org.eclipse.persistence.internal.oxm.ReferenceResolver Maven / Gradle / Ivy
Show all versions of eclipselink Show documentation
/*
* Copyright (c) 1998, 2018 Oracle and/or its affiliates. All rights reserved.
*
* This program and the accompanying materials are made available under the
* terms of the Eclipse Public License v. 2.0 which is available at
* http://www.eclipse.org/legal/epl-2.0,
* or the Eclipse Distribution License v. 1.0 which is available at
* http://www.eclipse.org/org/documents/edl-v10.php.
*
* SPDX-License-Identifier: EPL-2.0 OR BSD-3-Clause
*/
// Contributors:
// Oracle - initial API and implementation from Oracle TopLink
// Marcel Valovy - major speed up, major refurbishing.
package org.eclipse.persistence.internal.oxm;
import java.util.ArrayList;
import java.util.Collection;
import java.util.HashMap;
import java.util.Iterator;
import java.util.LinkedHashMap;
import java.util.List;
import java.util.Map;
import java.util.Vector;
import java.util.concurrent.Callable;
import org.eclipse.persistence.core.descriptors.CoreDescriptor;
import org.eclipse.persistence.core.descriptors.CoreInheritancePolicy;
import org.eclipse.persistence.core.mappings.CoreAttributeAccessor;
import org.eclipse.persistence.core.mappings.CoreMapping;
import org.eclipse.persistence.exceptions.ConversionException;
import org.eclipse.persistence.exceptions.XMLMarshalException;
import org.eclipse.persistence.internal.core.helper.CoreClassConstants;
import org.eclipse.persistence.internal.core.queries.CoreContainerPolicy;
import org.eclipse.persistence.internal.core.sessions.CoreAbstractSession;
import org.eclipse.persistence.internal.identitymaps.CacheId;
import org.eclipse.persistence.internal.oxm.mappings.CollectionReferenceMapping;
import org.eclipse.persistence.internal.oxm.mappings.Descriptor;
import org.eclipse.persistence.internal.oxm.mappings.Field;
import org.eclipse.persistence.internal.oxm.mappings.InverseReferenceMapping;
import org.eclipse.persistence.internal.oxm.mappings.Mapping;
import org.eclipse.persistence.internal.oxm.mappings.ObjectReferenceMapping;
import org.xml.sax.ErrorHandler;
import org.xml.sax.SAXException;
import org.xml.sax.SAXParseException;
/**
* This class is leveraged by reference mappings. It plays 3 roles:
*
* - Stores objects with an ID
* - Stores key based relationships
* - Resolves key based relationships based on the objects stored by ID
*
*/
public final class ReferenceResolver {
/**
* The default initial capacity of map - must be a power of two.
*/
private static final int MAP_INITIAL_CAPACITY = 1 << 6; // aka 64
/**
* The default initial capacity of list - must be a power of two.
*/
private static final int LIST_INITIAL_CAPACITY = 1 << 3; // aka 8
/**
* This field is in fact used locally only. We're just saving resources by reusing one instance of the object.
*/
private final ReferenceKey refKey;
/**
* Stores References.
*/
private LinkedHashMap referencesMap;
/**
* Stores References that have been kicked out of the {@link #referencesMap} because a Reference with same key
* was added to it.
*/
private List unluckyReferences;
/**
* Stores positions of References that have been kicked out of the {@link #referencesMap}. The position is
* the same this Reference would have if this code would run on a list instead of map.
*/
private List unluckyRefPositions;
/**
* Speed-up cache that was introduced in 2.5 instead of the previous speed-up mechanisms using session cache.
*/
private Map> cache;
/**
* The default constructor initializes the list of References.
*/
public ReferenceResolver() {
referencesMap = new LinkedHashMap<>(MAP_INITIAL_CAPACITY);
unluckyReferences = new ArrayList<>(MAP_INITIAL_CAPACITY);
unluckyRefPositions = new ArrayList<>(MAP_INITIAL_CAPACITY);
cache = new HashMap<>(MAP_INITIAL_CAPACITY);
refKey = new ReferenceKey();
}
/* Shows why the containers are not final. Keep this private method up here. */
/**
* Resets the references containers.
