Search in sources :

Example 6 with ProjectionMeasure

use of org.apache.carbondata.core.scan.model.ProjectionMeasure in project carbondata by apache.

the class AbstractScannedResultCollector method fillMeasureData.

protected void fillMeasureData(Object[] msrValues, int offset, BlockletScannedResult scannedResult) {
    int measureExistIndex = 0;
    for (short i = 0; i < measureInfo.getMeasureDataTypes().length; i++) {
        // data chunk to the collector
        if (measureInfo.getMeasureExists()[i]) {
            ProjectionMeasure queryMeasure = executionInfo.getProjectionMeasures()[measureExistIndex];
            msrValues[i + offset] = getMeasureData(scannedResult.getMeasureChunk(measureInfo.getMeasureOrdinals()[measureExistIndex]), scannedResult.getCurrentRowId(), queryMeasure.getMeasure());
            measureExistIndex++;
        } else {
            // if not then get the default value and use that value in aggregation
            Object defaultValue = measureInfo.getDefaultValues()[i];
            if (null != defaultValue && DataTypes.isDecimal(measureInfo.getMeasureDataTypes()[i])) {
                // convert data type as per the computing engine
                defaultValue = DataTypeUtil.getDataTypeConverter().convertFromBigDecimalToDecimal(defaultValue);
            }
            msrValues[i + offset] = defaultValue;
        }
    }
}
Also used : ProjectionMeasure(org.apache.carbondata.core.scan.model.ProjectionMeasure)

Example 7 with ProjectionMeasure

use of org.apache.carbondata.core.scan.model.ProjectionMeasure in project carbondata by apache.

the class RestructureUtil method createMeasureInfoAndGetCurrentBlockQueryMeasures.

/**
 * Below method will be used to prepare the measure info object
 * in this method some of the properties which will be extracted
 * from query measure and current block measures will be set
 *
 * @param blockExecutionInfo
 * @param queryMeasures        measures present in query
 * @param currentBlockMeasures current block measures
 * @return measures present in the block
 */
public static List<ProjectionMeasure> createMeasureInfoAndGetCurrentBlockQueryMeasures(BlockExecutionInfo blockExecutionInfo, List<ProjectionMeasure> queryMeasures, List<CarbonMeasure> currentBlockMeasures) {
    MeasureInfo measureInfo = new MeasureInfo();
    List<ProjectionMeasure> presentMeasure = new ArrayList<>(queryMeasures.size());
    int numberOfMeasureInQuery = queryMeasures.size();
    List<Integer> measureOrdinalList = new ArrayList<>(numberOfMeasureInQuery);
    Object[] defaultValues = new Object[numberOfMeasureInQuery];
    boolean[] measureExistsInCurrentBlock = new boolean[numberOfMeasureInQuery];
    int index = 0;
    for (ProjectionMeasure queryMeasure : queryMeasures) {
        // otherwise adding a default value of a measure
        for (CarbonMeasure carbonMeasure : currentBlockMeasures) {
            if (carbonMeasure.getColumnId().equals(queryMeasure.getMeasure().getColumnId())) {
                ProjectionMeasure currentBlockMeasure = new ProjectionMeasure(carbonMeasure);
                carbonMeasure.getColumnSchema().setDataType(queryMeasure.getMeasure().getDataType());
                carbonMeasure.getColumnSchema().setPrecision(queryMeasure.getMeasure().getPrecision());
                carbonMeasure.getColumnSchema().setScale(queryMeasure.getMeasure().getScale());
                carbonMeasure.getColumnSchema().setDefaultValue(queryMeasure.getMeasure().getDefaultValue());
                currentBlockMeasure.setOrdinal(queryMeasure.getOrdinal());
                presentMeasure.add(currentBlockMeasure);
                measureOrdinalList.add(carbonMeasure.getOrdinal());
                measureExistsInCurrentBlock[index] = true;
                break;
            }
        }
        if (!measureExistsInCurrentBlock[index]) {
            defaultValues[index] = getMeasureDefaultValue(queryMeasure.getMeasure().getColumnSchema(), queryMeasure.getMeasure().getDefaultValue());
            blockExecutionInfo.setRestructuredBlock(true);
        }
        index++;
    }
    int[] measureOrdinals = ArrayUtils.toPrimitive(measureOrdinalList.toArray(new Integer[measureOrdinalList.size()]));
    measureInfo.setDefaultValues(defaultValues);
    measureInfo.setMeasureOrdinals(measureOrdinals);
    measureInfo.setMeasureExists(measureExistsInCurrentBlock);
    blockExecutionInfo.setMeasureInfo(measureInfo);
    return presentMeasure;
}
Also used : MeasureInfo(org.apache.carbondata.core.scan.executor.infos.MeasureInfo) CarbonMeasure(org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure) ProjectionMeasure(org.apache.carbondata.core.scan.model.ProjectionMeasure) ArrayList(java.util.ArrayList)

