ValueHolder Indirection and TopLink Cache

We have a parent object A (main table) and child object B (a lookup table), mapped thought ValueHolder Indirection.
We use ReadAllQuery to build SQL and retrieve object A(s), and associated B(s).
The query result is used to populate a web page table.
The problem is that when the table is populated, the exact same lookup query against object B is repeated without checking cache. That data field is populated by object_A.object_b.description. For example, below same query would be repeated many times when loading up the web page table:
Select description_id, description from Table_B where description_id = 1
Why the query generated by this Indirection not checking cache? Anyway to force it to check TopLink cache first before querying against database?
Thanks for any help!
Jeffrey

Thanks for the reply.
We use JDeveloper 10.1.3.2. TopLink map in JDev is used to map all table objects and their relationships. So In TopLink map, object(table) A has a ValueHolder object(table) B through indirection. The primary key of B is: description_id, which is used in the table reference mapping.
We use the default TopLink settings in JDev, so object B has below settings in TopLink map:
Identity Map: FullIdentityMap
Size: 50 (there are only about 20 records in this lookup table)
Existence Checking: Check Cache
We don't have any other caching mechanism other than TopLink's. EJB 3.0 is used as service bean, and External Transaction Controller (OC4J) is used.
How to check if B is already in the TopLink cache? I heard ReadAllQuery always goes to database w/o checking TopLink cache, but in the case, the query generated by lazy loading indirection is after the ReadAllQuery execution (when the web page table is loading up).
Jeffrey

Similar Messages

Database Change Notification and TopLink Cache Invalidation

Has someone succeeed in implementing the How-to Database Change Notification and TopLink Cache Invalidation.
I have corrected some document errata about the pl/sql content and I manage to have messages in the 'notify_queue'.
I obtain the Topic in Java from this queue.
But the TopicSuscriber instances do not receive any message. Is there something to have in mind to make it work ?
Regards.

Reviving this thread again...
I am using DCN feature to build a middle-tier cache. I know oracle has problem sending physical rowid in case of 'Index Organized Table', however, in normal table also its not able to send proper rowid.
e.g, I have 2 records in Table A with rowid AAARIUAAGAAAV/uABw and AAARIUAAGAAAV/pAAX.
I have updated both the records. Strangely for the first record, oracle is sending INVALID rowid, although for the second record its sending the valid one.
Following is the output:
Row 1: (Wrong rowid being sent, AAARIUAAGAAAV/uABw is replaced with AAARIUAAGAAAXDCAAr)
Connection information : local=localhost.localdomain/127.0.0.1:47633, remote=localhost.localdomain/127.0.0.1:2278
Registration ID         : 2102
Notification version    : 1
Event type              : OBJCHANGE
Database name           : <sid>
Table Change Description (length=1)
    operation=[UPDATE], tableName=<table_name>, objectNumber=70164
    Row Change Description (length=1):
      ROW: operation=UPDATE, ROWID=AAARIUAAGAAAXDCAAr
Row 2: (Right rowid being sent, AAARIUAAGAAAV/pAAX)
Connection information : local=localhost.localdomain/127.0.0.1:47633, remote=localhost.localdomain/127.0.0.1:2278
Registration ID         : 2102
Notification version    : 1
Event type              : OBJCHANGE
Database name           : <sid>
Table Change Description (length=1)
    operation=[UPDATE], tableName=<table_name>, objectNumber=70164
    Row Change Description (length=1):
      ROW: operation=UPDATE, ROWID=AAARIUAAGAAAV/pAAX
Any idea ?

Child Objects and TopLink Cache

All,
I have a problem RE the TopLink cache:
Object A has a Vector of Object Bs (1:M) and Object B has a Vector of Object Cs(1:M). I am using ValueHolderInterface and indirection pattern for each of these Vectors.
When I update an Object C, it is not refreshed the next time I read Object A using the readObject(expression). I can see the changes in the database. Can someone tell me the best way to refresh the cache to get the updated Object Cs that belong to Object A.
What I am doing is updating the C objects that belong to object A (thru Vector B) and then retrieving them again in the very next method call. Hope this makes sense!
Thanks!
J

There is something wrong with your test case, I've seen this before -- if you update a C, then without fail the cached version of C is updated and if you have a handle to the cached A that has the B that has the C in question, then you will see the update. It sounds like perhaps you're not actually looking at the cached A, but instead looking at it from a UOW, etc.
Send me an email, it's simply my firstname . lastname at Oracle.com. I'll send you a UOW primer that should help better understand these semantics...
- Don

TopLink caching behaving strangly?

