No because I coudn't find which one uses lazy loading for a property. Do you know which one?

i've run other tests with lazy properties, it works for me too

OK, I found "instrument" test case that I think is supposed to test lazy loading of a property. In the mapping file that comes with HB test case the "name" property is not lazy loaded. The "summary" is. So my understanding is that when you load the object the summary property should not be in the first SELECT on the object. So either the instrumentation failed or there's some other kind of a problem, but the bottom line is that even though the test case doesn't fail the property is not lazily loaded. Or I'm messing something up.

UPDATED TEST CASE
s = openSession();
doc = (Document) s.createQuery("from Document").uniqueResult();

System.out.println("*** got doc = ");

doc.getName();
System.out.println("** got name = " + doc.getName());

assertEquals( doc.getText(), "blah blah" );
System.out.println("** got text = " + doc.getText());

assertEquals( doc.getSummary(), "blah" );

System.out.println("** summary = " + doc.getSummary());

s.connection().commit();
s.close();



DEBUG OUTPUT
16:07:13,085 DEBUG SQL:292 - select document0_.id as id, document0_.name as name2_, document0_.folder as folder2_ from documents document0_
*** got doc =
** got name = Hibernate in Action
16:07:13,596 DEBUG SQL:292 - select document_.summary as summary2_, document_.owner as owner2_, document_.text as text2_, document_.lastTextModification as lastText7_2_, upper(document_.name) as formula0_ from documents document_ where document_.id=?
16:07:13,907 DEBUG SQL:292 - select owner0_.id as id0_, owner0_.name as name1_0_ from owners owner0_ where owner0_.id=?
** got text = blah blah
** summary = blah

FYI, I am *quite* sure that this is working. And even your own SQL log seems to show it is working for *you*...

i've updated my cvs this morning, i'll run another test tonight.
But xplorem, confirm that you still have a bug with latest source (+homebuild, we don't know) + instrumentation

all tests run fine... even with my "crazy" little mapping file (lazy attribute in a join element):

I spend almost half a day trying to figure out why lazy loading doesn't work. Well, I finally got the sucker so check it out.

It appears that if an instance variable for the lazy property starts with the "_" then the lazy loading is not working. If it doesn't start with _, it's working fine! You can reproduce it with your own test case InstrumentTest by renaming field "text" of the Document class to "_text" and leaving methods getText() and setText(). The mapping file does not say "access=field" so as I understand by default it should be access=property.

Is this a bug that is going to be fixed or is some peculiar expected behavior?? Sounds like a bug to me.

It works as expected, enhancer transforms all methods in the same way (replaces field access with callback) it needs to know field names for this stuff, method names have no meaning for this type of transformation and there is no workaround (hibernate doe's not know your field names too).

Here is why I think it should be classified as a bug:

- The same code works when the lazy=false and doesn't when lazy=true. This can not be expected behavior.

- If lazy loading of properties operates directly on fields, then the enhancer should check the class for fields with an expected name and throw an exception if there is no field matching the property name.

So I guess what I'm saying is that either it should work for all access types (method, field etc.) or it should throw an exception if it knows that it wouldn't work. Makes sense?

There is no way to find field name by property name, but it is possible to validate this stuff.

Just out of curiosity, isn't it possible to intercept a getter method call to get the property, make the database call, and set the property value using the setter? Same thing as with the value but it would support all access strategies for a property - as recommended in Hibernate documentation: "Many other ORM tools directly persist instance variables. We believe it is far better to decouple this implementation detail from the persistence mechanism. Hibernate persists JavaBeans style properties, and recognizes method names of the form getFoo, isFoo and setFoo. You may however switch to direct field access for particular properties, if needed. "

Also, what are your guys thoughts on session reattachment? The way we are planning to use lazy properies is as follows. We have a table with large number of rows that shows only the summary inforamtion about the person (say name, age and data of birth). If a user clicks on the table row, we show the entire person's profile which includes the picture (BLOB), long text description fields and so on. I would think that's typical for lazy loading - to have some data retreived in one request and more data at some point later but NOT ON THE SAME SESSION (because we are going to close the session when the first request is processed). It sounds like to make lazy loading work we'd have to reattach the objects to the new session by calling update() or lock(). That is not very intuitive. Have you thought of associating the session with the thread and maybe using that for lazy loading?

No, it is not the same thing with field interception strategy, but probably it is too long story for forum message. I see current enhancer implementation doe's not use mapping files, it can not to validate too.

As remember "auto-reattachment" was discussed many times, it is trivial to implement it, but it is "dangerous". Probably it is not a problem to reread row in your case too, you have id and need to hit database/cache anyway. As I understand lazy properties do not use global cache, it can be important for this use case too.

<out-of-topic>
BTW I have added jar file support and filtering by "magic" number. "getTransformer" was changed to use array of class and interface names, it will help to remove custom class file visitor to detect duplicate enhancement (you need to remove it for jar file enhancement support).

</out-of-topic>