Examples with Dataset org.apache.asterix.metadata.entities.Dataset used on opensource projects

Example 26 with Dataset

use of org.apache.asterix.metadata.entities.Dataset 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);
    }
}

Example 27 with Dataset

use of org.apache.asterix.metadata.entities.Dataset in project asterixdb by apache.

the class MetadataProvider method getTokenizerRuntime.

@Override
public Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> getTokenizerRuntime(IDataSourceIndex<String, DataSourceId> dataSourceIndex, IOperatorSchema propagatedSchema, IOperatorSchema[] inputSchemas, IVariableTypeEnvironment typeEnv, List<LogicalVariable> primaryKeys, List<LogicalVariable> secondaryKeys, ILogicalExpression filterExpr, RecordDescriptor recordDesc, JobGenContext context, JobSpecification spec, boolean bulkload) throws AlgebricksException {
    String indexName = dataSourceIndex.getId();
    String dataverseName = dataSourceIndex.getDataSource().getId().getDataverseName();
    String datasetName = dataSourceIndex.getDataSource().getId().getDatasourceName();
    IOperatorSchema inputSchema;
    if (inputSchemas.length > 0) {
        inputSchema = inputSchemas[0];
    } else {
        throw new AlgebricksException("TokenizeOperator can not operate without any input variable.");
    }
    Dataset dataset = MetadataManagerUtil.findExistingDataset(mdTxnCtx, dataverseName, datasetName);
    Index secondaryIndex;
    try {
        secondaryIndex = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataset.getDataverseName(), dataset.getDatasetName(), indexName);
    } catch (MetadataException e) {
        throw new AlgebricksException(e);
    }
    // TokenizeOperator only supports a keyword or n-gram index.
    switch(secondaryIndex.getIndexType()) {
        case SINGLE_PARTITION_WORD_INVIX:
        case SINGLE_PARTITION_NGRAM_INVIX:
        case LENGTH_PARTITIONED_WORD_INVIX:
        case LENGTH_PARTITIONED_NGRAM_INVIX:
            return getBinaryTokenizerRuntime(dataverseName, datasetName, indexName, inputSchema, propagatedSchema, primaryKeys, secondaryKeys, recordDesc, spec, secondaryIndex.getIndexType());
        default:
            throw new AlgebricksException("Currently, we do not support TokenizeOperator for the index type: " + secondaryIndex.getIndexType());
    }
}

Example 28 with Dataset

use of org.apache.asterix.metadata.entities.Dataset in project asterixdb by apache.

the class MetadataProvider method getWriteResultRuntime.

@Override
public Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> getWriteResultRuntime(IDataSource<DataSourceId> dataSource, IOperatorSchema propagatedSchema, List<LogicalVariable> keys, LogicalVariable payload, List<LogicalVariable> additionalNonKeyFields, JobGenContext context, JobSpecification spec) throws AlgebricksException {
    String dataverseName = dataSource.getId().getDataverseName();
    String datasetName = dataSource.getId().getDatasourceName();
    Dataset dataset = MetadataManagerUtil.findExistingDataset(mdTxnCtx, dataverseName, datasetName);
    int numKeys = keys.size();
    int numFilterFields = DatasetUtil.getFilterField(dataset) == null ? 0 : 1;
    // move key fields to front
    int[] fieldPermutation = new int[numKeys + 1 + numFilterFields];
    int i = 0;
    for (LogicalVariable varKey : keys) {
        int idx = propagatedSchema.findVariable(varKey);
        fieldPermutation[i] = idx;
        i++;
    }
    fieldPermutation[numKeys] = propagatedSchema.findVariable(payload);
    if (numFilterFields > 0) {
        int idx = propagatedSchema.findVariable(additionalNonKeyFields.get(0));
        fieldPermutation[numKeys + 1] = idx;
    }
    try {
        boolean temp = dataset.getDatasetDetails().isTemp();
        isTemporaryDatasetWriteJob = isTemporaryDatasetWriteJob && temp;
        Pair<IFileSplitProvider, AlgebricksPartitionConstraint> splitsAndConstraint = getSplitProviderAndConstraints(dataset);
        long numElementsHint = getCardinalityPerPartitionHint(dataset);
        // TODO
        // figure out the right behavior of the bulkload and then give the
        // right callback
        // (ex. what's the expected behavior when there is an error during
        // bulkload?)
        IIndexDataflowHelperFactory indexHelperFactory = new IndexDataflowHelperFactory(storaegComponentProvider.getStorageManager(), splitsAndConstraint.first);
        TreeIndexBulkLoadOperatorDescriptor btreeBulkLoad = new TreeIndexBulkLoadOperatorDescriptor(spec, null, fieldPermutation, GlobalConfig.DEFAULT_TREE_FILL_FACTOR, false, numElementsHint, true, indexHelperFactory);
        return new Pair<>(btreeBulkLoad, splitsAndConstraint.second);
    } catch (MetadataException me) {
        throw new AlgebricksException(me);
    }
}

