private Map<Triple<DendrogramNode, RowKey, RowKey>, Number> visit(
final DendrogramNode root,
final Map<Triple<DendrogramNode, RowKey, RowKey>, Number> m,
final Map<RowKey, DistanceVectorDataValue> d,
final int allLeaves,
final ExecutionContext exec)
throws CanceledExecutionException {
if (root.isLeaf()) {
final RowKey key = RowKeyHelper.getKey(root);
return Collections.singletonMap(Triple.apply(root, key, key), (Number) Double.valueOf(0));
}
final DendrogramNode w = root.getFirstSubnode();
final Map<Triple<DendrogramNode, RowKey, RowKey>, Number> leftM =
visit(w, m, d, allLeaves, exec);
final DendrogramNode x = root.getSecondSubnode();
final Map<Triple<DendrogramNode, RowKey, RowKey>, Number> rightM =
visit(x, m, d, allLeaves, exec);
final Map<Triple<DendrogramNode, RowKey, RowKey>, Number> ret =
new HashMap<Triple<DendrogramNode, RowKey, RowKey>, Number>(leftM);
ret.putAll(rightM);
final Set<RowKey> leftKeys = computeLeaves(w);
final Set<RowKey> rightKeys = computeLeaves(x);
computeM(root, d, w, x, rightM, ret, leftKeys, rightKeys);
exec.checkCanceled();
computeM(root, d, x, w, leftM, ret, rightKeys, leftKeys);
exec.setProgress(((double) leftKeys.size() + rightKeys.size()) / allLeaves);
return ret;
}
@Override
protected PortObject[] execute(final PortObject[] inObjects, final ExecutionContext exec)
throws Exception {
// Initialising
exec.setMessage("initialising");
final ClusterTreeModel model = (ClusterTreeModel) inObjects[0];
final BufferedDataTable data = (BufferedDataTable) inObjects[1];
final int distanceColumnIdx =
data.getDataTableSpec().findColumnIndex(distanceColumnModel.getColumnName());
final Map<RowKey, DistanceVectorDataValue> distanceMatrix =
new HashMap<RowKey, DistanceVectorDataValue>();
for (final DataRow dataRow : data) {
final DistanceVectorDataValue distanceVector =
(DistanceVectorDataValue) dataRow.getCell(distanceColumnIdx);
distanceMatrix.put(dataRow.getKey(), distanceVector);
}
exec.setMessage("computing");
final DendrogramNode origRoot = model.getRoot();
final Map<Triple<DendrogramNode, RowKey, RowKey>, Number> m =
visit(
origRoot,
new HashMap<Triple<DendrogramNode, RowKey, RowKey>, Number>(),
distanceMatrix,
model.getClusterDistances().length + 1,
exec.createSilentSubExecutionContext(.9));
final Map<RowKey, Pair<DataRow, Integer>> rows = new HashMap<RowKey, Pair<DataRow, Integer>>();
int idx = 0;
for (final DataRow dataRow : data) {
rows.put(dataRow.getKey(), Pair.create(dataRow, Integer.valueOf(idx++)));
}
exec.setMessage("creating final tree");
final ClusterViewNode tree = buildNewTree(convertM(m), origRoot, rows, exec).getO1();
final ArrayList<DistanceVectorDataValue>
origList = new ArrayList<DistanceVectorDataValue>(model.getClusterDistances().length + 1),
newList = new ArrayList<DistanceVectorDataValue>(model.getClusterDistances().length + 1);
flatten(origRoot, origList, distanceMatrix);
exec.checkCanceled();
flatten(tree, newList, distanceMatrix);
logger.info("Before: " + sumDistance(origList));
logger.info("After: " + sumDistance(newList));
final ClusterTreeModel clusterTreeModel =
new ClusterTreeModel(
(DataTableSpec) model.getSpec(),
tree,
model.getClusterDistances(),
model.getClusterDistances().length + 1) {
@Override
public String getSummary() {
return "Before: " + sumDistance(origList) + "\nAfter: " + sumDistance(newList);
}
};
return new PortObject[] {clusterTreeModel};
}
@Override
protected BufferedDataTable[] execute(
final BufferedDataTable[] inData, final ExecutionContext exec) throws Exception {
final BufferedDataContainer con = exec.createDataContainer(m_spec);
final ImgPlusCellFactory imgCellFactory = new ImgPlusCellFactory(exec);
int id = 1;
for (final SerializableSetting<Img<T>[]> conf : m_kernelList.getObjects()) {
for (final Img<T> img : conf.get()) {
if (img instanceof ImgPlus<?>) {
con.addRowToTable(
new DefaultRow("Kernel " + id, imgCellFactory.createCell((ImgPlus<T>) img)));
} else {
con.addRowToTable(
new DefaultRow("Kernel " + id, imgCellFactory.createCell(new ImgPlus<T>(img))));
}
id++;
}
}
con.close();
// data for the table cell view
