Examples with IBinaryComparatorFactory org.apache.hyracks.api.dataflow.value.IBinaryComparatorFactory used on opensource projects

Example 76 with IBinaryComparatorFactory

use of org.apache.hyracks.api.dataflow.value.IBinaryComparatorFactory in project asterixdb by apache.

the class MetadataNode method searchIndex.

private <ResultType> void searchIndex(JobId jobId, IMetadataIndex index, ITupleReference searchKey, IValueExtractor<ResultType> valueExtractor, List<ResultType> results) throws MetadataException, HyracksDataException, RemoteException {
    IBinaryComparatorFactory[] comparatorFactories = index.getKeyBinaryComparatorFactory();
    if (index.getFile() == null) {
        throw new MetadataException("No file for Index " + index.getDataverseName() + "." + index.getIndexName());
    }
    String resourceName = index.getFile().getRelativePath();
    IIndex indexInstance = datasetLifecycleManager.get(resourceName);
    datasetLifecycleManager.open(resourceName);
    IIndexAccessor indexAccessor = indexInstance.createAccessor(NoOpOperationCallback.INSTANCE, NoOpOperationCallback.INSTANCE);
    ITreeIndexCursor rangeCursor = (ITreeIndexCursor) indexAccessor.createSearchCursor(false);
    IBinaryComparator[] searchCmps = null;
    MultiComparator searchCmp = null;
    RangePredicate rangePred = null;
    if (searchKey != null) {
        searchCmps = new IBinaryComparator[searchKey.getFieldCount()];
        for (int i = 0; i < searchKey.getFieldCount(); i++) {
            searchCmps[i] = comparatorFactories[i].createBinaryComparator();
        }
        searchCmp = new MultiComparator(searchCmps);
    }
    rangePred = new RangePredicate(searchKey, searchKey, true, true, searchCmp, searchCmp);
    indexAccessor.search(rangeCursor, rangePred);
    try {
        while (rangeCursor.hasNext()) {
            rangeCursor.next();
            ResultType result = valueExtractor.getValue(jobId, rangeCursor.getTuple());
            if (result != null) {
                results.add(result);
            }
        }
    } finally {
        rangeCursor.close();
    }
    datasetLifecycleManager.close(resourceName);
}

Example 77 with IBinaryComparatorFactory

use of org.apache.hyracks.api.dataflow.value.IBinaryComparatorFactory in project asterixdb by apache.

the class HashPartitionMergeExchangePOperator method createConnectorDescriptor.

@Override
public Pair<IConnectorDescriptor, TargetConstraint> createConnectorDescriptor(IConnectorDescriptorRegistry spec, ILogicalOperator op, IOperatorSchema opSchema, JobGenContext context) throws AlgebricksException {
    int[] keys = new int[partitionFields.size()];
    IBinaryHashFunctionFactory[] hashFunctionFactories = new IBinaryHashFunctionFactory[partitionFields.size()];
    IVariableTypeEnvironment env = context.getTypeEnvironment(op);
    {
        int i = 0;
        IBinaryHashFunctionFactoryProvider hashFunProvider = context.getBinaryHashFunctionFactoryProvider();
        for (LogicalVariable v : partitionFields) {
            keys[i] = opSchema.findVariable(v);
            hashFunctionFactories[i] = hashFunProvider.getBinaryHashFunctionFactory(env.getVarType(v));
            ++i;
        }
    }
    ITuplePartitionComputerFactory tpcf = new FieldHashPartitionComputerFactory(keys, hashFunctionFactories);
    int n = orderColumns.size();
    int[] sortFields = new int[n];
    IBinaryComparatorFactory[] comparatorFactories = new IBinaryComparatorFactory[n];
    INormalizedKeyComputerFactoryProvider nkcfProvider = context.getNormalizedKeyComputerFactoryProvider();
    INormalizedKeyComputerFactory nkcf = null;
    int j = 0;
    for (OrderColumn oc : orderColumns) {
        LogicalVariable var = oc.getColumn();
        sortFields[j] = opSchema.findVariable(var);
        Object type = env.getVarType(var);
        IBinaryComparatorFactoryProvider bcfp = context.getBinaryComparatorFactoryProvider();
        comparatorFactories[j] = bcfp.getBinaryComparatorFactory(type, oc.getOrder() == OrderKind.ASC);
        if (j == 0 && nkcfProvider != null && type != null) {
            nkcf = nkcfProvider.getNormalizedKeyComputerFactory(type, oc.getOrder() == OrderKind.ASC);
        }
        j++;
    }
    IConnectorDescriptor conn = new MToNPartitioningMergingConnectorDescriptor(spec, tpcf, sortFields, comparatorFactories, nkcf);
    return new Pair<IConnectorDescriptor, TargetConstraint>(conn, null);
}

