HSEARCH-114
by John Griffin
For the ResultTransformer, do you just want the user to pass a class as the
sole argument to indicate what they want as a result?
Ex. setResultTransformer(Class clazz) where clazz is Map.class or
Object.class
John G.
17 years, 11 months
Clustering and UpdateTimestampsCache
by Brian Stansberry
Wanted to raise a point about about timestamps cache handling in case
there's any desire to change the UpdateTimestampsCache API in 3.3.
AIUI, a goal of UpdateTimestampsCache is to ensure the cached timestamp
never moves backward in time *except* when a caller that has set the
timestamp to a far-in-the-future value in preInvalidate() later comes
back and calls invalidate(), passing the current time.
There's a race in UpdateTimestampsCache where this could break under
concurrent load. For example, you could see:
(now = 0)
tx1 : preInvalidate(60);
(now = 1)
tx2 : preInvalidate(61);
tx1 : cache queryA w/ timestamp 1
tx1 : invalidate(1)
tx2 : update entity in a way that would query A results
tx2 : read queryA; check timestamp; 1 == 1 so passes. Wrong!
To deal with this, there are some comments in UpdateTimestampsCache
about having preInvalidate() return some sort of Lock object, which
would then be returned as a param to invalidate(). Idea here is to
ensure that only the caller that most recently called preInvalidate is
allowed to call invalidate.
That could work if the backing TimestampsRegion isn't clustered, but it
doesn't address the fact that a clustered TimestampsRegion can be
getting updates not only via the local UpdateTimestampsCache, but also
asynchronously over the network. If a clustered TimestampsRegion gets a
replicated update that moves the timestamp back in time, it has no
simple way to know if this is because 1) a peer that earlier replicated
a high preinvalidate value is now replicating a normal invalidate
value or 2) an earlier change from peer A has arrived *after* a later
change from peer B.
This could be addressed with a change to the TimestampsRegion API.
Basically replace
public void put(Object key, Object value) throws CacheException;
with
public void preInvalidate(Object key, Object value) throws CacheException;
public void invalidate(Object key, Object value, Object
preInvalidateValue) throws CacheException;
Basically the value that is passed to preInvalidate is also passed as a
2nd param to invalidate. This gives the TimestampsRegion the
information it needs to properly track preinvalidations vs invalidations.
The UpdateTimestampsCache API is then changed to provide the caller with
the timestamp in preInvalidate() and take it back in invalidate():
public synchronized Object preinvalidate(Serializable[] spaces) throws
CacheException {
Long ts = new Long( region.nextTimestamp() + region.getTimeout() );
for ( int i=0; i<spaces.length; i++ ) {
region.preInvalidate( spaces[i], ts );
}
return ts;
}
public synchronized void invalidate(Serializable[] spaces, Object
preInvalidateValue) throws CacheException {
Long ts = new Long( region.nextTimestamp() );
for ( int i=0; i<spaces.length; i++ ) {
region.invalidate( spaces[i], ts, preInvalidateValue );
}
}
This is basically similar to the Lock concept in the
UpdateTimestampsCache comments; but the control over the update is
delegated to the TimestampsRegion.
The issue here is the UpdateTimestampsCache caller needs to hold onto
the value returned by preInvalidate() and then pass it back. Likely
requires a change to Executable to provide a holder for it.
A change to the TimestampsRegion API has no benefit without a
corresponding change in UpdateTimestampsCache and its caller.
--
Brian Stansberry
Lead, AS Clustering
JBoss, a division of Red Hat
brian.stansberry(a)redhat.com
18 years
HIbernate and 3.0 framework .Net
by S P
Does someone have any idea/ opinion on Nhibernate compatibility with framework 3.0 and/ or Visual Studio 2008. Any idea about nhibernate working with SQL Compact Edition.
Thanks,
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com
18 years
Query Extension: adding count method
by Gaetano Perrone
Developping a pagination module for an open source hibernate-struts based project
(http://smartweb.sourceforge.net)
i was envolved in the problem of how to extrapolate the results size of a given generic query (sql query, hql query or criteria)
My first solution consisted into a size() call on the returned collection but this means to extract the collection from the db
for knowing the result size.
