Examples with InstanceIndexService eu.esdihumboldt.hale.common.instance.index.InstanceIndexService used on opensource projects

Example 1 with InstanceIndexService

use of eu.esdihumboldt.hale.common.instance.index.InstanceIndexService in project hale by halestudio.

the class Transformation method transform.

/**
 * Transform the given instances, according to the given alignment.
 *
 * @param sources the collection of source instances
 * @param targetSink the target sink
 * @param exportJob the export job
 * @param validationJob the validation job, may be <code>null</code>
 * @param alignment the alignment, may not be changed outside this method
 * @param sourceSchema the source schema
 * @param reportHandler the report handler
 * @param serviceProvider the service provider in the transformation context
 * @param processId the identifier for the transformation process, may be
 *            <code>null</code> if grouping the jobs to a job family is not
 *            necessary
 * @return the future representing the successful completion of the
 *         transformation (note that a successful completion doesn't
 *         necessary mean there weren't any internal transformation errors)
 */
public static ListenableFuture<Boolean> transform(InstanceCollection sources, final TransformationSink targetSink, final ExportJob exportJob, final ValidationJob validationJob, final Alignment alignment, SchemaSpace sourceSchema, final ReportHandler reportHandler, final ServiceProvider serviceProvider, final Object processId) {
    final SettableFuture<Boolean> result = SettableFuture.create();
    final InstanceCollection sourceToUse;
    // Check whether to create a temporary database or not.
    // Currently do not create a temporary DB is there are Retypes/Creates
    // only.
    boolean useTempDatabase = false;
    final LocalOrientDB db;
    for (Cell cell : alignment.getActiveTypeCells()) if (!isStreamingTypeTransformation(cell.getTransformationIdentifier())) {
        useTempDatabase = true;
        break;
    }
    // Create temporary database if necessary.
    if (useTempDatabase) {
        // create db
        File tmpDir = Files.createTempDir();
        db = new LocalOrientDB(tmpDir);
        tmpDir.deleteOnExit();
        // get instance collection
        // sourceToUse = new BrowseOrientInstanceCollection(db, sourceSchema, DataSet.SOURCE);
        // only yield instances that were actually inserted
        // this is also done in OrientInstanceService
        // TODO make configurable?
        sourceToUse = FilteredInstanceCollection.applyFilter(new BrowseOrientInstanceCollection(db, sourceSchema, DataSet.SOURCE), new Filter() {

            @Override
            public boolean match(Instance instance) {
                if (instance instanceof OInstance) {
                    return ((OInstance) instance).isInserted();
                }
                return true;
            }
        });
    } else {
        sourceToUse = new StatsCountInstanceCollection(sources, reportHandler);
        db = null;
    }
    // create transformation job
    final AbstractTransformationJob transformJob = new AbstractTransformationJob("Transformation") {

        /**
         * @see org.eclipse.core.runtime.jobs.Job#run(org.eclipse.core.runtime.IProgressMonitor)
         */
        @Override
        protected IStatus run(IProgressMonitor monitor) {
            TransformationService transformationService = HalePlatform.getService(TransformationService.class);
            TransformationReport report = transformationService.transform(alignment, sourceToUse, targetSink, serviceProvider, new ProgressMonitorIndicator(monitor));
            try {
                // publish report
                reportHandler.publishReport(report);
                if (report.isSuccess()) {
                    return Status.OK_STATUS;
                } else {
                    return ERROR_STATUS;
                }
            } finally {
                // and may lead to the transformation report being lost
                if (monitor.isCanceled()) {
                    targetSink.done(true);
                    return Status.CANCEL_STATUS;
                } else {
                    targetSink.done(false);
                }
            }
        }
    };
    // set process IDs to group jobs in a job family
    if (processId != null) {
        transformJob.setProcessId(processId);
        exportJob.setProcessId(processId);
        if (validationJob != null) {
            validationJob.setProcessId(processId);
        }
    }
    exportJob.setUser(true);
    // the jobs should cancel each other
    transformJob.addJobChangeListener(new JobChangeAdapter() {