Example 8 with ProjectionMeasure

use of org.apache.carbondata.core.scan.model.ProjectionMeasure in project carbondata by apache.

the class AbstractQueryExecutor method getBlockExecutionInfoForBlock.

/**
 * Below method will be used to get the block execution info which is
 * required to execute any block  based on query model
 *
 * @param queryModel query model from user query
 * @param blockIndex block index
 * @return block execution info
 * @throws QueryExecutionException any failure during block info creation
 */
private BlockExecutionInfo getBlockExecutionInfoForBlock(QueryModel queryModel, AbstractIndex blockIndex, int startBlockletIndex, int numberOfBlockletToScan, String filePath, String[] deleteDeltaFiles, String segmentId) throws QueryExecutionException {
    BlockExecutionInfo blockExecutionInfo = new BlockExecutionInfo();
    SegmentProperties segmentProperties = blockIndex.getSegmentProperties();
    List<CarbonDimension> tableBlockDimensions = segmentProperties.getDimensions();
    // below is to get only those dimension in query which is present in the
    // table block
    List<ProjectionDimension> projectDimensions = RestructureUtil.createDimensionInfoAndGetCurrentBlockQueryDimension(blockExecutionInfo, queryModel.getProjectionDimensions(), tableBlockDimensions, segmentProperties.getComplexDimensions(), queryModel.getProjectionMeasures().size());
    blockExecutionInfo.setBlockId(CarbonUtil.getBlockId(queryModel.getAbsoluteTableIdentifier(), filePath, segmentId));
    blockExecutionInfo.setDeleteDeltaFilePath(deleteDeltaFiles);
    blockExecutionInfo.setStartBlockletIndex(startBlockletIndex);
    blockExecutionInfo.setNumberOfBlockletToScan(numberOfBlockletToScan);
    blockExecutionInfo.setProjectionDimensions(projectDimensions.toArray(new ProjectionDimension[projectDimensions.size()]));
    // get measures present in the current block
    List<ProjectionMeasure> currentBlockQueryMeasures = getCurrentBlockQueryMeasures(blockExecutionInfo, queryModel, blockIndex);
    blockExecutionInfo.setProjectionMeasures(currentBlockQueryMeasures.toArray(new ProjectionMeasure[currentBlockQueryMeasures.size()]));
    blockExecutionInfo.setDataBlock(blockIndex);
    // setting whether raw record query or not
    blockExecutionInfo.setRawRecordDetailQuery(queryModel.isForcedDetailRawQuery());
    // total number dimension
    blockExecutionInfo.setTotalNumberDimensionToRead(segmentProperties.getDimensionOrdinalToChunkMapping().size());
    blockExecutionInfo.setPrefetchBlocklet(!queryModel.isReadPageByPage());
    blockExecutionInfo.setTotalNumberOfMeasureToRead(segmentProperties.getMeasuresOrdinalToChunkMapping().size());
    blockExecutionInfo.setComplexDimensionInfoMap(QueryUtil.getComplexDimensionsMap(projectDimensions, segmentProperties.getDimensionOrdinalToChunkMapping(), segmentProperties.getEachComplexDimColumnValueSize(), queryProperties.columnToDictionaryMapping, queryProperties.complexFilterDimension));
    IndexKey startIndexKey = null;
    IndexKey endIndexKey = null;
    if (null != queryModel.getFilterExpressionResolverTree()) {
        // loading the filter executor tree for filter evaluation
        blockExecutionInfo.setFilterExecuterTree(FilterUtil.getFilterExecuterTree(queryModel.getFilterExpressionResolverTree(), segmentProperties, blockExecutionInfo.getComlexDimensionInfoMap()));
    }
    try {
        startIndexKey = FilterUtil.prepareDefaultStartIndexKey(segmentProperties);
        endIndexKey = FilterUtil.prepareDefaultEndIndexKey(segmentProperties);