*
* PERF:
* Allocating a new object may be faster than clearing old objects, especially in this case.
*
* As 'Stephen C' points out: ,,There are locality and cross-generational issues that could affect performance. When
* you repeatedly recycle an ArrayList, the object and its backing array are likely to be tenured. That means that:
*
* - The list objects and the objects representing list elements are likely to be in different areas of the
* heap, potentially increasing TLB misses and page traffic, especially at GC time.
* - Assignment of (young generation) references into the (tenured) list's backing array are likely to incur
* write barrier overheads ... depending on the GC implementation."
*
* from Stack Overflow.
*
* Taking last size can give approximate prediction of the next size.
* Halving provides convergence and increases efficiency. Since the list will be empty after, it's efficient.
*/
private void resetContainers() {
referencesMap = new LinkedHashMap<>(Math.max(referencesMap.size() / 2, MAP_INITIAL_CAPACITY));
unluckyReferences = new ArrayList<>(Math.max(unluckyReferences.size() / 2, LIST_INITIAL_CAPACITY));
unluckyRefPositions = new ArrayList<>(unluckyReferences.size());
cache = new HashMap<>(Math.max(cache.size() / 2, MAP_INITIAL_CAPACITY));
}
/**
* Add a Reference object to the list - these References will
* be resolved after unmarshalling is complete.
*
* #############################
* # Strategy - Hash Collision #
* #############################
* Suppose that hashing function is h(k) = k % 9;
* __________Input 9 entries_________________________
* Key k | 0, 8, 7, 5, 5, 5, 1, 14, 3 |
* Position# p | 0, 1, 2, 3, 4, 5, 6, 7, 8 |
* Value = p * 3 | 0, 3, 6, 9, 12, 15, 18, 21, 24 |
* --------------------------------------------------
* e.g. eighth entry is Entry#7{ key = 14, value = 21 }
*
* ####################################
* # Insert element - O(1) guaranteed #
* ####################################
* Processing the 9th key:
*
* 1. Attempt to insert Entry#7 with key '14' into map of references.
* > h(14) = 5;
* HashMap buckets:
* position | 0 1 2 3 4 5 6 7 8
* entry(key) | 0 1 3 5 7 8
* ^
* > Bucket 5 is taken.
*
* 2. Store the entry in a separate list.
* List for unlucky references:
* position | 0 1 2
* entry(key) | 5 5
* ^
*
* position | 0 1 2
* entry(key) | 5 5 14
* ^
* 3. Store the position # p of this element, i.e. what spot it would have
* taken if all entries were stored in a position list, counting from zero.
*
* List storing position # of unlucky references:
* position | 0 1 2
* entry # (p) | 4 5
* ^
*
* position | 0 1 2
* entry # (p) | 4 5 7
* ^
*
* #####################################################
* # Retrieve element - O(1) expected, O(n) worst case #
* #####################################################
* Retrieve entry with key '14'
* 1. Attempt to retrieve it from map
* > h(14) = 5;
* HashMap buckets:
* position | 0 1 2 3 4 5 6 7 8
* entry(key) | 0 1 3 5 7 8
* ^
* Hash function points to bucket # 5. Stored key is 5.
* > key 5 != 14.
*
* 2. Iterate through list of unluckyReferences, comparing
* key to all keys in the list.
*
* position | 0 1 2
* entry(key) | 5 5 14
* ^
* > key 5 != 14
*
* position | 0 1 2
* entry(key) | 5 5 14
* ^
*
* > key 5 != 14
*
* position | 0 1 2
* entry(key) | 5 5 14
* ^
*
* > key 14 = 14, retrieve entry.
*
* ##################################################
* # Iterate through all elements - O(n) guaranteed #
* ##################################################
*
* 1. Create boolean array of size n that keeps track
* of unlucky positions:
* > boolean[] a = new boolean[lastPosition + 1];
*
* 2. Set a[p] = true for elements that did not fit into
* hash map, p = position # of element.