Hi folks,
Realizing that my previous post titled 'Session#acquireUnitOfWork() resets identiy maps' at Session#acquireUnitOfWork() resets identiy maps may be a bit muddleheaded I'll take a different approach at the same problem.
I'm not yet very familiar with TopLink caching (links to good articles greatly appreciated) but the behaviour we see here still does not make sense to me. Caching is provided by identity maps referenced by the server session. That much I understand. So, every time an application reads/writes data from/to the database TopLink would update those identity maps - if necessary. Based on that I assume that those maps remain untouched as long as I do not modify a certain object and commit those modifications.
[Pseudo code]
1: TopLink.save(obj1)
2: obj2 = TopLink.read(queryForObj1)
3: obj1 == obj2
It all boils down to the simple question why obj1 and obj2 are two different instances? I would expect them to be one and the same instance.
I would greatly appreciate if someone could provide some clarification.
Regards,
Marcel

Hello Marcel,
Your question in this and the other thread touch on how the UnitOfWork and TopLink caching works. Basically what you are seeing is probably because you are reading instances from different UnitOfWork objects. Each UOW has its own cache to maintain transactional issolation. Using a UOW will return a cloned copy of the object contained in the central cache so that you can modify it without changes being seen outside the transaction. On commit, the changes in the clone are put in the database and merged into the object in the main cache. Only changes made to the UnitOfWork working copy of an object will be persisted, but you can read from the session or ClientSession directly instead of the UnitOfWork - this will return the copy from the central cache.
When you commit/release a UnitOfWork, all working copies become detached; changes will no longer be persisted to the database.
I would recommend looking over the white paper on the UnitOfWork at:
http://www.oracle.com/technology/products/ias/toplink/technical/unitOfWorkWP.pdf
Best Regards,
Chris

Hibernate's dual-layer cach and TopLink's caching strategy

Dear members,
I understand that caching between hibernate and toplink is implemented (or utilized) differently. Hibernate seems to have 'dual-layer caching' (which may imply they have two layers of cache) whereas TopLink has session cache and shared cache. The way I see it, they seem to be aiming for the same thing. Are there any differences between (obviously there are, only that I do not know them) those two caching architectures, and how different are they?
Howard

Yes there are differences :) For details check out
TopLink vs Hibernate... revisited... again :)
and
Indirection - how are references resolved after session has been closed?

TopLink caching policy and inheritance

Just looking for some confirmation on TopLink caching policy behavior for descriptors involving inheritance.
The TopLink workbench allows me to specify the "Identity" policy advanced property for descriptor that involve inheritance (subclasses).
Subclass descriptors can have different identity policies right?
I could have some subclasses (descriptors) which I wish to use a FullIdentityMap and other subclasses (descriptors) that I wish to have a SoftCacheWeakIdentityMap?
Thanks for clarification. The manuals don't really clarify this behavior.
...Steve

TopLink maintains one identity map (cache) per root class. That means that all settings are shared for an inheritance tree and cannot be overridden on a per class basis.
Doug

Toplink Caching issue

I have three Entities say C1 , P1 and card. Relation between these
entities are like this
P1 to Card ( one - one mapping )
C1 to Card ( one - many mapping )
I have three usecases. Usecase 1 uses the mail Entity C1. It will
create one record using C1. ( keep in mind that there will not be any data for entity card at that time ).
Second usecase , Usecase 2 , uses Entity P1 and creates data in P1 as
well as Card.
Third usecase will again take C1 and try to get the card data which
were inserted through P1.
But unfortunately , it is coming as null ... But if I restart the server and access the third usecase then I'm getting all the card data from C1 Entity which were saved through P1 Entity .
To summarise my finding , issue is with Toplink caching. When i
accessed the first usecase, C1 is cached. In the second usecase , I updated Entity card using P1. When i accessed the third usecase , toplink is returing the cached entity. Toplink is not able to identify whether related entity got changed or not....
Workaround is not to use cache for C1 entity. It is working but it will
not be the correct solution.
any one faced this issue in Toplink ?
Correct me if my findings are not correct...

