EDIT

// $Id$

/**

* Copyright (C) 2013 EDIT

* European Distributed Institute of Taxonomy

* http://www.e-taxonomy.eu

*

* The contents of this file are subject to the Mozilla Public License Version 1.1

* See LICENSE.TXT at the top of this package for the full license terms.

*/

package eu.etaxonomy.cdm.api.service.description;

import java.util.ArrayList;

import java.util.Arrays;

import java.util.HashMap;

import java.util.HashSet;

import java.util.Iterator;

import java.util.LinkedList;

import java.util.List;

import java.util.Map;

import java.util.Set;

import java.util.UUID;

import org.apache.log4j.Logger;

import org.hibernate.FlushMode;

import org.hibernate.HibernateException;

import org.hibernate.Session;

import org.hibernate.engine.spi.SessionFactoryImplementor;

import org.hibernate.search.Search;

import org.springframework.beans.factory.annotation.Autowired;

import org.springframework.orm.hibernate5.HibernateTransactionManager;

import org.springframework.stereotype.Service;

import org.springframework.transaction.TransactionDefinition;

import org.springframework.transaction.TransactionStatus;

import org.springframework.transaction.support.DefaultTransactionDefinition;

import eu.etaxonomy.cdm.api.service.IClassificationService;

import eu.etaxonomy.cdm.api.service.IDescriptionService;

import eu.etaxonomy.cdm.api.service.INameService;

import eu.etaxonomy.cdm.api.service.ITaxonService;

import eu.etaxonomy.cdm.api.service.ITermService;

import eu.etaxonomy.cdm.common.DynamicBatch;

import eu.etaxonomy.cdm.common.JvmLimitsException;

import eu.etaxonomy.cdm.common.monitor.IProgressMonitor;

import eu.etaxonomy.cdm.common.monitor.NullProgressMonitor;

import eu.etaxonomy.cdm.common.monitor.SubProgressMonitor;

import eu.etaxonomy.cdm.model.common.CdmBase;

import eu.etaxonomy.cdm.model.common.DefinedTermBase;

import eu.etaxonomy.cdm.model.common.Extension;

import eu.etaxonomy.cdm.model.common.ExtensionType;

import eu.etaxonomy.cdm.model.common.Marker;

import eu.etaxonomy.cdm.model.common.MarkerType;

import eu.etaxonomy.cdm.model.common.OrderedTermBase;

import eu.etaxonomy.cdm.model.description.DescriptionElementBase;

import eu.etaxonomy.cdm.model.description.DescriptionElementSource;

import eu.etaxonomy.cdm.model.description.Distribution;

import eu.etaxonomy.cdm.model.description.PresenceAbsenceTerm;

import eu.etaxonomy.cdm.model.description.TaxonDescription;

import eu.etaxonomy.cdm.model.location.NamedArea;

import eu.etaxonomy.cdm.model.name.Rank;

import eu.etaxonomy.cdm.model.taxon.Classification;

import eu.etaxonomy.cdm.model.taxon.Taxon;

import eu.etaxonomy.cdm.model.taxon.TaxonBase;

import eu.etaxonomy.cdm.persistence.dto.ClassificationLookupDTO;

/**

 *

 * <h2>GENERAL NOTES </h2>

 * <em>TODO: These notes are directly taken from original Transmission Engine Occurrence

 * version 14 written in Visual Basic and still need to be

 * adapted to the java version of the transmission engine!</em>

 *

 * <h3>summaryStatus</h3>

 *

 *   Each distribution information has a summaryStatus, this is an summary of the status codes

 *   as stored in the fields of emOccurrence native, introduced, cultivated, ...

 *   The summaryStatus seems to be equivalent to  the CDM DistributionStatus

 *

 * <h3>map generation</h3>

 *

 *   When generating maps from the accumulated distribution information some special cases have to be handled:

 * <ol>

 *   <li>if a entered or imported status information exist for the same area for which calculated (accumulated)

 *       data is available, the calculated data has to be given preference over other data.

 *   </li>

 *   <li>If there is an area with a sub area and both areas have the same calculated status only the subarea

 *       status should be shown in the map, whereas the super area should be ignored.