Example 78 with IBinaryComparatorFactory

use of org.apache.hyracks.api.dataflow.value.IBinaryComparatorFactory in project asterixdb by apache.

the class ExternalGroupByPOperator 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("pre-sorted 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("External 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("External 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();
    ISerializedAggregateEvaluatorFactory[] aff = new ISerializedAggregateEvaluatorFactory[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.createSerializableAggregateFunctionFactory(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[] comparatorFactories = JobGenHelper.variablesToAscBinaryComparatorFactories(gbyCols, aggOpInputEnv, context);
    RecordDescriptor recordDescriptor = JobGenHelper.mkRecordDescriptor(context.getTypeEnvironment(op), opSchema, context);
    IBinaryHashFunctionFamily[] hashFunctionFactories = JobGenHelper.variablesToBinaryHashFunctionFamilies(gbyCols, aggOpInputEnv, context);
    ISerializedAggregateEvaluatorFactory[] merges = new ISerializedAggregateEvaluatorFactory[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.createSerializableAggregateFunctionFactory(mergeFun, aggOpInputEnv, localInputSchemas, context);
    }
    IAggregatorDescriptorFactory aggregatorFactory = new SerializableAggregatorDescriptorFactory(aff);
    IAggregatorDescriptorFactory mergeFactory = new SerializableAggregatorDescriptorFactory(merges);
    INormalizedKeyComputerFactory normalizedKeyFactory = JobGenHelper.variablesToAscNormalizedKeyComputerFactory(gbyCols, aggOpInputEnv, context);
    // Calculates the hash table size (# of unique hash values) based on the budget and a tuple size.
    int memoryBudgetInBytes = context.getFrameSize() * frameLimit;
    int groupByColumnsCount = gby.getGroupByList().size() + numFds;
    int hashTableSize = ExternalGroupOperatorDescriptor.calculateGroupByTableCardinality(memoryBudgetInBytes, groupByColumnsCount, context.getFrameSize());
    ExternalGroupOperatorDescriptor gbyOpDesc = new ExternalGroupOperatorDescriptor(spec, hashTableSize, inputSize, keyAndDecFields, frameLimit, comparatorFactories, normalizedKeyFactory, aggregatorFactory, mergeFactory, recordDescriptor, recordDescriptor, new HashSpillableTableFactory(hashFunctionFactories));
    contributeOpDesc(builder, gby, gbyOpDesc);
    ILogicalOperator src = op.getInputs().get(0).getValue();
    builder.contributeGraphEdge(src, 0, op, 0);
}

Example 79 with IBinaryComparatorFactory

use of org.apache.hyracks.api.dataflow.value.IBinaryComparatorFactory in project asterixdb by apache.