Hello,
Its not that the cache isn't working, its that you are updating relationships in the database but not in the cache. Take the case of a 1:M from B->C used in a previous response. It looks like the foreign key used in this relationship (in the C table) is mapped to the C object either as a direct to field mapping, or through a 1:1 to the A object. When you change this so that it now somehow points to the B object, you haven't updated the B object to include the C in its collection. Because B has a 1:M mapping collection of Cs, when you next get that B object from the cache, its collection will not contain the newly referenced C (assuming indirection is not used or has been previously triggered). This is a result of you mapping the 1:M in the first place - since it is storing the collection in the object.
If you are going to make changes that affect relationships and yet not update the relationships directly I would suggest either:
A) refreshing affected objects from the database when done. In this case, refreshing the B object will result in TopLink requerying the B->C relation and picking up any changes (depending on your refresh options on the B descriptor though - please check the docs if you need more info).
B) Not mapping these relationships. Instead, when you need to find C objects referenced from B, query for them.
Both options are less efficient as they cause more hits to the database. This is why it would be preferred if you updated the relationships when you change fields that affect those relationships, but it is up to you to decide which will work best for your application.
Please note that the cache is working, in fact the situation you are seeing is due to it working. I am not sure of how your situation would work with DAO as you mentioned in a previous post, but it sounds like it would always hit the database. This situation is a result of you making changes in the database but not changing the object model to reflect those changes, proving that the cache is indeed working. While you mention that some objects are not in your use case, they are in the database but just not mapped in TopLink. In the case of a 1:M mapping, the database shows a 1:1 back relationship that you have not mapped in your object model. So this would be more a case of your object model not matching your database model, and not being taken into account in your use cases.
Best Regards,
Chris

Understanding ValueHolder Indirection

Hello Everyone:
I have a question regarding ValueHolder Indirection. I have encountered a situation for which I cannot adequately explain the outcome.
Consider that I have an object named A for which the optimistic locking policy is set to version locking. A has attributes B, C, and D that are declared as type ValueHolderInterface. All objects map to different database tables. In Toplink, the descriptor for object A has one-to-one mappings for each of the attributes B, C, and D and each object is set to use indirection.
In my code, I register object A and set new instances of B, C, and D in object A. Entries in the database are being properly created for these three objects, but the version is not being updated for object A. How does utilizing ValueHolder Indirection affect the setting of the version number? When I set an attribute that is a direct-to-field mapping, the version is updated. Am I missing something? I would think that the parent object A would be updated when B, C, and D are set.
Any insight would be appreciated.
Thanks.

Hello,
This happens regardless of Indirection being involved. The way versioning works is that it only gets updated if a field in A's table requires an update. This is a performance enhancement so that if something in A changes, the version is not automatically incremented to avoid forcing an unneccessary update to the A table. This has the bigest impact in 1:M and M:M mappings, where say adding a new object to a collection really does not require a change to the parent object. In these cases, you can force the version field update using the uow.forceUpdateToVersionField(object,true) api.
It can also happen in bi-directional 1:1s, on the side of the relationship that is set as the target foreign key. For example, if A has a 1:1 to B, but the foreign key is in the B table, changing the A->B relationship does not result in an update (as A does not own the foreign key). The foreign key is only affected from the B side, and needs to be mapped as either a 1:1 or a DirectToFieldMapping.
Hope this helps,
Chris

