use of org.apache.hyracks.dataflow.std.sort.TopKSorterOperatorDescriptor in project asterixdb by apache.
the class HeapSortMergeTest method optimizedSortMergeTest01.
@Test
public void optimizedSortMergeTest01() throws Exception {
JobSpecification spec = new JobSpecification();
FileSplit[] ordersSplits = new FileSplit[] { new ManagedFileSplit(NC1_ID, "data" + File.separator + "tpch0.001" + File.separator + "orders-part1.tbl"), new ManagedFileSplit(NC2_ID, "data" + File.separator + "tpch0.001" + File.separator + "orders-part2.tbl") };
IFileSplitProvider ordersSplitProvider = new ConstantFileSplitProvider(ordersSplits);
RecordDescriptor ordersDesc = new RecordDescriptor(new ISerializerDeserializer[] { new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer() });
FileScanOperatorDescriptor ordScanner = new FileScanOperatorDescriptor(spec, ordersSplitProvider, new DelimitedDataTupleParserFactory(new IValueParserFactory[] { UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE }, '|'), ordersDesc);
PartitionConstraintHelper.addAbsoluteLocationConstraint(spec, ordScanner, NC1_ID, NC2_ID);
// larger than the total record numbers.
int outputLimit = 5;
TopKSorterOperatorDescriptor sorter = new TopKSorterOperatorDescriptor(spec, 4, outputLimit, new int[] { 1, 0 }, null, new IBinaryComparatorFactory[] { PointableBinaryComparatorFactory.of(UTF8StringPointable.FACTORY), PointableBinaryComparatorFactory.of(UTF8StringPointable.FACTORY) }, ordersDesc);
PartitionConstraintHelper.addAbsoluteLocationConstraint(spec, sorter, NC1_ID, NC2_ID);
ResultSetId rsId = new ResultSetId(1);
spec.addResultSetId(rsId);
FileSplit fs = createFile(nc1);
IFileSplitProvider outputSplitProvider = new ConstantFileSplitProvider(new FileSplit[] { fs });
IOperatorDescriptor printer = new PlainFileWriterOperatorDescriptor(spec, outputSplitProvider, "|");
PartitionConstraintHelper.addAbsoluteLocationConstraint(spec, printer, NC1_ID);
spec.connect(new OneToOneConnectorDescriptor(spec), ordScanner, 0, sorter, 0);
spec.connect(new MToNPartitioningMergingConnectorDescriptor(spec, new FieldHashPartitionComputerFactory(new int[] { 1, 0 }, new IBinaryHashFunctionFactory[] { PointableBinaryHashFunctionFactory.of(UTF8StringPointable.FACTORY), PointableBinaryHashFunctionFactory.of(UTF8StringPointable.FACTORY) }), new int[] { 1, 0 }, new IBinaryComparatorFactory[] { PointableBinaryComparatorFactory.of(UTF8StringPointable.FACTORY), PointableBinaryComparatorFactory.of(UTF8StringPointable.FACTORY) }, new UTF8StringNormalizedKeyComputerFactory()), sorter, 0, printer, 0);
runTest(spec);
System.out.println("Result write into :" + fs.getPath() + " in node: " + fs.getNodeName());
}
use of org.apache.hyracks.dataflow.std.sort.TopKSorterOperatorDescriptor in project asterixdb by apache.
the class HeapSortMergeTest method createSortMergeJobSpec.