 *   </li>

 * </ol>

 *

 * @author Anton Güntsch (author of original Transmission Engine Occurrence version 14 written in Visual Basic)

 * @author Andreas Kohlbecker (2013, porting Transmission Engine Occurrence to Java)

 * @date Feb 22, 2013

 */

@Service

public class TransmissionEngineDistribution { //TODO extends IoBase?

    public static final String EXTENSION_VALUE_PREFIX = "transmissionEngineDistribution.priority:";

    public static final Logger logger = Logger.getLogger(TransmissionEngineDistribution.class);

    /**

     * only used for performance testing

     */

    final boolean ONLY_FISRT_BATCH = false;

    protected static final List<String> TAXONDESCRIPTION_INIT_STRATEGY = Arrays.asList(new String [] {

            "description.markers.markerType",

            "description.elements.markers.markerType",

            "description.elements.area",

            "description.elements.status",

            "description.elements.sources.citation.authorship",

//            "description.elements.sources.nameUsedInSource",

//            "description.elements.multilanguageText",

//            "name.status.type",

    });

    /**

     * A map which contains the status terms as key and the priority as value

     * The map will contain both, the PresenceTerms and the AbsenceTerms

     */

    private Map<PresenceAbsenceTerm, Integer> statusPriorityMap = null;

    @Autowired

    private IDescriptionService descriptionService;

    @Autowired

    private ITermService termService;

    @Autowired

    private ITaxonService taxonService;

    @Autowired

    private IClassificationService classificationService;

    @Autowired

    private INameService mameService;

    @Autowired

    private HibernateTransactionManager transactionManager;

    private List<PresenceAbsenceTerm> byAreaIgnoreStatusList = null;

    private List<PresenceAbsenceTerm> byRankIgnoreStatusList = null;

    private final Map<NamedArea, Set<NamedArea>> subAreaMap = new HashMap<NamedArea, Set<NamedArea>>();

    int byRankTicks = 300;

    int byAreasTicks = 100;

    private static final long BATCH_MIN_FREE_HEAP = 800  * 1024 * 1024;

    /**

     * ratio of the initially free heap which should not be used

     * during the batch processing. This amount of the heap is reserved

     * for the flushing of the session and to the index

     */

    private static final double BATCH_FREE_HEAP_RATIO = 0.9;

    private static final int BATCH_SIZE_BY_AREA = 1000;

    private static final int BATCH_SIZE_BY_RANK = 500;

    private long batchMinFreeHeap = BATCH_MIN_FREE_HEAP;

    /**

     * byAreaIgnoreStatusList contains by default:

     *  <ul>

     *    <li>AbsenceTerm.CULTIVATED_REPORTED_IN_ERROR()</li>

     *    <li>AbsenceTerm.INTRODUCED_REPORTED_IN_ERROR()</li>

     *    <li>AbsenceTerm.INTRODUCED_FORMERLY_INTRODUCED()</li>

     *    <li>AbsenceTerm.NATIVE_REPORTED_IN_ERROR()</li>

     *    <li>AbsenceTerm.NATIVE_FORMERLY_NATIVE()</li>

     *  </ul>

     *

     * @return the byAreaIgnoreStatusList

     */

    public List<PresenceAbsenceTerm> getByAreaIgnoreStatusList() {

        if(byAreaIgnoreStatusList == null ){

            byAreaIgnoreStatusList = Arrays.asList(

                    new PresenceAbsenceTerm[] {

                    		PresenceAbsenceTerm.CULTIVATED_REPORTED_IN_ERROR(),

                    		PresenceAbsenceTerm.INTRODUCED_REPORTED_IN_ERROR(),

                    		PresenceAbsenceTerm.NATIVE_REPORTED_IN_ERROR(),

                    		PresenceAbsenceTerm.INTRODUCED_FORMERLY_INTRODUCED(),

                    		PresenceAbsenceTerm.NATIVE_FORMERLY_NATIVE()

                            // TODO what about PresenceAbsenceTerm.ABSENT() also ignore?

