Javadogs,
I'd like to get feedback from jdo vendors who have implemented detach
and attach. I'm really keen on getting user feedback or use-cases for
attach.
In particular, how do implementations handle transient instances? Are
there use-cases that demand that transient instances are copied? This
would be the case if the same transient instance was intended to be
attached to multiple persistence managers.
Please let's hear from you.
Thanks,
Craig
P.S. If we agree to remove attachCopy, there are still some questions.
1. What do we do about the "makeTransactional" boolean flag on
attachCopy? Do we automatically make the attached instances
transactional? That is, default the flag to true, or do we add
another variant on makePersistent?
2. What do we do about the attach callbacks? We could preserve them
and just change the documentation to say that the callbacks are
called during makePersistent for detached objects.
On Oct 18, 2005, at 11:21 AM, Wes Biggs wrote:
I think this is an excellent idea.
I'd be happy to remove the copy semantic on transient instances in
order to eliminate the need for a separate method. My feeling is
that transient instances are likely to be retained as references,
and it's actually less confusing if those references become
persistent-new.
Wes
Matthew T. Adams wrote:
I like the convergence in the api, and I also like the suggestion
to return
the root object(s) from makePersistent, as it increases symmetry.
Could attachCopy and makePersistent become one, just
makePersistent? The
only difference in the proposal as you suggest, Craig, is the
handling of
transient instances. attachCopy copies instances and makes the
copies
persistent-new, whereas makePersistent just makes them persistent-
new; all
other behavior is the same. If we got rid of attachCopy, in favor of
makePersistent, and changed makePersistent such that it returns
the root
object(s) given, then the behavior is covered, except for
attachCopy's copy
semantic for transient instances. The only difference in the API
if we keep
attachCopy is the copy semantic for transient instances, which I
don't think
is all that important.
Thoughts?
--matthew
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Sent:
Sunday, October 16, 2005 12:26 PM
To: JDO Expert Group; [email protected]
Subject: Re: RETRY: Transient instance referencing a detached
instance? (was: Question about attachCopy, transient & detached
instances)
Javadogs,
I've reviewed this material and agree with the general
conclusion, that we can add the ability to include detached
objects in the closure of instances for both makePersistent and
attachCopy.
The spec as of today requires treating these three types of
objects in the closure differently:
persistent instances
attachCopy throws an exception; makePersistent ignores them
transient instances
attachCopy copies them and creates persistent-new instances;
makePersistent makes them persistent-new
detached instances
attachCopy copies them; makePersistent throws an exception
Here's the change as I would propose it:
persistent instances
attachCopy ignores them; makePersistent ignores them
transient instances
attachCopy copies them and creates persistent-new instances;
makePersistent makes them persistent-new
detached instances
attachCopy copies them; makePersistent copies them
If we make this change, the behavior of makePersistent and
attachCopy are more similar than today. The differences are in
the treatment of transient instances and whether or not the root
objects are returned. If we like, we can have makePersistent
also return the root objects by making a source-compatible
change to the signatures of makePersistent.
Craig
On Oct 5, 2005, at 11:59 AM, Matthew T. Adams wrote:
"Wes Biggs" <[EMAIL PROTECTED]> wrote in message
news:<[EMAIL PROTECTED]>...
Would it be cleaner to not allow transient instances to be
included in
attachCopy() graphs at all? Sounds that way to me.
No, I'd like to continue to allow transient instances to be
included in
attachCopy graphs. I'd like to **add** the ability for
detached objects to
be included in makePersistent graphs.
This use case fell out very naturally for me while working on a
website that
uses detached objects. I ended up with a transient User object
referencing
a detached Country object, and, of course, when I called
pm.makePersistent(user), I got an JDOUserException as I should
have, per the
spec as it stands right now.
--matthew
