com.dooapp.gaedo.blueprints.operations.Updater Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of gaedo-blueprints Show documentation
Show all versions of gaedo-blueprints Show documentation
Implementation of gaedo mechanisms backed by blueprints graph layer
package com.dooapp.gaedo.blueprints.operations;
import java.util.Collection;
import java.util.HashMap;
import java.util.HashSet;
import java.util.LinkedHashMap;
import java.util.LinkedList;
import java.util.List;
import java.util.Map;
import java.util.Map.Entry;
import java.util.Set;
import java.util.TreeSet;
import java.util.logging.Level;
import java.util.logging.Logger;
import javax.persistence.CascadeType;
import com.dooapp.gaedo.blueprints.AbstractBluePrintsBackedFinderService;
import com.dooapp.gaedo.blueprints.CantCreateAVertexForALiteralException;
import com.dooapp.gaedo.blueprints.GraphDatabaseDriver;
import com.dooapp.gaedo.blueprints.GraphUtils;
import com.dooapp.gaedo.blueprints.ObjectCache;
import com.dooapp.gaedo.blueprints.Properties;
import com.dooapp.gaedo.blueprints.transformers.LiteralHelper;
import com.dooapp.gaedo.blueprints.transformers.Literals;
import com.dooapp.gaedo.properties.AbstractPropertyAdapter;
import com.dooapp.gaedo.properties.Property;
import com.dooapp.gaedo.utils.CollectionUtils;
import com.tinkerpop.blueprints.Direction;
import com.tinkerpop.blueprints.Edge;
import com.tinkerpop.blueprints.Graph;
import com.tinkerpop.blueprints.Vertex;
public class Updater {
public class UpdateProperties extends AbstractCardinalityDistinguishingOperation {
private final GraphDatabaseDriver driver;
private final Graph database;
private final AbstractBluePrintsBackedFinderService service;
private final ObjectCache objectsBeingAccessed;
private final Object toUpdate;
private final Vertex objectVertex;
private UpdateProperties(GraphDatabaseDriver driver, Graph database, AbstractBluePrintsBackedFinderService service,
ObjectCache objectsBeingAccessed, Object toUpdate, Vertex objectVertex) {
this.driver = driver;
this.database = database;
this.service = service;
this.objectsBeingAccessed = objectsBeingAccessed;
this.toUpdate = toUpdate;
this.objectVertex = objectVertex;
}
@Override
protected void operateOnSingle(Property p, CascadeType cascade) {
updateSingle(service, driver, database, p, toUpdate, objectVertex, cascade, objectsBeingAccessed);
}
@Override
protected void operateOnMap(Property p, CascadeType cascade) {
updateMap(service, driver, database, p, toUpdate, objectVertex, cascade, objectsBeingAccessed);
}
@Override
protected void operateOnCollection(Property p, CascadeType cascade) {
updateCollection(service, driver, database, p, toUpdate, objectVertex, cascade, objectsBeingAccessed);
}
}
public static final Boolean ELEMENT_IN_COLLECTION_MARKER = Boolean.TRUE;
public static final String ELEMENT_IN_COLLECTION_MARKER_GRAPH_VALUE = Literals.get(ELEMENT_IN_COLLECTION_MARKER.getClass()).toString(ELEMENT_IN_COLLECTION_MARKER);
private static final Logger logger = Logger.getLogger(Updater.class.getName());
/**
* Create or update given object
*
* @param service
* source of modification
* @param driver
* TODO
* @param objectVertexId
* object expected vertex id
* @param objectVertex
* vertex corresponding to object to update
* @param valueClass
* class of the value to be updated here
* @param containedProperties
* list of contained properties
* @param toUpdate
* object to update
* @param cascade
* kind of cascade used for dependent properties
* @param objectsBeingUpdated
* map containing subgraph of obejcts currently being updated,
* this is used to avoid loops, and NOT as a cache
* @return updated object
*/
public Object performUpdate(AbstractBluePrintsBackedFinderService service, GraphDatabaseDriver driver,
String objectVertexId, Vertex objectVertex, Class valueClass, Map> containedProperties,
Object toUpdate, CascadeType cascade, ObjectCache objectsBeingUpdated) {
Graph database = service.getDatabase();
// it's in fact an object creation
if (objectVertex == null) {
if (logger.isLoggable(Level.FINER)) {
logger.log(Level.FINER, "object " + objectVertexId.toString() + " has never before been seen in graph, so create central node for it");
}
objectVertex = driver.createEmptyVertex(valueClass, objectVertexId, toUpdate);
// Create a value for that node (useful for RDF export)
driver.setValue(objectVertex, objectVertexId);
}
// Here come the caching !