*
* > for (Integer p : unluckyRefPositions) {
* > a[p] = true;
* > }
*
* 3. Iterate through LinkedMap and List as if they were one joined collection
* of size s = map.size() + list.size(), ordered by p = position # of element:
* > for (p = 0; p < s; p ++) {
* > if a[p] = false, take next element from linked map iterator,
* > if a[p] = true, take next element from list iterator,
* > }
*
*/
public final void addReference(final Reference ref) {
final ReferenceKey key = new ReferenceKey(ref);
/* If an entry with equal key is already present in map, we preserve the original entry and
* note the position of the new entry into the list of positions and put the new misfit value
* into the list of unluckyReferences. */
final Reference previous = referencesMap.get(key);
if (previous != null || ref.getSourceObject() instanceof Collection) {
unluckyReferences.add(ref);
// The input integer represents the position (starting from 0) of the new element that didn't fit into the map.
unluckyRefPositions.add(referencesMap.size() + unluckyReferences.size() - 1);
} else {
referencesMap.put(key, ref);
}
}
/**
* Retrieve the reference for a given mapping instance. If more keys match, returns the first occurrence.
*/
public final Reference getReference(final ObjectReferenceMapping mapping, final Object sourceObject) {
refKey.setMapping(mapping);
refKey.setSourceObject(sourceObject);
final Reference reference = referencesMap.get(refKey);
if (reference != null) {
return reference;
}
// Search for unlucky references that were kicked out of hashMap by entries with equal key.
for (Reference reference1 : unluckyReferences) {
if (reference1.getMapping() == mapping && reference1.getSourceObject() == sourceObject) {
return reference1;
}
}
return null;
}
/**
* Return a reference for the given mapping and source object, that doesn't already
* contain an entry for the provided field.
*/
public final Reference getReference(final ObjectReferenceMapping mapping, final Object sourceObject,
final Field xmlField) {
final Field targetField = (Field) mapping.getSourceToTargetKeyFieldAssociations().get(xmlField);
String tgtXpath = null;
if (!(mapping.getReferenceClass() == null || mapping.getReferenceClass() == Object.class)) {
if (targetField != null) {
tgtXpath = targetField.getXPath();
}
}
final ReferenceKey key = new ReferenceKey(sourceObject, mapping);
Reference reference = referencesMap.get(key);
if (reference != null && reference.getPrimaryKeyMap().get(tgtXpath) == null) {
return reference;
}
// Search for unlucky references that were kicked out of hashMap by entries with equal key.
for (Reference reference1 : unluckyReferences) {
if (reference1.getMapping() == mapping && reference1.getSourceObject() == sourceObject) {
if (reference1.getPrimaryKeyMap().get(tgtXpath) == null) {
return reference1;
}
}
}
return null;
}
/**
* Store an instance by key based on a mapped class. These values will be
* used when it comes time to resolve the references.
*
* @since EclipseLink 2.5.0
*/
public final void putValue(final Class clazz, final Object key, final Object object) {
Map keyToObject = cache.get(clazz);
if (null == keyToObject) {
keyToObject = new HashMap<>();
cache.put(clazz, keyToObject);
}
keyToObject.put(key, object);
}
/**
* INTERNAL:
* Iterates through all references. Resolves them. Resets containers.
*
* @param session typically will be a unit of work
* @param userSpecifiedResolver a user-provided subclass of IDResolver, may be null
*/
public final void resolveReferences(final CoreAbstractSession session, final IDResolver userSpecifiedResolver,
final ErrorHandler handler) {
final Collection luckyReferences = referencesMap.values();
final Iterator mapIterator = luckyReferences.iterator();
final Iterator listIterator = unluckyReferences.iterator();
/**
* Speed up array which lowers time complexity by a factor of n.
*/
boolean[] a = null;
/**
* Represents position of last Reference that did not fit into hash map.
*/
Integer lastPosition;
if (!unluckyReferences.isEmpty()) {
lastPosition = unluckyRefPositions.get(unluckyRefPositions.size() - 1);
a = new boolean[lastPosition + 1];
for (Integer position : unluckyRefPositions) {
a[position] = true;
}
} else {
lastPosition = -1; // for the condition "i <= lastPosition"
}
for (int i = 0, totalSize = luckyReferences.size() + unluckyReferences.size(); i < totalSize; i++) {
final Reference reference;
// Fast check to see if position [i] should have
// contained Reference that did not fit in into map.