                    });

        }

        return byAreaIgnoreStatusList;

    }

    /**

     * @param byAreaIgnoreStatusList the byAreaIgnoreStatusList to set

     */

    public void setByAreaIgnoreStatusList(List<PresenceAbsenceTerm> byAreaIgnoreStatusList) {

        this.byAreaIgnoreStatusList = byAreaIgnoreStatusList;

    }

    /**

     * byRankIgnoreStatusList contains by default

     *  <ul>

     *    <li>PresenceTerm.ENDEMIC_FOR_THE_RELEVANT_AREA()</li>

     *  </ul>

     *

     * @return the byRankIgnoreStatusList

     */

    public List<PresenceAbsenceTerm> getByRankIgnoreStatusList() {

        if (byRankIgnoreStatusList == null) {

            byRankIgnoreStatusList = Arrays.asList(

                    new PresenceAbsenceTerm[] {

                    		PresenceAbsenceTerm.ENDEMIC_FOR_THE_RELEVANT_AREA()

                    });

        }

        return byRankIgnoreStatusList;

    }

    /**

     * @param byRankIgnoreStatusList the byRankIgnoreStatusList to set

     */

    public void setByRankIgnoreStatusList(List<PresenceAbsenceTerm> byRankIgnoreStatusList) {

        this.byRankIgnoreStatusList = byRankIgnoreStatusList;

    }

    /**

     *

     * @param superAreas

     */

    public TransmissionEngineDistribution() {

    }

    /**

     * initializes the map which contains the status terms as key and the priority as value

     * The map will contain both, the PresenceTerms and the AbsenceTerms

     */

    private void initializeStatusPriorityMap() {

        statusPriorityMap = new HashMap<PresenceAbsenceTerm, Integer>();

        Integer priority;

        // PresenceTerms

        for(PresenceAbsenceTerm term : termService.list(PresenceAbsenceTerm.class, null, null, null, null)){

            priority = getPriorityFor(term);

            if(priority != null){

                statusPriorityMap.put(term, priority);

            }

        }

    }

    /**

     * Compares the PresenceAbsenceTermBase terms contained in <code>a.status</code> and <code>b.status</code> after

     * the priority as stored in the statusPriorityMap. The StatusAndSources object with

     * the higher priority is returned. In the case of <code>a == b</code> the sources of b will be added to the sources

     * of a.

     *

     * If either a or b or the status are null b or a is returned.

     *

     * @see initializeStatusPriorityMap()

     *

     * @param a

     * @param b

     * @param sourcesForWinnerB

     *  In the case when <code>b</code> is preferred over <code>a</code> these Set of sources will be added to the sources of <code>b</code>

     * @return

     */

    private StatusAndSources choosePreferred(StatusAndSources a, StatusAndSources b, Set<DescriptionElementSource> sourcesForWinnerB){

        if (statusPriorityMap == null) {

            initializeStatusPriorityMap();

        }

        if (b == null || b.status == null) {

            return a;

        }

        if (a == null || a.status == null) {

            return b;

        }

        if (statusPriorityMap.get(a.status) == null) {

            logger.warn("No priority found in map for " + a.status.getLabel());

            return b;

        }

        if (statusPriorityMap.get(b.status) == null) {

            logger.warn("No priority found in map for " + b.status.getLabel());

            return a;

        }

        if(statusPriorityMap.get(a.status) < statusPriorityMap.get(b.status)){

            if(sourcesForWinnerB != null) {

                b.addSources(sourcesForWinnerB);

            }

            return b;

        } else if (statusPriorityMap.get(a.status) == statusPriorityMap.get(b.status)){

            a.addSources(b.sources);

            return a;

        } else {

            return a;

        }

    }

    /**

     * reads the priority for the given status term from the extensions.

     *

     * @param term

     * @return the priority value

     */

    private Integer getPriorityFor(DefinedTermBase<?> term) {

        Set<Extension> extensions = term.getExtensions();

        for(Extension extension : extensions){

            if(!extension.getType().equals(ExtensionType.ORDER())) {

                continue;

            }

            int pos = extension.getValue().indexOf(EXTENSION_VALUE_PREFIX);

            if(pos == 0){ // if starts with EXTENSION_VALUE_PREFIX

                try {

                    Integer priority = Integer.valueOf(extension.getValue().substring(EXTENSION_VALUE_PREFIX.length()));

                    return priority;

                } catch (NumberFormatException e) {

                    logger.warn("Invalid number format in Extension:" + extension.getValue());

                }

            }

        }

        logger.warn("no priority defined for '" + term.getLabel() + "'");

        return null;

    }

    /**

     * runs both steps

     * <ul>

     * <li>Step 1: Accumulate occurrence records by area</li>

     * <li>Step 2: Accumulate by ranks starting from lower rank to upper rank,

     * the status of all children are accumulated on each rank starting from

     * lower rank to upper rank.</li>

     * </ul>

     *

     * @param superAreas

     *            the areas to which the subordinate areas should be projected.

