMerge(AverageAggregationBuffer<Double> aggregation, Long partialCount,
ObjectInspector sumFieldOI, Object partialSum) {
double value = ((DoubleObjectInspector)sumFieldOI).get(partialSum);
aggregation.count += partialCount;
aggregation.sum += value;
}
@Override
protected void doMergeAdd(Double sum,
ObjectInspectorObject obj) {
sum += PrimitiveObjectInspectorUtils.getDouble(obj.getValues()[0], copiedOI);
}
@Override
protected void doTerminatePartial(AverageAggregationBuffer<Double> aggregation) {
if(partialResult[1] == null) {
partialResult[1] = new DoubleWritable(0);
}
((LongWritable) partialResult[0]).set(aggregation.count);
((DoubleWritable) partialResult[1]).set(aggregation.sum);
}
@Override
protected Object doTerminate(AverageAggregationBuffer<Double> aggregation) {
if(aggregation.count == 0) {
return null;
} else {
DoubleWritable result = new DoubleWritable(0);
result.set(aggregation.sum / aggregation.count);
return result;
}
}
@Override
public AggregationBuffer getNewAggregationBuffer() throws HiveException {
AverageAggregationBuffer<Double> result = new AverageAggregationBuffer<Double>();
reset(result);
return result;
}
@Override
public GenericUDAFEvaluator getWindowingEvaluator(WindowFrameDef wFrameDef) {
// Don't use streaming for distinct cases
if (isWindowingDistinct()) {
return null;
}
return new GenericUDAFStreamingEvaluator.SumAvgEnhancer<DoubleWritable, Object[]>(this, wFrameDef) {
@Override
protected DoubleWritable getNextResult(
org.apache.hadoop.hive.ql.udf.generic.GenericUDAFStreamingEvaluator.SumAvgEnhancer<DoubleWritable, Object[]>.SumAvgStreamingState ss)
throws HiveException {
AverageAggregationBuffer<Double> myagg = (AverageAggregationBuffer<Double>) ss.wrappedBuf;
Double r = myagg.count == 0 ? null : myagg.sum;
long cnt = myagg.count;
Object[] o = ss.retrieveNextIntermediateValue();
if (o != null) {
Double d = (Double) o[0];
r = r == null ? null : r - d;
cnt = cnt - ((Long) o[1]);
}
return r == null ? null : new DoubleWritable(r / cnt);
}
@Override
protected Object[] getCurrentIntermediateResult(
org.apache.hadoop.hive.ql.udf.generic.GenericUDAFStreamingEvaluator.SumAvgEnhancer<DoubleWritable, Object[]>.SumAvgStreamingState ss)
throws HiveException {
AverageAggregationBuffer<Double> myagg = (AverageAggregationBuffer<Double>) ss.wrappedBuf;
return myagg.count == 0 ? null : new Object[] {
new Double(myagg.sum), myagg.count };
}
};
}
@Override
protected BasePartitionEvaluator createPartitionEvaluator(
WindowFrameDef winFrame,
PTFPartition partition,
List<PTFExpressionDef> parameters,
ObjectInspector outputOI) {
try {
return new BasePartitionEvaluator.AvgPartitionDoubleEvaluator(this, winFrame, partition, parameters, inputOI, outputOI);
} catch(HiveException e) {
return super.createPartitionEvaluator(winFrame, partition, parameters, outputOI);
}
}
}
public static class GenericUDAFAverageEvaluatorDecimal extends AbstractGenericUDAFAverageEvaluator<HiveDecimal> {
@Override
public void doReset(AverageAggregationBuffer<HiveDecimal> aggregation) throws HiveException {
aggregation.count = 0;
aggregation.sum = HiveDecimal.ZERO;
aggregation.uniqueObjects = new HashSet<ObjectInspectorObject>();
}
@Override
protected ObjectInspector getSumFieldJavaObjectInspector() {
DecimalTypeInfo typeInfo = deriveResultDecimalTypeInfo();
return PrimitiveObjectInspectorFactory.getPrimitiveJavaObjectInspector(typeInfo);
}
@Override
protected ObjectInspector getSumFieldWritableObjectInspector() {
DecimalTypeInfo typeInfo = deriveResultDecimalTypeInfo();
return PrimitiveObjectInspectorFactory.getPrimitiveWritableObjectInspector(typeInfo);
}
private DecimalTypeInfo deriveResultDecimalTypeInfo() {
int prec = inputOI.precision();
int scale = inputOI.scale();
if (mode == Mode.FINAL || mode == Mode.COMPLETE) {
int intPart = prec - scale;
// The avg() result type has the same number of integer digits and 4 more decimal digits.