DataType updated = (DataType) objectsBeingUpdated.get(objectVertexId);
if (updated == null) {
try {
objectsBeingUpdated.put(objectVertexId, toUpdate);
updateProperties(service, driver, database, toUpdate, objectVertex, containedProperties, cascade, objectsBeingUpdated);
return toUpdate;
} finally {
objectsBeingUpdated.remove(objectVertexId);
}
} else {
return updated;
}
}
/**
* Update all properties of given object
*
* @param driver
* TODO
* @param toUpdate
* object to update
* @param objectVertex
* object root vertex
* @param containedProperties
* map linking each object property to the cascade types
* associated to it (allows us to easily see if there is any
* cascade to perform on object)
* @param cascade
* cascade type used to perform this operation, depend if this
* method is called from a {@link #create(Object)} or an
* {@link #update(Object)}
* @param objectsBeingAccessed
* cache of objects being accessed during that write
*/
private void updateProperties(
final AbstractBluePrintsBackedFinderService service,
final GraphDatabaseDriver driver,
final Graph database,
final Object toUpdate,
final Vertex objectVertex,
Map> containedProperties,
CascadeType cascade,
final ObjectCache objectsBeingAccessed) {
new OperateOnProperties().execute(containedProperties, cascade, new UpdateProperties(driver, database, service, objectsBeingAccessed, toUpdate, objectVertex));
// Migrator property has been added to object if needed
// it's also the case of classes list
}
/**
* Persisting a map consist into considering each map entry as an object of
* the map entries collection, then associating each entry object to its
* contained key and value. To make this association as easy (and readable
* as posisble) map entries keys are their keys objects ids (if managed) or
* values) elsewhere, and values are their values ids (if managed) or values
* (elsewhere). Notice a link is always made between a map entry and both
* its key and value.
*
* @param driver
* TODO
* @param p
* property containing that map
* @param toUpdate
* map to update
* @param rootVertex
* object root vertex
* @param cascade
* used cascade type, can be either {@link CascadeType#PERSIST}
* or {@link CascadeType#MERGE}
*/
private void updateMap(AbstractBluePrintsBackedFinderService service, GraphDatabaseDriver driver, Graph database,
Property p, Object toUpdate, Vertex rootVertex, CascadeType cascade, ObjectCache objectsBeingAccessed) {
// Cast should work like a charm
Map value = (Map) p.get(toUpdate);
// As a convention, null values are never stored
if (value != null /*
* && value.size()>0 that case precisely created
* https://github.com/Riduidel/gaedo/issues/13
*/) {
// Get previously existing vertices
Iterable existingIterator = service.getStrategy().getOutEdgesFor(rootVertex, p);
// Do not change previously existing vertices if they correspond to
// new ones
// Which is done in that call : as vertex is always looked up before
// creation, there is little duplication risk
// or at last that risk should be covered by selected Blueprints
// implementation
Collection newVertices = createMapVerticesFor(service, value, cascade, objectsBeingAccessed);
Map oldVertices = new HashMap();
for (Edge e : existingIterator) {
Vertex inVertex = e.getVertex(Direction.IN);
if (newVertices.contains(inVertex)) {
newVertices.remove(inVertex);
} else {
oldVertices.put(inVertex, e);
}
}
// Now the have been collected, remove all old vertices
for (Map.Entry entry : oldVertices.entrySet()) {
GraphUtils.removeSafely(database, entry.getValue());
// TODO also remove map entry vertex associated edges
}
// And finally add new vertices
for (Vertex newVertex : newVertices) {
driver.createEdgeFor(rootVertex, newVertex, p);
}
}
}
/**
* Update given collection by creating a set of edges/vertices for each
* element
*
* @param driver
* TODO
* @param p
* properties to update associated vertices for
* @param toUpdate
* source object to update
* @param rootVertex
* vertex associated to toUpdate
* @param cascade
* used cascade type, can be either {@link CascadeType#PERSIST}
* or {@link CascadeType#MERGE}
*
* @category update
*/
private void updateCollection(AbstractBluePrintsBackedFinderService service, GraphDatabaseDriver driver,
Graph database, Property p, Object toUpdate, Vertex rootVertex, CascadeType cascade, ObjectCache objectsBeingAccessed) {
// Cast should work like a charm
Collection value = (Collection) p.get(toUpdate);
// As a convention, null values are never stored
if (value != null /*
* && value.size()>0 that case precisely created
* https://github.com/Riduidel/gaedo/issues/13
*/) {
// Get previously existing vertices
Iterable previousEdges = service.getStrategy().getOutEdgesFor(rootVertex, p);
// Get the new, updated Collection of vertices (which is already
// sorted).