     * @param lowerRank

     * @param upperRank

     * @param classification

     * @param classification

     *            limit the accumulation process to a specific classification

     *            (not yet implemented)

     * @param monitor

     *            the progress monitor to use for reporting progress to the

     *            user. It is the caller's responsibility to call done() on the

     *            given monitor. Accepts null, indicating that no progress

     *            should be reported and that the operation cannot be cancelled.

     */

    public void accumulate(AggregationMode mode, List<NamedArea> superAreas, Rank lowerRank, Rank upperRank,

            Classification classification, IProgressMonitor monitor) throws JvmLimitsException {

        if (monitor == null) {

            monitor = new NullProgressMonitor();

        }

        // only for debugging:

        //logger.setLevel(Level.TRACE); // TRACE will slow down a lot since it forces loading all term representations

        //Logger.getLogger("org.hibernate.SQL").setLevel(Level.DEBUG);

        logger.info("Hibernate JDBC Batch size: "

                + ((SessionFactoryImplementor) getSession().getSessionFactory()).getSettings().getJdbcBatchSize());

        Set<Classification> classifications = new HashSet<Classification>();

        if(classification == null) {

            classifications.addAll(classificationService.listClassifications(null, null, null, null));

        } else {

            classifications.add(classification);

        }

        int aggregationWorkTicks;

        switch(mode){

        case byAreasAndRanks:

            aggregationWorkTicks = byAreasTicks + byRankTicks;

            break;

        case byAreas:

            aggregationWorkTicks = byAreasTicks;

            break;

        case byRanks:

            aggregationWorkTicks = byRankTicks;

            break;

        default:

            aggregationWorkTicks = 0;

            break;

        }

        // take start time for performance testing

        // NOTE: use ONLY_FISRT_BATCH = true to measure only one batch

        double start = System.currentTimeMillis();

        monitor.beginTask("Accumulating distributions", (classifications.size() * aggregationWorkTicks) + 1 );

        updatePriorities();

        List<Rank> ranks = rankInterval(lowerRank, upperRank);

        monitor.worked(1);

        for(Classification _classification : classifications) {

            ClassificationLookupDTO classificationLookupDao = classificationService.classificationLookup(_classification);

            classificationLookupDao.filterInclude(ranks);

            double end1 = System.currentTimeMillis();

            logger.info("Time elapsed for classificationLookup() : " + (end1 - start) / (1000) + "s");

            double start2 = System.currentTimeMillis();

            monitor.subTask("Accumulating distributions to super areas for " + _classification.getTitleCache());

            if (mode.equals(AggregationMode.byAreas) || mode.equals(AggregationMode.byAreasAndRanks)) {

                accumulateByArea(superAreas, classificationLookupDao, new SubProgressMonitor(monitor, byAreasTicks), true);

            }

            monitor.subTask("Accumulating distributions to higher ranks for " + _classification.getTitleCache());

            double end2 = System.currentTimeMillis();

            logger.info("Time elapsed for accumulateByArea() : " + (end2 - start2) / (1000) + "s");

            double start3 = System.currentTimeMillis();

            if (mode.equals(AggregationMode.byRanks) || mode.equals(AggregationMode.byAreasAndRanks)) {

                accumulateByRank(ranks, classificationLookupDao, new SubProgressMonitor(monitor, byRankTicks), mode.equals(AggregationMode.byRanks));

            }

            double end3 = System.currentTimeMillis();

            logger.info("Time elapsed for accumulateByRank() : " + (end3 - start3) / (1000) + "s");

            logger.info("Time elapsed for accumulate(): " + (end3 - start) / (1000) + "s");

            if(ONLY_FISRT_BATCH) {

                monitor.done();

                break;

            }

        }

        monitor.done();

    }

    /**

     * Step 1: Accumulate occurrence records by area

     * <ul>

     * <li>areas are projected to super areas e.g.:  HS <-- HS(A), HS(G), HS(S)</li>

     * <li>super areas do initially not have a status set ==> Prerequisite to check in CDM</li>

     * <li>areas having a summary status of summary value different from {@link #getByAreaIgnoreStatusList()} are ignored</li>

     * <li>areas have a priority value, the status of the area with highest priority determines the status of the super area</li>

     * <li>the source references of the accumulated distributions are also accumulated into the new distribution,,</li>