scale = Math.min(scale + 4, HiveDecimal.MAX_SCALE - intPart);
return TypeInfoFactory.getDecimalTypeInfo(intPart + scale, scale);
} else {
// For intermediate sum field
return GenericUDAFAverage.deriveSumFieldTypeInfo(prec, scale);
}
}
@Override
protected void doIterate(AverageAggregationBuffer<HiveDecimal> aggregation,
PrimitiveObjectInspector oi, Object parameter) {
HiveDecimal value = PrimitiveObjectInspectorUtils.getHiveDecimal(parameter, oi);
aggregation.count++;
if (aggregation.sum != null) {
aggregation.sum = aggregation.sum.add(value);
}
}
@Override
protected void doMerge(AverageAggregationBuffer<HiveDecimal> aggregation, Long partialCount,
ObjectInspector sumFieldOI, Object partialSum) {
HiveDecimal value = ((HiveDecimalObjectInspector)sumFieldOI).getPrimitiveJavaObject(partialSum);
if (value == null) {
aggregation.sum = null;
}
aggregation.count += partialCount;
if (aggregation.sum != null) {
aggregation.sum = aggregation.sum.add(value);
}
}
@Override
protected void doMergeAdd(
HiveDecimal sum,
ObjectInspectorObject obj) {
sum.add(PrimitiveObjectInspectorUtils.getHiveDecimal(obj.getValues()[0], copiedOI));
}
@Override
protected void doTerminatePartial(AverageAggregationBuffer<HiveDecimal> aggregation) {
if(partialResult[1] == null && aggregation.sum != null) {
partialResult[1] = new HiveDecimalWritable(HiveDecimal.ZERO);
}
((LongWritable) partialResult[0]).set(aggregation.count);
if (aggregation.sum != null) {
((HiveDecimalWritable) partialResult[1]).set(aggregation.sum);
} else {
partialResult[1] = null;
}
}
@Override
protected Object doTerminate(AverageAggregationBuffer<HiveDecimal> aggregation) {
if(aggregation.count == 0 || aggregation.sum == null) {
return null;
} else {
HiveDecimalWritable result = new HiveDecimalWritable(HiveDecimal.ZERO);
result.set(aggregation.sum.divide(HiveDecimal.create(aggregation.count)));
return result;
}
}
@Override
public AggregationBuffer getNewAggregationBuffer() throws HiveException {
AverageAggregationBuffer<HiveDecimal> result = new AverageAggregationBuffer<HiveDecimal>();
reset(result);
return result;
}
@Override
public GenericUDAFEvaluator getWindowingEvaluator(WindowFrameDef wFrameDef) {
// Don't use streaming for distinct cases
if (isWindowingDistinct()) {
return null;
}
return new GenericUDAFStreamingEvaluator.SumAvgEnhancer<HiveDecimalWritable, Object[]>(
this, wFrameDef) {
@Override
protected HiveDecimalWritable getNextResult(
org.apache.hadoop.hive.ql.udf.generic.GenericUDAFStreamingEvaluator.SumAvgEnhancer<HiveDecimalWritable, Object[]>.SumAvgStreamingState ss)
throws HiveException {
AverageAggregationBuffer<HiveDecimal> myagg = (AverageAggregationBuffer<HiveDecimal>) ss.wrappedBuf;
HiveDecimal r = myagg.count == 0 ? null : myagg.sum;
long cnt = myagg.count;
Object[] o = ss.retrieveNextIntermediateValue();
if (o != null) {
HiveDecimal d = (HiveDecimal) o[0];
r = r == null ? null : r.subtract(d);
cnt = cnt - ((Long) o[1]);
}
return r == null ? null : new HiveDecimalWritable(
r.divide(HiveDecimal.create(cnt)));
}
@Override
protected Object[] getCurrentIntermediateResult(
org.apache.hadoop.hive.ql.udf.generic.GenericUDAFStreamingEvaluator.SumAvgEnhancer<HiveDecimalWritable, Object[]>.SumAvgStreamingState ss)
throws HiveException {
AverageAggregationBuffer<HiveDecimal> myagg = (AverageAggregationBuffer<HiveDecimal>) ss.wrappedBuf;