//noinspection ConstantConditions
if (i <= lastPosition && a[i]) {
reference = listIterator.next();
} else {
reference = mapIterator.next();
}
perform(session, userSpecifiedResolver, handler, reference);
}
resetContainers();
}
/**
* Add java doc if you understand this code.
*/
private void perform(final CoreAbstractSession session, final IDResolver userSpecifiedResolver,
final ErrorHandler handler, final Reference reference) {
final Object referenceSourceObject = reference.getSourceObject();
if (reference.getMapping() instanceof CollectionReferenceMapping) {
final CollectionReferenceMapping mapping = (CollectionReferenceMapping) reference.getMapping();
final CoreContainerPolicy cPolicy = mapping.getContainerPolicy();
//container should never be null
final Object container = reference.getContainer();
// create vectors of primary key values - one vector per reference instance
createPKVectorsFromMap(reference, mapping);
// if the we could not generate the primary key for the reference, it will not resolve - skip it
if (reference.getPrimaryKey() == null) {
return;
}
// loop over each pk vector and get object from cache - then add to collection and set on object
Object value = null;
if (!mapping.isWriteOnly()) {
for (Object o : ((Vector) reference.getPrimaryKey())) {
final CacheId primaryKey = (CacheId) o;
if (userSpecifiedResolver != null) {
final Callable c;
try {
if (primaryKey.getPrimaryKey().length > 1) {
final Map idWrapper = new HashMap<>();
for (int y = 0; y < primaryKey.getPrimaryKey().length; y++) {
final ObjectReferenceMapping refMapping =
(ObjectReferenceMapping) reference.getMapping();
final String idName = (String) refMapping.getReferenceDescriptor().
getPrimaryKeyFieldNames().get(y);
final Object idValue = primaryKey.getPrimaryKey()[y];
idWrapper.put(idName, idValue);
}
c = userSpecifiedResolver.resolve(idWrapper, reference.getTargetClass());
} else {
c = userSpecifiedResolver.resolve(primaryKey.getPrimaryKey()[0],
reference.getTargetClass());
}
if (c != null) {
value = c.call();
}
} catch (Exception e) {
throw XMLMarshalException.unmarshalException(e);
}
} else {
value = getValue(session, reference, primaryKey, handler);
}
if (value != null) {
cPolicy.addInto(value, container, session);
}
}
}
// for each reference, get the source object and add it to the container policy
// when finished, set the policy on the mapping
mapping.setAttributeValueInObject(referenceSourceObject, container);
final InverseReferenceMapping inverseReferenceMapping = mapping.getInverseReferenceMapping();
if (inverseReferenceMapping != null && value != null) {
final CoreAttributeAccessor backpointerAccessor = inverseReferenceMapping.getAttributeAccessor();
final CoreContainerPolicy backpointerContainerPolicy = inverseReferenceMapping.getContainerPolicy();
if (backpointerContainerPolicy == null) {
backpointerAccessor.setAttributeValueInObject(value, referenceSourceObject);
} else {
Object backpointerContainer = backpointerAccessor.getAttributeValueFromObject(value);
if (backpointerContainer == null) {
backpointerContainer = backpointerContainerPolicy.containerInstance();
backpointerAccessor.setAttributeValueInObject(value, backpointerContainer);
}
backpointerContainerPolicy.addInto(referenceSourceObject, backpointerContainer, session);
}
}
} else if (reference.getMapping() instanceof ObjectReferenceMapping) {
final CacheId primaryKey = (CacheId) reference.getPrimaryKey();
Object value = null;
if (userSpecifiedResolver != null) {
final Callable c;
try {
if (primaryKey.getPrimaryKey().length > 1) {
final Map idWrapper = new HashMap<>();
for (int y = 0; y < primaryKey.getPrimaryKey().length; y++) {