        @Override
        public void done(IJobChangeEvent event) {
            if (!event.getResult().isOK()) {
                // log transformation job error (because it otherwise gets
                // lost)
                String msg = "Error during transformation";
                if (event.getResult().getMessage() != null) {
                    msg = ": " + event.getResult().getMessage();
                }
                log.error(msg, event.getResult().getException());
                // failing transformation is done by cancelling the export
                exportJob.cancel();
            }
            if (db != null) {
                db.delete();
            }
        }
    });
    // after export is done, validation should run
    exportJob.addJobChangeListener(new JobChangeAdapter() {

        @Override
        public void done(IJobChangeEvent event) {
            if (!event.getResult().isOK()) {
                transformJob.cancel();
                // failure
                failure(result, event);
            } else {
                if (validationJob == null) {
                    // success
                    result.set(true);
                } else {
                    // schedule the validation job
                    validationJob.schedule();
                }
            }
        }
    });
    // validation ends the process
    if (validationJob != null) {
        validationJob.addJobChangeListener(new JobChangeAdapter() {

            @Override
            public void done(IJobChangeEvent event) {
                if (!event.getResult().isOK()) {
                    // failure
                    failure(result, event);
                } else {
                    // success
                    result.set(true);
                }
            }
        });
    }
    if (useTempDatabase) {
        // Initialize instance index with alignment
        InstanceIndexService indexService = serviceProvider.getService(InstanceIndexService.class);
        indexService.addPropertyMappings(alignment.getActiveTypeCells(), serviceProvider);
        // run store instance job first...
        Job storeJob = new StoreInstancesJob("Load source instances into temporary database", db, sources, serviceProvider, reportHandler, true) {

            @Override
            protected void onComplete() {
            // onComplete is also called if monitor is cancelled...
            }

            @Override
            public boolean belongsTo(Object family) {
                if (processId == null) {
                    return super.belongsTo(family);
                }
                return AbstractTransformationJob.createFamily(processId).equals(family);
            }
        };
        // and schedule jobs on successful completion
        storeJob.addJobChangeListener(new JobChangeAdapter() {

            @Override
            public void done(IJobChangeEvent event) {
                if (event.getResult().isOK()) {
                    exportJob.schedule();
                    transformJob.schedule();
                } else {
                    failure(result, event);
                }
            }
        });
        storeJob.schedule();
    } else {
        // otherwise feed InstanceProcessors directly from the
        // InstanceCollection...
        // TODO Implement differently, not w/ PseudoInstanceReference which
        // will cause memory problems
        // final InstanceProcessingExtension ext = new InstanceProcessingExtension(
        // serviceProvider);
        // final List<InstanceProcessor> processors = ext.getInstanceProcessors();
        // 
        // ResourceIterator<Instance> it = sourceToUse.iterator();
        // try {
        // while (it.hasNext()) {
        // Instance instance = it.next();
        // 
        // ResolvableInstanceReference resolvableRef = new ResolvableInstanceReference(
        // new PseudoInstanceReference(instance), sourceToUse);
        // processors.forEach(p -> p.process(instance, resolvableRef));
        // 
        // }
        // } finally {
        // it.close();
        // }
        // ...and schedule jobs
        exportJob.schedule();
        transformJob.schedule();
    }
    return result;
}

Example 2 with InstanceIndexService

use of eu.esdihumboldt.hale.common.instance.index.InstanceIndexService in project hale by halestudio.

the class StoreInstancesJob method run.