    } catch (KeyGenException e) {
        throw new QueryExecutionException(e);
    }
    // setting the start index key of the block node
    blockExecutionInfo.setStartKey(startIndexKey);
    // setting the end index key of the block node
    blockExecutionInfo.setEndKey(endIndexKey);
    // expression dimensions
    List<CarbonDimension> expressionDimensions = new ArrayList<CarbonDimension>(CarbonCommonConstants.DEFAULT_COLLECTION_SIZE);
    // expression measure
    List<CarbonMeasure> expressionMeasures = new ArrayList<CarbonMeasure>(CarbonCommonConstants.DEFAULT_COLLECTION_SIZE);
    // setting all the dimension chunk indexes to be read from file
    int numberOfElementToConsider = 0;
    // list of dimensions to be projected
    Set<Integer> allProjectionListDimensionIdexes = new LinkedHashSet<>();
    // create a list of filter dimensions present in the current block
    Set<CarbonDimension> currentBlockFilterDimensions = getCurrentBlockFilterDimensions(queryProperties.complexFilterDimension, segmentProperties);
    int[] dimensionChunkIndexes = QueryUtil.getDimensionChunkIndexes(projectDimensions, segmentProperties.getDimensionOrdinalToChunkMapping(), expressionDimensions, currentBlockFilterDimensions, allProjectionListDimensionIdexes);
    int numberOfColumnToBeReadInOneIO = Integer.parseInt(CarbonProperties.getInstance().getProperty(CarbonV3DataFormatConstants.NUMBER_OF_COLUMN_TO_READ_IN_IO, CarbonV3DataFormatConstants.NUMBER_OF_COLUMN_TO_READ_IN_IO_DEFAULTVALUE));
    if (dimensionChunkIndexes.length > 0) {
        numberOfElementToConsider = dimensionChunkIndexes[dimensionChunkIndexes.length - 1] == segmentProperties.getBlockTodimensionOrdinalMapping().size() - 1 ? dimensionChunkIndexes.length - 1 : dimensionChunkIndexes.length;
        blockExecutionInfo.setAllSelectedDimensionColumnIndexRange(CarbonUtil.getRangeIndex(dimensionChunkIndexes, numberOfElementToConsider, numberOfColumnToBeReadInOneIO));
    } else {
        blockExecutionInfo.setAllSelectedDimensionColumnIndexRange(new int[0][0]);
    }
    // get the list of updated filter measures present in the current block
    Set<CarbonMeasure> filterMeasures = getCurrentBlockFilterMeasures(queryProperties.filterMeasures, segmentProperties);
    // list of measures to be projected
    List<Integer> allProjectionListMeasureIndexes = new ArrayList<>();
    int[] measureChunkIndexes = QueryUtil.getMeasureChunkIndexes(currentBlockQueryMeasures, expressionMeasures, segmentProperties.getMeasuresOrdinalToChunkMapping(), filterMeasures, allProjectionListMeasureIndexes);
    if (measureChunkIndexes.length > 0) {
        numberOfElementToConsider = measureChunkIndexes[measureChunkIndexes.length - 1] == segmentProperties.getMeasures().size() - 1 ? measureChunkIndexes.length - 1 : measureChunkIndexes.length;
        // setting all the measure chunk indexes to be read from file
        blockExecutionInfo.setAllSelectedMeasureIndexRange(CarbonUtil.getRangeIndex(measureChunkIndexes, numberOfElementToConsider, numberOfColumnToBeReadInOneIO));
    } else {
        blockExecutionInfo.setAllSelectedMeasureIndexRange(new int[0][0]);
    }
    // setting the indexes of list of dimension in projection list
    blockExecutionInfo.setProjectionListDimensionIndexes(ArrayUtils.toPrimitive(allProjectionListDimensionIdexes.toArray(new Integer[allProjectionListDimensionIdexes.size()])));
    // setting the indexes of list of measures in projection list