Spring and TopLink/JPA on OC4J 10.1.3.4.0

Our configuration worked fine on OC4J 10.1.3.1 (linux) and 10.1.3.3 (windows). We have a problem with JPA (toplink essentials--not eclipselink) on OC4J 10.1.3.4.0 using Spring (2.0.7 and 2.5.6). We found that it was trying to connect to the default OracleDS data source(there were 2 connections in toplink essentials for some reason), but once we defined an OracleDS, it gave us this error:
javax.servlet.ServletException: javax.faces.el.EvaluationException: Error getting property 'experiments' from bean of type gov.llnl.nif.dataviz.mssar.web.ExperimentListBean: org.springframework.transaction.CannotCreateTransactionException: Could not open JPA EntityManager for transaction; nested exception is java.lang.IllegalStateException:
Exception Description: Cannot use an EntityTransaction while using JTA.
Our persistence.xml looks like:
<?xml version="1.0" encoding="UTF-8"?>
<persistence version="1.0" xmlns="http://java.sun.com/xml/ns/persistence" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://java.sun.com/xml/ns/persistence http://java.sun.com/xml/ns/persistence/persistence_1_0.xsd">
     <persistence-unit name="mssar">
          <class>gov.llnl.nif.dataviz.mssar.model.ExperimentInfo</class>
          <class>gov.llnl.nif.dataviz.mssar.model.UsePlan</class>
     <properties>
     <property name="toplink.cache.shared.gov.llnl.nif.dataviz.mssar.model.ExperimentInfo" value="false"/>
     <property name="toplink.cache.shared.gov.llnl.nif.dataviz.mssar.model.UsePlan" value="false"/>
     </properties>
     </persistence-unit>
</persistence>
Our orm.xml looks like:
<?xml version="1.0" encoding="UTF-8"?>
<entity-mappings xmlns="http://java.sun.com/xml/ns/persistence/orm" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://java.sun.com/xml/ns/persistence/orm http://java.sun.com/xml/ns/persistence/orm_1_0.xsd" version="1.0">
</entity-mappings>
Our spring configuration looks like:
<?xml version="1.0" encoding="UTF-8"?>
<beans xmlns="http://www.springframework.org/schema/beans"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xmlns:jee="http://www.springframework.org/schema/jee"
xmlns:aop="http://www.springframework.org/schema/aop"
     xmlns:tx="http://www.springframework.org/schema/tx"
xsi:schemaLocation="
http://www.springframework.org/schema/beans http://www.springframework.org/schema/beans/spring-beans-2.0.xsd
http://www.springframework.org/schema/jee http://www.springframework.org/schema/jee/spring-jee-2.0.xsd
http://www.springframework.org/schema/tx http://www.springframework.org/schema/tx/spring-tx-2.0.xsd
     http://www.springframework.org/schema/aop http://www.springframework.org/schema/aop/spring-aop-2.0.xsd"
     default-autowire="byName">
<bean id="entityManagerFactory" class="org.springframework.orm.jpa.LocalContainerEntityManagerFactoryBean">
<property name="dataSource" ref="dataSource"/>
<property name="jpaVendorAdapter">
<bean class="org.springframework.orm.jpa.vendor.TopLinkJpaVendorAdapter">
<property name="showSql" value="true"/>
