3.0.5 to 3.1.1 migration: problem with many-to-one lazy

frankgrimes · **Joined:** Tue Mar 22, 2005 2:27 am **Posts:** 62

Hi All,

I'm trying to upgrade from hibernate-3.0.5 to hibernate-3.1.1 and my application's regression tests have caught a problem with lazily loaded many-to-one associations.

In Hibernate-3.0.5, I had the following mapping fragment.

Code:

<many-to-one name="customer" column="PRIMARY_KEY_COLUMN" unique="true" lazy="true" fetch="select" insert="false" update="false"/>

I *am* doing the necessary build-time instrumentation and with hibernate-3.0.5 "customer" is not loaded when loading the associated entity through an HQL query.
(The HQL query specifies no eager fetching of the customer association.)

For Hibernate-3.1.1, I applied the following change (i.e. lazy="true" --> lazy="no-proxy"):

Code:

<many-to-one name="customer" column="PRIMARY_KEY_COLUMN" unique="true" lazy="no-proxy" fetch="select" insert="false" update="false"/>

Now "customer" is always being loaded when executing the query.

Is this a hibernate bug?
Or perhaps was my previous usage wrong and only working by chance?

I don't suppose there is a regression test in the Hibernate test suite for many-to-one unique lazy associations that someone could point me to?

If not, I'd be glad to write up a simple test case to reproduce this behaviour.

Thanks!

frankgrimes · **Joined:** Tue Mar 22, 2005 2:27 am **Posts:** 62

Just a quick follow-up... changing the loading from an HQL query to a simple Session.load yields the same (seemingly) buggy behaviour.

frankgrimes · **Joined:** Tue Mar 22, 2005 2:27 am **Posts:** 62

I've found a test case for many-to-one no-proxy, but it seems that the actual test code was commented out between hibernate-3.0.5 and hibernate-3.1.1.

org.hibernate.test.instrument.InstrumentTest line 160:

Code:

//assertFalse(Hibernate.isPropertyInitialized(doc, "owner"));

http://cvs.sourceforge.net/viewcvs.py/h ... 16&r2=1.17

If I uncomment the line and re-run the test, it fails.
This is equivalent to my own application's test failure.

So, it looks like this was a known issue at some point... are there any plans to address this regression?
I've already searched JIRA and didn't find anything related to this.

frankgrimes · **Joined:** Tue Mar 22, 2005 2:27 am **Posts:** 62

FYI, I've created a JIRA task for this issue:

http://opensource2.atlassian.com/projects/hibernate/browse/HHH-1435

frankgrimes · **Joined:** Tue Mar 22, 2005 2:27 am **Posts:** 62

Hmm... still no replies either here or on JIRA.
Has nobody else encountered this issue?

dogu · **Joined:** Thu Jan 05, 2006 4:13 pm **Posts:** 4

Frank,

I had what could be a similar problem.

I was getting a StackOverflowException with a large linked list of Items.

Code:

<many-to-one name="next" class="Item" cascade="none" column="nextItemID" lazy="proxy" outer-join="false" />
<many-to-one name="prev" class="Item" cascade="none" column="prevItemID" lazy="proxy" outer-join="false"/>

It looked like it was trying to load all of the objects, despite the lazy="proxy" After 500 or so, I'd get a StackOverflowException.

Today, we upgraded to Hibernate 3.2. That seems to have corrected the problem.

I hope this helps.

DougP

frankgrimes · **Joined:** Tue Mar 22, 2005 2:27 am **Posts:** 62

Hi Doug,

Did you mean hibernate-3.1.2?
We did test with hibernate-3.1.2 and the problem remains. (at least with the no-proxy option)
I don't recall if I tested lazy="proxy" with 3.1.1 or 3.1.2.
If lazy="proxy" is working for you in 3.1.2, maybe I'll double check that option for us.

Thanks!

dogu · **Joined:** Thu Jan 05, 2006 4:13 pm **Posts:** 4

Yes, it was Hibernate 3.1.2. I don't see in the bug list where it was fixed, but it appears to have helped up out.

Since I have it handy, here's the list of fixes in 3.1.2

Changes in version 3.1.2 (01.27.2006)

-------------------------------------------
** Bug
* [HHH-73] - Statistics for HQL queries use pre-processed query string
* [HHH-1306] - HQL parsing problem with join fetching of arrays/collections of values
* [HHH-1370] - Warning in hibernate-mapping-3.0.dtd
* [HHH-1371] - MappingException is thrown when the same column is referenced with different case
* [HHH-1386] - Numeric (long) literals not properly handled by HQL parser
* [HHH-1390] - Session.isOpen() throws exeception when the session is closed with ThreadLocalSessionContext
* [HHH-1391] - Invalid parameter index SQLException when using named parameters after positional parameters
* [HHH-1392] - Proxies cannot be serialized after session is closed
* [HHH-1398] - extends and entity-names broken with mapping-level package attribute
* [HHH-1407] - return-join broken for entity collections

** Improvement

* [HHH-1364] - Defensive check of isClosed when obtaining a connection from ConnectionManager
* [HHH-1367] - warn level log "this operation breaks ==" may be disturbing

** New Feature
* [HHH-1372] - Support for MySQL5 new varchar length

** Patch
* [HHH-1005] - Criteria LEFT JOIN capability when adding Order to associations

** Task
* [HHH-1373] - Document update versioned

esword · **Joined:** Tue Jun 07, 2005 11:36 pm **Posts:** 22

For anyone else who discovers this thread in the future - another caveat that I just figured out after pounding my head on the same issue is that if you have default-lazy="true" set on the class which is the single-end of a many-to-one association, the lazy settings for the association will be ignored and the whole object will be loaded when the object at the multi-end is loaded. This is with 3.1.2, at least.

It makes sense now that I realize it. It would be nice if there was a caveat to this affect in section 19.1 of the (usually great) manual.

johnh · **Joined:** Thu Aug 04, 2005 12:45 pm **Posts:** 4

I have been trying to figure this one out, and your comment really helps!

So I have a one-to-many and the many side class has lazy="false". This forces the lazy one side to load the many side using a subsequent select.

Why is this? I've tested against the latest code in the head.

whoover · **Joined:** Thu Nov 02, 2006 8:26 pm **Posts:** 6

I cannot believe that this issue is still not resolved in 3.2.6 ga