/**
 * @see Job#run(IProgressMonitor)
 */
@Override
public IStatus run(IProgressMonitor monitor) {
    boolean exactProgress = instances.hasSize();
    monitor.beginTask("Store instances in database", (exactProgress) ? (instances.size()) : (IProgressMonitor.UNKNOWN));
    AtomicInteger count = new AtomicInteger(0);
    TObjectIntHashMap<QName> typeCount = new TObjectIntHashMap<>();
    if (report != null) {
        // set the correct start time
        report.setStartTime(new Date());
    }
    // get database connection
    DatabaseReference<ODatabaseDocumentTx> ref = database.openWrite();
    ODatabaseDocumentTx db = ref.getDatabase();
    ATransaction trans = log.begin("Store instances in database");
    try {
        // use intent
        db.declareIntent(new OIntentMassiveInsert());
        // Find all the InstanceProcessors to feed them the stored Instances
        final List<InstanceProcessor> processors;
        if (doProcessing) {
            final InstanceProcessingExtension ext = new InstanceProcessingExtension(serviceProvider);
            processors = ext.getInstanceProcessors();
        } else {
            processors = Collections.emptyList();
        }
        BrowseOrientInstanceCollection browser = new BrowseOrientInstanceCollection(database, null, DataSet.SOURCE);
        final InstanceIndexService indexService;
        if (doProcessing) {
            indexService = serviceProvider.getService(InstanceIndexService.class);
        } else {
            indexService = null;
        }
        // TODO decouple next() and save()?
        SimpleLogContext.withLog(report, () -> {
            if (report != null && instances instanceof LogAware) {
                ((LogAware) instances).setLog(report);
            }
            ResourceIterator<Instance> it = instances.iterator();
            int size = instances.size();
            try {
                while (it.hasNext() && !monitor.isCanceled()) {
                    // last count update
                    long lastUpdate = 0;
                    if (report != null && instances instanceof LogAware) {
                        ((LogAware) instances).setLog(report);
                    }
                    Instance instance = it.next();
                    // further processing before storing
                    processInstance(instance);
                    // get/create OInstance
                    OInstance conv = ((instance instanceof OInstance) ? ((OInstance) instance) : (new OInstance(instance)));
                    conv.setInserted(true);
                    // update the instance to store, e.g. generating
                    // metadata
                    updateInstance(conv);
                    ODatabaseRecordThreadLocal.INSTANCE.set(db);
                    // configure the document
                    ODocument doc = conv.configureDocument(db);
                    // and save it
                    doc.save();
                    // Create an InstanceReference for the saved instance
                    // and
                    // feed it to all known InstanceProcessors. The
                    // decoration
                    // with ResolvableInstanceReference allows the
                    // InstanceProcessors to resolve the instances if
                    // required.
                    OrientInstanceReference oRef = new OrientInstanceReference(doc.getIdentity(), conv.getDataSet(), conv.getDefinition());
                    IdentifiableInstanceReference idRef = new IdentifiableInstanceReference(oRef, doc.getIdentity());
                    ResolvableInstanceReference resolvableRef = new ResolvableInstanceReference(idRef, browser);
                    processors.forEach(p -> p.process(instance, resolvableRef));
                    if (indexService != null) {
                        indexService.add(instance, resolvableRef);
                    }
                    count.incrementAndGet();
                    TypeDefinition type = instance.getDefinition();
                    if (type != null) {
                        typeCount.adjustOrPutValue(type.getName(), 1, 1);
                    }
                    if (exactProgress) {
                        monitor.worked(1);
                    }
                    long now = System.currentTimeMillis();
                    if (now - lastUpdate > 100) {
                        // only update every 100
                        // milliseconds
                        monitor.subTask(MessageFormat.format("{0}{1} instances processed", String.valueOf(count.get()), size != InstanceCollection.UNKNOWN_SIZE ? "/" + String.valueOf(size) : ""));
                        lastUpdate = now;
                    }
                }
            } finally {
                it.close();
                if (report != null && instances instanceof LogAware) {
                    ((LogAware) instances).setLog(null);
                }
            }
        });
        db.declareIntent(null);
    } catch (RuntimeException e) {
        if (report != null) {
            reportTypeCount(report, typeCount);
            report.error(new MessageImpl("Error storing instances in database", e));
            report.setSuccess(false);
            reportHandler.publishReport(report);
        }
        throw e;
    } finally {
        ref.dispose();
        trans.end();
        /*
			 * Reset instances to prevent memory leak. It seems Eclipse
			 * internally holds a reference to the job (in JobInfo and/or
			 * ProgressMonitorFocusJobDialog) and this results in the instance
			 * collection not being garbage collected. This is especially bad,
			 * if an in-memory instance collection is used, e.g. a
			 * DefaultInstanceCollection that is used when loading a Shapefile.
			 */
        instances = null;
    }
    try {
        onComplete();
    } catch (RuntimeException e) {
        String message = "Error while post processing stored instances";
        if (report != null) {
            report.error(new MessageImpl(message, e));
        } else {
            log.error(message, e);
        }
    }
    String message = MessageFormat.format("Stored {0} instances in the database.", count);
    if (monitor.isCanceled()) {
        String warn = "Loading instances was canceled, incomplete data set in the database.";
        if (report != null) {
            report.warn(new MessageImpl(warn, null));
        } else {
            log.warn(warn);
        }
    }
    if (report != null) {
        reportTypeCount(report, typeCount);
        report.setSuccess(true);
        report.setSummary(message);
        reportHandler.publishReport(report);
    } else {
        log.info(message);
    }
    monitor.done();
    return new Status((monitor.isCanceled()) ? (IStatus.CANCEL) : (IStatus.OK), "eu.esdihumboldt.hale.common.instance.orient", message);
}

