/** Extracts document and cluster lists before serialization. */
@Persist
private void beforeSerialization() {
/*
* See http://issues.carrot2.org/browse/CARROT-693; this monitor does not save us
* in multi-threaded environment anyway. A better solution would be to prepare
* this eagerly in the constructor, but we try to balance overhead and full
* correctness here.
*/
synchronized (this) {
query = (String) attributes.get(AttributeNames.QUERY);
if (getDocuments() != null) {
documents = Lists.newArrayList(getDocuments());
} else {
documents = null;
}
if (getClusters() != null) {
clusters = Lists.newArrayList(getClusters());
} else {
clusters = null;
}
otherAttributesForSerialization = MapUtils.asHashMap(SimpleXmlWrappers.wrap(attributes));
otherAttributesForSerialization.remove(AttributeNames.QUERY);
otherAttributesForSerialization.remove(AttributeNames.CLUSTERS);
otherAttributesForSerialization.remove(AttributeNames.DOCUMENTS);
if (otherAttributesForSerialization.isEmpty()) {
otherAttributesForSerialization = null;
}
}
}
@Commit
@SuppressWarnings("unused")
private void afterDeserialization() throws Exception {
if (otherAttributesForSerialization != null) {
attributes.putAll(SimpleXmlWrappers.unwrap(otherAttributesForSerialization));
}
phrasesView = Collections.unmodifiableList(phrases);
subclustersView = Collections.unmodifiableList(subclusters);
// Documents will be restored on the ProcessingResult level
}
/** Replace document refids with the actual references upon deserialization. */
private void documentIdToReference(Cluster cluster, Map<String, Document> documents) {
if (cluster.documentIds != null) {
for (Cluster.DocumentRefid documentRefid : cluster.documentIds) {
cluster.addDocuments(documents.get(documentRefid.refid));
}
}
for (Cluster subcluster : cluster.getSubclusters()) {
documentIdToReference(subcluster, documents);
}
}
/**
* Creates a {@link ProcessingResult} with the provided <code>attributes</code>. Assigns unique
* document identifiers if documents are present in the <code>attributes</code> map (under the key
* {@link AttributeNames#DOCUMENTS}).
*/
@SuppressWarnings("unchecked")
ProcessingResult(Map<String, Object> attributes) {
this.attributes = attributes;
// Replace a modifiable collection of documents with an unmodifiable one
final List<Document> documents = (List<Document>) attributes.get(AttributeNames.DOCUMENTS);
if (documents != null) {
Document.assignDocumentIds(documents);
attributes.put(AttributeNames.DOCUMENTS, Collections.unmodifiableList(documents));
}
// Replace a modifiable collection of clusters with an unmodifiable one
final List<Cluster> clusters = (List<Cluster>) attributes.get(AttributeNames.CLUSTERS);
if (clusters != null) {
Cluster.assignClusterIds(clusters);
attributes.put(AttributeNames.CLUSTERS, Collections.unmodifiableList(clusters));
}
// Store a reference to attributes as an unmodifiable map
this.attributesView = Collections.unmodifiableMap(attributes);
}
/** Transfers document and cluster lists to the attributes map after deserialization. */
@Commit
private void afterDeserialization() throws Exception {
if (otherAttributesForSerialization != null) {
attributes = SimpleXmlWrappers.unwrap(otherAttributesForSerialization);
}
attributesView = Collections.unmodifiableMap(attributes);
attributes.put(AttributeNames.QUERY, query != null ? query.trim() : null);
attributes.put(AttributeNames.DOCUMENTS, documents);
attributes.put(AttributeNames.CLUSTERS, clusters);
// Convert document ids to the actual references
if (clusters != null && documents != null) {
final Map<String, Document> documentsById = Maps.newHashMap();
for (Document document : documents) {
documentsById.put(document.getStringId(), document);
}
for (Cluster cluster : clusters) {
documentIdToReference(cluster, documentsById);
}
}
}
/**
* Prepares a temporary attributes map for serialization purposes. Includes only the requested
* elements in the map.
*/
private Map<String, Object> prepareAttributesForSerialization(
boolean saveDocuments, boolean saveClusters, boolean saveOtherAttributes) {
final Map<String, Object> tempAttributes = Maps.newHashMap();
if (saveOtherAttributes) {
tempAttributes.putAll(attributes);
tempAttributes.remove(AttributeNames.DOCUMENTS);
tempAttributes.remove(AttributeNames.CLUSTERS);
} else {
tempAttributes.put(AttributeNames.QUERY, attributes.get(AttributeNames.QUERY));
}
if (saveDocuments) {
tempAttributes.put(AttributeNames.DOCUMENTS, getDocuments());
}
if (saveClusters) {
tempAttributes.put(AttributeNames.CLUSTERS, getClusters());
}
return tempAttributes;
}
/** For JSON and XML serialization only. */
@JsonProperty("attributes")
@SuppressWarnings("unused")
private Map<String, Object> getOtherAttributes() {
final Map<String, Object> otherAttributes = Maps.newHashMap(attributesView);
return otherAttributes.isEmpty() ? null : otherAttributes;
}
/**
* Associates an attribute with this cluster.
*
* @param key for the attribute
* @param value for the attribute
* @return this cluster for convenience
*/
public <T> Cluster setAttribute(String key, T value) {
synchronized (attributes) {
attributes.put(key, value);
}
return this;
}
/**
* Returns the attribute associated with this cluster under the provided <code>key</code>. If
* there is no attribute under the provided <code>key</code>, <code>null</code> will be returned.
*
* @param key of the attribute
* @return attribute value of <code>null</code>
*/
@SuppressWarnings("unchecked")
public <T> T getAttribute(String key) {
return (T) attributes.get(key);
}
/**
* Returns the clusters that have been created during processing. The returned list is
* unmodifiable.
*
* @return clusters created during processing or <code>null</code> if no clusters were present in
* the result.
*/
@SuppressWarnings("unchecked")
public List<Cluster> getClusters() {
return (List<Cluster>) attributes.get(AttributeNames.CLUSTERS);
}
/**
* Returns the documents that have been processed. The returned collection is unmodifiable.
*
* @return documents that have been processed or <code>null</code> if no documents are present in
* the result.
*/
@SuppressWarnings("unchecked")
public List<Document> getDocuments() {
return (List<Document>) attributes.get(AttributeNames.DOCUMENTS);
}