Search in sources :

Example 16 with SegmentProperties

use of org.apache.carbondata.core.datastore.block.SegmentProperties in project carbondata by apache.

the class InMemoryBTreeIndex method getDataBlocksOfIndex.

/**
 * get data blocks of given btree
 */
private List<DataRefNode> getDataBlocksOfIndex(AbstractIndex abstractIndex) {
    List<DataRefNode> blocks = new LinkedList<DataRefNode>();
    SegmentProperties segmentProperties = abstractIndex.getSegmentProperties();
    try {
        IndexKey startIndexKey = FilterUtil.prepareDefaultStartIndexKey(segmentProperties);
        IndexKey endIndexKey = FilterUtil.prepareDefaultEndIndexKey(segmentProperties);
        // Add all blocks of btree into result
        DataRefNodeFinder blockFinder = new BTreeDataRefNodeFinder(segmentProperties.getEachDimColumnValueSize(), segmentProperties.getNumberOfSortColumns(), segmentProperties.getNumberOfNoDictSortColumns());
        DataRefNode startBlock = blockFinder.findFirstDataBlock(abstractIndex.getDataRefNode(), startIndexKey);
        DataRefNode endBlock = blockFinder.findLastDataBlock(abstractIndex.getDataRefNode(), endIndexKey);
        while (startBlock != endBlock) {
            blocks.add(startBlock);
            startBlock = startBlock.getNextDataRefNode();
        }
        blocks.add(endBlock);
    } catch (KeyGenException e) {
        LOG.error("Could not generate start key", e);
    }
    return blocks;
}
Also used : IndexKey(org.apache.carbondata.core.datastore.IndexKey) KeyGenException(org.apache.carbondata.core.keygenerator.KeyGenException) DataRefNode(org.apache.carbondata.core.datastore.DataRefNode) SegmentProperties(org.apache.carbondata.core.datastore.block.SegmentProperties) BTreeDataRefNodeFinder(org.apache.carbondata.core.datastore.impl.btree.BTreeDataRefNodeFinder) DataRefNodeFinder(org.apache.carbondata.core.datastore.DataRefNodeFinder) LinkedList(java.util.LinkedList) BTreeDataRefNodeFinder(org.apache.carbondata.core.datastore.impl.btree.BTreeDataRefNodeFinder)

Example 17 with SegmentProperties

use of org.apache.carbondata.core.datastore.block.SegmentProperties in project carbondata by apache.

the class RestructureBasedRawResultCollector method initCurrentBlockKeyGenerator.

/**
 * This method will initialize the block key generator for the current block based on the
 * dictionary columns present in the current block
 */
private void initCurrentBlockKeyGenerator() {
    SegmentProperties segmentProperties = executionInfo.getDataBlock().getSegmentProperties();
    int[] dictionaryColumnBlockIndex = executionInfo.getDictionaryColumnChunkIndex();
    int[] updatedColumnCardinality = new int[dictionaryColumnBlockIndex.length];
    int[] updatedDimensionPartitioner = new int[dictionaryColumnBlockIndex.length];
    for (int i = 0; i < dictionaryColumnBlockIndex.length; i++) {
        // get the dictionary key ordinal as column cardinality in segment properties
        // will only be for dictionary encoded columns
        CarbonDimension currentBlockDimension = segmentProperties.getDimensions().get(dictionaryColumnBlockIndex[i]);
        updatedColumnCardinality[i] = segmentProperties.getDimColumnsCardinality()[currentBlockDimension.getKeyOrdinal()];
        updatedDimensionPartitioner[i] = segmentProperties.getDimensionPartitions()[currentBlockDimension.getKeyOrdinal()];
    }
    if (dictionaryColumnBlockIndex.length > 0) {
        int[] dimensionBitLength = CarbonUtil.getDimensionBitLength(updatedColumnCardinality, updatedDimensionPartitioner);
        updatedCurrentBlockKeyGenerator = new MultiDimKeyVarLengthGenerator(dimensionBitLength);
    }
}
Also used : MultiDimKeyVarLengthGenerator(org.apache.carbondata.core.keygenerator.mdkey.MultiDimKeyVarLengthGenerator) SegmentProperties(org.apache.carbondata.core.datastore.block.SegmentProperties) CarbonDimension(org.apache.carbondata.core.metadata.schema.table.column.CarbonDimension)

Example 18 with SegmentProperties

use of org.apache.carbondata.core.datastore.block.SegmentProperties in project carbondata by apache.

the class RestructureBasedRawResultCollector method initRestructuredKeyGenerator.

