private DataTableSpec[] getOutSpec(DataType type) {
DataTableSpec[] ret = new DataTableSpec[1];
DataColumnSpec[] allColSpecs = new DataColumnSpec[1];
allColSpecs[0] =
new DataColumnSpecCreator("column 0", ListCell.getCollectionType(type)).createSpec();
DataTableSpec outputSpec = new DataTableSpec(allColSpecs);
ret[0] = outputSpec;
return ret;
}
/** {@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};
}