<property name="databasePlatform" value="oracle.toplink.essentials.platform.database.oracle.OraclePlatform"/>
</bean>
</property>
<property name="loadTimeWeaver">
     <bean class="org.springframework.instrument.classloading.InstrumentationLoadTimeWeaver"/>
</property>
</bean>
<jee:jndi-lookup id="dataSource" jndi-name="jdbc/SSARDS"/>
<bean id="transactionManager" class="org.springframework.orm.jpa.JpaTransactionManager">
<property name="entityManagerFactory" ref="entityManagerFactory" />
     <property name="dataSource" ref="dataSource" />
</bean>
     <bean id="experimentsQueryBuilder" class="gov.llnl.nif.dataviz.mssar.dao.jpa.ExperimentsQueryBuilderJpa"/>
     <bean id="experimentsDao" class="gov.llnl.nif.dataviz.mssar.dao.jpa.ExperimentsDaoJpa"/>
     <bean id="experimentListManager" class="gov.llnl.nif.dataviz.mssar.service.impl.ExperimentListManagerImpl"/>
</beans>
ExperimentInfo.java mapping looks like:
@Entity
@Table(name = "EXPERIMENT_IDS")
public class ExperimentInfo implements ExperimentInformation {
     @Id
     @Column(name="EXPERIMENT_ID")
     String experimentId;
     @Column(name="EXPERIMENT_DESCRIPTION")
     String expDescription;
     @Column(name="STATUS")
     String status;
     @Column(name="EXPERIMENT_PI")
     String expPI;
     //Date lpomDate;
     //Date settingsDate;
     @Column(name="LPOM_DATE")
     Timestamp lpomDate;
     @Column(name="SETTINGS_DATE")
     Timestamp settingsDate;
UsePlan.java looks like:
@Entity
@Table(name="USE_PLAN")
public class UsePlan extends BasePlan implements PlanData, Serializable {
     static final long serialVersionUID = -1555327955428351924L;
     @Id
     @Column(name="EXP_SHOT_ID")
     String expShotId;
     @Id
     @Column(name="LOCATION")
     String location;
     @Id
     @Column(name="INFO_VAR")
     String infoVar;
     @Id
     @Column(name="USE_ITEM_NAME")
     String useItemName;
     @Column(name="USE_PRIORITY")
     String usePriority;
ExperimentListManagerImpl looks like:
@Transactional(readOnly=true)
public class ExperimentListManagerImpl implements ExperimentListManager {
This is a read-only app but it should not use the cache.
Our Dao Looks like:
public class ExperimentsDaoJpa extends JpaDaoSupport implements ExperimentsDao {
private QueryBuilder experimentsQueryBuilder;
public List<ExperimentInfo> shotsByQueryBean(ExperimentsQueryBean queryBean, String sortColumn, boolean ascending) {
     try {
          getJpaTemplate().flush();
          experimentsQueryBuilder.buildQuery(queryBean);
     experimentsQueryBuilder.setSortColumn(sortColumn, ascending);
     List<ExperimentInfo> result = getJpaTemplate().findByNamedParams(experimentsQueryBuilder.getQuery().toString(), experimentsQueryBuilder.getParams());
     // pull new results from database
     /*for (ExperimentInfo ei : result) {
          getJpaTemplate().refresh(ei);
     return result;
     } catch (Throwable t) {
          t.printStackTrace();
          return null;
     public void setExperimentsQueryBuilder(QueryBuilder experimentsQueryBuilder) {
          this.experimentsQueryBuilder = experimentsQueryBuilder;
Thanks,
John Carlson

