Search in sources :

Example 1 with CarbonMeasure

use of org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure 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
   */
protected BlockExecutionInfo getBlockExecutionInfoForBlock(QueryModel queryModel, AbstractIndex blockIndex, int startBlockletIndex, int numberOfBlockletToScan, String filePath) throws QueryExecutionException {
    BlockExecutionInfo blockExecutionInfo = new BlockExecutionInfo();
    SegmentProperties segmentProperties = blockIndex.getSegmentProperties();
    List<CarbonDimension> tableBlockDimensions = segmentProperties.getDimensions();
    KeyGenerator blockKeyGenerator = segmentProperties.getDimensionKeyGenerator();
    // below is to get only those dimension in query which is present in the
    // table block
    List<QueryDimension> currentBlockQueryDimensions = RestructureUtil.createDimensionInfoAndGetCurrentBlockQueryDimension(blockExecutionInfo, queryModel.getQueryDimension(), tableBlockDimensions, segmentProperties.getComplexDimensions());
    int tableFactPathLength = CarbonStorePath.getCarbonTablePath(queryModel.getAbsoluteTableIdentifier().getStorePath(), queryModel.getAbsoluteTableIdentifier().getCarbonTableIdentifier()).getFactDir().length() + 1;
    blockExecutionInfo.setBlockId(filePath.substring(tableFactPathLength));
    blockExecutionInfo.setStartBlockletIndex(startBlockletIndex);
    blockExecutionInfo.setNumberOfBlockletToScan(numberOfBlockletToScan);
    blockExecutionInfo.setQueryDimensions(currentBlockQueryDimensions.toArray(new QueryDimension[currentBlockQueryDimensions.size()]));
    // get measures present in the current block
    List<QueryMeasure> currentBlockQueryMeasures = getCurrentBlockQueryMeasures(blockExecutionInfo, queryModel, blockIndex);
    blockExecutionInfo.setQueryMeasures(currentBlockQueryMeasures.toArray(new QueryMeasure[currentBlockQueryMeasures.size()]));
    blockExecutionInfo.setDataBlock(blockIndex);
    blockExecutionInfo.setBlockKeyGenerator(blockKeyGenerator);
    // setting whether raw record query or not
    blockExecutionInfo.setRawRecordDetailQuery(queryModel.isForcedDetailRawQuery());
    // total number dimension
    blockExecutionInfo.setTotalNumberDimensionBlock(segmentProperties.getDimensionOrdinalToBlockMapping().size());
    blockExecutionInfo.setTotalNumberOfMeasureBlock(segmentProperties.getMeasuresOrdinalToBlockMapping().size());
    blockExecutionInfo.setAbsoluteTableIdentifier(queryModel.getAbsoluteTableIdentifier());
    blockExecutionInfo.setComplexDimensionInfoMap(QueryUtil.getComplexDimensionsMap(currentBlockQueryDimensions, segmentProperties.getDimensionOrdinalToBlockMapping(), segmentProperties.getEachComplexDimColumnValueSize(), queryProperties.columnToDictionayMapping, queryProperties.complexFilterDimension));
    IndexKey startIndexKey = null;
    IndexKey endIndexKey = null;
    if (null != queryModel.getFilterExpressionResolverTree()) {
        // loading the filter executer tree for filter evaluation
        blockExecutionInfo.setFilterExecuterTree(FilterUtil.getFilterExecuterTree(queryModel.getFilterExpressionResolverTree(), segmentProperties, blockExecutionInfo.getComlexDimensionInfoMap()));
        List<IndexKey> listOfStartEndKeys = new ArrayList<IndexKey>(2);
        FilterUtil.traverseResolverTreeAndGetStartAndEndKey(segmentProperties, queryModel.getFilterExpressionResolverTree(), listOfStartEndKeys);
        startIndexKey = listOfStartEndKeys.get(0);
        endIndexKey = listOfStartEndKeys.get(1);
    } else {
        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[] dimensionsBlockIndexes = QueryUtil.getDimensionsBlockIndexes(currentBlockQueryDimensions, segmentProperties.getDimensionOrdinalToBlockMapping(), 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 (dimensionsBlockIndexes.length > 0) {
        numberOfElementToConsider = dimensionsBlockIndexes[dimensionsBlockIndexes.length - 1] == segmentProperties.getBlockTodimensionOrdinalMapping().size() - 1 ? dimensionsBlockIndexes.length - 1 : dimensionsBlockIndexes.length;