return myagg.count == 0 ? null : new Object[] { myagg.sum,
myagg.count };
}
};
}
@Override
protected BasePartitionEvaluator createPartitionEvaluator(
WindowFrameDef winFrame,
PTFPartition partition,
List<PTFExpressionDef> parameters,
ObjectInspector outputOI) {
try {
return new BasePartitionEvaluator.AvgPartitionHiveDecimalEvaluator(this, winFrame, partition, parameters, inputOI, outputOI);
} catch(HiveException e) {
return super.createPartitionEvaluator(winFrame, partition, parameters, outputOI);
}
}
}
@AggregationType(estimable = true)
private static class AverageAggregationBuffer<TYPE> extends AbstractAggregationBuffer {
private HashSet<ObjectInspectorObject> uniqueObjects; // Unique rows.
private long count;
private TYPE sum;
@Override
public int estimate() {
return 2*JavaDataModel.PRIMITIVES2;
}
};
@SuppressWarnings("unchecked")
public static abstract class AbstractGenericUDAFAverageEvaluator<TYPE> extends GenericUDAFEvaluator {
protected boolean isWindowing;
protected boolean avgDistinct;
// For PARTIAL1 and COMPLETE
protected transient PrimitiveObjectInspector inputOI;
protected transient PrimitiveObjectInspector copiedOI;
// For PARTIAL2 and FINAL
private transient StructObjectInspector soi;
private transient StructField countField;
private transient StructField sumField;
private LongObjectInspector countFieldOI;
protected ObjectInspector sumFieldOI;
// For PARTIAL1 and PARTIAL2
protected transient Object[] partialResult;
private boolean warned = false;
protected abstract ObjectInspector getSumFieldJavaObjectInspector();
protected abstract ObjectInspector getSumFieldWritableObjectInspector();
protected abstract void doIterate(AverageAggregationBuffer<TYPE> aggregation,
PrimitiveObjectInspector inputOI, Object parameter);
protected abstract void doMerge(AverageAggregationBuffer<TYPE> aggregation, Long partialCount,
ObjectInspector sumFieldOI, Object partialSum);
protected abstract void doMergeAdd(TYPE sum, ObjectInspectorObject obj);
protected abstract void doTerminatePartial(AverageAggregationBuffer<TYPE> aggregation);
protected abstract Object doTerminate(AverageAggregationBuffer<TYPE> aggregation);
protected abstract void doReset(AverageAggregationBuffer<TYPE> aggregation) throws HiveException;
@Override
public ObjectInspector init(Mode m, ObjectInspector[] parameters)
throws HiveException {
assert (parameters.length == 1);
super.init(m, parameters);
// init input
partialResult = new Object[2];
partialResult[0] = new LongWritable(0);
if (mode == Mode.PARTIAL1 || mode == Mode.COMPLETE) {
inputOI = (PrimitiveObjectInspector) parameters[0];
copiedOI = (PrimitiveObjectInspector)ObjectInspectorUtils.getStandardObjectInspector(inputOI,
ObjectInspectorCopyOption.JAVA);
} else {
soi = (StructObjectInspector) parameters[0];
countField = soi.getStructFieldRef("count");
sumField = soi.getStructFieldRef("sum");
countFieldOI = (LongObjectInspector) countField.getFieldObjectInspector();
sumFieldOI = sumField.getFieldObjectInspector();
inputOI = (PrimitiveObjectInspector) soi.getStructFieldRef("input").getFieldObjectInspector();
}
// init output
if (mode == Mode.PARTIAL1 || mode == Mode.PARTIAL2) {
// The output of a partial aggregation is a struct containing
// a "long" count and a "double" sum.