     * <li>this has been especially implemented for the EuroMed Checklist Vol2 and might not be a general requirement</li>

     * </ul>

     *

     * @param superAreas

     *      the areas to which the subordinate areas should be projected

     * @param classificationLookupDao

     * @throws JvmLimitsException

     *

     */

    protected void accumulateByArea(List<NamedArea> superAreas, ClassificationLookupDTO classificationLookupDao,  IProgressMonitor subMonitor, boolean doClearDescriptions) throws JvmLimitsException {

        DynamicBatch batch = new DynamicBatch(BATCH_SIZE_BY_AREA, batchMinFreeHeap);

        batch.setRequiredFreeHeap(BATCH_FREE_HEAP_RATIO);

        TransactionStatus txStatus = startTransaction(false);

        // reload superAreas TODO is it faster to getSession().merge(object) ??

        Set<UUID> superAreaUuids = new HashSet<UUID>(superAreas.size());

        for (NamedArea superArea : superAreas){

            superAreaUuids.add(superArea.getUuid());

        }

        // visit all accepted taxa

        subMonitor.beginTask("Accumulating by area ",  classificationLookupDao.getTaxonIds().size());

        Iterator<Integer> taxonIdIterator = classificationLookupDao.getTaxonIds().iterator();

        while (taxonIdIterator.hasNext() || batch.hasUnprocessedItems()) {

            if(txStatus == null) {

                // transaction has been comitted at the end of this batch, start a new one

                txStatus = startTransaction(false);

            }

            // the session is cleared after each batch, so load the superAreaList for each batch

            List<NamedArea> superAreaList = (List)termService.find(superAreaUuids);

            // load taxa for this batch

            List<Integer> taxonIds = batch.nextItems(taxonIdIterator);

//            logger.debug("accumulateByArea() - taxon " + taxonPager.getFirstRecord() + " to " + taxonPager.getLastRecord() + " of " + taxonPager.getCount() + "]");

            List<TaxonBase> taxa = taxonService.loadByIds(taxonIds, TAXONDESCRIPTION_INIT_STRATEGY);

            // iterate over the taxa and accumulate areas

            // start processing the new batch

            for(TaxonBase taxonBase : taxa) {

                if(logger.isDebugEnabled()){

                    logger.debug("accumulateByArea() - taxon :" + taxonToString(taxonBase));

                }

                batch.incementCounter();

                Taxon taxon = (Taxon)taxonBase;

                TaxonDescription description = findComputedDescription(taxon, doClearDescriptions);

                List<Distribution> distributions = distributionsFor(taxon);

                // Step through superAreas for accumulation of subAreas

                for (NamedArea superArea : superAreaList){

                    // accumulate all sub area status

                    StatusAndSources accumulatedStatusAndSources = null;

                    // TODO consider using the TermHierarchyLookup (only in local branch a.kohlbecker)

                    Set<NamedArea> subAreas = getSubAreasFor(superArea);

                    for(NamedArea subArea : subAreas){

                        if(logger.isTraceEnabled()){

                            logger.trace("accumulateByArea() - \t\t" + termToString(subArea));

                        }

                        // step through all distributions for the given subArea

                        for(Distribution distribution : distributions){

                            if(distribution.getArea() != null && distribution.getArea().equals(subArea) && distribution.getStatus() != null) {

                                PresenceAbsenceTerm status = distribution.getStatus();

                                if(logger.isTraceEnabled()){

                                    logger.trace("accumulateByArea() - \t\t" + termToString(subArea) + ": " + termToString(status));

                                }

                                // skip all having a status value different of those in byAreaIgnoreStatusList

                                if (getByAreaIgnoreStatusList().contains(status)){

                                    continue;

                                }

                                StatusAndSources subStatusAndSources = new StatusAndSources(status, distribution.getSources());

                                accumulatedStatusAndSources = choosePreferred(accumulatedStatusAndSources, subStatusAndSources, null);

                            }

                        }

                    } // next sub area

                    if (accumulatedStatusAndSources != null) {

                        if(logger.isDebugEnabled()){

                            logger.debug("accumulateByArea() - \t >> " + termToString(superArea) + ": " + termToString(accumulatedStatusAndSources.status));

                        }

                        // store new distribution element for superArea in taxon description

                        Distribution newDistribitionElement = Distribution.NewInstance(superArea, accumulatedStatusAndSources.status);