Example 29 with Dataset

use of org.apache.asterix.metadata.entities.Dataset in project asterixdb by apache.

the class MetadataProvider method getBinaryTokenizerRuntime.

// Get a Tokenizer for the bulk-loading data into a n-gram or keyword index.
private Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> getBinaryTokenizerRuntime(String dataverseName, String datasetName, String indexName, IOperatorSchema inputSchema, IOperatorSchema propagatedSchema, List<LogicalVariable> primaryKeys, List<LogicalVariable> secondaryKeys, RecordDescriptor recordDesc, JobSpecification spec, IndexType indexType) throws AlgebricksException {
    // Sanity checks.
    if (primaryKeys.size() > 1) {
        throw new AlgebricksException("Cannot tokenize composite primary key.");
    }
    if (secondaryKeys.size() > 1) {
        throw new AlgebricksException("Cannot tokenize composite secondary key fields.");
    }
    boolean isPartitioned;
    if (indexType == IndexType.LENGTH_PARTITIONED_WORD_INVIX || indexType == IndexType.LENGTH_PARTITIONED_NGRAM_INVIX) {
        isPartitioned = true;
    } else {
        isPartitioned = false;
    }
    // Number of Keys that needs to be propagated
    int numKeys = inputSchema.getSize();
    // Get the rest of Logical Variables that are not (PK or SK) and each
    // variable's positions.
    // These variables will be propagated through TokenizeOperator.
    List<LogicalVariable> otherKeys = new ArrayList<>();
    if (inputSchema.getSize() > 0) {
        for (int k = 0; k < inputSchema.getSize(); k++) {
            boolean found = false;
            for (LogicalVariable varKey : primaryKeys) {
                if (varKey.equals(inputSchema.getVariable(k))) {
                    found = true;
                    break;
                } else {
                    found = false;
                }
            }
            if (!found) {
                for (LogicalVariable varKey : secondaryKeys) {
                    if (varKey.equals(inputSchema.getVariable(k))) {
                        found = true;
                        break;
                    } else {
                        found = false;
                    }
                }
            }
            if (!found) {
                otherKeys.add(inputSchema.getVariable(k));
            }
        }
    }
    // For tokenization, sorting and loading.
    // One token (+ optional partitioning field) + primary keys + secondary
    // keys + other variables
    // secondary keys and other variables will be just passed to the
    // IndexInsertDelete Operator.
    int numTokenKeyPairFields = (!isPartitioned) ? 1 + numKeys : 2 + numKeys;
    // generate field permutations for the input
    int[] fieldPermutation = new int[numKeys];
    int[] modificationCallbackPrimaryKeyFields = new int[primaryKeys.size()];
    int i = 0;
    int j = 0;
    for (LogicalVariable varKey : primaryKeys) {
        int idx = propagatedSchema.findVariable(varKey);
        fieldPermutation[i] = idx;
        modificationCallbackPrimaryKeyFields[j] = i;
        i++;
        j++;
    }
    for (LogicalVariable varKey : otherKeys) {
        int idx = propagatedSchema.findVariable(varKey);
        fieldPermutation[i] = idx;
        i++;
    }
    for (LogicalVariable varKey : secondaryKeys) {
        int idx = propagatedSchema.findVariable(varKey);
        fieldPermutation[i] = idx;
        i++;
    }
    Dataset dataset = MetadataManagerUtil.findExistingDataset(mdTxnCtx, dataverseName, datasetName);
    String itemTypeName = dataset.getItemTypeName();
    IAType itemType;
    try {
        itemType = MetadataManager.INSTANCE.getDatatype(mdTxnCtx, dataset.getItemTypeDataverseName(), itemTypeName).getDatatype();
        if (itemType.getTypeTag() != ATypeTag.OBJECT) {
            throw new AlgebricksException("Only record types can be tokenized.");
        }
