use of org.apache.hyracks.dataflow.std.group.IAggregatorDescriptorFactory in project asterixdb by apache.
the class PreSortedDistinctByPOperator method contributeRuntimeOperator.
@Override
public void contributeRuntimeOperator(IHyracksJobBuilder builder, JobGenContext context, ILogicalOperator op, IOperatorSchema opSchema, IOperatorSchema[] inputSchemas, IOperatorSchema outerPlanSchema) throws AlgebricksException {
IOperatorDescriptorRegistry spec = builder.getJobSpec();
int[] keys = JobGenHelper.variablesToFieldIndexes(columnList, inputSchemas[0]);
int sz = inputSchemas[0].getSize();
int fdSz = sz - columnList.size();
int[] fdColumns = new int[fdSz];
int j = 0;
for (LogicalVariable v : inputSchemas[0]) {
if (!columnList.contains(v)) {
fdColumns[j++] = inputSchemas[0].findVariable(v);
}
}
int[] keysAndDecs = new int[keys.length + fdColumns.length];
for (int i = 0; i < keys.length; i++) {
keysAndDecs[i] = keys[i];
}
for (int i = 0; i < fdColumns.length; i++) {
keysAndDecs[i + keys.length] = fdColumns[i];
}
IBinaryComparatorFactory[] comparatorFactories = JobGenHelper.variablesToAscBinaryComparatorFactories(columnList, context.getTypeEnvironment(op), context);
IAggregateEvaluatorFactory[] aggFactories = new IAggregateEvaluatorFactory[] {};
IAggregatorDescriptorFactory aggregatorFactory = new SimpleAlgebricksAccumulatingAggregatorFactory(aggFactories, keysAndDecs);
RecordDescriptor recordDescriptor = JobGenHelper.mkRecordDescriptor(context.getTypeEnvironment(op), opSchema, context);
/** make fd columns part of the key but the comparator only compares the distinct key columns */
PreclusteredGroupOperatorDescriptor opDesc = new PreclusteredGroupOperatorDescriptor(spec, keysAndDecs, comparatorFactories, aggregatorFactory, recordDescriptor);
contributeOpDesc(builder, (AbstractLogicalOperator) op, opDesc);
ILogicalOperator src = op.getInputs().get(0).getValue();
builder.contributeGraphEdge(src, 0, op, 0);
}
use of org.apache.hyracks.dataflow.std.group.IAggregatorDescriptorFactory in project asterixdb by apache.
the class PreclusteredGroupByPOperator method contributeRuntimeOperator.
@Override
public void contributeRuntimeOperator(IHyracksJobBuilder builder, JobGenContext context, ILogicalOperator op, IOperatorSchema opSchema, IOperatorSchema[] inputSchemas, IOperatorSchema outerPlanSchema) throws AlgebricksException {
int[] keys = JobGenHelper.variablesToFieldIndexes(columnList, inputSchemas[0]);
GroupByOperator gby = (GroupByOperator) op;
int[] fdColumns = getFdColumns(gby, inputSchemas[0]);
// compile subplans and set the gby op. schema accordingly
AlgebricksPipeline[] subplans = compileSubplans(inputSchemas[0], gby, opSchema, context);
IAggregatorDescriptorFactory aggregatorFactory;
if (gby.getNestedPlans().get(0).getRoots().get(0).getValue().getOperatorTag() == LogicalOperatorTag.RUNNINGAGGREGATE) {
aggregatorFactory = new NestedPlansRunningAggregatorFactory(subplans, keys, fdColumns);
} else {
aggregatorFactory = new NestedPlansAccumulatingAggregatorFactory(subplans, keys, fdColumns);
}
IOperatorDescriptorRegistry spec = builder.getJobSpec();
IBinaryComparatorFactory[] comparatorFactories = JobGenHelper.variablesToAscBinaryComparatorFactories(columnList, context.getTypeEnvironment(op), context);
RecordDescriptor recordDescriptor = JobGenHelper.mkRecordDescriptor(context.getTypeEnvironment(op), opSchema, context);
PreclusteredGroupOperatorDescriptor opDesc = new PreclusteredGroupOperatorDescriptor(spec, keys, comparatorFactories, aggregatorFactory, recordDescriptor, groupAll);
contributeOpDesc(builder, (AbstractLogicalOperator) op, opDesc);
ILogicalOperator src = op.getInputs().get(0).getValue();
builder.contributeGraphEdge(src, 0, op, 0);
}
use of org.apache.hyracks.dataflow.std.group.IAggregatorDescriptorFactory in project asterixdb by apache.
the class SortGroupByPOperator method contributeRuntimeOperator.