public static JobSpecification createSortMergeJobSpec() throws Exception {
JobSpecification spec = new JobSpecification();
FileSplit[] ordersSplits = new FileSplit[] { new ManagedFileSplit(NC1_ID, "data" + File.separator + "tpch0.001" + File.separator + "orders-part1.tbl"), new ManagedFileSplit(NC2_ID, "data" + File.separator + "tpch0.001" + File.separator + "orders-part2.tbl") };
IFileSplitProvider ordersSplitProvider = new ConstantFileSplitProvider(ordersSplits);
RecordDescriptor ordersDesc = new RecordDescriptor(new ISerializerDeserializer[] { new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer(), new UTF8StringSerializerDeserializer() });
FileScanOperatorDescriptor ordScanner = new FileScanOperatorDescriptor(spec, ordersSplitProvider, new DelimitedDataTupleParserFactory(new IValueParserFactory[] { UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE, UTF8StringParserFactory.INSTANCE }, '|'), ordersDesc);
PartitionConstraintHelper.addAbsoluteLocationConstraint(spec, ordScanner, NC1_ID, NC2_ID);
int outputLimit = 20;
TopKSorterOperatorDescriptor sorter = new TopKSorterOperatorDescriptor(spec, 4, outputLimit, new int[] { 1, 0 }, null, new IBinaryComparatorFactory[] { PointableBinaryComparatorFactory.of(UTF8StringPointable.FACTORY), PointableBinaryComparatorFactory.of(UTF8StringPointable.FACTORY) }, ordersDesc);
PartitionConstraintHelper.addAbsoluteLocationConstraint(spec, sorter, NC1_ID, NC2_ID);
LimitOperatorDescriptor filter = new LimitOperatorDescriptor(spec, ordersDesc, outputLimit);
PartitionConstraintHelper.addAbsoluteLocationConstraint(spec, filter, NC1_ID);
ResultSetId rsId = new ResultSetId(1);
spec.addResultSetId(rsId);
IOperatorDescriptor printer = new ResultWriterOperatorDescriptor(spec, rsId, false, false, ResultSerializerFactoryProvider.INSTANCE.getResultSerializerFactoryProvider());
PartitionConstraintHelper.addAbsoluteLocationConstraint(spec, printer, NC1_ID);
spec.connect(new OneToOneConnectorDescriptor(spec), ordScanner, 0, sorter, 0);
spec.connect(new MToNPartitioningMergingConnectorDescriptor(spec, new FieldHashPartitionComputerFactory(new int[] { 1, 0 }, new IBinaryHashFunctionFactory[] { PointableBinaryHashFunctionFactory.of(UTF8StringPointable.FACTORY), PointableBinaryHashFunctionFactory.of(UTF8StringPointable.FACTORY) }), new int[] { 1, 0 }, new IBinaryComparatorFactory[] { PointableBinaryComparatorFactory.of(UTF8StringPointable.FACTORY), PointableBinaryComparatorFactory.of(UTF8StringPointable.FACTORY) }, new UTF8StringNormalizedKeyComputerFactory()), sorter, 0, filter, 0);
spec.connect(new OneToOneConnectorDescriptor(spec), filter, 0, printer, 0);
return spec;
}
use of org.apache.hyracks.dataflow.std.sort.TopKSorterOperatorDescriptor in project asterixdb by apache.
the class StableSortPOperator method contributeRuntimeOperator.
@Override
public void contributeRuntimeOperator(IHyracksJobBuilder builder, JobGenContext context, ILogicalOperator op, IOperatorSchema opSchema, IOperatorSchema[] inputSchemas, IOperatorSchema outerPlanSchema) throws AlgebricksException {
IOperatorDescriptorRegistry spec = builder.getJobSpec();
RecordDescriptor recDescriptor = JobGenHelper.mkRecordDescriptor(context.getTypeEnvironment(op), opSchema, context);
int n = sortColumns.length;
int[] sortFields = new int[n];
IBinaryComparatorFactory[] comps = new IBinaryComparatorFactory[n];
INormalizedKeyComputerFactoryProvider nkcfProvider = context.getNormalizedKeyComputerFactoryProvider();
INormalizedKeyComputerFactory nkcf = null;
IVariableTypeEnvironment env = context.getTypeEnvironment(op);
int i = 0;
for (OrderColumn oc : sortColumns) {
LogicalVariable var = oc.getColumn();
sortFields[i] = opSchema.findVariable(var);
Object type = env.getVarType(var);
OrderKind order = oc.getOrder();
if (i == 0 && nkcfProvider != null && type != null) {
nkcf = nkcfProvider.getNormalizedKeyComputerFactory(type, order == OrderKind.ASC);
}
IBinaryComparatorFactoryProvider bcfp = context.getBinaryComparatorFactoryProvider();
comps[i] = bcfp.getBinaryComparatorFactory(type, oc.getOrder() == OrderKind.ASC);
i++;
}
// topK == -1 means that a topK value is not provided.