    blockExecutionInfo.setProjectionListMeasureIndexes(ArrayUtils.toPrimitive(allProjectionListMeasureIndexes.toArray(new Integer[allProjectionListMeasureIndexes.size()])));
    // setting the size of fixed key column (dictionary column)
    blockExecutionInfo.setFixedLengthKeySize(getKeySize(projectDimensions, segmentProperties));
    Set<Integer> dictionaryColumnChunkIndex = new HashSet<Integer>();
    List<Integer> noDictionaryColumnChunkIndex = new ArrayList<Integer>();
    // get the block index to be read from file for query dimension
    // for both dictionary columns and no dictionary columns
    QueryUtil.fillQueryDimensionChunkIndexes(projectDimensions, segmentProperties.getDimensionOrdinalToChunkMapping(), dictionaryColumnChunkIndex, noDictionaryColumnChunkIndex);
    int[] queryDictionaryColumnChunkIndexes = ArrayUtils.toPrimitive(dictionaryColumnChunkIndex.toArray(new Integer[dictionaryColumnChunkIndex.size()]));
    // need to sort the dictionary column as for all dimension
    // column key will be filled based on key order
    Arrays.sort(queryDictionaryColumnChunkIndexes);
    blockExecutionInfo.setDictionaryColumnChunkIndex(queryDictionaryColumnChunkIndexes);
    // setting the no dictionary column block indexes
    blockExecutionInfo.setNoDictionaryColumnChunkIndexes(ArrayUtils.toPrimitive(noDictionaryColumnChunkIndex.toArray(new Integer[noDictionaryColumnChunkIndex.size()])));
    // setting each column value size
    blockExecutionInfo.setEachColumnValueSize(segmentProperties.getEachDimColumnValueSize());
    blockExecutionInfo.setComplexColumnParentBlockIndexes(getComplexDimensionParentBlockIndexes(projectDimensions));
    blockExecutionInfo.setVectorBatchCollector(queryModel.isVectorReader());
    try {
        // to set column group and its key structure info which will be used
        // to
        // for getting the column group column data in case of final row
        // and in case of dimension aggregation
        blockExecutionInfo.setColumnGroupToKeyStructureInfo(QueryUtil.getColumnGroupKeyStructureInfo(projectDimensions, segmentProperties));
    } catch (KeyGenException e) {
        throw new QueryExecutionException(e);
    }
    // set actual query dimensions and measures. It may differ in case of restructure scenarios
    blockExecutionInfo.setActualQueryDimensions(queryModel.getProjectionDimensions().toArray(new ProjectionDimension[queryModel.getProjectionDimensions().size()]));
    blockExecutionInfo.setActualQueryMeasures(queryModel.getProjectionMeasures().toArray(new ProjectionMeasure[queryModel.getProjectionMeasures().size()]));
    DataTypeUtil.setDataTypeConverter(queryModel.getConverter());
    return blockExecutionInfo;
}
Also used : LinkedHashSet(java.util.LinkedHashSet) IndexKey(org.apache.carbondata.core.datastore.IndexKey) ArrayList(java.util.ArrayList) ProjectionDimension(org.apache.carbondata.core.scan.model.ProjectionDimension) CarbonDimension(org.apache.carbondata.core.metadata.schema.table.column.CarbonDimension) QueryExecutionException(org.apache.carbondata.core.scan.executor.exception.QueryExecutionException) CarbonMeasure(org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure) ProjectionMeasure(org.apache.carbondata.core.scan.model.ProjectionMeasure) KeyGenException(org.apache.carbondata.core.keygenerator.KeyGenException) BlockExecutionInfo(org.apache.carbondata.core.scan.executor.infos.BlockExecutionInfo) SegmentProperties(org.apache.carbondata.core.datastore.block.SegmentProperties) HashSet(java.util.HashSet) LinkedHashSet(java.util.LinkedHashSet)