Map allVertices = createCollectionVerticesFor(service, value, cascade, objectsBeingAccessed);
// Keep track of the edges that correspond to each vertex (for later
// ordering)...
Map> savedEdges = new HashMap>();
// ...and put the old, invalid vertices aside for later deletion.
Set edgesToRemove = new HashSet();
// Add previous edges
for (Edge e : previousEdges) {
Vertex inVertex = e.getVertex(Direction.IN);
if (allVertices.values().contains(inVertex)) {
if (!savedEdges.containsKey(inVertex))
savedEdges.put(inVertex, new LinkedList());
savedEdges.get(inVertex).add(e);
} else {
edgesToRemove.add(e);
}
}
// And previous properties
String propertyNamePrefix = GraphUtils.getEdgeNameFor(p);
Collection suspectProperties = new TreeSet();
for(String propertyName : rootVertex.getPropertyKeys()) {
if(propertyName.startsWith(propertyNamePrefix)) {
suspectProperties.add(propertyName);
}
}
// Delete the edges that we don't need anymore.
for (Edge edge : edgesToRemove) {
GraphUtils.removeSafely(database, edge);
}
// Then, go through the updated Vertices. Create edges if necessary,
// then always set the order property.
// This is possible since #createCollectionVerticesFor maintains the
// ordering.
int order = 0;
for(Object element : value) {
if(allVertices.containsKey(element)) {
Vertex vertex = allVertices.get(element);
Edge edgeForVertex;
if (savedEdges.containsKey(vertex)) {
List edges = savedEdges.get(vertex);
edgeForVertex = edges.remove(0);
if (edges.size() == 0)
savedEdges.remove(vertex);
} else
edgeForVertex = driver.createEdgeFor(rootVertex, vertex, p);
// Add a fancy-schmancy property to maintain order in this town (this property is NOT indexed)
edgeForVertex.setProperty(Properties.collection_index.name(), order++);
} else {
// Element is a literal value, so it won't require a vertex, but a literal saving
AbstractPropertyAdapter elementByIndexProperty = new CollectionAccessByIndexProperty(p, order, element);
updateLiteralPropertyIn(service.getDatabase(), toUpdate, rootVertex, elementByIndexProperty, element);
suspectProperties.remove(GraphUtils.getEdgeNameFor(elementByIndexProperty));
// We also add an inverted property allowing fast query of containing collection
// Value here is not signifiant : we only want to mark collection as containing value. And for that, one simple string is enough
AbstractPropertyAdapter elementByValueProperty = new CollectionAccessByValueProperty(p, element, ELEMENT_IN_COLLECTION_MARKER);
elementByValueProperty.setGenericType(ELEMENT_IN_COLLECTION_MARKER.getClass());
updateLiteralPropertyIn(service.getDatabase(), toUpdate, rootVertex, elementByValueProperty, ELEMENT_IN_COLLECTION_MARKER);
suspectProperties.remove(GraphUtils.getEdgeNameFor(elementByValueProperty));
order++;
}
}
// Finally, delete any remaining edges.
for (Vertex vertex : savedEdges.keySet())
for (Edge edge : savedEdges.get(vertex))
GraphUtils.removeSafely(database, edge);
// and remaining suspect properties
for(String suspect : suspectProperties) {
rootVertex.removeProperty(suspect);
}
// and don't forget to write collection size property, for easier querying
CollectionSizeProperty sizeProperty = new CollectionSizeProperty(p);
updateLiteralPropertyIn(service.getDatabase(), toUpdate, rootVertex, sizeProperty, value.size());
}
}
/**
* Create a collection of vertices for the given collection of values
*
* @param value
* collection of values to create vertices for
* @param cascade
* used cascade type, can be either {@link CascadeType#PERSIST}
* or {@link CascadeType#MERGE}
* @return map linking objects to vertices created by {@link #getVertexFor(Object)}.
* order of vertices is guaranteed to be the same as input value
* one.
*/
private Map createCollectionVerticesFor(AbstractBluePrintsBackedFinderService service, Collection value,
CascadeType cascade, ObjectCache objectsBeingAccessed) {
Map returned = new LinkedHashMap();
for (Object o : value) {
// already heard about null-containing collections ? I do know them,
// and they're pure EVIL
if (o != null) {
try {
returned.put(o, service.getVertexFor(o, cascade, objectsBeingAccessed));
} catch(CantCreateAVertexForALiteralException e) {
// Literal objects have no vertices created for them. They
// will be stored as "special" property values, which may
// put a mess in many code blocks
}
}
}
return returned;
}
/**
* Create a collection of map vertices (each representing one map entry) for
* each entry of the input map.