We also had this problem. We have changed the TransactionManager.
Replace the JPATransactionManager with the JTATransactionManager.
<bean id="transactionManager" class="org.springframework.transaction.jta.JtaTransactionManager" />
If you are using JTA you can't use the JPA Transaction Manager. It doesn't support JTA.
Found it on a forum of Spring. (http://forum.springframework.org/showthread.php?t=48191)
Dennis

Querying the toplink cache under high-load

We've had some interesting experiences with "querying" the TopLink Cache lately.
It was recently discovered that our "read a single object" method was incorrectly
setting query.checkCacheThenDB() for all ReadObjectQueries. This was brought to light
when we upgraded our production servers from 4 cores to 8. We immediatly started
experiencing very long response times under load.
We traced this down to the following stack: (TopLink version 10.1.3.1.0)
at java.lang.Object.wait(Native Method)
- waiting on <0x00002aab08fd26d8> (a oracle.toplink.internal.helper.ConcurrencyManager)
at java.lang.Object.wait(Object.java:474)
at oracle.toplink.internal.helper.ConcurrencyManager.acquireReadLock(ConcurrencyManager.java:179)
- locked <0x00002aab08fd26d8> (a oracle.toplink.internal.helper.ConcurrencyManager)
at oracle.toplink.internal.helper.ConcurrencyManager.checkReadLock(ConcurrencyManager.java:167)
at oracle.toplink.internal.identitymaps.CacheKey.checkReadLock(CacheKey.java:122)
at oracle.toplink.internal.identitymaps.IdentityMapKeyEnumeration.nextElement(IdentityMapKeyEnumeration.java:31)
at oracle.toplink.internal.identitymaps.IdentityMapManager.getFromIdentityMap(IdentityMapManager.java:530)
at oracle.toplink.internal.queryframework.ExpressionQueryMechanism.checkCacheForObject(ExpressionQueryMechanism.java:412)
at oracle.toplink.queryframework.ReadObjectQuery.checkEarlyReturnImpl(ReadObjectQuery.java:223)
at oracle.toplink.queryframework.ObjectLevelReadQuery.checkEarlyReturn(ObjectLevelReadQuery.java:504)
at oracle.toplink.queryframework.DatabaseQuery.execute(DatabaseQuery.java:564)
at oracle.toplink.queryframework.ObjectLevelReadQuery.execute(ObjectLevelReadQuery.java:779)
at oracle.toplink.queryframework.ReadObjectQuery.execute(ReadObjectQuery.java:388)
We moved the query back to the default, query.checkByPrimaryKey() and this issue went away.
The bottleneck seemed to stem from the read lock on the CacheKey from IdenityMapKeyEnumeration
public Object nextElement() {
if (this.nextKey == null) {
throw new NoSuchElementException("IdentityMapKeyEnumeration nextElement");
// CR#... Must check the read lock to avoid
// returning half built objects.
this.nextKey.checkReadLock();
return this.nextKey;
We had many threads that were having contention while searching the cache for a particular query.
From the stack we know that the contention was limited to one class. We've since refactored that code
not to use a query in that code path.
Question:
Armed with this better knowledge of how TopLink queries the cache, we do have a few objects that we
frequently read by something other than the primary key. A natural key, but not the oid.
We have some other caching mechanisms in place (JBoss TreeCache) to help eliminate queries to the DB
for these objects. But the TreeCache also tries to acquire a read lock when accessing the cache.
Presumably a read lock over the network to the cluster.
Is there anything that can be done about the read lock on CacheKey when querying the cache in a high load
situation?

CheckCacheThenDatabase will check the entire cache for a match using a linear search. This can be efficient if the cache is very large. Typically it is more efficient to access the database if your cache is large and the field you are querying in the database is indexed in the table.
The cache concurrency was greatly improved in TopLink 11g/EclipseLink, so you may wish to try it out.
Supporting indexes in the TopLink/EclipseLink cache is something desirable (feel free to log the enhancement request on EclipseLink). You can simulate this to some degree using a named query and a query cache.
-- James : http://www.eclipselink.org

Refresh toplink cache from tirgger

How to update the toplink cache due to change in the database by some external process or procedure?
This question was posted some time back and one of the suggestion was to create a trigger on the table holding the data and implement callout to the toplink cache to refresh. I will appreciate if any one can let me know where I can I find more information to implement such a callout method from trigger on the database table.
We are accessing the toplink objects from OC4J container from where a singleton is managing the calls to the toplink objects. We already have methods in place to refresh the cached object based on timeouts but now the new requirements are to refresh the objects only if the data is changed in the database.
Thanks
Ahmad

I have url error on this thread : How to refresh cache in TopLink, turn off cache
[b]Discussion Forums Error
We cannot process your request at this time. Please try again later.
Thank's

Session Facade design and TopLink objects

Hi, we are just about to start using Toplink as our ORM tool. Ideally we want to wrap all business logic behind stateless session bean facades. So instead of BMP entity beans we are thinking of encapsulating the relevant Toplink objects in each facade and querying, updating that object when the session facade bean is instantiated.
We are using Weblogic 70. Example code:
e.g. MySessionBean implements sessionbean...
     MyTopLinkObject myObject = null;
     public void populateObject(String myID) throws java.rmi.RemoteException {
          try {
               MyTopLinkObject myObject = new MyTopLinkObject(myID);
          catch (Exception e) {
     public void setMyTopLinkObjectsName(String name) {
     //persist this change via the toplink object...
     myObject.setName(name);
Then in MyTopLinkObject:
public class MyTopLinkObject {
     private String myID;
     private String myName;
public MyTopLinkObject(String myID) {
          build(myID);
public void build(String myID) {
          //read TOPLINK cache and populate this MyTopLinkObject name etc
public void setName(String myName) {
     //persist this change...
     this.myName = myName;
// more gets sets
I will then want to update MyTopLinkObject data in for example in calling. MySessionBean.setMyTopLinkObjectsName("my name") from the remote interface on the client.
So what I want to know is how do I can I do this efficiently. I'm too too versant with TopLink caches etc. For example to read the cache do I have to pass thru a TopLink client/server Session to each Session Facade bean so the cache is available for that read, update etc?
Any other tips is using this design would be appreciated!
Thanks in advance,
J