With hibernate 3 the solution for criteria was too easy and consists into add Projection on the given generic Criteria object
that otherwise of the session.iterate() solution doesn't need to extract any raws . The only result extracted is the Integer result of the count wrapped query.
Otherwise an analoughe solution for Hql/SQL Query is not very easy because of the protected visibility of the getNamedParams() method of org.hibernate.impl.AbstractQueryImpl
class.
this deny any users to access to the query parameters needed to wrap and execute query.
To resolve this problem i tought a generich solution that improves query class adding a count() public method
that allows to ask for a query result preview count.
You can ask a given prepared query:
- for uniqueresult: with the uniqueresult method
- for raw list: with the list method
- for raw count: with the count method
My implementation class are attached.
I hope i could support the project with this little work...
Gaetano Perrone
18 years
'INSERT' transactions are always rolled back on PostgreSQL when partitioning is used.
by Julius Stroffek
Hi All,
I have created a simple java application running on glassfish and using
a hibernate as a persistence provider on PostgreSQL. I just simply
insert display and delete rows from a table.
Everything worked fine but when I use partitioning on tables as described at
http://www.postgresql.org/docs/8.2/static/ddl-partitioning.html
The application started to report every insert to the table as
"Transaction marked for rollback" and throwing an exception
javax.persistence.OptimisticLockException:
org.hibernate.StaleStateException: Batch update returned unexpected row
count from update [0]; actual row count: 0; expected: 1
at
org.hibernate.ejb.AbstractEntityManagerImpl.wrapStaleStateException(AbstractEntityManagerImpl.java:654)
at
org.hibernate.ejb.AbstractEntityManagerImpl.throwPersistenceException(AbstractEntityManagerImpl.java:600)
at
org.hibernate.ejb.AbstractEntityManagerImpl$1.beforeCompletion(AbstractEntityManagerImpl.java:525)
.... (plenty of more lines)
Caused by: org.hibernate.StaleStateException: Batch update returned
unexpected row count from update [0]; actual row count: 0; expected: 1
at
org.hibernate.jdbc.Expectations$BasicExpectation.checkBatched(Expectations.java:61)
at
org.hibernate.jdbc.Expectations$BasicExpectation.verifyOutcome(Expectations.java:46)
at
org.hibernate.jdbc.BatchingBatcher.checkRowCounts(BatchingBatcher.java:68)
at
org.hibernate.jdbc.BatchingBatcher.doExecuteBatch(BatchingBatcher.java:48)
at
org.hibernate.jdbc.AbstractBatcher.executeBatch(AbstractBatcher.java:246)
at
org.hibernate.engine.ActionQueue.executeActions(ActionQueue.java:237)
at
org.hibernate.engine.ActionQueue.executeActions(ActionQueue.java:141)
at
org.hibernate.event.def.AbstractFlushingEventListener.performExecutions(AbstractFlushingEventListener.java:298)
at
org.hibernate.event.def.DefaultFlushEventListener.onFlush(DefaultFlushEventListener.java:27)
at org.hibernate.impl.SessionImpl.flush(SessionImpl.java:1000)
at org.hibernate.impl.SessionImpl.managedFlush(SessionImpl.java:338)
at
org.hibernate.ejb.AbstractEntityManagerImpl$1.beforeCompletion(AbstractEntityManagerImpl.java:516)
... 50 more
This is due that Statement.executeBatch returns the incorrect number of
rows affected by the transaction. This number returned by PostgreSQL is
defined at
http://www.postgresql.org/docs/8.2/static/rules-status.html
Since the rule system in general is too complex in PostgreSQL it is not
possible to define the command status string so it would contain the
number of rows affected by the transaction. I think that JDBC driver
should return Statement.SUCCESS_NO_INFO in this case, however, it is not
doing so now and the JDBC spec does not allow to return
Statement.SUCCESS_NO_INFO in Statement.executeUpdate.
I had a look at hibernate source where there are couple of
"Expectations" implementations but I have not found how it is possible
to choose different implementation in my application. A workaround for
this could be that I can use Expectation.NONE as an "Expectations"
instance. Is it possible to set this up in an application
code/configuration? I have not found anything like this in documentation.
Thanks for your comments.
Cheers
Julo
18 years