@Override
public void contributeRuntimeOperator(IHyracksJobBuilder builder, JobGenContext context, ILogicalOperator op, IOperatorSchema opSchema, IOperatorSchema[] inputSchemas, IOperatorSchema outerPlanSchema) throws AlgebricksException {
List<LogicalVariable> gbyCols = getGbyColumns();
int[] keys = JobGenHelper.variablesToFieldIndexes(gbyCols, inputSchemas[0]);
GroupByOperator gby = (GroupByOperator) op;
int numFds = gby.getDecorList().size();
int[] fdColumns = new int[numFds];
int j = 0;
for (Pair<LogicalVariable, Mutable<ILogicalExpression>> p : gby.getDecorList()) {
ILogicalExpression expr = p.second.getValue();
if (expr.getExpressionTag() != LogicalExpressionTag.VARIABLE) {
throw new AlgebricksException("Sort group-by expects variable references.");
}
VariableReferenceExpression v = (VariableReferenceExpression) expr;
LogicalVariable decor = v.getVariableReference();
fdColumns[j++] = inputSchemas[0].findVariable(decor);
}
if (gby.getNestedPlans().size() != 1) {
throw new AlgebricksException("Sort group-by currently works only for one nested plan with one root containing" + "an aggregate and a nested-tuple-source.");
}
ILogicalPlan p0 = gby.getNestedPlans().get(0);
if (p0.getRoots().size() != 1) {
throw new AlgebricksException("Sort group-by currently works only for one nested plan with one root containing" + "an aggregate and a nested-tuple-source.");
}
Mutable<ILogicalOperator> r0 = p0.getRoots().get(0);
AggregateOperator aggOp = (AggregateOperator) r0.getValue();
IPartialAggregationTypeComputer partialAggregationTypeComputer = context.getPartialAggregationTypeComputer();
List<Object> intermediateTypes = new ArrayList<Object>();
int n = aggOp.getExpressions().size();
IAggregateEvaluatorFactory[] aff = new IAggregateEvaluatorFactory[n];
int i = 0;
IExpressionRuntimeProvider expressionRuntimeProvider = context.getExpressionRuntimeProvider();
IVariableTypeEnvironment aggOpInputEnv = context.getTypeEnvironment(aggOp.getInputs().get(0).getValue());
IVariableTypeEnvironment outputEnv = context.getTypeEnvironment(op);
for (Mutable<ILogicalExpression> exprRef : aggOp.getExpressions()) {
AggregateFunctionCallExpression aggFun = (AggregateFunctionCallExpression) exprRef.getValue();
aff[i++] = expressionRuntimeProvider.createAggregateFunctionFactory(aggFun, aggOpInputEnv, inputSchemas, context);
intermediateTypes.add(partialAggregationTypeComputer.getType(aggFun, aggOpInputEnv, context.getMetadataProvider()));
}
int[] keyAndDecFields = new int[keys.length + fdColumns.length];
for (i = 0; i < keys.length; ++i) {
keyAndDecFields[i] = keys[i];
}
for (i = 0; i < fdColumns.length; i++) {
keyAndDecFields[keys.length + i] = fdColumns[i];
}
List<LogicalVariable> keyAndDecVariables = new ArrayList<LogicalVariable>();
for (Pair<LogicalVariable, Mutable<ILogicalExpression>> p : gby.getGroupByList()) {
keyAndDecVariables.add(p.first);
}
for (Pair<LogicalVariable, Mutable<ILogicalExpression>> p : gby.getDecorList()) {
keyAndDecVariables.add(GroupByOperator.getDecorVariable(p));
}
for (LogicalVariable var : keyAndDecVariables) {
aggOpInputEnv.setVarType(var, outputEnv.getVarType(var));
}
compileSubplans(inputSchemas[0], gby, opSchema, context);
IOperatorDescriptorRegistry spec = builder.getJobSpec();
IBinaryComparatorFactory[] compFactories = new IBinaryComparatorFactory[gbyCols.size()];
IBinaryComparatorFactoryProvider bcfProvider = context.getBinaryComparatorFactoryProvider();
i = 0;
for (LogicalVariable v : gbyCols) {
Object type = aggOpInputEnv.getVarType(v);
if (orderColumns[i].getOrder() == OrderKind.ASC) {
compFactories[i] = bcfProvider.getBinaryComparatorFactory(type, true);
} else {