if (topK == -1) {
ExternalSortOperatorDescriptor sortOpDesc = new ExternalSortOperatorDescriptor(spec, maxNumberOfFrames, sortFields, nkcf, comps, recDescriptor);
contributeOpDesc(builder, (AbstractLogicalOperator) op, sortOpDesc);
ILogicalOperator src = op.getInputs().get(0).getValue();
builder.contributeGraphEdge(src, 0, op, 0);
} else {
// Since topK value is provided, topK optimization is possible.
// We call topKSorter instead of calling ExternalSortOperator.
TopKSorterOperatorDescriptor sortOpDesc = new TopKSorterOperatorDescriptor(spec, maxNumberOfFrames, topK, sortFields, nkcf, comps, recDescriptor);
contributeOpDesc(builder, (AbstractLogicalOperator) op, sortOpDesc);
ILogicalOperator src = op.getInputs().get(0).getValue();
builder.contributeGraphEdge(src, 0, op, 0);
}
}
use of org.apache.hyracks.dataflow.std.sort.TopKSorterOperatorDescriptor in project asterixdb by apache.
the class Sort method createJob.
private static JobSpecification createJob(FileSplit[] ordersSplits, FileSplit[] outputSplit, String memBufferAlg, int frameLimit, int frameSize, int limit, boolean usingHeapSorter) {
JobSpecification spec = new JobSpecification();
spec.setFrameSize(frameSize);
IFileSplitProvider ordersSplitProvider = new ConstantFileSplitProvider(ordersSplits);
FileScanOperatorDescriptor ordScanner = new FileScanOperatorDescriptor(spec, ordersSplitProvider, new DelimitedDataTupleParserFactory(orderParserFactories, '|'), ordersDesc);
createPartitionConstraint(spec, ordScanner, ordersSplits);
AbstractSorterOperatorDescriptor sorter;
if (usingHeapSorter && limit < Integer.MAX_VALUE) {
sorter = new TopKSorterOperatorDescriptor(spec, frameLimit, limit, SortFields, null, SortFieldsComparatorFactories, ordersDesc);
} else {
if (memBufferAlg.equalsIgnoreCase("bestfit")) {
sorter = new ExternalSortOperatorDescriptor(spec, frameLimit, SortFields, null, SortFieldsComparatorFactories, ordersDesc, Algorithm.MERGE_SORT, EnumFreeSlotPolicy.SMALLEST_FIT, limit);
} else if (memBufferAlg.equalsIgnoreCase("biggestfit")) {
sorter = new ExternalSortOperatorDescriptor(spec, frameLimit, SortFields, null, SortFieldsComparatorFactories, ordersDesc, Algorithm.MERGE_SORT, EnumFreeSlotPolicy.BIGGEST_FIT, limit);
} else {
sorter = new ExternalSortOperatorDescriptor(spec, frameLimit, SortFields, null, SortFieldsComparatorFactories, ordersDesc, Algorithm.MERGE_SORT, EnumFreeSlotPolicy.LAST_FIT, limit);
}
}
createPartitionConstraint(spec, sorter, ordersSplits);
IFileSplitProvider outputSplitProvider = new ConstantFileSplitProvider(outputSplit);
IOperatorDescriptor printer = new PlainFileWriterOperatorDescriptor(spec, outputSplitProvider, "|");
createPartitionConstraint(spec, printer, outputSplit);
spec.connect(new OneToOneConnectorDescriptor(spec), ordScanner, 0, sorter, 0);
spec.connect(new MToNPartitioningMergingConnectorDescriptor(spec, new FieldHashPartitionComputerFactory(SortFields, orderBinaryHashFunctionFactories), SortFields, SortFieldsComparatorFactories, new UTF8StringNormalizedKeyComputerFactory()), sorter, 0, printer, 0);
spec.addRoot(printer);
return spec;
}
Aggregations