        ARecordType recType = (ARecordType) itemType;
        // Index parameters.
        Index secondaryIndex = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataset.getDataverseName(), dataset.getDatasetName(), indexName);
        List<List<String>> secondaryKeyExprs = secondaryIndex.getKeyFieldNames();
        List<IAType> secondaryKeyTypeEntries = secondaryIndex.getKeyFieldTypes();
        int numTokenFields = (!isPartitioned) ? secondaryKeys.size() : secondaryKeys.size() + 1;
        ITypeTraits[] tokenTypeTraits = new ITypeTraits[numTokenFields];
        ITypeTraits[] invListsTypeTraits = new ITypeTraits[primaryKeys.size()];
        // Find the key type of the secondary key. If it's a derived type,
        // return the derived type.
        // e.g. UNORDERED LIST -> return UNORDERED LIST type
        IAType secondaryKeyType;
        Pair<IAType, Boolean> keyPairType = Index.getNonNullableOpenFieldType(secondaryKeyTypeEntries.get(0), secondaryKeyExprs.get(0), recType);
        secondaryKeyType = keyPairType.first;
        List<List<String>> partitioningKeys = dataset.getPrimaryKeys();
        i = 0;
        for (List<String> partitioningKey : partitioningKeys) {
            IAType keyType = recType.getSubFieldType(partitioningKey);
            invListsTypeTraits[i] = TypeTraitProvider.INSTANCE.getTypeTrait(keyType);
            ++i;
        }
        tokenTypeTraits[0] = NonTaggedFormatUtil.getTokenTypeTrait(secondaryKeyType);
        if (isPartitioned) {
            // The partitioning field is hardcoded to be a short *without*
            // an Asterix type tag.
            tokenTypeTraits[1] = ShortPointable.TYPE_TRAITS;
        }
        IBinaryTokenizerFactory tokenizerFactory = NonTaggedFormatUtil.getBinaryTokenizerFactory(secondaryKeyType.getTypeTag(), indexType, secondaryIndex.getGramLength());
        Pair<IFileSplitProvider, AlgebricksPartitionConstraint> splitsAndConstraint = getSplitProviderAndConstraints(dataset, secondaryIndex.getIndexName());
        // Generate Output Record format
        ISerializerDeserializer<?>[] tokenKeyPairFields = new ISerializerDeserializer[numTokenKeyPairFields];
        ITypeTraits[] tokenKeyPairTypeTraits = new ITypeTraits[numTokenKeyPairFields];
        ISerializerDeserializerProvider serdeProvider = FormatUtils.getDefaultFormat().getSerdeProvider();
        // #1. propagate all input variables
        for (int k = 0; k < recordDesc.getFieldCount(); k++) {
            tokenKeyPairFields[k] = recordDesc.getFields()[k];
            tokenKeyPairTypeTraits[k] = recordDesc.getTypeTraits()[k];
        }
        int tokenOffset = recordDesc.getFieldCount();
        // #2. Specify the token type
        tokenKeyPairFields[tokenOffset] = serdeProvider.getSerializerDeserializer(secondaryKeyType);
        tokenKeyPairTypeTraits[tokenOffset] = tokenTypeTraits[0];
        tokenOffset++;
        // #3. Specify the length-partitioning key: number of token
        if (isPartitioned) {
            tokenKeyPairFields[tokenOffset] = ShortSerializerDeserializer.INSTANCE;
            tokenKeyPairTypeTraits[tokenOffset] = tokenTypeTraits[1];
        }
        RecordDescriptor tokenKeyPairRecDesc = new RecordDescriptor(tokenKeyPairFields, tokenKeyPairTypeTraits);
        IOperatorDescriptor tokenizerOp;
        // Keys to be tokenized : SK
        int docField = fieldPermutation[fieldPermutation.length - 1];
        // Keys to be propagated
        int[] keyFields = new int[numKeys];
        for (int k = 0; k < keyFields.length; k++) {
            keyFields[k] = k;
        }
        tokenizerOp = new BinaryTokenizerOperatorDescriptor(spec, tokenKeyPairRecDesc, tokenizerFactory, docField, keyFields, isPartitioned, true);
        return new Pair<>(tokenizerOp, splitsAndConstraint.second);
    } catch (Exception e) {
        throw new AlgebricksException(e);
    }
}