compFactories[i] = bcfProvider.getBinaryComparatorFactory(type, false);
}
i++;
}
RecordDescriptor recordDescriptor = JobGenHelper.mkRecordDescriptor(context.getTypeEnvironment(op), opSchema, context);
IAggregateEvaluatorFactory[] merges = new IAggregateEvaluatorFactory[n];
List<LogicalVariable> usedVars = new ArrayList<LogicalVariable>();
IOperatorSchema[] localInputSchemas = new IOperatorSchema[1];
localInputSchemas[0] = new OperatorSchemaImpl();
for (i = 0; i < n; i++) {
AggregateFunctionCallExpression aggFun = (AggregateFunctionCallExpression) aggOp.getMergeExpressions().get(i).getValue();
aggFun.getUsedVariables(usedVars);
}
i = 0;
for (Object type : intermediateTypes) {
aggOpInputEnv.setVarType(usedVars.get(i++), type);
}
for (LogicalVariable keyVar : keyAndDecVariables) {
localInputSchemas[0].addVariable(keyVar);
}
for (LogicalVariable usedVar : usedVars) {
localInputSchemas[0].addVariable(usedVar);
}
for (i = 0; i < n; i++) {
AggregateFunctionCallExpression mergeFun = (AggregateFunctionCallExpression) aggOp.getMergeExpressions().get(i).getValue();
merges[i] = expressionRuntimeProvider.createAggregateFunctionFactory(mergeFun, aggOpInputEnv, localInputSchemas, context);
}
RecordDescriptor partialAggRecordDescriptor = JobGenHelper.mkRecordDescriptor(context.getTypeEnvironment(op), localInputSchemas[0], context);
IAggregatorDescriptorFactory aggregatorFactory = new SimpleAlgebricksAccumulatingAggregatorFactory(aff, keyAndDecFields);
IAggregatorDescriptorFactory mergeFactory = new SimpleAlgebricksAccumulatingAggregatorFactory(merges, keyAndDecFields);
INormalizedKeyComputerFactory normalizedKeyFactory = null;
INormalizedKeyComputerFactoryProvider nkcfProvider = context.getNormalizedKeyComputerFactoryProvider();
if (nkcfProvider == null) {
normalizedKeyFactory = null;
}
Object type = aggOpInputEnv.getVarType(gbyCols.get(0));
normalizedKeyFactory = orderColumns[0].getOrder() == OrderKind.ASC ? nkcfProvider.getNormalizedKeyComputerFactory(type, true) : nkcfProvider.getNormalizedKeyComputerFactory(type, false);
SortGroupByOperatorDescriptor gbyOpDesc = new SortGroupByOperatorDescriptor(spec, frameLimit, keys, keyAndDecFields, normalizedKeyFactory, compFactories, aggregatorFactory, mergeFactory, partialAggRecordDescriptor, recordDescriptor, false);
contributeOpDesc(builder, gby, gbyOpDesc);
ILogicalOperator src = op.getInputs().get(0).getValue();
builder.contributeGraphEdge(src, 0, op, 0);
}
use of org.apache.hyracks.dataflow.std.group.IAggregatorDescriptorFactory in project asterixdb by apache.
the class ExternalSortGroupByRunMerger method prepareIntermediateMergeResultWriter.
@Override
protected IFrameWriter prepareIntermediateMergeResultWriter(RunFileWriter mergeFileWriter) throws HyracksDataException {
IAggregatorDescriptorFactory aggregatorFactory = localSide ? mergeAggregatorFactory : partialAggregatorFactory;
boolean outputPartial = true;
return new PreclusteredGroupWriter(ctx, mergeGroupFields, groupByComparators, aggregatorFactory, partialAggRecordDesc, partialAggRecordDesc, mergeFileWriter, outputPartial);
}
use of org.apache.hyracks.dataflow.std.group.IAggregatorDescriptorFactory in project asterixdb by apache.
the class ExternalSortGroupByRunMerger method prepareSkipMergingFinalResultWriter.
@Override
protected IFrameWriter prepareSkipMergingFinalResultWriter(IFrameWriter nextWriter) throws HyracksDataException {
IAggregatorDescriptorFactory aggregatorFactory = localSide ? partialAggregatorFactory : mergeAggregatorFactory;
boolean outputPartial = false;
return new PreclusteredGroupWriter(ctx, groupFields, groupByComparators, aggregatorFactory, inputRecordDesc, outRecordDesc, nextWriter, outputPartial);
}
Aggregations