        blockExecutionInfo.setAllSelectedDimensionBlocksIndexes(CarbonUtil.getRangeIndex(dimensionsBlockIndexes, numberOfElementToConsider, numberOfColumnToBeReadInOneIO));
    } else {
        blockExecutionInfo.setAllSelectedDimensionBlocksIndexes(new int[0][0]);
    }
    // get the list of updated filter measures present in the current block
    Set<CarbonMeasure> currentBlockFilterMeasures = getCurrentBlockFilterMeasures(queryProperties.filterMeasures, segmentProperties);
    // list of measures to be projected
    List<Integer> allProjectionListMeasureIndexes = new ArrayList<>();
    int[] measureBlockIndexes = QueryUtil.getMeasureBlockIndexes(currentBlockQueryMeasures, expressionMeasures, segmentProperties.getMeasuresOrdinalToBlockMapping(), currentBlockFilterMeasures, allProjectionListMeasureIndexes);
    if (measureBlockIndexes.length > 0) {
        numberOfElementToConsider = measureBlockIndexes[measureBlockIndexes.length - 1] == segmentProperties.getMeasures().size() - 1 ? measureBlockIndexes.length - 1 : measureBlockIndexes.length;
        // setting all the measure chunk indexes to be read from file
        blockExecutionInfo.setAllSelectedMeasureBlocksIndexes(CarbonUtil.getRangeIndex(measureBlockIndexes, numberOfElementToConsider, numberOfColumnToBeReadInOneIO));
    } else {
        blockExecutionInfo.setAllSelectedMeasureBlocksIndexes(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(currentBlockQueryDimensions, segmentProperties));
    Set<Integer> dictionaryColumnBlockIndex = new HashSet<Integer>();
    List<Integer> noDictionaryColumnBlockIndex = new ArrayList<Integer>();
    // get the block index to be read from file for query dimension
    // for both dictionary columns and no dictionary columns
    QueryUtil.fillQueryDimensionsBlockIndexes(currentBlockQueryDimensions, segmentProperties.getDimensionOrdinalToBlockMapping(), dictionaryColumnBlockIndex, noDictionaryColumnBlockIndex);
    int[] queryDictionaryColumnBlockIndexes = ArrayUtils.toPrimitive(dictionaryColumnBlockIndex.toArray(new Integer[dictionaryColumnBlockIndex.size()]));
    // need to sort the dictionary column as for all dimension
    // column key will be filled based on key order
    Arrays.sort(queryDictionaryColumnBlockIndexes);
    blockExecutionInfo.setDictionaryColumnBlockIndex(queryDictionaryColumnBlockIndexes);
    // setting the no dictionary column block indexes
    blockExecutionInfo.setNoDictionaryBlockIndexes(ArrayUtils.toPrimitive(noDictionaryColumnBlockIndex.toArray(new Integer[noDictionaryColumnBlockIndex.size()])));
    // setting column id to dictionary mapping
    blockExecutionInfo.setColumnIdToDcitionaryMapping(queryProperties.columnToDictionayMapping);
    // setting each column value size
    blockExecutionInfo.setEachColumnValueSize(segmentProperties.getEachDimColumnValueSize());
    blockExecutionInfo.setComplexColumnParentBlockIndexes(getComplexDimensionParentBlockIndexes(currentBlockQueryDimensions));
    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(currentBlockQueryDimensions, 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.getQueryDimension().toArray(new QueryDimension[queryModel.getQueryDimension().size()]));
    blockExecutionInfo.setActualQueryMeasures(queryModel.getQueryMeasures().toArray(new QueryMeasure[queryModel.getQueryMeasures().size()]));
    return blockExecutionInfo;
}
Also used : LinkedHashSet(java.util.LinkedHashSet) IndexKey(org.apache.carbondata.core.datastore.IndexKey) ArrayList(java.util.ArrayList) QueryMeasure(org.apache.carbondata.core.scan.model.QueryMeasure) BlockExecutionInfo(org.apache.carbondata.core.scan.executor.infos.BlockExecutionInfo) KeyGenerator(org.apache.carbondata.core.keygenerator.KeyGenerator) HashSet(java.util.HashSet) LinkedHashSet(java.util.LinkedHashSet) 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) KeyGenException(org.apache.carbondata.core.keygenerator.KeyGenException) SegmentProperties(org.apache.carbondata.core.datastore.block.SegmentProperties) QueryDimension(org.apache.carbondata.core.scan.model.QueryDimension)