/**
 * This method will create a new key generator for generating mdKey according to latest schema
 */
private void initRestructuredKeyGenerator() {
    SegmentProperties segmentProperties = executionInfo.getDataBlock().getSegmentProperties();
    ProjectionDimension[] queryDimensions = executionInfo.getActualQueryDimensions();
    List<Integer> updatedColumnCardinality = new ArrayList<>(queryDimensions.length);
    List<Integer> updatedDimensionPartitioner = new ArrayList<>(queryDimensions.length);
    int[] dictionaryColumnBlockIndex = executionInfo.getDictionaryColumnChunkIndex();
    int dimCounterInCurrentBlock = 0;
    for (int i = 0; i < queryDimensions.length; i++) {
        if (queryDimensions[i].getDimension().hasEncoding(Encoding.DICTIONARY)) {
            if (executionInfo.getDimensionInfo().getDimensionExists()[i]) {
                // get the dictionary key ordinal as column cardinality in segment properties
                // will only be for dictionary encoded columns
                CarbonDimension currentBlockDimension = segmentProperties.getDimensions().get(dictionaryColumnBlockIndex[dimCounterInCurrentBlock]);
                updatedColumnCardinality.add(segmentProperties.getDimColumnsCardinality()[currentBlockDimension.getKeyOrdinal()]);
                updatedDimensionPartitioner.add(segmentProperties.getDimensionPartitions()[currentBlockDimension.getKeyOrdinal()]);
                dimCounterInCurrentBlock++;
            } else {
                // partitioner index will be 1 every column will be in columnar format
                updatedDimensionPartitioner.add(1);
                // for direct dictionary 4 bytes need to be allocated else 1
                if (queryDimensions[i].getDimension().hasEncoding(Encoding.DIRECT_DICTIONARY)) {
                    updatedColumnCardinality.add(Integer.MAX_VALUE);
                } else {
                    // cardinality will be 2 will user has provided a default value
                    byte[] defaultValue = queryDimensions[i].getDimension().getDefaultValue();
                    if (null != defaultValue) {
                        updatedColumnCardinality.add(CarbonCommonConstants.DICTIONARY_DEFAULT_CARDINALITY + 1);
                    } else {
                        updatedColumnCardinality.add(CarbonCommonConstants.DICTIONARY_DEFAULT_CARDINALITY);
                    }
                }
            }
        }
    }
    if (!updatedColumnCardinality.isEmpty()) {
        int[] latestColumnCardinality = ArrayUtils.toPrimitive(updatedColumnCardinality.toArray(new Integer[updatedColumnCardinality.size()]));
        int[] latestColumnPartitioner = ArrayUtils.toPrimitive(updatedDimensionPartitioner.toArray(new Integer[updatedDimensionPartitioner.size()]));
        int[] dimensionBitLength = CarbonUtil.getDimensionBitLength(latestColumnCardinality, latestColumnPartitioner);
        restructuredKeyGenerator = new MultiDimKeyVarLengthGenerator(dimensionBitLength);
    }
}
Also used : MultiDimKeyVarLengthGenerator(org.apache.carbondata.core.keygenerator.mdkey.MultiDimKeyVarLengthGenerator) ArrayList(java.util.ArrayList) SegmentProperties(org.apache.carbondata.core.datastore.block.SegmentProperties) ProjectionDimension(org.apache.carbondata.core.scan.model.ProjectionDimension) CarbonDimension(org.apache.carbondata.core.metadata.schema.table.column.CarbonDimension)

Example 19 with SegmentProperties

use of org.apache.carbondata.core.datastore.block.SegmentProperties 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)

Aggregations

SegmentProperties (org.apache.carbondata.core.datastore.block.SegmentProperties)19 ArrayList (java.util.ArrayList)11 ColumnSchema (org.apache.carbondata.core.metadata.schema.table.column.ColumnSchema)8 IndexKey (org.apache.carbondata.core.datastore.IndexKey)6 HashMap (java.util.HashMap)5 Test (org.junit.Test)5 KeyGenException (org.apache.carbondata.core.keygenerator.KeyGenException)4 CarbonDimension (org.apache.carbondata.core.metadata.schema.table.column.CarbonDimension)4 Map (java.util.Map)3 MockUp (mockit.MockUp)3 DataFileFooter (org.apache.carbondata.core.metadata.blocklet.DataFileFooter)3 ColumnSchema (org.apache.carbondata.format.ColumnSchema)3 HashSet (java.util.HashSet)2 LinkedHashSet (java.util.LinkedHashSet)2 AbstractDictionaryCacheTest (org.apache.carbondata.core.cache.dictionary.AbstractDictionaryCacheTest)2 DataRefNode (org.apache.carbondata.core.datastore.DataRefNode)2 DataRefNodeFinder (org.apache.carbondata.core.datastore.DataRefNodeFinder)2 TableBlockInfo (org.apache.carbondata.core.datastore.block.TableBlockInfo)2 BTreeDataRefNodeFinder (org.apache.carbondata.core.datastore.impl.btree.BTreeDataRefNodeFinder)2 MultiDimKeyVarLengthGenerator (org.apache.carbondata.core.keygenerator.mdkey.MultiDimKeyVarLengthGenerator)2