Examples with BlobSupportDataCellIterator org.knime.core.data.collection.BlobSupportDataCellIterator used on opensource projects

Example 1 with BlobSupportDataCellIterator

use of org.knime.core.data.collection.BlobSupportDataCellIterator in project knime-core by knime.

the class SplitCellFactory method getCells.

/**
 * {@inheritDoc}
 */
@Override
public DataCell[] getCells(final DataRow row) {
    DataCell inCell = row.getCell(m_colIndex);
    DataCell[] result = new DataCell[m_colSpecs.length];
    Arrays.fill(result, DataType.getMissingCell());
    if (inCell.isMissing()) {
        if (m_warnMessage == null) {
            m_warnMessage = "Some rows contain missing values";
        }
        return result;
    }
    CollectionDataValue v = (CollectionDataValue) inCell;
    Iterator<DataCell> it = v.iterator();
    for (int i = 0; i < m_colSpecs.length && it.hasNext(); i++) {
        DataCell next;
        DataType type;
        if (it instanceof BlobSupportDataCellIterator) {
            next = ((BlobSupportDataCellIterator) it).nextWithBlobSupport();
            if (next instanceof BlobWrapperDataCell) {
                // try to not access the cell (will get deserialized)
                BlobWrapperDataCell bw = (BlobWrapperDataCell) next;
                type = DataType.getType(bw.getBlobClass());
            } else {
                type = next.getType();
            }
        } else {
            next = it.next();
            type = next.getType();
        }
        if (m_commonTypes[i] == null) {
            m_commonTypes[i] = type;
        } else {
            m_commonTypes[i] = DataType.getCommonSuperType(m_commonTypes[i], type);
        }
        result[i] = next;
    }
    if (it.hasNext()) {
        m_warnMessage = "At least one row had more elements than " + "specified; row was truncated.";
    }
    m_domainCreator.updateDomain(new DefaultRow(row.getKey(), result));
    return result;
}

Example 2 with BlobSupportDataCellIterator

use of org.knime.core.data.collection.BlobSupportDataCellIterator in project knime-core by knime.

the class Buffer method handleIncomingBlob.