m_data = con.getTable();
return new BufferedDataTable[] {m_data};
}
@Test
public void shouldTransform() throws CanceledExecutionException {
OutputTransformer<ITSDataType> transformer = new ITSAdapterTransformer(createDataColumnSpec());
List<ITSDataType> entries = Lists.newArrayList(create("Bug1"), create("Bug2"));
BufferedDataContainer container = mock(BufferedDataContainer.class);
ExecutionContext exec = mock(ExecutionContext.class);
when(exec.createDataContainer(Mockito.any(DataTableSpec.class))).thenReturn(container);
// when
transformer.transform(entries, exec);
// then
verify(container, times(2)).addRowToTable(Mockito.any(DataRow.class));
}
protected void waitForCompletion(MuscleClient cli, ExecutionContext exec, String jobId)
throws Exception {
if (jobId.length() > 0) {
int check_period = 20 * 1000; // every 10s
String status = "PENDING";
int retry = 0;
while (status.equals("PENDING") || status.equals("RUNNING")) {
try {
logger.info("Waiting for " + jobId);
status = cli.checkStatus(jobId);
if (status.equals("RUNNING") || status.equals("PENDING")) {
logger.info(jobId + " " + status + ", sleeping for " + check_period + " milliseconds");
// check ten times each check_period to see if the user pressed cancel
for (int i = 0; i < 10; i++) {
Thread.sleep(check_period / 10);
exec.checkCanceled();
}
// each time job is still going, we double check_period to reduce likelihood of
// overloading EBI
check_period *= 2;
if (check_period > 200000) {
check_period = 200000;
}
}
if (status == "FAILED") {
logger.error("MUSCLE job failed: " + jobId);
}
} catch (IOException e) {
if (retry < 3) {
logger.warn(
"Unable to check job " + jobId + " retrying (after linear-backoff delay)... ");
Thread.sleep(((420 * retry) + 120) * 1000);
status = "PENDING";
retry++;
} else {
throw new Exception("Cannot check job " + jobId + " via MUSCLE (EBI)... aborting" + e);
}
}
}
} else {
throw new Exception("Bogus EBI job id... aborting!");
}
}
@Override
protected BufferedDataTable[] execute(BufferedDataTable[] inData, ExecutionContext exec)
throws Exception {
BufferedDataContainer cont = exec.createDataContainer(outspec[0]);
List<DataCell> cells = new ArrayList<DataCell>();
CloseableRowIterator iter = inData[0].iterator();
while (iter.hasNext()) {
cells.add(iter.next().getCell(0));
}
ListCell lc = CollectionCellFactory.createListCell(cells);
DataRow row = new DefaultRow("Row 1", lc);
cont.addRowToTable(row);
cont.close();
return new BufferedDataTable[] {cont.getTable()};
}
private static Triple<ClusterViewNode, RowKey, RowKey> buildNewTree(
final Map<DendrogramNode, Map<Pair<RowKey, RowKey>, Number>> m,
final DendrogramNode root,
final Map<RowKey, Pair<DataRow, Integer>> rows,
ExecutionContext exec)
throws CanceledExecutionException {
if (root.isLeaf()) {
final Pair<DataRow, Integer> leafRow = rows.get(RowKeyHelper.getKey(root));
return Triple.apply(
new ClusterViewNode(leafRow.getFirst().getKey()),
leafRow.getFirst().getKey(),
leafRow.getFirst().getKey());
}
final Triple<ClusterViewNode, RowKey, RowKey> firstTree =
buildNewTree(m, root.getFirstSubnode(), rows, exec);
exec.checkCanceled();
final Triple<ClusterViewNode, RowKey, RowKey> secondTree =
buildNewTree(m, root.getSecondSubnode(), rows, exec);
final Map<Pair<RowKey, RowKey>, Number> map = m.get(root);
Pair<RowKey, RowKey> pairNoChange = Pair.create(firstTree.getO3(), secondTree.getO2());
if (!map.containsKey(pairNoChange)) {
pairNoChange = flip(pairNoChange);
}
Pair<RowKey, RowKey> pairChange = Pair.create(secondTree.getO3(), firstTree.getO2());
if (!map.containsKey(pairChange)) {
pairChange = flip(pairChange);
}
assert map.containsKey(pairNoChange);
assert map.containsKey(pairChange);
final double noChangeValue = map.get(pairNoChange).doubleValue();
final double changeValue = map.get(pairChange).doubleValue();
if (noChangeValue > changeValue) {
return Triple.apply(
new ClusterViewNode(firstTree.getO1(), secondTree.getO1(), root.getDist()),
firstTree.getO2(),
secondTree.getO3());
}
return Triple.apply(
new ClusterViewNode(secondTree.getO1(), firstTree.getO1(), root.getDist()),