Don, thanks for ur advice yet again.
I think now I am getting a better grasp of how TopLink should be used. However there is very little doco out there in terms of using TOPLInk in as EJB session facade scenario.
After your advice my thoughts are to provide something like the TOPLInkFactory object as you say which will encapsulate the ServerSession TOPLink API and provide more fine grain access to the TOPLInk API. e.g. instead of having to create a ClientSession in each client call b4 getting a UnitOfWOrk, this factory class would provide a method to do this directly:
public final class TOPLinkFactory {
     private static ServerSession serverSession = null;
     private static final String TOPLINK_CONFIG_FILE = "C:\\test.xml";
     private static final String TOPLINK_SESSION = "toplink";
     public TOPLinkFactory() throws Exception {
          super();
     private static ServerSession createServerSession(ClassLoader cl) throws Exception {
          try {
          Project builderProject = XMLProjectReader.read(TOPLINK_CONFIG_FILE, cl);
     serverSession = (ServerSession)builderProject.createServerSession();
          serverSession.login();
               return serverSession;
          catch (Exception e) {
          return null;
     public static UnitOfWork getUnitOfWork(ClassLoader cl) {
          try {
               return createClientSession(cl).acquireUnitOfWork();
     public static ClientSession createClientSession(ClassLoader cl) throws Exception {
          try {
               if (serverSession == null) {
serverSession = createServerSession(cl);
ServerSession s = (ServerSession)SessionManager.getManager().getSession(TOPLINK_SESSION, cl);
               return serverSession.acquireClientSession();
          catch (Exception e) {
          return null;
Then in my Session EJBs I could write a method like this where Thing is a TOPLInk object mapped to the thing table:
     public boolean setThingStatus(String thingID) {
ExpressionBuilder builder = new ExpressionBuilder();
Expression expression = builder.get("THING_ID").equal( Integer.parseInt(thingID));
ClientSession s = TOPLinkFactory.createClientSession(this.getClass().getClassLoader());
Thing thing = (Thing) s.readObject(Thing.class, expression);
UnitOfWork uow = TOPLinkFactory.getUnitOfWork(this.getClass().getClassLoader());
Thing thingClone = (Thing)uow.registerObject(thing);
thingClone.setStatus("WITH JASON");
TOPLinkFactory.commitAndReleaseUnitOfWork(uow);
return true;
Now if the above looks OK and there are no Thread/Session issues that you could make me aware of (or perhaps you would encourage using a full Singleton pattern instead).
I have another question in relation to ClassLoaders. In this architecture I have been passing around the ClassLoader (as above) to the TOPLInkFactory methods to create the unit of works etc... Is this necessary? Or could I avoid it?
Also I think some people in my team are thinking its a good idea for all TOPLink business objects to extend an abstract base class that contains this behaviour and provides these TOPLinkFactory methods. I no expert on TOPLInk but I would have thought this code should not be coupled with the TOPLink business objects for reasons such as:
1) Why should a TOPLink Person object (attributes name, age etc) e.g. extend a class with TOPLink API methods and attributes - I just don't see the logic there as I thought TOPLink was to be non-intrusive. For example a Person object should not encapsulate a ServerSession object and nor does it need to. To this end they are then coupled and if for example we wanted to move these Person objects to work with another persistence framework then they would have to be rewritten to remove all TOPLInk attribute/method references
2) I would have thought accessing the TOPLInk API is only necessary when the client code needs to invoke calls to the database for reads and updates. i.e. that's where you start getting your sessions, unit of works etc. Like in the setThingStatus Session EJB method (i.e. the client code) above and NOT in your business methods of for example the TOPLink Person object?
Please tell me if I am mistaken!
If you could clear some of this up it would be great.
Regards,
Jason

How do I sync Toplink Cache with my Database?