private DataCell handleIncomingBlob(final DataCell cell, final int col, final int totalColCount, final boolean copyForVersionHop, final boolean forceCopyOfBlobsArg) throws IOException {
    if (!(m_outputFormat instanceof DefaultTableStoreFormat)) {
        return cell;
    }
    // whether the content of the argument row needs to be copied
    // into a new BlobSupportDataRow (will do that when either this
    // flag is true or cellCopies != null)
    boolean isWrapperCell = cell instanceof BlobWrapperDataCell;
    BlobAddress ad;
    final CellClassInfo cl;
    BlobWrapperDataCell wc;
    if (isWrapperCell) {
        wc = (BlobWrapperDataCell) cell;
        ad = wc.getAddress();
        cl = wc.getBlobClassInfo();
    } else if (cell instanceof BlobDataCell) {
        wc = null;
        cl = CellClassInfo.get(cell);
        ad = ((BlobDataCell) cell).getBlobAddress();
    } else if (cell instanceof CellCollection) {
        CellCollection cdv = (CellCollection) cell;
        if (cdv.containsBlobWrapperCells()) {
            Iterator<DataCell> it = cdv.iterator();
            if (!(it instanceof BlobSupportDataCellIterator)) {
                LOGGER.coding("(Collection) DataCell of class \"" + cell.getClass().getSimpleName() + "\" contains Blobs, but does not " + "return an iterator supporting those " + "(expected " + BlobSupportDataCellIterator.class.getName() + ", got " + it.getClass().getName() + ")");
            }
            while (it.hasNext()) {
                DataCell n = it instanceof BlobSupportDataCellIterator ? ((BlobSupportDataCellIterator) it).nextWithBlobSupport() : it.next();
                DataCell correctedCell = handleIncomingBlob(n, col, totalColCount, copyForVersionHop, forceCopyOfBlobsArg);
                if (correctedCell != n) {
                    if (it instanceof BlobSupportDataCellIterator) {
                        BlobSupportDataCellIterator bsdi = (BlobSupportDataCellIterator) it;
                        bsdi.replaceLastReturnedWithWrapperCell(correctedCell);
                    } else {
                    // coding problem was reported above.
                    }
                }
            }
        }
        return cell;
    } else {
        // ordinary cell (e.g. double cell)
        return cell;
    }
    boolean forceCopyOfBlobs = forceCopyOfBlobsArg;
    Buffer ownerBuffer;
    if (ad != null) {
        // either copying from or to an isolated buffer (or both)
        forceCopyOfBlobs |= ad.getBufferID() == -1 || getBufferID() == -1;
        // (and this is not an ordinary buffer (but a BufferedDataCont.)
        if (ad.getBufferID() == getBufferID() && getBufferID() != -1) {
            ownerBuffer = this;
        } else {
            // table that's been created somewhere in the workflow
            ContainerTable t = m_globalRepository.get(ad.getBufferID());
            ownerBuffer = t != null ? t.getBuffer() : null;
        }
        /* this can only be true if the argument row contains wrapper
             * cells for blobs that do not have a buffer set; that is,
             * someone took a BlobDataCell from a predecessor node
             * (ad != null) and put it manually into a new wrapper cell
             * (wc != null) - by doing that you loose the buffer info
             * (wc.getBuffer == null) */
        if (isWrapperCell && wc.getBuffer() == null) {
            wc.setAddressAndBuffer(ad, ownerBuffer);
        }
    } else {
        ownerBuffer = null;
    }
    // if we have to make a clone of the blob cell (true if
    // isCopyOfExisting is true and the blob address corresponds to the next
    // assignable m_indicesOfBlobInColumns[col])
    boolean isToCloneForVersionHop = false;
    if (copyForVersionHop) {
        isToCloneForVersionHop = ad != null && ad.getBufferID() == getBufferID();
        // buffer multiple times -- don't copy the duplicates
        if (isToCloneForVersionHop && m_indicesOfBlobInColumns == null) {
            // first to assign
            isToCloneForVersionHop = ad.getIndexOfBlobInColumn() == 0;
            assert isToCloneForVersionHop : "Clone of buffer does not return blobs in order";
        } else if (isToCloneForVersionHop && m_indicesOfBlobInColumns != null) {
            isToCloneForVersionHop = ad.getIndexOfBlobInColumn() == m_indicesOfBlobInColumns[col];
        }
    }
    // if we have to clone the blob because the forceCopyOfBlobs flag is
    // on (e.g. because the owning node is a loop end node)
    boolean isToCloneDueToForceCopyOfBlobs = false;
    // don't overwrite the deep-clone
    if (forceCopyOfBlobs && !isToCloneForVersionHop) {
        if (m_copiedBlobsMap == null) {
            m_copiedBlobsMap = new HashMap<BlobAddress, BlobAddress>();
        }
        // if not previously copied into this buffer
        if (ad != null) {
            BlobAddress previousCopyAddress = m_copiedBlobsMap.get(ad);
            if (previousCopyAddress == null) {
                isToCloneDueToForceCopyOfBlobs = true;
                if (isWrapperCell && ownerBuffer == null) {
                    ownerBuffer = ((BlobWrapperDataCell) cell).getBuffer();
                }
            } else {
                return new BlobWrapperDataCell(this, previousCopyAddress, cl);
            }
        }
    }
    // we have to make a clone
    if (ownerBuffer == null || isToCloneForVersionHop || isToCloneDueToForceCopyOfBlobs) {
        // need to set ownership if this blob was not assigned yet
        // or has been assigned to an unlinked (i.e. local) buffer
        boolean isCompress = ad != null ? ad.isUseCompression() : isUseCompressionForBlobs(cl);
        BlobAddress rewrite = new BlobAddress(m_bufferID, col, isCompress);
        if (ad == null) {
            // take ownership
            if (isWrapperCell) {
                ((BlobWrapperDataCell) cell).setAddressAndBuffer(rewrite, this);
            } else {
                ((BlobDataCell) cell).setBlobAddress(rewrite);
            }
            ad = rewrite;
        }
        if (m_indicesOfBlobInColumns == null) {
            m_indicesOfBlobInColumns = new int[totalColCount];
        }
        // to buffer to copy the blob from (if at all)
        Buffer b = null;
        if (isToCloneDueToForceCopyOfBlobs) {
            b = ownerBuffer;
            m_copiedBlobsMap.put(ad, rewrite);
        } else {
            ContainerTable tbl = m_localRepository.get(ad.getBufferID());
            b = tbl == null ? null : tbl.getBuffer();
        }
        if (b != null && !isToCloneForVersionHop) {
            int indexBlobInCol = m_indicesOfBlobInColumns[col]++;
            rewrite.setIndexOfBlobInColumn(indexBlobInCol);
            File source = b.getBlobFile(ad.getIndexOfBlobInColumn(), ad.getColumn(), false, ad.isUseCompression());
            File dest = getBlobFile(indexBlobInCol, col, true, ad.isUseCompression());
            FileUtil.copy(source, dest);
            wc = new BlobWrapperDataCell(this, rewrite, cl);
        } else {
            BlobDataCell bc;
            if (isWrapperCell) {
                DataCell c = ((BlobWrapperDataCell) cell).getCell();
                bc = c.isMissing() ? null : (BlobDataCell) c;
            } else {
                bc = (BlobDataCell) cell;
            }
            // to take an error along
            if (bc != null) {
                if (m_outputWriter == null) {
                    ensureTempFileExists();
                    initOutputWriter(m_binFile);
                }
                writeBlobDataCell(bc, rewrite);
                wc = new BlobWrapperDataCell(this, rewrite, cl, bc);
            } else {
                wc = new BlobWrapperDataCell(this, rewrite, cl);
            }
        }
        m_containsBlobs = true;
    } else {
        // blob has been saved in one of the predecessor nodes
        if (isWrapperCell) {
            wc = (BlobWrapperDataCell) cell;
        } else {
            wc = new BlobWrapperDataCell(ownerBuffer, ad, cl);
        }
    }
    return wc;
}