secondTree.getO2(),
firstTree.getO3());
}
/** {@inheritDoc} */
@Override
protected BufferedDataTable[] execute(
final BufferedDataTable[] inData, final ExecutionContext exec) throws Exception {
if (m_email.equals(DEFAULT_EMAIL)) {
throw new Exception(
"You must set a valid E-Mail for EBI to contact you in the event of problems with the service!");
}
int n_rows = inData[0].getRowCount();
int seq_idx =
inData[0].getSpec().findColumnIndex(((SettingsModelString) m_seq_col).getStringValue());
int accsn_idx =
inData[0].getSpec().findColumnIndex(((SettingsModelString) m_accsn_col).getStringValue());
if (seq_idx < 0 || accsn_idx < 0) {
throw new Exception("Cannot find columns... valid data?");
}
int done = 0;
// create the output columns (raw format for use with R)
DataTableSpec outputSpec = new DataTableSpec(inData[0].getDataTableSpec(), make_output_spec());
BufferedDataContainer container = exec.createDataContainer(outputSpec, false, 0);
// instantiate MUSCLE client
MuscleClient cli = new MuscleClient();
// each row is a separate MUSCLE job, the sequences are in one collection cell, the accessions
// (IDs) in the other
RowIterator it = inData[0].iterator();
while (it.hasNext()) {
DataRow r = it.next();
ListCell seqs = (ListCell) r.getCell(seq_idx);
ListCell accsns = (ListCell) r.getCell(accsn_idx);
if (seqs.size() != accsns.size()) {
throw new Exception(
"Every sequence must have a corresponding accession: error at row "
+ r.getKey().getString());
}
if (seqs.size() < 1) {
throw new Exception("Cannot MUSCLE zero sequences: error at row " + r.getKey().getString());
}
if (seqs.size() > 1000) {
throw new Exception("Too many sequences in row " + r.getKey().getString());
}
// dummy a fake "FASTA" file (in memory) and then submit that to MUSCLE@EBI along with other
// necessary parameters
StringBuffer seq_as_fasta = new StringBuffer();
for (int i = 0; i < seqs.size(); i++) {
seq_as_fasta.append(">");
seq_as_fasta.append(accsns.get(i).toString());
seq_as_fasta.append("\n");
seq_as_fasta.append(seqs.get(i).toString());
seq_as_fasta.append("\n");
}
// System.err.println(seq_as_fasta);
// lodge the muscle job and store the results in the output table
InputParameters ip = new InputParameters();
ip.setSequence(seq_as_fasta.toString());
// start the job
String jobId = cli.runApp(m_email.getStringValue(), r.getKey().getString(), ip);
exec.checkCanceled();
exec.setProgress(((double) done) / n_rows, "Executing " + jobId);
Thread.sleep(20 * 1000); // 20 seconds
waitForCompletion(cli, exec, jobId);
done++;
// process results and add them into the table...
// 1. fasta alignment data
byte[] bytes = cli.getSrvProxy().getResult(jobId, "aln-fasta", null);
DataCell[] cells = new DataCell[3];
cells[0] = new StringCell(jobId);
// compute the base64 encoded phylip aligned sequences suitable for use by R's phangorn
// package
String fasta = new String(bytes);
String ret = fasta2phylip(fasta);
// it must be encoded (I chose base64) as it is common to both Java and R and it must be
// encoded due to containing multiple lines, which confuses the CSV passed between KNIME and R
String rk = r.getKey().getString();
DataCell mac = AlignmentCellFactory.createCell(fasta, AlignmentType.AL_AA);
if (mac instanceof MultiAlignmentCell) m_muscle_map.put(rk, (MultiAlignmentCell) mac);
cells[1] = mac;
bytes = cli.getSrvProxy().getResult(jobId, "out", null);
cells[2] = new StringCell("<html><pre>" + new String(bytes));
container.addRowToTable(new JoinedRow(r, new DefaultRow(r.getKey(), cells)));
}
container.close();
BufferedDataTable out = container.getTable();
return new BufferedDataTable[] {out};
}
@Override
protected BufferedDataTable[] execute(BufferedDataTable[] inData, ExecutionContext exec)
throws Exception {
List<File> inputFiles =
FileSelectPanel.getInputFiles(propInputDir.getStringValue(), getAllowedFileExtensions());
if (inputFiles.isEmpty()) {
throw new RuntimeException("No files selected");
}
// first group files into plate-groups
Map<String, List<File>> plateFiles = splitFilesIntoPlates(inputFiles);
if (inputFiles.isEmpty()) {