Example 9 with ProjectionMeasure

use of org.apache.carbondata.core.scan.model.ProjectionMeasure in project carbondata by apache.

the class PrestoCarbonVectorizedRecordReader method initBatch.

/**
 * Returns the ColumnarBatch object that will be used for all rows returned by this reader.
 * This object is reused. Calling this enables the vectorized reader. This should be called
 * before any calls to nextKeyValue/nextBatch.
 */
private void initBatch() {
    List<ProjectionDimension> queryDimension = queryModel.getProjectionDimensions();
    List<ProjectionMeasure> queryMeasures = queryModel.getProjectionMeasures();
    StructField[] fields = new StructField[queryDimension.size() + queryMeasures.size()];
    for (int i = 0; i < queryDimension.size(); i++) {
        ProjectionDimension dim = queryDimension.get(i);
        if (dim.getDimension().hasEncoding(Encoding.DIRECT_DICTIONARY)) {
            DirectDictionaryGenerator generator = DirectDictionaryKeyGeneratorFactory.getDirectDictionaryGenerator(dim.getDimension().getDataType());
            fields[dim.getOrdinal()] = new StructField(dim.getColumnName(), generator.getReturnType());
        } else if (!dim.getDimension().hasEncoding(Encoding.DICTIONARY)) {
            fields[dim.getOrdinal()] = new StructField(dim.getColumnName(), dim.getDimension().getDataType());
        } else if (dim.getDimension().isComplex()) {
            fields[dim.getOrdinal()] = new StructField(dim.getColumnName(), dim.getDimension().getDataType());
        } else {
            fields[dim.getOrdinal()] = new StructField(dim.getColumnName(), DataTypes.INT);
        }
    }
    for (ProjectionMeasure msr : queryMeasures) {
        DataType dataType = msr.getMeasure().getDataType();
        if (dataType == DataTypes.BOOLEAN || dataType == DataTypes.SHORT || dataType == DataTypes.INT || dataType == DataTypes.LONG) {
            fields[msr.getOrdinal()] = new StructField(msr.getColumnName(), msr.getMeasure().getDataType());
        } else if (DataTypes.isDecimal(dataType)) {
            fields[msr.getOrdinal()] = new StructField(msr.getColumnName(), msr.getMeasure().getDataType());
        } else {
            fields[msr.getOrdinal()] = new StructField(msr.getColumnName(), DataTypes.DOUBLE);
        }
    }
    columnarBatch = CarbonVectorBatch.allocate(fields);
    CarbonColumnVector[] vectors = new CarbonColumnVector[fields.length];
    boolean[] filteredRows = new boolean[columnarBatch.capacity()];
    for (int i = 0; i < fields.length; i++) {
        vectors[i] = new CarbonColumnVectorWrapper(columnarBatch.column(i), filteredRows);
    }
    carbonColumnarBatch = new CarbonColumnarBatch(vectors, columnarBatch.capacity(), filteredRows);
}
Also used : CarbonColumnarBatch(org.apache.carbondata.core.scan.result.vector.CarbonColumnarBatch) CarbonColumnVector(org.apache.carbondata.core.scan.result.vector.CarbonColumnVector) ProjectionDimension(org.apache.carbondata.core.scan.model.ProjectionDimension) StructField(org.apache.carbondata.core.metadata.datatype.StructField) ProjectionMeasure(org.apache.carbondata.core.scan.model.ProjectionMeasure) DataType(org.apache.carbondata.core.metadata.datatype.DataType) DirectDictionaryGenerator(org.apache.carbondata.core.keygenerator.directdictionary.DirectDictionaryGenerator)

Aggregations

ProjectionMeasure (org.apache.carbondata.core.scan.model.ProjectionMeasure)9 ArrayList (java.util.ArrayList)4 CarbonMeasure (org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure)4 ProjectionDimension (org.apache.carbondata.core.scan.model.ProjectionDimension)4 DataType (org.apache.carbondata.core.metadata.datatype.DataType)3 BlockExecutionInfo (org.apache.carbondata.core.scan.executor.infos.BlockExecutionInfo)3 DirectDictionaryGenerator (org.apache.carbondata.core.keygenerator.directdictionary.DirectDictionaryGenerator)2 CarbonDimension (org.apache.carbondata.core.metadata.schema.table.column.CarbonDimension)2 ColumnSchema (org.apache.carbondata.core.metadata.schema.table.column.ColumnSchema)2 MeasureInfo (org.apache.carbondata.core.scan.executor.infos.MeasureInfo)2 CarbonColumnVector (org.apache.carbondata.core.scan.result.vector.CarbonColumnVector)2 CarbonColumnarBatch (org.apache.carbondata.core.scan.result.vector.CarbonColumnarBatch)2 Test (org.junit.Test)2 HashSet (java.util.HashSet)1 LinkedHashMap (java.util.LinkedHashMap)1 LinkedHashSet (java.util.LinkedHashSet)1 List (java.util.List)1 CacheProvider (org.apache.carbondata.core.cache.CacheProvider)1 BlockIndexStore (org.apache.carbondata.core.datastore.BlockIndexStore)1 IndexKey (org.apache.carbondata.core.datastore.IndexKey)1