*
* @param value
* map of values to create vertices for
* @param cascade
* used cascade type, can be either {@link CascadeType#PERSIST}
* or {@link CascadeType#MERGE}
* @return collection of vertices created by {@link #getVertexFor(Object)}
* @see #getVertexFor(AbstractBluePrintsBackedFinderService, CascadeType, Map) for details about the way to
* generate a vertex for a Map.Entry node
*/
private Collection createMapVerticesFor(AbstractBluePrintsBackedFinderService service, Map value, CascadeType cascade,
ObjectCache objectsBeingAccessed) {
Collection returned = new HashSet();
// Strangely, the entrySet is not seen as a Set
for (Entry o : (Set) value.entrySet()) {
Vertex newVertex = service.getVertexFor(o, cascade, objectsBeingAccessed);
returned.add(newVertex);
}
return returned;
}
/**
* Update single-valued property by changing target of edge used to
* represent the property
*
* @param driver
* TODO
* @param p
* updated property
* @param toUpdate
* updated object
* @param rootVertex
* vertex representing the object
* @param cascade
* used cascade type, can be either {@link CascadeType#PERSIST}
* or {@link CascadeType#MERGE}
*
* @category update
*/
private void updateSingle(AbstractBluePrintsBackedFinderService service, GraphDatabaseDriver driver,
Graph database, Property p, Object toUpdate, Vertex rootVertex, CascadeType cascade, ObjectCache objectsBeingAccessed) {
Object value = p.get(toUpdate);
// As a convention, null values are never stored but they may replace
// existing ones, in which case previous values must be removed
// as a consequence, valueVertex is loaded only for non null values
Vertex valueVertex = null;
if (value != null) {
if (Literals.containsKey(value.getClass())) {
updateLiteralPropertyIn(service.getDatabase(), toUpdate, rootVertex, p, value);
return;
} else {
valueVertex = service.getVertexFor(value, cascade, objectsBeingAccessed);
}
}
/*
* If vertex has a property named from the property name, remove it : it
* is due to a previous call to update with a literal non null property
* value, which may no more be the case if property type is ... Object
* (which is a bad idea)
*/
if (rootVertex.getPropertyKeys().contains(GraphUtils.getEdgeNameFor(p))) {
rootVertex.removeProperty(GraphUtils.getEdgeNameFor(p));
}
/*
* Totally crazy confident non-nullity lack of test : this method is
* only called when cascade type is either PERSIST or MERGE. In both
* cases the call to getVertexFor will create the vertex if missing. As
* a consequence there is no need for nullity check.
*/
Edge link = null;
// Get previously existing vertex
List matching = CollectionUtils.asList(service.getStrategy().getOutEdgesFor(rootVertex, p));
// property is single-valued, so iteration can be done at most one
if (matching.size() == 1) {
// There is an existing edge, change its target and maybe delete
// previous one
Edge existing = matching.get(0);
if (valueVertex != null && existing.getVertex(Direction.IN).equals(valueVertex)) {
// Nothing to do
link = existing;
} else {
// delete old edge (if it exists)
GraphUtils.removeSafely(database, existing);
if (value != null)
link = driver.createEdgeFor(rootVertex, valueVertex, p);
}
} else if (matching.size() > 1) {
if (logger.isLoggable(Level.SEVERE)) {
// There is some incoherent data in graph .. log it !
StringBuilder sOut = new StringBuilder("An object with the following monovalued property\n").append(p.toGenericString()).append(
" is linked to more than one vertex :");
for (Edge e : matching) {
sOut.append("\n\t").append(e.getVertex(Direction.IN).toString());
}
logger.log(Level.SEVERE, "Graph contains some incoherence :" + sOut.toString());
}
// absolutly all edges are removed, including the first one. As a
// consequence, initial edge will have to be re-created
for (Edge e : matching) {
GraphUtils.removeSafely(database, e);
}
}
if (link == null && value != null)
link = driver.createEdgeFor(rootVertex, valueVertex, p);
}
public void updateLiteralPropertyIn(Graph database, Object toUpdate, Vertex vertexToUpdate, Property property, Object propertyValue) {
Class propertyClass = property.getType();
GraphUtils.setIndexedProperty(database, vertexToUpdate, GraphUtils.getEdgeNameFor(property),
LiteralHelper.getLiteralTextFor(propertyClass, propertyValue));
}
}
© 2015 - 2025 Weber Informatics LLC | Privacy Policy