OC,
see comments below.
On 02/03/2015 12:39, OC wrote:
Bogdan,
On 2. 3. 2015, at 12:05, Bogdan Zlatanov <[email protected]> wrote:
ERXGenericRecord has permanentGlobalID() which returns a EOKeyGlobalID that
might give you the unique instance you want.
Aha... is that guaranteed? That would be easier, but quickly checking the
Wonder sources I got an impression that different instances of the EO in
different ECs might return different (although reliably equal) instances of
global ID.
But as always, I might have missed something of importance.
eo.permanentGlobalID() is the same as ec.globalIDForObject(eo), provided
the EOGlobalID returned from ec.globalIDForObject(eo) is not
EOTemporaryGlobalID() which only happens for newly inserted but not yet
saved EOs. Nevertheless, if you consistently use eo.permanentGlobalID()
for both saved and unsaved EOs, you should get the same instance.
I did a quick test to verify this, but I'm sure if I'm wrong, someone
will point that out.
Is there something that prevents you from isolating the processing code in its
own method/class that is called by all threads and is guarded by a lock of your
choice?
I can use a lock of my choice; my problem was to find a lock which would have
the desired feature that it is reliably the same instance for all EOs which
represent the same object (regardless the EC).
What I meant by isolating the processing logic is this:
1. Create a method that does all the processing steps which must be atomic;
2. Use it all threads so that whenever a thread needs to do the
processing it calls. Could be static, could be an instance method.
That would allow you to use a lock that is independent from EOs, e.g.
(for the non-static case)
class MySharedProcessor {
public void doWorkAtomically(...) {
...
synchronized (this) { // doesn't have to be 'this', could
be a dedicated lock object
// do sensitive processing
}
...
}
}
Regards, Bogdan
Nevertheless, meantime I've created the following code (based on the blogspot
linked below), which preliminarily seems to work well -- note that I lock (a)
only objects of one entity, (b) already saved to DB, and thus can rely on PK --
can't bump into different entities, can't bump into null PK.
Still, if someone happens to see any problem there, I'll appreciate a nudge :)
===
static private __locks=new WeakHashMap()
static private auctionLock(DBAuction auction) {
Integer id=auction.primaryKeyNumericValue // has been saved -> is
unique and nonzero
synchronized(this.@__locks) {
def lockwr=this.@__locks[id]
if (lockwr) return lockwr.get()
def lock=new Integer(id)
this.@__locks[lock]=new java.lang.ref.WeakReference(lock)
return lock
}
}
...
synchronize (auctionLock(auction)) {
... this code never runs concurrently for same auction, regardless its
EC ...
}
===
Thanks a big lot,
OC
On 02/03/2015 11:29, OC wrote:
Bogdan,
On 2. 3. 2015, at 11:00, Bogdan Zlatanov <[email protected]> wrote:
This response will not be very helpful, but this sounds like an X-Y problem.
Could you elaborate more on what exactly you'd like to achieve?
To have a code into which "no two threads would get with same EO, not even if the EO
is in different ECs".
Namely, to perform some processing based on particular EO (or its global ID or
its PK), which processing needs to be either atomic, or covered by optimistic
locks and rejectable. Since the latter would be rather difficult, and since the
processing is not terribly time-consuming, and since the app is
single-instance, I would opt for the atomicity, if I knew how.
(Well meantime I have found
http://illegalargumentexception.blogspot.cz/2008/04/java-synchronizing-on-transient-id.html
which probably would help with PK, just seems unnecessarily convoluted...)
Thanks,
OC
On 02/03/2015 10:39, OC wrote:
Hello there,
is there an object whose _instance_ (not just value like it seems to be with
the global ID) would consistently represent one EO, regardless the EC in which
the EO just happens to be?
I would need to lock on an EO, but I would need that it worked over all ECs,
i.e.
===
synchronized (eo) {
// no two threads would get there with same EO
// _not even if the EO is in different ECs_
// which is why 'synchronized (eo)' does not work
}
===
Of course, I could make a thread-safe map, insert lock objects there and use
globalids for index, but it seems to me rather convoluted -- not speaking of
that I would not really know when to flush the data (unless I override EOs
finalize, ick...)
Isn't there a better solution?
Thanks a lot,
OC
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list ([email protected])
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/webobjects-dev/bogdan.zlatanov%40gmail.com
This email sent to [email protected]
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list ([email protected])
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/webobjects-dev/ocs%40ocs.cz
This email sent to [email protected]
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Webobjects-dev mailing list ([email protected])
Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/webobjects-dev/archive%40mail-archive.com
This email sent to [email protected]
