Question about Finder-Load-Beans flag

Question

Hi all,
I've read that the Finder-Load-Beans flag could yield some valuable gains in performance
but:
1) why is it suggested to do individual gets of methods within the same Transaction
? (tx-Required).
2) this strategy is useful only for small sets of data, isn't it? I imagine I
would choose Finder-Load-Beans to false (or JDBC) for larger sets of data.
3) A last question: its default value is true or false ?
Thanks
Francesco

Answer 1

Because if there are different transactions where the get method is called
then the state/data of the bean would most be reloaded from the database. A
new transactions causes the ejbLoad method to be invoked in the beginning
and the ejbStore at the end. That is the usual case but there are other ways
to modify this behavior.
Thanks
Gaurav
"Francesco" <[email protected]> wrote in message
news:[email protected]...
>
Hi thorick,
I have found this in the newsgroup. It's from R.Woolen answering
a question about Finder-Load-Beans flag.
"Consider this case:
tx.begin();
Collection c = findAllEmployeesNamed("Rob");
Iterator it = c.iterator();
while (it.hasNext()) {
Employee e = (Employee) it.next(); System.out.println("Favorite color is:"+ e.getFavColor());
tx.commit();
With CMP (and finders-load-beans set to its default true value), thefindAllEmployeesNamed
finder will load all the employees with the name of rob. The getFavColormethods
do not hit the db because they are in the same tx, and the beans arealready loaded
in the cache.
It's the big CMP performance advantage."
So I wonder why this performance gain can be achieved when the iterationis inside
a transaction.
Thanks
regards
Francesco
thorick <[email protected]> wrote:
1) why is it suggested to do individual gets of methods within thesame Transaction
? (tx-Required).I'm not sure about the context of this question (in what document,
paragraph
is this
mentioned).
2) this strategy is useful only for small sets of data, isn't it? Iimagine I
would choose Finder-Load-Beans to false (or JDBC) for larger sets ofdata.
>
If you know that you will be accessing the fields of all the Beans that
you get back from a
finder,
then you will realize a significant performance gain. If one selects
100s or more beans
using
a finder, but only accesses the fields for a few, then there may be some
performance cost.
It could
depend on how large some of the fields are. I'd guess that the cost
of 1 hit to the DB per
bean vs.
the cost of 1 + maybe 1 more hit to the DB per bean, would usually be
less. A performance
test using
your actual apps beans would be the only way to know for sure.
3) A last question: its default value is true or false ?The default is 'True'
-thorick