Example 3 with InstanceIndexService

use of eu.esdihumboldt.hale.common.instance.index.InstanceIndexService in project hale by halestudio.

the class IndexJoinHandler method partitionInstances.

/**
 * @see eu.esdihumboldt.hale.common.align.transformation.function.InstanceHandler#partitionInstances(eu.esdihumboldt.hale.common.instance.model.InstanceCollection,
 *      java.lang.String,
 *      eu.esdihumboldt.hale.common.align.transformation.engine.TransformationEngine,
 *      com.google.common.collect.ListMultimap, java.util.Map,
 *      eu.esdihumboldt.hale.common.align.transformation.report.TransformationLog)
 */
@Override
public ResourceIterator<FamilyInstance> partitionInstances(InstanceCollection instances, String transformationIdentifier, TransformationEngine engine, ListMultimap<String, ParameterValue> transformationParameters, Map<String, String> executionParameters, TransformationLog log) throws TransformationException {
    if (transformationParameters == null || !transformationParameters.containsKey(PARAMETER_JOIN) || transformationParameters.get(PARAMETER_JOIN).isEmpty()) {
        throw new TransformationException("No join parameter defined");
    }
    JoinHandler fallbackHandler = new JoinHandler();
    InstanceIndexService indexService = serviceProvider.getService(InstanceIndexService.class);
    if (indexService == null) {
        log.warn(MessageFormat.format("Index service not available, falling back to join handler {0}", fallbackHandler.getClass().getCanonicalName()));
        return fallbackHandler.partitionInstances(instances, transformationIdentifier, engine, transformationParameters, executionParameters, log);
    }
    JoinParameter joinParameter = transformationParameters.get(PARAMETER_JOIN).get(0).as(JoinParameter.class);
    String validation = joinParameter.validate();
    if (validation != null) {
        throw new TransformationException("Join parameter invalid: " + validation);
    }
    List<TypeEntityDefinition> types = joinParameter.getTypes();
    JoinDefinition joinDefinition = JoinUtil.getJoinDefinition(joinParameter);
    // remember instances of first type to start join afterwards
    Collection<ResolvableInstanceReference> startInstances = new LinkedList<ResolvableInstanceReference>();
    List<Object> inputInstanceIds = new ArrayList<>();
    try (ResourceIterator<Instance> it = instances.iterator()) {
        while (it.hasNext()) {
            Instance i = InstanceDecorator.getRoot(it.next());
            // remember instances of first type
            if (i.getDefinition().equals(types.get(0).getDefinition())) {
                startInstances.add(new ResolvableInstanceReference(instances.getReference(i), instances));
            }
            if (!Identifiable.is(i)) {
                log.warn(MessageFormat.format("At least one instance does not have an ID, falling back to join handler {0}", fallbackHandler.getClass().getCanonicalName()));
                return fallbackHandler.partitionInstances(instances, transformationIdentifier, engine, transformationParameters, executionParameters, log);
            }
            inputInstanceIds.add(Identifiable.getId(i));
        }
    }
    return new IndexJoinIterator(startInstances, joinDefinition, indexService);
}

Example 4 with InstanceIndexService

use of eu.esdihumboldt.hale.common.instance.index.InstanceIndexService in project hale by halestudio.

the class IndexMergeHandler method partitionInstances.