                        newDistribitionElement.getSources().addAll(accumulatedStatusAndSources.sources);

                        newDistribitionElement.addMarker(Marker.NewInstance(MarkerType.COMPUTED(), true));

                        description.addElement(newDistribitionElement);

                    }

                } // next super area ....

                descriptionService.saveOrUpdate(description);

                taxonService.saveOrUpdate(taxon);

                subMonitor.worked(1);

                if(!batch.isWithinJvmLimits()) {

                    break; // flushAndClear and start with new batch

                }

            } // next taxon

            flushAndClear();

            // commit for every batch, otherwise the persistent context

            // may grow too much and eats up all the heap

            commitTransaction(txStatus);

            txStatus = null;

            if(ONLY_FISRT_BATCH) {

                break;

            }

        } // next batch of taxa

        subMonitor.done();

    }

   /**

    * Step 2: Accumulate by ranks starting from lower rank to upper rank, the status of all children

    * are accumulated on each rank starting from lower rank to upper rank.

    * <ul>

    * <li>aggregate distribution of included taxa of the next lower rank for any rank level starting from the lower rank (e.g. sub species)

    *    up to upper rank (e.g. Genus)</li>

    *  <li>the accumulation id done for each distribution area found in the included taxa</li>

    *  <li>areas of subtaxa with status endemic are ignored</li>

    *  <li>the status with the highest priority determines the value for the accumulated distribution</li>

    *  <li>the source reference of the accumulated distributions are also accumulated into the new distribution,

    *    this has been especially implemented for the EuroMed Checklist Vol2 and might not be a general requirement</li>

    *</ul>

 * @throws JvmLimitsException

    */

    protected void accumulateByRank(List<Rank> rankInterval, ClassificationLookupDTO classificationLookupDao,  IProgressMonitor subMonitor, boolean doClearDescriptions) throws JvmLimitsException {

        DynamicBatch batch = new DynamicBatch(BATCH_SIZE_BY_RANK, batchMinFreeHeap);

        batch.setRequiredFreeHeap(BATCH_FREE_HEAP_RATIO);

        batch.setMaxAllowedGcIncreases(10);

        int ticksPerRank = 100;

        TransactionStatus txStatus = startTransaction(false);

        // the loadRankSpecificRootNodes() method not only finds

        // taxa of the specified rank but also taxa of lower ranks

        // if no taxon of the specified rank exists, so we need to

        // remember which taxa have been processed already

        Set<Integer> taxaProcessedIds = new HashSet<Integer>();

        List<TaxonBase> taxa = null;

        List<TaxonBase> childTaxa = null;

        List<Rank> ranks = rankInterval;

        subMonitor.beginTask("Accumulating by rank", ranks.size() * ticksPerRank);

        for (Rank rank : ranks) {

            if(logger.isDebugEnabled()){

                logger.debug("accumulateByRank() - at Rank '" + termToString(rank) + "'");

            }

            Set<Integer> taxonIdsPerRank = classificationLookupDao.getTaxonIdByRank().get(rank);

            int taxonCountperRank = taxonIdsPerRank != null ? taxonIdsPerRank.size() : 0;

            SubProgressMonitor taxonSubMonitor = new SubProgressMonitor(subMonitor, ticksPerRank);

            taxonSubMonitor.beginTask("Accumulating by rank " + termToString(rank), taxonCountperRank);

            if(taxonCountperRank == 0) {

                taxonSubMonitor.done();

                continue;

            }

            Iterator<Integer> taxonIdIterator = taxonIdsPerRank.iterator();

            while (taxonIdIterator.hasNext() || batch.hasUnprocessedItems()) {

                if(txStatus == null) {

                    // transaction has been committed at the end of this batch, start a new one

                    txStatus = startTransaction(false);

                }

                // load taxa for this batch

                List<Integer> taxonIds = batch.nextItems(taxonIdIterator);

                taxa = taxonService.loadByIds(taxonIds, null);

//                if(logger.isDebugEnabled()){

//                           logger.debug("accumulateByRank() - taxon " + taxonPager.getFirstRecord() + " to " + taxonPager.getLastRecord() + " of " + taxonPager.getCount() + "]");

//                }

                for(TaxonBase taxonBase : taxa) {

                    batch.incementCounter();