Example 2 with CarbonMeasure

use of org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure in project carbondata by apache.

the class QueryUtil method resolveQueryModel.

/**
   * Below method will be used to resolve the query model
   * resolve will be setting the actual dimension and measure object
   * as from driver only column name will be passes to avoid the heavy object
   * serialization
   *
   * @param queryModel query model
   */
public static void resolveQueryModel(QueryModel queryModel) {
    CarbonMetadata.getInstance().addCarbonTable(queryModel.getTable());
    // TODO need to load the table from table identifier
    CarbonTable carbonTable = queryModel.getTable();
    String tableName = queryModel.getAbsoluteTableIdentifier().getCarbonTableIdentifier().getTableName();
    // resolve query dimension
    for (QueryDimension queryDimension : queryModel.getQueryDimension()) {
        queryDimension.setDimension(carbonTable.getDimensionByName(tableName, queryDimension.getColumnName()));
    }
    // resolve query measure
    for (QueryMeasure queryMeasure : queryModel.getQueryMeasures()) {
        //never come false but if in future we can remove so not removing first if check
        if (queryMeasure.getColumnName().equals("count(*)")) {
            if (carbonTable.getMeasureByTableName(tableName).size() > 0 && !carbonTable.getMeasureByTableName(tableName).get(0).getColName().equals(CarbonCommonConstants.DEFAULT_INVISIBLE_DUMMY_MEASURE)) {
                queryMeasure.setMeasure(carbonTable.getMeasureByTableName(tableName).get(0));
            } else {
                CarbonMeasure dummyMeasure = new CarbonMeasure(carbonTable.getDimensionByTableName(tableName).get(0).getColumnSchema(), 0);
                queryMeasure.setMeasure(dummyMeasure);
            }
        } else {
            queryMeasure.setMeasure(carbonTable.getMeasureByName(tableName, queryMeasure.getColumnName()));
        }
    }
}
Also used : CarbonTable(org.apache.carbondata.core.metadata.schema.table.CarbonTable) CarbonMeasure(org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure) QueryMeasure(org.apache.carbondata.core.scan.model.QueryMeasure) QueryDimension(org.apache.carbondata.core.scan.model.QueryDimension)

Example 3 with CarbonMeasure

use of org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure in project carbondata by apache.

the class RowLevelFilterExecuterImpl method initMeasureBlockIndexes.

/**
   * This method will initialize the measure info for the current block to be
   * used for filtering the data
   */
private void initMeasureBlockIndexes() {
    for (int i = 0; i < msrColEvalutorInfoList.size(); i++) {
        // find the measure in the current block measures list
        CarbonMeasure measureFromCurrentBlock = segmentProperties.getMeasureFromCurrentBlock(msrColEvalutorInfoList.get(i).getCarbonColumn().getColumnId());
        if (null != measureFromCurrentBlock) {
            msrColEvalutorInfoList.get(i).setColumnIndex(measureFromCurrentBlock.getOrdinal());
            this.measureBlocksIndex[i] = segmentProperties.getMeasuresOrdinalToBlockMapping().get(measureFromCurrentBlock.getOrdinal());
            isMeasurePresentInCurrentBlock[i] = true;
        }
    }
}
Also used : CarbonMeasure(org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure)

Example 4 with CarbonMeasure

use of org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure in project carbondata by apache.

the class CarbonCompactionExecutor method prepareQueryModel.