final ObjectReferenceMapping refMapping = (ObjectReferenceMapping) reference.getMapping();
final String idName = (String) refMapping.getReferenceDescriptor()
.getPrimaryKeyFieldNames().get(y);
final Object idValue = primaryKey.getPrimaryKey()[y];
idWrapper.put(idName, idValue);
}
c = userSpecifiedResolver.resolve(idWrapper, reference.getTargetClass());
} else {
c = userSpecifiedResolver.resolve(primaryKey.getPrimaryKey()[0], reference.getTargetClass());
}
if (c != null) {
value = c.call();
}
} catch (Exception e) {
throw XMLMarshalException.unmarshalException(e);
}
} else {
value = getValue(session, reference, primaryKey, handler);
}
ObjectReferenceMapping mapping = (ObjectReferenceMapping) reference.getMapping();
if (value != null) {
mapping.setAttributeValueInObject(reference.getSourceObject(), value);
}
if (null != reference.getSetting()) {
reference.getSetting().setValue(value);
}
InverseReferenceMapping inverseReferenceMapping = mapping.getInverseReferenceMapping();
if (inverseReferenceMapping != null) {
CoreAttributeAccessor backpointerAccessor = inverseReferenceMapping.getAttributeAccessor();
CoreContainerPolicy backpointerContainerPolicy = inverseReferenceMapping.getContainerPolicy();
if (backpointerContainerPolicy == null) {
backpointerAccessor.setAttributeValueInObject(value, referenceSourceObject);
} else {
Object backpointerContainer = backpointerAccessor.getAttributeValueFromObject(value);
if (backpointerContainer == null) {
backpointerContainer = backpointerContainerPolicy.containerInstance();
backpointerAccessor.setAttributeValueInObject(value, backpointerContainer);
}
backpointerContainerPolicy.addInto(reference.getSourceObject(), backpointerContainer, session);
}
}
}
}
/**
* INTERNAL:
* Create primary key values to be used for cache lookup. The map
* of primary keys on the reference is keyed on the reference descriptors primary
* key field names. Each of these primary keys contains all of the values for a
* particular key - in the order that they we read in from the document. For
* example, if the key field names are A, B, and C, and there are three reference
* object instances, then the hashmap would have the following:
* (A=[1,2,3], B=[X,Y,Z], C=[Jim, Joe, Jane]). If the primary key field names on
* the reference descriptor contained [B, C, A], then the result of this method call
* would be reference.primaryKeys=([X, Jim, 1], [Y, Joe, 2], [Z, Jane, 3]).
*/
private void createPKVectorsFromMap(final Reference reference, final CollectionReferenceMapping mapping) {
final CoreDescriptor referenceDescriptor = mapping.getReferenceDescriptor();
final Vector pks = new Vector<>();
if (null == referenceDescriptor) {
final CacheId pkVals = (CacheId) reference.getPrimaryKeyMap().get(null);
if (null == pkVals) {
return;
}
for (int x = 0; x < pkVals.getPrimaryKey().length; x++) {
final Object[] values = new Object[1];
values[0] = pkVals.getPrimaryKey()[x];
pks.add(new CacheId(values));
}
} else {
final List pkFields = referenceDescriptor.getPrimaryKeyFieldNames();
if (pkFields.isEmpty()) {
return;
}
boolean init = true;
// for each primary key field name
for (Object pkField : pkFields) {
final CacheId pkVals = (CacheId) reference.getPrimaryKeyMap().get(pkField);
if (pkVals == null) {
return;
}
// initialize the list of pk vectors once and only once
if (init) {
for (int i = 0; i < pkVals.getPrimaryKey().length; i++) {
pks.add(new CacheId(new Object[0]));
}
init = false;
}
// now add each value for the current target key to it's own vector
for (int i = 0; i < pkVals.getPrimaryKey().length; i++) {
final Object val = pkVals.getPrimaryKey()[i];
(pks.get(i)).add(val);
}
}
}
reference.setPrimaryKey(pks);
}
/**
* Add java doc if you understand this code.