throw new RuntimeException("No valid envision-files in selection " + inputFiles);
}
// split files
List<String> allAttributes = mergeAttributes(plateFiles);
List<Attribute> colAttributes = compileColumnModel(allAttributes);
DataTableSpec outputSpec = AttributeUtils.compileTableSpecs(colAttributes);
BufferedDataContainer container = exec.createDataContainer(outputSpec);
// populate the table
int fileCounter = 0, rowCounter = 0;
for (String barcode : plateFiles.keySet()) {
logger.info("Processing plate " + barcode);
Plate plate = new Plate();
// invalidate plate-dims as these become fixed in the loop
plate.setNumColumns(-1);
plate.setNumRows(-1);
for (File file : plateFiles.get(barcode)) {
String attributeName = getAttributeNameOfEnvisionFile(file);
parseFile(plate, attributeName, file);
BufTableUtils.updateProgress(exec, fileCounter++, inputFiles.size());
}
// now create the data-rows for this table
for (Well well : plate.getWells()) {
if (well.getReadOutNames().isEmpty()) {
continue;
}
DataCell[] knimeRow = new DataCell[colAttributes.size()];
// first add the barcode-column
knimeRow[0] = new StringCell(barcode);
knimeRow[1] = colAttributes.get(1).createCell(well.getPlateRow());
knimeRow[2] = colAttributes.get(2).createCell(well.getPlateColumn());
for (String attributeName : allAttributes) {
int rowIndex = allAttributes.indexOf(attributeName);
Double value = well.getReadout(attributeName);
if (value != null) {
knimeRow[3 + rowIndex] = new DoubleCell(value);
} else {
knimeRow[3 + rowIndex] = DataType.getMissingCell();
}
}
DataRow tableRow = new DefaultRow(new RowKey("" + rowCounter++), knimeRow);
container.addRowToTable(tableRow);
}
}
container.close();
return new BufferedDataTable[] {container.getTable()};
}
@Override
protected BufferedDataTable[] execute(BufferedDataTable[] inData, ExecutionContext exec)
throws Exception {
BufferedDataTable input = inData[0];
// Get the condition attribute
Attribute treatmentAttribute =
new InputTableAttribute(this.treatmentAttribute.getStringValue(), input);
// Get the library and reference condition names
String libraryName = AbstractScreenTrafoModel.getAndValidateTreatment(reference);
String referenceName = AbstractScreenTrafoModel.getAndValidateTreatment(library);
// Get the parameter and make sure there all double value columns
List<Attribute> parameters = getParameterList(input);
// Split the columns according to groups contained in the condition column
Map<String, List<DataRow>> groupedRows = AttributeUtils.splitRows(input, treatmentAttribute);
List<DataRow> libraryRows = groupedRows.get(libraryName);
List<DataRow> referenceRows = groupedRows.get(referenceName);
int progress = parameters.size();
BufTableUtils.updateProgress(exec, progress / 2, progress);
// Initialize
BufferedDataContainer container = exec.createDataContainer(new DataTableSpec(getListSpec()));
MutualInformation mutualinfo = new MutualInformation();
mutualinfo.set_base(logbase.getDoubleValue());
mutualinfo.set_method(method.getStringValue());
mutualinfo.set_axeslinking(linkaxes.getBooleanValue());
DataCell[] cells = new DataCell[container.getTableSpec().getNumColumns()];
int p = 0;
// Calculate mutual information
for (Attribute parameter : parameters) {
Double[] x = getDataVec(libraryRows, parameter);
Double[] y = getDataVec(referenceRows, parameter);
mutualinfo.set_vectors(x, y);
if (binning.getIntValue() == 0) {
mutualinfo.set_binning();
} else {
mutualinfo.set_binning(binning.getIntValue());
}
int[] bins = mutualinfo.get_binning();
Double[] res = mutualinfo.calculate();
cells[0] = new StringCell(parameter.getName());
cells[1] = new DoubleCell(res[0]);
cells[2] = new DoubleCell(res[1]);
cells[3] = new DoubleCell(res[2]);
cells[4] = new IntCell(bins[0]);
cells[5] = new IntCell(bins[1]);
cells[6] = new DoubleCell(mutualinfo.get_logbase());
cells[7] = new StringCell(mutualinfo.get_method());
container.addRowToTable(new DefaultRow("row" + p, cells));
BufTableUtils.updateProgress(exec, (progress + p++) / 2, progress);
exec.checkCanceled();
}
container.close();
return new BufferedDataTable[] {container.getTable()};
}