/**
   * Preparing of the query model.
   *
   * @param blockList
   * @return
   */
private QueryModel prepareQueryModel(List<TableBlockInfo> blockList) {
    QueryModel model = new QueryModel();
    model.setTableBlockInfos(blockList);
    model.setForcedDetailRawQuery(true);
    model.setFilterExpressionResolverTree(null);
    List<QueryDimension> dims = new ArrayList<>(CarbonCommonConstants.DEFAULT_COLLECTION_SIZE);
    List<CarbonDimension> dimensions = carbonTable.getDimensionByTableName(carbonTable.getFactTableName());
    for (CarbonDimension dim : dimensions) {
        // check if dimension is deleted
        QueryDimension queryDimension = new QueryDimension(dim.getColName());
        queryDimension.setDimension(dim);
        dims.add(queryDimension);
    }
    model.setQueryDimension(dims);
    List<QueryMeasure> msrs = new ArrayList<>(CarbonCommonConstants.DEFAULT_COLLECTION_SIZE);
    List<CarbonMeasure> measures = carbonTable.getMeasureByTableName(carbonTable.getFactTableName());
    for (CarbonMeasure carbonMeasure : measures) {
        // check if measure is deleted
        QueryMeasure queryMeasure = new QueryMeasure(carbonMeasure.getColName());
        queryMeasure.setMeasure(carbonMeasure);
        msrs.add(queryMeasure);
    }
    model.setQueryMeasures(msrs);
    model.setQueryId(System.nanoTime() + "");
    model.setAbsoluteTableIdentifier(carbonTable.getAbsoluteTableIdentifier());
    model.setTable(carbonTable);
    return model;
}
Also used : CarbonMeasure(org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure) QueryMeasure(org.apache.carbondata.core.scan.model.QueryMeasure) ArrayList(java.util.ArrayList) QueryModel(org.apache.carbondata.core.scan.model.QueryModel) QueryDimension(org.apache.carbondata.core.scan.model.QueryDimension) CarbonDimension(org.apache.carbondata.core.metadata.schema.table.column.CarbonDimension)

Example 5 with CarbonMeasure

use of org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure in project carbondata by apache.

the class FilterUtil method getIncludeFilterExecuter.