Example 30 with Dataset

use of org.apache.asterix.metadata.entities.Dataset in project asterixdb by apache.

the class MetadataProvider method getInsertOrDeleteRuntime.

private Pair<IOperatorDescriptor, AlgebricksPartitionConstraint> getInsertOrDeleteRuntime(IndexOperation indexOp, IDataSource<DataSourceId> dataSource, IOperatorSchema propagatedSchema, List<LogicalVariable> keys, LogicalVariable payload, List<LogicalVariable> additionalNonKeyFields, RecordDescriptor inputRecordDesc, JobGenContext context, JobSpecification spec, boolean bulkload, List<LogicalVariable> additionalNonFilteringFields) throws AlgebricksException {
    String datasetName = dataSource.getId().getDatasourceName();
    Dataset dataset = MetadataManagerUtil.findExistingDataset(mdTxnCtx, dataSource.getId().getDataverseName(), datasetName);
    boolean temp = dataset.getDatasetDetails().isTemp();
    isTemporaryDatasetWriteJob = isTemporaryDatasetWriteJob && temp;
    int numKeys = keys.size();
    int numFilterFields = DatasetUtil.getFilterField(dataset) == null ? 0 : 1;
    // Move key fields to front.
    int[] fieldPermutation = new int[numKeys + 1 + numFilterFields + (additionalNonFilteringFields == null ? 0 : additionalNonFilteringFields.size())];
    int[] bloomFilterKeyFields = new int[numKeys];
    int i = 0;
    for (LogicalVariable varKey : keys) {
        int idx = propagatedSchema.findVariable(varKey);
        fieldPermutation[i] = idx;
        bloomFilterKeyFields[i] = i;
        i++;
    }
    fieldPermutation[i++] = propagatedSchema.findVariable(payload);
    if (numFilterFields > 0) {
        int idx = propagatedSchema.findVariable(additionalNonKeyFields.get(0));
        fieldPermutation[i++] = idx;
    }
    if (additionalNonFilteringFields != null) {
        for (LogicalVariable variable : additionalNonFilteringFields) {
            int idx = propagatedSchema.findVariable(variable);
            fieldPermutation[i++] = idx;
        }
    }
    try {
        Index primaryIndex = MetadataManager.INSTANCE.getIndex(mdTxnCtx, dataset.getDataverseName(), dataset.getDatasetName(), dataset.getDatasetName());
        Pair<IFileSplitProvider, AlgebricksPartitionConstraint> splitsAndConstraint = getSplitProviderAndConstraints(dataset);
        // prepare callback
        int[] primaryKeyFields = new int[numKeys];
        for (i = 0; i < numKeys; i++) {
            primaryKeyFields[i] = i;
        }
        IModificationOperationCallbackFactory modificationCallbackFactory = dataset.getModificationCallbackFactory(storaegComponentProvider, primaryIndex, jobId, indexOp, primaryKeyFields);
        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, true, numElementsHint, true, idfh);
        } else {
            op = new LSMTreeInsertDeleteOperatorDescriptor(spec, inputRecordDesc, fieldPermutation, indexOp, idfh, null, true, modificationCallbackFactory);
        }
        return new Pair<>(op, splitsAndConstraint.second);
    } catch (MetadataException me) {
        throw new AlgebricksException(me);
    }
}