Examples with IFileSplitProvider org.apache.hyracks.dataflow.std.file.IFileSplitProvider used on opensource projects

Example 1 with IFileSplitProvider

use of org.apache.hyracks.dataflow.std.file.IFileSplitProvider in project asterixdb by apache.

the class InvertedIndexPOperator method buildInvertedIndexRuntime.

public Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> buildInvertedIndexRuntime(MetadataProvider metadataProvider, JobGenContext context, JobSpecification jobSpec, AbstractUnnestMapOperator unnestMap, IOperatorSchema opSchema, boolean retainInput, boolean retainMissing, String datasetName, Dataset dataset, String indexName, ATypeTag searchKeyType, int[] keyFields, SearchModifierType searchModifierType, IAlgebricksConstantValue similarityThreshold, int[] minFilterFieldIndexes, int[] maxFilterFieldIndexes, boolean isFullTextSearchQuery) throws AlgebricksException {
    try {
        IAObject simThresh = ((AsterixConstantValue) similarityThreshold).getObject();
        int numPrimaryKeys = dataset.getPrimaryKeys().size();
        Index secondaryIndex = MetadataManager.INSTANCE.getIndex(metadataProvider.getMetadataTxnContext(), dataset.getDataverseName(), dataset.getDatasetName(), indexName);
        if (secondaryIndex == null) {
            throw new AlgebricksException("Code generation error: no index " + indexName + " for dataset " + datasetName);
        }
        IVariableTypeEnvironment typeEnv = context.getTypeEnvironment(unnestMap);
        RecordDescriptor outputRecDesc = JobGenHelper.mkRecordDescriptor(typeEnv, opSchema, context);
        Pair<IFileSplitProvider, AlgebricksPartitionConstraint> secondarySplitsAndConstraint = metadataProvider.getSplitProviderAndConstraints(dataset, indexName);
        // TODO: Here we assume there is only one search key field.
        int queryField = keyFields[0];
        // Get tokenizer and search modifier factories.
        IInvertedIndexSearchModifierFactory searchModifierFactory = InvertedIndexAccessMethod.getSearchModifierFactory(searchModifierType, simThresh, secondaryIndex);
        IBinaryTokenizerFactory queryTokenizerFactory = InvertedIndexAccessMethod.getBinaryTokenizerFactory(searchModifierType, searchKeyType, secondaryIndex);
        IIndexDataflowHelperFactory dataflowHelperFactory = new IndexDataflowHelperFactory(metadataProvider.getStorageComponentProvider().getStorageManager(), secondarySplitsAndConstraint.first);
        LSMInvertedIndexSearchOperatorDescriptor invIndexSearchOp = new LSMInvertedIndexSearchOperatorDescriptor(jobSpec, outputRecDesc, queryField, dataflowHelperFactory, queryTokenizerFactory, searchModifierFactory, retainInput, retainMissing, context.getMissingWriterFactory(), dataset.getSearchCallbackFactory(metadataProvider.getStorageComponentProvider(), secondaryIndex, ((JobEventListenerFactory) jobSpec.getJobletEventListenerFactory()).getJobId(), IndexOperation.SEARCH, null), minFilterFieldIndexes, maxFilterFieldIndexes, isFullTextSearchQuery, numPrimaryKeys, false);
        return new Pair<>(invIndexSearchOp, secondarySplitsAndConstraint.second);
    } catch (MetadataException e) {
        throw new AlgebricksException(e);
    }
}

Example 2 with IFileSplitProvider

use of org.apache.hyracks.dataflow.std.file.IFileSplitProvider in project asterixdb by apache.

the class FeedOperations method buildRemoveFeedStorageJob.

public static JobSpecification buildRemoveFeedStorageJob(MetadataProvider metadataProvider, Feed feed) throws AsterixException {
    JobSpecification spec = RuntimeUtils.createJobSpecification(metadataProvider.getApplicationContext());
    AlgebricksAbsolutePartitionConstraint allCluster = ClusterStateManager.INSTANCE.getClusterLocations();
    Set<String> nodes = new TreeSet<>();
    for (String node : allCluster.getLocations()) {
        nodes.add(node);
    }
    AlgebricksAbsolutePartitionConstraint locations = new AlgebricksAbsolutePartitionConstraint(nodes.toArray(new String[nodes.size()]));
    FileSplit[] feedLogFileSplits = FeedUtils.splitsForAdapter(feed.getDataverseName(), feed.getFeedName(), locations);
    org.apache.hyracks.algebricks.common.utils.Pair<IFileSplitProvider, AlgebricksPartitionConstraint> spC = StoragePathUtil.splitProviderAndPartitionConstraints(feedLogFileSplits);
    FileRemoveOperatorDescriptor frod = new FileRemoveOperatorDescriptor(spec, spC.first, true);
    AlgebricksPartitionConstraintHelper.setPartitionConstraintInJobSpec(spec, frod, spC.second);
    spec.addRoot(frod);
    return spec;
}

Example 3 with IFileSplitProvider

use of org.apache.hyracks.dataflow.std.file.IFileSplitProvider in project asterixdb by apache.

the class MetadataProvider method buildExternalDataLookupRuntime.

public Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> buildExternalDataLookupRuntime(JobSpecification jobSpec, Dataset dataset, int[] ridIndexes, boolean retainInput, IVariableTypeEnvironment typeEnv, IOperatorSchema opSchema, JobGenContext context, MetadataProvider metadataProvider, boolean retainMissing) throws AlgebricksException {
    try {
        // Get data type
        ARecordType itemType = (ARecordType) MetadataManager.INSTANCE.getDatatype(metadataProvider.getMetadataTxnContext(), dataset.getDataverseName(), dataset.getItemTypeName()).getDatatype();
        ExternalDatasetDetails datasetDetails = (ExternalDatasetDetails) dataset.getDatasetDetails();
        LookupAdapterFactory<?> adapterFactory = AdapterFactoryProvider.getLookupAdapterFactory(getApplicationContext().getServiceContext(), datasetDetails.getProperties(), itemType, ridIndexes, retainInput, retainMissing, context.getMissingWriterFactory());
        String fileIndexName = IndexingConstants.getFilesIndexName(dataset.getDatasetName());
        Pair<IFileSplitProvider, AlgebricksPartitionConstraint> spPc = metadataProvider.getSplitProviderAndConstraints(dataset, fileIndexName);
        Index fileIndex = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataset.getDataverseName(), dataset.getDatasetName(), fileIndexName);
        // Create the file index data flow helper
        IIndexDataflowHelperFactory indexDataflowHelperFactory = new IndexDataflowHelperFactory(storaegComponentProvider.getStorageManager(), spPc.first);
        // Create the out record descriptor, appContext and fileSplitProvider for the files index
        RecordDescriptor outRecDesc = JobGenHelper.mkRecordDescriptor(typeEnv, opSchema, context);
        ISearchOperationCallbackFactory searchOpCallbackFactory = dataset.getSearchCallbackFactory(storaegComponentProvider, fileIndex, jobId, IndexOperation.SEARCH, null);
        // Create the operator
        ExternalLookupOperatorDescriptor op = new ExternalLookupOperatorDescriptor(jobSpec, adapterFactory, outRecDesc, indexDataflowHelperFactory, searchOpCallbackFactory, ExternalDatasetsRegistry.INSTANCE.getAndLockDatasetVersion(dataset, this));
        return new Pair<>(op, spPc.second);
    } catch (Exception e) {
        throw new AlgebricksException(e);
    }
}

Example 4 with IFileSplitProvider

use of org.apache.hyracks.dataflow.std.file.IFileSplitProvider in project asterixdb by apache.

the class MetadataProvider method getBTreeRuntime.

private Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> getBTreeRuntime(String dataverseName, String datasetName, String indexName, IOperatorSchema propagatedSchema, List<LogicalVariable> primaryKeys, List<LogicalVariable> secondaryKeys, List<LogicalVariable> additionalNonKeyFields, AsterixTupleFilterFactory filterFactory, RecordDescriptor inputRecordDesc, JobGenContext context, JobSpecification spec, IndexOperation indexOp, boolean bulkload, List<LogicalVariable> prevSecondaryKeys, List<LogicalVariable> prevAdditionalFilteringKeys) throws AlgebricksException {
    Dataset dataset = MetadataManagerUtil.findExistingDataset(mdTxnCtx, dataverseName, datasetName);
    boolean temp = dataset.getDatasetDetails().isTemp();
    isTemporaryDatasetWriteJob = isTemporaryDatasetWriteJob && temp;
    int numKeys = primaryKeys.size() + secondaryKeys.size();
    int numFilterFields = DatasetUtil.getFilterField(dataset) == null ? 0 : 1;
    // generate field permutations
    int[] fieldPermutation = new int[numKeys + numFilterFields];
    int[] modificationCallbackPrimaryKeyFields = new int[primaryKeys.size()];
    int i = 0;
    int j = 0;
    for (LogicalVariable varKey : secondaryKeys) {
        int idx = propagatedSchema.findVariable(varKey);
        fieldPermutation[i] = idx;
        i++;
    }
    for (LogicalVariable varKey : primaryKeys) {
        int idx = propagatedSchema.findVariable(varKey);
        fieldPermutation[i] = idx;
        modificationCallbackPrimaryKeyFields[j] = i;
        i++;
        j++;
    }
    if (numFilterFields > 0) {
        int idx = propagatedSchema.findVariable(additionalNonKeyFields.get(0));
        fieldPermutation[numKeys] = idx;
    }
    int[] prevFieldPermutation = null;
    if (indexOp == IndexOperation.UPSERT) {
        // generate field permutations for prev record
        prevFieldPermutation = new int[numKeys + numFilterFields];
        int k = 0;
        for (LogicalVariable varKey : prevSecondaryKeys) {
            int idx = propagatedSchema.findVariable(varKey);
            prevFieldPermutation[k] = idx;
            k++;
        }
        for (LogicalVariable varKey : primaryKeys) {
            int idx = propagatedSchema.findVariable(varKey);
            prevFieldPermutation[k] = idx;
            k++;
        }
        // Filter can only be one field!
        if (numFilterFields > 0) {
            int idx = propagatedSchema.findVariable(prevAdditionalFilteringKeys.get(0));
            prevFieldPermutation[numKeys] = idx;
        }
    }
    try {
        // Index parameters.
        Index secondaryIndex = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataset.getDataverseName(), dataset.getDatasetName(), indexName);
        Pair<IFileSplitProvider, AlgebricksPartitionConstraint> splitsAndConstraint = getSplitProviderAndConstraints(dataset, secondaryIndex.getIndexName());
        // prepare callback
        JobId jobId = ((JobEventListenerFactory) spec.getJobletEventListenerFactory()).getJobId();
        IModificationOperationCallbackFactory modificationCallbackFactory = dataset.getModificationCallbackFactory(storaegComponentProvider, secondaryIndex, jobId, indexOp, modificationCallbackPrimaryKeyFields);
        IIndexDataflowHelperFactory idfh = new IndexDataflowHelperFactory(storaegComponentProvider.getStorageManager(), splitsAndConstraint.first);
        IOperatorDescriptor op;
        if (bulkload) {
            long numElementsHint = getCardinalityPerPartitionHint(dataset);
            op = new TreeIndexBulkLoadOperatorDescriptor(spec, inputRecordDesc, fieldPermutation, GlobalConfig.DEFAULT_TREE_FILL_FACTOR, false, numElementsHint, false, idfh);
        } else if (indexOp == IndexOperation.UPSERT) {
            op = new LSMSecondaryUpsertOperatorDescriptor(spec, inputRecordDesc, fieldPermutation, idfh, filterFactory, modificationCallbackFactory, prevFieldPermutation);
        } else {
            op = new LSMTreeInsertDeleteOperatorDescriptor(spec, inputRecordDesc, fieldPermutation, indexOp, idfh, filterFactory, false, modificationCallbackFactory);
        }
        return new Pair<>(op, splitsAndConstraint.second);
    } catch (Exception e) {
        throw new AlgebricksException(e);
    }
}

Example 5 with IFileSplitProvider

use of org.apache.hyracks.dataflow.std.file.IFileSplitProvider in project asterixdb by apache.

the class MetadataProvider method getRTreeRuntime.

private Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> getRTreeRuntime(String dataverseName, String datasetName, String indexName, IOperatorSchema propagatedSchema, List<LogicalVariable> primaryKeys, List<LogicalVariable> secondaryKeys, List<LogicalVariable> additionalNonKeyFields, AsterixTupleFilterFactory filterFactory, RecordDescriptor recordDesc, JobGenContext context, JobSpecification spec, IndexOperation indexOp, boolean bulkload, List<LogicalVariable> prevSecondaryKeys, List<LogicalVariable> prevAdditionalFilteringKeys) throws AlgebricksException {
    try {
        Dataset dataset = MetadataManagerUtil.findExistingDataset(mdTxnCtx, dataverseName, datasetName);
        boolean temp = dataset.getDatasetDetails().isTemp();
        isTemporaryDatasetWriteJob = isTemporaryDatasetWriteJob && temp;
        String itemTypeName = dataset.getItemTypeName();
        IAType itemType = MetadataManager.INSTANCE.getDatatype(mdTxnCtx, dataset.getItemTypeDataverseName(), itemTypeName).getDatatype();
        validateRecordType(itemType);
        ARecordType recType = (ARecordType) itemType;
        Index secondaryIndex = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataset.getDataverseName(), dataset.getDatasetName(), indexName);
        List<List<String>> secondaryKeyExprs = secondaryIndex.getKeyFieldNames();
        List<IAType> secondaryKeyTypes = secondaryIndex.getKeyFieldTypes();
        Pair<IAType, Boolean> keyPairType = Index.getNonNullableOpenFieldType(secondaryKeyTypes.get(0), secondaryKeyExprs.get(0), recType);
        IAType spatialType = keyPairType.first;
        int dimension = NonTaggedFormatUtil.getNumDimensions(spatialType.getTypeTag());
        int numSecondaryKeys = dimension * 2;
        int numPrimaryKeys = primaryKeys.size();
        int numKeys = numSecondaryKeys + numPrimaryKeys;
        int numFilterFields = DatasetUtil.getFilterField(dataset) == null ? 0 : 1;
        int[] fieldPermutation = new int[numKeys + numFilterFields];
        int[] modificationCallbackPrimaryKeyFields = new int[primaryKeys.size()];
        int i = 0;
        int j = 0;
        for (LogicalVariable varKey : secondaryKeys) {
            int idx = propagatedSchema.findVariable(varKey);
            fieldPermutation[i] = idx;
            i++;
        }
        for (LogicalVariable varKey : primaryKeys) {
            int idx = propagatedSchema.findVariable(varKey);
            fieldPermutation[i] = idx;
            modificationCallbackPrimaryKeyFields[j] = i;
            i++;
            j++;
        }
        if (numFilterFields > 0) {
            int idx = propagatedSchema.findVariable(additionalNonKeyFields.get(0));
            fieldPermutation[numKeys] = idx;
        }
        int[] prevFieldPermutation = null;
        if (indexOp == IndexOperation.UPSERT) {
            // Get field permutation for previous value
            prevFieldPermutation = new int[numKeys + numFilterFields];
            i = 0;
            // Get field permutation for new value
            for (LogicalVariable varKey : prevSecondaryKeys) {
                int idx = propagatedSchema.findVariable(varKey);
                prevFieldPermutation[i] = idx;
                i++;
            }
            for (int k = 0; k < numPrimaryKeys; k++) {
                prevFieldPermutation[k + i] = fieldPermutation[k + i];
                i++;
            }
            if (numFilterFields > 0) {
                int idx = propagatedSchema.findVariable(prevAdditionalFilteringKeys.get(0));
                prevFieldPermutation[numKeys] = idx;
            }
        }
        Pair<IFileSplitProvider, AlgebricksPartitionConstraint> splitsAndConstraint = getSplitProviderAndConstraints(dataset, secondaryIndex.getIndexName());
        // prepare callback
        JobId jobId = ((JobEventListenerFactory) spec.getJobletEventListenerFactory()).getJobId();
        IModificationOperationCallbackFactory modificationCallbackFactory = dataset.getModificationCallbackFactory(storaegComponentProvider, secondaryIndex, jobId, indexOp, modificationCallbackPrimaryKeyFields);
        IIndexDataflowHelperFactory indexDataflowHelperFactory = new IndexDataflowHelperFactory(storaegComponentProvider.getStorageManager(), splitsAndConstraint.first);
        IOperatorDescriptor op;
        if (bulkload) {
            long numElementsHint = getCardinalityPerPartitionHint(dataset);
            op = new TreeIndexBulkLoadOperatorDescriptor(spec, recordDesc, fieldPermutation, GlobalConfig.DEFAULT_TREE_FILL_FACTOR, false, numElementsHint, false, indexDataflowHelperFactory);
        } else if (indexOp == IndexOperation.UPSERT) {
            op = new LSMSecondaryUpsertOperatorDescriptor(spec, recordDesc, fieldPermutation, indexDataflowHelperFactory, filterFactory, modificationCallbackFactory, prevFieldPermutation);
        } else {
            op = new LSMTreeInsertDeleteOperatorDescriptor(spec, recordDesc, fieldPermutation, indexOp, indexDataflowHelperFactory, filterFactory, false, modificationCallbackFactory);
        }
        return new Pair<>(op, splitsAndConstraint.second);
    } catch (MetadataException e) {
        throw new AlgebricksException(e);
    }
}