Hey guys,
We are running macro's through an excel sheet that connects to the database and performs Updates and Inserts.
Since this database update is NOT taking place through TopLink (in a Unit of Work) - we do not see the database changes through the web app on the front end unless we bounce our webserver. Presumably the Toplink cache is re-built on start up...so then we can see the changes.
My question is, what can I do to make sure the TopLink cache is aware of the database changes we have made through the macro without having to bounce the server? Is there a re-fresh or sync command that can be run?
This task is sort of a one time thing, so I don't want a solution that involves the cache going to sync itself on a schedule or anything like that. Maybe bouncing is the best solution?
Thoughts?
We are using Toplink 9.0.4
Thanks.

Hello,
Because it is a one time thing, if you can make sure no other TopLink process are going on, you can probably get away with initializeAllIdentityMaps() or the intitializeIdentityMap(Class) methods on the session. These will clear the identity maps, having the obvious draw back of removing all object identity which will cause problems with running processes though. It might be better than bouncing the server - it depends on the application. Logging out of the session and logging back in has the same effect of clearing the cache, but with a bit more overhead. Benifit is that running processes will get errors if they continue to use the session, rather than strange behaviior if they continue to use objects after identity is lost.
Another alternative is to run refresh queries on the data you know might be in the cache or that might have been affected. Drawbacks are that this brings them into the cache if they are not already there.
TopLink 9.0.4 is a quire a few versions back, and in newer versions there is cache invalidation. An object marked as invalid is not removed, so object identity is maintained, but on the next query the data will be refreshed - ensuring subsequent queries will get results from the database without having to be told explicietly to refresh or being set to always refresh. Invalidation can be triggered on particular objects, classes or the entire cache or different policies can be set to set a time to live etc.
Except for bouncing the server or logging out of the session, all of the above leave some possibility that a concurrent user will still have a reference to stale data and continue to use it after the process has run on the database. So I hope you use optimistic locking and that your batch process updates the version to avoid other process from overwriting with stale data.
Best Regards,
Chris

Querying by PK against Toplink Cache

Lets say I have table A that has PK id.
If toplink generates SQL like this:
"select id from A where 1=1 and id=3"
Assuming the data is already cached against that PK, will Toplink retrieve the data from the Toplink cache with this SQL? or since i used 1=1, will it not find it in the cache and require a database retrieval?
Thanks,
Zev.

Doug, thanks for the response. I expected it since we have seen a large increase in SQL. Let me explain the reason for this and maybe you can point us in a better direction.
We generate a Map of CriteriaValue Objects. The in our base object, we iterate over the map and based on the criteria, we build the appropriate expression.
Here is the code for the iteration.
          CriterionValue criterionValueObj;
          selectionExpression = builder.prefixSQL("1=1");
          while (keyIterator.hasNext()) {
               String key = (String) keyIterator.next();
               criterionValueObj = (CriterionValue) criteriaHash.get(key);
               selectionExpression = getExpression(builder, selectionExpression, criterionValueObj, key, firstTimeThroughLoop);
               firstTimeThroughLoop = false;
As you can see, we add a prefix of 1=1. The reason for this is that if this is the first time through the look, we need to say:
ExpressionBuilder builder; //just showing the declaration type
selection = builder.get(key).containsSubstring(value.toString());
if it is the second time through (with conjunction)
ExpressionBuilder builder; //just showing the declaration type
selection = selection.and (builder.get(key).containsSubstring(value.toString()));
so we need two seperate code blocks in this case. A tremendous code minimization in our base class is to say 1=1 AND then it is as if we are always accessing it the second time.
Do you have any suggestions for us? I do not want modify the caching options programitically.
Thanks,
Zev.

Cluster config. - toplink cache

hi all!
A simple question...
I've built an application using ADF and toplink.
Actually the application runs on a single IAS.
now, for load reasons, i'll need to migrate to a cluster configuration.
Is there any kind of problem with the toplink cache?
Thanks.
Luca

hi all!
A simple question...
I've built an application using ADF and toplink.
Actually the application runs on a single IAS.
now, for load reasons, i'll need to migrate to a cluster configuration.
Is there any kind of problem with the toplink cache?
Thanks.
Luca

ValueHolder Indirection and TopLink Cache

Similar Messages

Maybe you are looking for