Re: a HUGE Core Data bug
Re: a HUGE Core Data bug
- Subject: Re: a HUGE Core Data bug
- From: "I. Savant" <email@hidden>
- Date: Mon, 12 Mar 2007 12:24:50 -0400
On Mar 12, 2007, at 5:35 AM, Aurélien Hugelé wrote:
I think that your reactions guys are quite surprising. Saying that
"I didn't experience any problems either. Everything works as
expected" is a typical user answer, not a serious developer answer.
We all know, as developers, that sometimes bugs are difficult to
reproduce, and difficult to isolate. I've demonstrated a Core Data
problem (that is documented), so the problem exists, there is no
reason to deny it.
Now *that's* surprising. Nobody said there weren't any problems.
We've all acknowledged that your example demonstrates a problem. The
part you seem to refuse to accept is that it's your error, not a bug.
As Jim explained, the behavior is undefined when there is no inverse
relationship. That is *not* a bug.
Moreover, you seem to have overlooked the details: the problem only
arise with SQL store, not with the 2 other stores... so the
behavior of Core Data, regarding graph coherency, is not consistent
and depends on the store type. This is what I call a major
drawback: performance by type of store are well commented, but the
coherency is not...
That to me highlights the reason *why* the behavior is undefined.
Consider that a SQL store has its own boundaries and limitations
versus an XML store. The underlying mechanism for writing that data
to disk is completely different. It would stand to reason that a
general "save this to disk" case wouldn't exactly fit all store types
without defining some very exact rules, etc. Some relationships are
tricky to persist.
If adding an inverse relationship is a compulsory for Core Data,
then i'll file a bug to make Xcode not only warn me when there is
no inverse, but generate a compiler error by default (and generate
a warning only if the developer set the right parameter...)
What part of this do you keep misunderstanding? It's *not* a
compulsory. It's necessary if you're not going to (or don't know how
to) manage the relationship's specifics yourself, so making this an
outright error is just plain wrong.
And please try my program both on Mac OS X 10.5 and 10.4, the
behavior is definitely not the same, so I think that Core Data
engineers have probably found a way to enhance relationships
without inverses integrity.
Nobody can tell you outright whether this works or not on 10.5
without violating their NDA (as you're doing repeatedly in this
thread), so though it may be frustrating, nobody can discuss this
point at the moment.
I'm mystified by the continued confusion some people have on this
subject ... *YOU CANNOT DISCUSS 10.5 DETAILS THAT HAVE NOT BEEN
RELEASED TO THE PUBLIC.* It can't be stated any more plainly than
that. It's the binding legal agreement you signed for an early look
at Leopard and you're obligated to obey it.
--
I.S.
_______________________________________________
Cocoa-dev mailing list (email@hidden)
Do not post admin requests or moderator comments to the list.
Contact the moderators at cocoa-dev-admins(at)lists.apple.com
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden