Examples with ColumnGroupDimensionDataChunk org.apache.carbondata.core.datastore.chunk.impl.ColumnGroupDimensionDataChunk used on opensource projects

Example 1 with ColumnGroupDimensionDataChunk

use of org.apache.carbondata.core.datastore.chunk.impl.ColumnGroupDimensionDataChunk in project carbondata by apache.

the class CompressedDimensionChunkFileBasedReaderV3 method convertToDimensionChunk.

/**
   * Below method will be used to convert the compressed dimension chunk raw data to actual data
   *
   * @param dimensionRawColumnChunk dimension raw chunk
   * @param pageNumber              number
   * @return DimensionColumnDataChunk
   */
@Override
public DimensionColumnDataChunk convertToDimensionChunk(DimensionRawColumnChunk dimensionRawColumnChunk, int pageNumber) throws IOException {
    byte[] dataPage = null;
    int[] invertedIndexes = null;
    int[] invertedIndexesReverse = null;
    int[] rlePage = null;
    // data chunk of page
    DataChunk2 dimensionColumnChunk = null;
    // data chunk of blocklet column
    DataChunk3 dataChunk3 = dimensionRawColumnChunk.getDataChunkV3();
    // get the data buffer
    ByteBuffer rawData = dimensionRawColumnChunk.getRawData();
    dimensionColumnChunk = dataChunk3.getData_chunk_list().get(pageNumber);
    // calculating the start point of data
    // as buffer can contain multiple column data, start point will be datachunkoffset +
    // data chunk length + page offset
    int copySourcePoint = dimensionRawColumnChunk.getOffSet() + dimensionChunksLength.get(dimensionRawColumnChunk.getBlockletId()) + dataChunk3.getPage_offset().get(pageNumber);
    // first read the data and uncompressed it
    dataPage = COMPRESSOR.unCompressByte(rawData.array(), copySourcePoint, dimensionColumnChunk.data_page_length);
    copySourcePoint += dimensionColumnChunk.data_page_length;
    // if row id block is present then read the row id chunk and uncompress it
    if (hasEncoding(dimensionColumnChunk.encoders, Encoding.INVERTED_INDEX)) {
        invertedIndexes = CarbonUtil.getUnCompressColumnIndex(dimensionColumnChunk.rowid_page_length, rawData, copySourcePoint);
        copySourcePoint += dimensionColumnChunk.rowid_page_length;
        // get the reverse index
        invertedIndexesReverse = getInvertedReverseIndex(invertedIndexes);
    }
    //then actual data based on rle block
    if (hasEncoding(dimensionColumnChunk.encoders, Encoding.RLE)) {
        rlePage = CarbonUtil.getIntArray(rawData, copySourcePoint, dimensionColumnChunk.rle_page_length);
        // uncompress the data with rle indexes
        dataPage = UnBlockIndexer.uncompressData(dataPage, rlePage, eachColumnValueSize[dimensionRawColumnChunk.getBlockletId()]);
        rlePage = null;
    }
    // fill chunk attributes
    DimensionColumnDataChunk columnDataChunk = null;
    if (dimensionColumnChunk.isRowMajor()) {
        // to store fixed length column chunk values
        columnDataChunk = new ColumnGroupDimensionDataChunk(dataPage, eachColumnValueSize[dimensionRawColumnChunk.getBlockletId()], dimensionRawColumnChunk.getRowCount()[pageNumber]);
    } else // and set to data chunk instance
    if (!hasEncoding(dimensionColumnChunk.encoders, Encoding.DICTIONARY)) {
        columnDataChunk = new VariableLengthDimensionDataChunk(dataPage, invertedIndexes, invertedIndexesReverse, dimensionRawColumnChunk.getRowCount()[pageNumber]);
    } else {
        // to store fixed length column chunk values
        columnDataChunk = new FixedLengthDimensionDataChunk(dataPage, invertedIndexes, invertedIndexesReverse, dimensionRawColumnChunk.getRowCount()[pageNumber], eachColumnValueSize[dimensionRawColumnChunk.getBlockletId()]);
    }
    return columnDataChunk;
}

Example 2 with ColumnGroupDimensionDataChunk

use of org.apache.carbondata.core.datastore.chunk.impl.ColumnGroupDimensionDataChunk in project carbondata by apache.

the class CompressedDimensionChunkFileBasedReaderV2 method convertToDimensionChunk.

public DimensionColumnDataChunk convertToDimensionChunk(DimensionRawColumnChunk dimensionRawColumnChunk, int pageNumber) throws IOException {
    byte[] dataPage = null;
    int[] invertedIndexes = null;
    int[] invertedIndexesReverse = null;
    int[] rlePage = null;
    DataChunk2 dimensionColumnChunk = null;
    int copySourcePoint = dimensionRawColumnChunk.getOffSet();
    int blockIndex = dimensionRawColumnChunk.getBlockletId();
    ByteBuffer rawData = dimensionRawColumnChunk.getRawData();
    if (dimensionChunksOffset.size() - 1 == blockIndex) {
        dimensionColumnChunk = CarbonUtil.readDataChunk(rawData, copySourcePoint, dimensionRawColumnChunk.getLength());
        int totalDimensionDataLength = dimensionColumnChunk.data_page_length + dimensionColumnChunk.rle_page_length + dimensionColumnChunk.rowid_page_length;
        synchronized (dimensionRawColumnChunk.getFileReader()) {
            rawData = dimensionRawColumnChunk.getFileReader().readByteBuffer(filePath, dimensionChunksOffset.get(blockIndex) + dimensionChunksLength.get(blockIndex), totalDimensionDataLength);
        }
    } else {
        dimensionColumnChunk = CarbonUtil.readDataChunk(rawData, copySourcePoint, dimensionChunksLength.get(blockIndex));
        copySourcePoint += dimensionChunksLength.get(blockIndex);
    }
    // first read the data and uncompressed it
    dataPage = COMPRESSOR.unCompressByte(rawData.array(), copySourcePoint, dimensionColumnChunk.data_page_length);
    copySourcePoint += dimensionColumnChunk.data_page_length;
    // if row id block is present then read the row id chunk and uncompress it
    if (hasEncoding(dimensionColumnChunk.encoders, Encoding.INVERTED_INDEX)) {
        byte[] dataInv = new byte[dimensionColumnChunk.rowid_page_length];
        rawData.position(copySourcePoint);
        rawData.get(dataInv);
        invertedIndexes = CarbonUtil.getUnCompressColumnIndex(dimensionColumnChunk.rowid_page_length, dataInv, numberComressor, 0);
        copySourcePoint += dimensionColumnChunk.rowid_page_length;
        // get the reverse index
        invertedIndexesReverse = getInvertedReverseIndex(invertedIndexes);
    }
    //then actual data based on rle block
    if (hasEncoding(dimensionColumnChunk.encoders, Encoding.RLE)) {
        byte[] dataRle = new byte[dimensionColumnChunk.rle_page_length];
        rawData.position(copySourcePoint);
        rawData.get(dataRle);
        rlePage = numberComressor.unCompress(dataRle, 0, dimensionColumnChunk.rle_page_length);
        // uncompress the data with rle indexes
        dataPage = UnBlockIndexer.uncompressData(dataPage, rlePage, eachColumnValueSize[blockIndex]);
    }
    // fill chunk attributes
    DimensionColumnDataChunk columnDataChunk = null;
    if (dimensionColumnChunk.isRowMajor()) {
        // to store fixed length column chunk values
        columnDataChunk = new ColumnGroupDimensionDataChunk(dataPage, eachColumnValueSize[blockIndex], numberOfRows);
    } else // and set to data chunk instance
    if (!hasEncoding(dimensionColumnChunk.encoders, Encoding.DICTIONARY)) {
        columnDataChunk = new VariableLengthDimensionDataChunk(dataPage, invertedIndexes, invertedIndexesReverse, numberOfRows);
    } else {
        // to store fixed length column chunk values
        columnDataChunk = new FixedLengthDimensionDataChunk(dataPage, invertedIndexes, invertedIndexesReverse, numberOfRows, eachColumnValueSize[blockIndex]);
    }
    return columnDataChunk;
}

Example 3 with ColumnGroupDimensionDataChunk

use of org.apache.carbondata.core.datastore.chunk.impl.ColumnGroupDimensionDataChunk in project carbondata by apache.

the class CompressedDimensionChunkFileBasedReaderV1 method convertToDimensionChunk.

@Override
public DimensionColumnDataChunk convertToDimensionChunk(DimensionRawColumnChunk dimensionRawColumnChunk, int pageNumber) throws IOException {
    int blockIndex = dimensionRawColumnChunk.getBlockletId();
    byte[] dataPage = null;
    int[] invertedIndexes = null;
    int[] invertedIndexesReverse = null;
    int[] rlePage = null;
    FileHolder fileReader = dimensionRawColumnChunk.getFileReader();
    ByteBuffer rawData = dimensionRawColumnChunk.getRawData();
    dataPage = COMPRESSOR.unCompressByte(rawData.array(), dimensionRawColumnChunk.getOffSet(), dimensionRawColumnChunk.getLength());
    // if row id block is present then read the row id chunk and uncompress it
    DataChunk dataChunk = dimensionColumnChunk.get(blockIndex);
    if (CarbonUtil.hasEncoding(dataChunk.getEncodingList(), Encoding.INVERTED_INDEX)) {
        byte[] columnIndexData;
        synchronized (fileReader) {
            columnIndexData = fileReader.readByteArray(filePath, dataChunk.getRowIdPageOffset(), dataChunk.getRowIdPageLength());
        }
        invertedIndexes = CarbonUtil.getUnCompressColumnIndex(dataChunk.getRowIdPageLength(), columnIndexData, numberComressor, 0);
        // get the reverse index
        invertedIndexesReverse = getInvertedReverseIndex(invertedIndexes);
    }
    //then actual data based on rle block
    if (CarbonUtil.hasEncoding(dataChunk.getEncodingList(), Encoding.RLE)) {
        // read and uncompress the rle block
        byte[] key;
        synchronized (fileReader) {
            key = fileReader.readByteArray(filePath, dataChunk.getRlePageOffset(), dataChunk.getRlePageLength());
        }
        rlePage = numberComressor.unCompress(key, 0, dataChunk.getRlePageLength());
        // uncompress the data with rle indexes
        dataPage = UnBlockIndexer.uncompressData(dataPage, rlePage, eachColumnValueSize[blockIndex]);
        rlePage = null;
    }
    // fill chunk attributes
    DimensionColumnDataChunk columnDataChunk = null;
    if (dataChunk.isRowMajor()) {
        // to store fixed length column chunk values
        columnDataChunk = new ColumnGroupDimensionDataChunk(dataPage, eachColumnValueSize[blockIndex], numberOfRows);
    } else // and set to data chunk instance
    if (!CarbonUtil.hasEncoding(dataChunk.getEncodingList(), Encoding.DICTIONARY)) {
        columnDataChunk = new VariableLengthDimensionDataChunk(dataPage, invertedIndexes, invertedIndexesReverse, numberOfRows);
    } else {
        // to store fixed length column chunk values
        columnDataChunk = new FixedLengthDimensionDataChunk(dataPage, invertedIndexes, invertedIndexesReverse, numberOfRows, eachColumnValueSize[blockIndex]);
    }
    return columnDataChunk;
}