                    Taxon taxon = (Taxon)taxonBase;

                    if (taxaProcessedIds.contains(taxon.getId())) {

                        if(logger.isDebugEnabled()){

                            logger.debug("accumulateByRank() - skipping already processed taxon :" + taxonToString(taxon));

                        }

                        continue;

                    }

                    taxaProcessedIds.add(taxon.getId());

                    if(logger.isDebugEnabled()){

                        logger.debug("accumulateByRank() [" + rank.getLabel() + "] - taxon :" + taxonToString(taxon));

                    }

                    // Step through direct taxonomic children for accumulation

                    Map<NamedArea, StatusAndSources> accumulatedStatusMap = new HashMap<NamedArea, StatusAndSources>();

                    List<Integer> childTaxonIds = new ArrayList<>();

                    Set<Integer> childSet = classificationLookupDao.getChildTaxonMap().get(taxon.getId());

                    if(childSet != null) {

                        childTaxonIds.addAll(childSet);

                    }

                    if(!childTaxonIds.isEmpty()) {

                        childTaxa = taxonService.loadByIds(childTaxonIds, TAXONDESCRIPTION_INIT_STRATEGY);

                        LinkedList<TaxonBase> childStack = new LinkedList<TaxonBase>(childTaxa);

                        childTaxa = null; // allow to be garbage collected

                        while(childStack.size() > 0){

                            TaxonBase childTaxonBase = childStack.pop();

                            getSession().setReadOnly(childTaxonBase, true);

                            Taxon childTaxon = (Taxon) childTaxonBase;

                            getSession().setReadOnly(childTaxon, true);

                            if(logger.isTraceEnabled()){

                                logger.trace("                   subtaxon :" + taxonToString(childTaxon));

                            }

                            for(Distribution distribution : distributionsFor(childTaxon) ) {

                                PresenceAbsenceTerm status = distribution.getStatus();

                                NamedArea area = distribution.getArea();

                                if (status == null || getByRankIgnoreStatusList().contains(status)){

                                  continue;

                                }

                                StatusAndSources subStatusAndSources = new StatusAndSources(status, distribution.getSources());

                                accumulatedStatusMap.put(area, choosePreferred(accumulatedStatusMap.get(area), subStatusAndSources, null));

                             }

                            // evict all initialized entities of the childTaxon

                            // TODO consider using cascade="evict" in the model classes

//                            for( TaxonDescription description : ((Taxon)childTaxonBase).getDescriptions()) {

//                                for (DescriptionElementBase deb : description.getElements()) {

//                                    getSession().evict(deb);

//                                }

//                                getSession().evict(description); // this causes in some cases the taxon object to be detached from the session

//                            }

                            getSession().evict(childTaxonBase); // no longer needed, save heap

                        }

                        if(accumulatedStatusMap.size() > 0) {

                            TaxonDescription description = findComputedDescription(taxon, doClearDescriptions);

                            for (NamedArea area : accumulatedStatusMap.keySet()) {

                                Distribution distribition = findDistribution(description, area, accumulatedStatusMap.get(area).status);

                                if(distribition == null) {

                                    // create a new distribution element

                                    distribition = Distribution.NewInstance(area, accumulatedStatusMap.get(area).status);

                                    distribition.addMarker(Marker.NewInstance(MarkerType.COMPUTED(), true));

                                }

                                addSourcesDeduplicated(distribition.getSources(), accumulatedStatusMap.get(area).sources);

                                description.addElement(distribition);

                            }

                            taxonService.saveOrUpdate(taxon);

                            descriptionService.saveOrUpdate(description);

                        }

                    }

                    taxonSubMonitor.worked(1); // one taxon worked

                    if(!batch.isWithinJvmLimits()) {

                        break; // flushAndClear and start with new batch

                    }

                } // next taxon ....

                flushAndClear();

                // commit for every batch, otherwise the persistent context

                // may grow too much and eats up all the heap

                commitTransaction(txStatus);

                txStatus = null;

                // flushing the session and to the index (flushAndClear() ) can impose a

                // massive heap consumption. therefore we explicitly do a check after the

                // flush to detect these situations and to reduce the batch size.

                if(batch.getJvmMonitor().getGCRateSiceLastCheck() > 0.05) {

                    batch.reduceSize(0.5);

                }

                if(ONLY_FISRT_BATCH) {

                    break;

                }

            } // next batch

            taxonSubMonitor.done();

            subMonitor.worked(1);

            if(ONLY_FISRT_BATCH) {

                break;