the class MicroPreclusteredGroupByPOperator 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;
    IVariableTypeEnvironment env = context.getTypeEnvironment(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 = new NestedPlansAccumulatingAggregatorFactory(subplans, keys, fdColumns);
    IBinaryComparatorFactory[] comparatorFactories = JobGenHelper.variablesToAscBinaryComparatorFactories(columnList, env, context);
    RecordDescriptor recordDescriptor = JobGenHelper.mkRecordDescriptor(context.getTypeEnvironment(op), opSchema, context);
    RecordDescriptor inputRecordDesc = JobGenHelper.mkRecordDescriptor(context.getTypeEnvironment(op.getInputs().get(0).getValue()), inputSchemas[0], context);
    MicroPreClusteredGroupRuntimeFactory runtime = new MicroPreClusteredGroupRuntimeFactory(keys, comparatorFactories, aggregatorFactory, inputRecordDesc, recordDescriptor, null);
    builder.contributeMicroOperator(gby, runtime, recordDescriptor);
    ILogicalOperator src = op.getInputs().get(0).getValue();
    builder.contributeGraphEdge(src, 0, op, 0);
}

Example 80 with IBinaryComparatorFactory

use of org.apache.hyracks.api.dataflow.value.IBinaryComparatorFactory in project asterixdb by apache.

the class InMemoryHashJoinPOperator method contributeRuntimeOperator.

@Override
public void contributeRuntimeOperator(IHyracksJobBuilder builder, JobGenContext context, ILogicalOperator op, IOperatorSchema propagatedSchema, IOperatorSchema[] inputSchemas, IOperatorSchema outerPlanSchema) throws AlgebricksException {
    int[] keysLeft = JobGenHelper.variablesToFieldIndexes(keysLeftBranch, inputSchemas[0]);
    int[] keysRight = JobGenHelper.variablesToFieldIndexes(keysRightBranch, inputSchemas[1]);
    IVariableTypeEnvironment env = context.getTypeEnvironment(op);
    IBinaryHashFunctionFactory[] hashFunFactories = JobGenHelper.variablesToBinaryHashFunctionFactories(keysLeftBranch, env, context);
    IBinaryComparatorFactory[] comparatorFactories = new IBinaryComparatorFactory[keysLeft.length];
    int i = 0;
    IBinaryComparatorFactoryProvider bcfp = context.getBinaryComparatorFactoryProvider();
    for (LogicalVariable v : keysLeftBranch) {
        Object t = env.getVarType(v);
        comparatorFactories[i++] = bcfp.getBinaryComparatorFactory(t, true);
    }
    IPredicateEvaluatorFactoryProvider predEvaluatorFactoryProvider = context.getPredicateEvaluatorFactoryProvider();
    IPredicateEvaluatorFactory predEvaluatorFactory = (predEvaluatorFactoryProvider == null ? null : predEvaluatorFactoryProvider.getPredicateEvaluatorFactory(keysLeft, keysRight));
    RecordDescriptor recDescriptor = JobGenHelper.mkRecordDescriptor(context.getTypeEnvironment(op), propagatedSchema, context);
    IOperatorDescriptorRegistry spec = builder.getJobSpec();
    IOperatorDescriptor opDesc = null;
    switch(kind) {
        case INNER:
            {
                opDesc = new InMemoryHashJoinOperatorDescriptor(spec, keysLeft, keysRight, hashFunFactories, comparatorFactories, recDescriptor, tableSize, predEvaluatorFactory, memSizeInFrames);
                break;
            }
        case LEFT_OUTER:
            {
                IMissingWriterFactory[] nonMatchWriterFactories = new IMissingWriterFactory[inputSchemas[1].getSize()];
                for (int j = 0; j < nonMatchWriterFactories.length; j++) {
                    nonMatchWriterFactories[j] = context.getMissingWriterFactory();
                }
                opDesc = new InMemoryHashJoinOperatorDescriptor(spec, keysLeft, keysRight, hashFunFactories, comparatorFactories, predEvaluatorFactory, recDescriptor, true, nonMatchWriterFactories, tableSize, memSizeInFrames);
                break;
            }
        default:
            {
                throw new NotImplementedException();
            }
    }
    contributeOpDesc(builder, (AbstractLogicalOperator) op, opDesc);
    ILogicalOperator src1 = op.getInputs().get(0).getValue();
    builder.contributeGraphEdge(src1, 0, op, 0);
    ILogicalOperator src2 = op.getInputs().get(1).getValue();
    builder.contributeGraphEdge(src2, 0, op, 1);
}