*/
private Object getValue(final CoreAbstractSession session, final Reference reference, final CacheId primaryKey,
final ErrorHandler handler) {
final Class referenceTargetClass = reference.getTargetClass();
if (null == referenceTargetClass || referenceTargetClass == CoreClassConstants.OBJECT) {
for (Object entry : session.getDescriptors().values()) {
Object value = null;
final Descriptor targetDescriptor = (Descriptor) entry;
final List pkFields = targetDescriptor.getPrimaryKeyFields();
if (null != pkFields && 1 == pkFields.size()) {
Field pkField = (Field) pkFields.get(0);
pkField = (Field) targetDescriptor.getTypedField(pkField);
final Class targetType = pkField.getType();
if (targetType == CoreClassConstants.STRING || targetType == CoreClassConstants.OBJECT) {
value = getValue(targetDescriptor.getJavaClass(), primaryKey);
} else {
try {
final Object[] pkValues = primaryKey.getPrimaryKey();
final Object[] convertedPkValues = new Object[pkValues.length];
for (int x = 0; x < pkValues.length; x++) {
convertedPkValues[x] = session.getDatasourcePlatform().getConversionManager()
.convertObject(pkValues[x], targetType);
}
value = getValue(targetDescriptor.getJavaClass(), new CacheId(convertedPkValues));
} catch (ConversionException ignored) {
}
}
if (null != value) {
return value;
}
}
}
if (primaryKey.getPrimaryKey()[0] != null) {
final XMLMarshalException e = XMLMarshalException.missingIDForIDRef(
Object.class.getName(), primaryKey.getPrimaryKey());
if (handler != null) {
final SAXParseException saxParseException = new SAXParseException(e.getLocalizedMessage(), null, e);
try {
handler.warning(saxParseException);
} catch (SAXException saxException) {
throw e;
}
}
}
return null;
} else {
Object value = getValue(referenceTargetClass, primaryKey);
if (null == value) {
final CoreMapping mapping = (CoreMapping) reference.getMapping();
final CoreDescriptor targetDescriptor = mapping.getReferenceDescriptor();
if (targetDescriptor.hasInheritance()) {
final CoreInheritancePolicy inheritancePolicy = targetDescriptor.getInheritancePolicy();
final List childDescriptors = inheritancePolicy.getAllChildDescriptors();
for (CoreDescriptor childDescriptor : childDescriptors) {
value = getValue(childDescriptor.getJavaClass(), primaryKey);
if (null != value) {
return value;
}
}
}
}
if (value == null && (primaryKey.getPrimaryKey()[0] != null)) {
final XMLMarshalException e = XMLMarshalException.missingIDForIDRef(
referenceTargetClass.getName(), primaryKey.getPrimaryKey());
if (handler != null) {
SAXParseException saxParseException = new SAXParseException(e.getLocalizedMessage(), null, e);
try {
handler.warning(saxParseException);
} catch (SAXException saxException) {
throw e;
}
}
}
return value;
}
}
/**
* Retrieves value from {@link #cache}.
*/
private Object getValue(Class clazz, CacheId primaryKey) {
Map keyToObject = cache.get(clazz);
if (null != keyToObject) {
return keyToObject.get(primaryKey);
}
return null;
}
/**
* This class serves to represent a key for {@link org.eclipse.persistence.internal.oxm.Reference} when used
* in maps.
*/
private static final class ReferenceKey {
private Object sourceObject;
private Mapping mapping;
public ReferenceKey() {
}
public ReferenceKey(final Object sourceObject, final Mapping mapping) {
this.sourceObject = sourceObject;
this.mapping = mapping;
}
public ReferenceKey(final Reference ref) {
this.sourceObject = ref.getSourceObject();
this.mapping = ref.getMapping();
}
public void setMapping(final ObjectReferenceMapping mapping) {
this.mapping = mapping;
}
public void setSourceObject(final Object sourceObject) {
this.sourceObject = sourceObject;
}
@Override
public final int hashCode() {
int result = System.identityHashCode(sourceObject);
result = 31 * result + System.identityHashCode(mapping);
return result;
}
@Override
public final boolean equals(final Object o) {
if (this == o) return true;
if (!(o instanceof ReferenceKey)) return false;
ReferenceKey that = (ReferenceKey) o;
return sourceObject == that.sourceObject && mapping == that.mapping;
}
}
}