ArrayList<ObjectInspector> foi = new ArrayList<ObjectInspector>();
foi.add(PrimitiveObjectInspectorFactory.writableLongObjectInspector);
foi.add(getSumFieldWritableObjectInspector());
// We need to "remember" the input object inspector so that we need to know the input type
// in order to determine the sum field type (precision/scale) for Mode.PARTIAL2 and Mode.FINAL.
foi.add(inputOI);
ArrayList<String> fname = new ArrayList<String>();
fname.add("count");
fname.add("sum");
fname.add("input");
// index 1 set by child
return ObjectInspectorFactory.getStandardStructObjectInspector(fname, foi);
} else {
return getSumFieldWritableObjectInspector();
}
}
public boolean isWindowingDistinct() {
return isWindowing && avgDistinct;
}
@AggregationType(estimable = true)
static class AverageAgg extends AbstractAggregationBuffer {
long count;
double sum;
@Override
public int estimate() { return JavaDataModel.PRIMITIVES2 * 2; }
};
@Override
public void reset(AggregationBuffer aggregation) throws HiveException {
doReset((AverageAggregationBuffer<TYPE>)aggregation);
}
@Override
public void iterate(AggregationBuffer aggregation, Object[] parameters)
throws HiveException {
assert (parameters.length == 1);
Object parameter = parameters[0];
if (parameter != null) {
AverageAggregationBuffer<TYPE> averageAggregation = (AverageAggregationBuffer<TYPE>) aggregation;
try {
// Skip the same value if avgDistinct is true
if (isWindowingDistinct()) {
ObjectInspectorObject obj = new ObjectInspectorObject(
ObjectInspectorUtils.copyToStandardObject(parameter, inputOI, ObjectInspectorCopyOption.JAVA),
copiedOI);
if (averageAggregation.uniqueObjects.contains(obj)) {
return;
}
averageAggregation.uniqueObjects.add(obj);
}
doIterate(averageAggregation, inputOI, parameter);
} catch (NumberFormatException e) {
if (!warned) {
warned = true;
LOG.warn("Ignoring similar exceptions: " + StringUtils.stringifyException(e));
}
}
}
}
@Override
public Object terminatePartial(AggregationBuffer aggregation) throws HiveException {
if (isWindowingDistinct()) {
throw new HiveException("Distinct windowing UDAF doesn't support merge and terminatePartial");
}
doTerminatePartial((AverageAggregationBuffer<TYPE>) aggregation);
return partialResult;
}
@Override
public void merge(AggregationBuffer aggregation, Object partial)
throws HiveException {
if (partial != null) {
if (isWindowingDistinct()) {
throw new HiveException("Distinct windowing UDAF doesn't support merge and terminatePartial");
} else {
doMerge((AverageAggregationBuffer<TYPE>)aggregation,
countFieldOI.get(soi.getStructFieldData(partial, countField)),
sumFieldOI, soi.getStructFieldData(partial, sumField));
}
}
}
@Override
public Object terminate(AggregationBuffer aggregation) throws HiveException {
return doTerminate((AverageAggregationBuffer<TYPE>)aggregation);
}
}
/**
* The intermediate sum field has 10 more integer digits with the same scale.
* This is exposed as static so that the vectorized AVG operator use the same logic
* @param precision
* @param scale
* @return
*/
public static DecimalTypeInfo deriveSumFieldTypeInfo(int precision, int scale) {
int intPart = precision - scale;
intPart = Math.min(intPart + 10, HiveDecimal.MAX_PRECISION - scale);
return TypeInfoFactory.getDecimalTypeInfo(intPart + scale, scale);
}
}