/**
 * It gives filter executer based on columnar or column group
 *
 * @param dimColResolvedFilterInfo
 * @param segmentProperties
 * @return
 */
private static FilterExecuter getIncludeFilterExecuter(DimColumnResolvedFilterInfo dimColResolvedFilterInfo, MeasureColumnResolvedFilterInfo msrColResolvedFilterInfo, SegmentProperties segmentProperties) {
    if (null != msrColResolvedFilterInfo && msrColResolvedFilterInfo.getMeasure().isColumnar()) {
        CarbonMeasure measuresFromCurrentBlock = segmentProperties.getMeasureFromCurrentBlock(msrColResolvedFilterInfo.getMeasure().getColumnId());
        if (null != measuresFromCurrentBlock) {
            // update dimension and column index according to the dimension position in current block
            MeasureColumnResolvedFilterInfo msrColResolvedFilterInfoCopyObject = msrColResolvedFilterInfo.getCopyObject();
            msrColResolvedFilterInfoCopyObject.setMeasure(measuresFromCurrentBlock);
            msrColResolvedFilterInfoCopyObject.setColumnIndex(measuresFromCurrentBlock.getOrdinal());
            msrColResolvedFilterInfoCopyObject.setType(measuresFromCurrentBlock.getDataType());
            return new IncludeFilterExecuterImpl(null, msrColResolvedFilterInfoCopyObject, segmentProperties, true);
        } else {
            return new RestructureIncludeFilterExecutorImpl(dimColResolvedFilterInfo, msrColResolvedFilterInfo, true);
        }
    }
    if (null != dimColResolvedFilterInfo) {
        CarbonDimension dimension = dimColResolvedFilterInfo.getDimension();
        if (dimension.hasEncoding(Encoding.IMPLICIT)) {
            return new ImplicitIncludeFilterExecutorImpl(dimColResolvedFilterInfo);
        } else if (dimension.isColumnar()) {
            CarbonDimension dimensionFromCurrentBlock = segmentProperties.getDimensionFromCurrentBlock(dimColResolvedFilterInfo.getDimension());
            if (null != dimensionFromCurrentBlock) {
                // update dimension and column index according to the dimension position in current block
                DimColumnResolvedFilterInfo dimColResolvedFilterInfoCopyObject = dimColResolvedFilterInfo.getCopyObject();
                dimColResolvedFilterInfoCopyObject.setDimension(dimensionFromCurrentBlock);
                dimColResolvedFilterInfoCopyObject.setColumnIndex(dimensionFromCurrentBlock.getOrdinal());
                return new IncludeFilterExecuterImpl(dimColResolvedFilterInfoCopyObject, null, segmentProperties, false);
            } else {
                return new RestructureIncludeFilterExecutorImpl(dimColResolvedFilterInfo, msrColResolvedFilterInfo, false);
            }
        } else {
            return new IncludeColGroupFilterExecuterImpl(dimColResolvedFilterInfo, segmentProperties);
        }
    } else {
        return new IncludeColGroupFilterExecuterImpl(null, segmentProperties);
    }
}
Also used : MeasureColumnResolvedFilterInfo(org.apache.carbondata.core.scan.filter.resolver.resolverinfo.MeasureColumnResolvedFilterInfo) DimColumnResolvedFilterInfo(org.apache.carbondata.core.scan.filter.resolver.resolverinfo.DimColumnResolvedFilterInfo) CarbonMeasure(org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure) ImplicitIncludeFilterExecutorImpl(org.apache.carbondata.core.scan.filter.executer.ImplicitIncludeFilterExecutorImpl) RestructureIncludeFilterExecutorImpl(org.apache.carbondata.core.scan.filter.executer.RestructureIncludeFilterExecutorImpl) IncludeColGroupFilterExecuterImpl(org.apache.carbondata.core.scan.filter.executer.IncludeColGroupFilterExecuterImpl) IncludeFilterExecuterImpl(org.apache.carbondata.core.scan.filter.executer.IncludeFilterExecuterImpl) CarbonDimension(org.apache.carbondata.core.metadata.schema.table.column.CarbonDimension)

Aggregations

CarbonMeasure (org.apache.carbondata.core.metadata.schema.table.column.CarbonMeasure)47 CarbonDimension (org.apache.carbondata.core.metadata.schema.table.column.CarbonDimension)31 ArrayList (java.util.ArrayList)16 ColumnSchema (org.apache.carbondata.core.metadata.schema.table.column.ColumnSchema)10 DataType (org.apache.carbondata.core.metadata.datatype.DataType)8 CarbonColumn (org.apache.carbondata.core.metadata.schema.table.column.CarbonColumn)7 HashSet (java.util.HashSet)5 CarbonTable (org.apache.carbondata.core.metadata.schema.table.CarbonTable)4 BlockExecutionInfo (org.apache.carbondata.core.scan.executor.infos.BlockExecutionInfo)4 ProjectionMeasure (org.apache.carbondata.core.scan.model.ProjectionMeasure)4 Test (org.junit.Test)4 MeasureInfo (org.apache.carbondata.core.scan.executor.infos.MeasureInfo)3 QueryDimension (org.apache.carbondata.core.scan.model.QueryDimension)3 QueryMeasure (org.apache.carbondata.core.scan.model.QueryMeasure)3 GenericDataType (org.apache.carbondata.processing.datatypes.GenericDataType)3 BufferedReader (java.io.BufferedReader)2 FileReader (java.io.FileReader)2 ByteBuffer (java.nio.ByteBuffer)2 HashMap (java.util.HashMap)2 LinkedHashSet (java.util.LinkedHashSet)2