/**
 * @see eu.esdihumboldt.cst.functions.core.merge.AbstractMergeHandler#partitionInstances(eu.esdihumboldt.hale.common.instance.model.InstanceCollection,
 *      java.lang.String,
 *      eu.esdihumboldt.hale.common.align.transformation.engine.TransformationEngine,
 *      com.google.common.collect.ListMultimap, java.util.Map,
 *      eu.esdihumboldt.hale.common.align.transformation.report.TransformationLog)
 */
@Override
public ResourceIterator<FamilyInstance> partitionInstances(InstanceCollection instances, String transformationIdentifier, TransformationEngine engine, ListMultimap<String, ParameterValue> transformationParameters, Map<String, String> executionParameters, TransformationLog log) throws TransformationException {
    PropertiesMergeHandler fallbackHandler = new PropertiesMergeHandler();
    InstanceIndexService indexService = serviceProvider.getService(InstanceIndexService.class);
    if (indexService == null) {
        log.warn(MessageFormat.format("Index service not available, falling back to merge handler {0}", fallbackHandler.getClass().getCanonicalName()));
        return fallbackHandler.partitionInstances(instances, transformationIdentifier, engine, transformationParameters, executionParameters, log);
    }
    final IndexMergeConfig mergeConfig = createMergeConfiguration(transformationParameters);
    QName typeName;
    try (ResourceIterator<Instance> it = instances.iterator()) {
        if (it.hasNext()) {
            typeName = it.next().getDefinition().getName();
        } else {
            // Nothing to partition
            return new ResourceIterator<FamilyInstance>() {

                @Override
                public boolean hasNext() {
                    return false;
                }

                @Override
                public FamilyInstance next() {
                    return null;
                }

                @Override
                public void close() {
                // Do nothing
                }
            };
        }
    }
    // Querying the index will yield a result over all instances. We must,
    // however, be able to operate only on the given input instances instead
    // of all instances.
    // We must, therefore, be able to uniquely identify every instance in
    // the index, so that we can retain from the index query only the
    // relevant instances.
    List<Object> inputInstanceIds = new ArrayList<>();
    try (ResourceIterator<Instance> it = instances.iterator()) {
        while (it.hasNext()) {
            Instance i = InstanceDecorator.getRoot(it.next());
            if (!Identifiable.is(i)) {
                log.warn(MessageFormat.format("At least one instance does not have an ID, falling back to merge handler {0}", fallbackHandler.getClass().getCanonicalName()));
                return fallbackHandler.partitionInstances(instances, transformationIdentifier, engine, transformationParameters, executionParameters, log);
            }
            inputInstanceIds.add(Identifiable.getId(i));
        }
    }
    Collection<Collection<ResolvableInstanceReference>> partitionedIndex = indexService.groupBy(typeName, mergeConfig.keyProperties);
    // Remove instance groups from the partitioned index where none of the
    // instances in the group are in the processed instances.
    partitionedIndex.removeIf(part -> !part.stream().map(ref -> ref.getId()).anyMatch(id -> inputInstanceIds.contains(id)));
    Iterator<Collection<ResolvableInstanceReference>> it = partitionedIndex.iterator();
    return new ResourceIterator<FamilyInstance>() {

        @Override
        public boolean hasNext() {
            return it.hasNext();
        }

        @Override
        public FamilyInstance next() {
            Collection<ResolvableInstanceReference> instanceRefs = it.next();
            InstanceCollection instancesToBeMerged = new DefaultInstanceCollection(instanceRefs.stream().map(ref -> ref.resolve()).collect(Collectors.toList()));
            return new FamilyInstanceImpl(merge(instancesToBeMerged, mergeConfig));
        }

        @Override
        public void close() {
        // TODO Auto-generated method stub
        }
    };
}