            }

        } // next Rank

        logger.info("accumulateByRank() - done");

        subMonitor.done();

    }

/**

 * @param description

 * @param area

 * @param status

 * @return

 */

private Distribution findDistribution(TaxonDescription description, NamedArea area, PresenceAbsenceTerm status) {

    for(DescriptionElementBase item : description.getElements()) {

        if(!(item instanceof Distribution)) {

            continue;

        }

        Distribution distribution = ((Distribution)item);

        if(distribution.getArea().equals(area) && distribution.getStatus().equals(status)) {

            return distribution;

        }

    }

    return null;

}

/**

 * @param lowerRank

 * @param upperRank

 * @return

 */

private List<Rank> rankInterval(Rank lowerRank, Rank upperRank) {

    TransactionStatus txStatus = startTransaction(false);

    Rank currentRank = lowerRank;

    List<Rank> ranks = new ArrayList<Rank>();

    ranks.add(currentRank);

    while (!currentRank.isHigher(upperRank)) {

        currentRank = findNextHigherRank(currentRank);

        ranks.add(currentRank);

    }

    commitTransaction(txStatus);

    txStatus = null;

    return ranks;

}

    /**

     * @return

     */

    private Session getSession() {

        return descriptionService.getSession();

    }

    /**

     *

     */

    private void flush() {

        logger.debug("flushing session ...");

        getSession().flush();

        try {

            logger.debug("flushing to indexes ...");

            Search.getFullTextSession(getSession()).flushToIndexes();

        } catch (HibernateException e) {

            /* IGNORE - Hibernate Search Event listeners not configured ... */

            if(!e.getMessage().startsWith("Hibernate Search Event listeners not configured")){

                throw e;

            }

        }

    }

    /**

    *

    */

   private void flushAndClear() {

       flush();

       logger.debug("clearing session ...");

       getSession().clear();

   }

    // TODO merge with CdmApplicationDefaultConfiguration#startTransaction() into common base class

    public TransactionStatus startTransaction(Boolean readOnly) {

        DefaultTransactionDefinition defaultTxDef = new DefaultTransactionDefinition();

        defaultTxDef.setReadOnly(readOnly);

        TransactionDefinition txDef = defaultTxDef;

        // Log some transaction-related debug information.

        if (logger.isTraceEnabled()) {

            logger.trace("Transaction name = " + txDef.getName());

            logger.trace("Transaction facets:");

            logger.trace("Propagation behavior = " + txDef.getPropagationBehavior());

            logger.trace("Isolation level = " + txDef.getIsolationLevel());

            logger.trace("Timeout = " + txDef.getTimeout());

            logger.trace("Read Only = " + txDef.isReadOnly());

            // org.springframework.orm.hibernate5.HibernateTransactionManager

            // provides more transaction/session-related debug information.

        }

        TransactionStatus txStatus = transactionManager.getTransaction(txDef);

        getSession().setFlushMode(FlushMode.COMMIT);

        return txStatus;

    }

    // TODO merge with CdmApplicationDefaultConfiguration#startTransaction() into common base class

    public void commitTransaction(TransactionStatus txStatus){

        logger.debug("commiting transaction ...");

        transactionManager.commit(txStatus);

        return;

    }

    /**

     * returns the next higher rank

     *

     * TODO better implement OrderedTermBase.getNextHigherTerm() and OrderedTermBase.getNextLowerTerm()?

     *

     * @param rank

     * @return

     */

    private Rank findNextHigherRank(Rank rank) {

        rank = (Rank) termService.load(rank.getUuid());

        return rank.getNextHigherTerm();

//        OrderedTermVocabulary<Rank> rankVocabulary = mameService.getRankVocabulary();;

//        return rankVocabulary.getNextHigherTerm(rank);

    }

    /**

     * Either finds an existing taxon description of the given taxon or creates a new one.

     * If the doClear is set all existing description elements will be cleared.

     *

     * @param taxon

     * @param doClear will remove all existing Distributions if the taxon already

     * has a MarkerType.COMPUTED() TaxonDescription

     * @return

     */

    private TaxonDescription findComputedDescription(Taxon taxon, boolean doClear) {

        String descriptionTitle = this.getClass().getSimpleName();

        // find existing one

        for (TaxonDescription description : taxon.getDescriptions()) {