How to unload a singleton

[Saruman]

Unfortunately I am not able to find a direct answer to this question
neither in documentation nor in newsgroups. I have an application
biggest part of which is dynamically reloaded in certain circumstances
- kind of webstart, but dynamic (you have to restart webstart to get a
new version of an application) and without a GUI.

The problem is that I have some singletons among my objects. Turns out
that garbage collector has some trouble with affected classes. When I
load new versions of my classes, the instances of the old versions are
not finalized. Accordingly their classloader stays in memory, and all
old classes do too. The classes are therefore never unloaded. The only
workaround I've found is to explcitly to to every class and set the
singleton instance inside it to null. Then the famed Java unloading
kicks in. I understand that class unloading is optional for JVMs and
depends on implementation (mine is HotSpot), but still - is that how
it's supposed to work? Is Singleton pattern flawed in these
circumstances (not that I insist that my usage of it is always
justified...)?

[SPG]

Re: How to unload a singleton

My suggestion is never to directly access the object itself, always
reference it by the getter method, this way you do not hold references to
the signleton throughout your code, and garbage collection will happen as
you expect:

public class MySingleton{
private MySingleton _instance;

public MySingleton getInstance(){
if( _instance == null){
_instance = new MySingleton();
}
return _instance;
}

public String getMessage(){
return "Happy Days";
}
}


public class ReferenceMySing leton{
public static void main(String args[] ){
//Do not do this:
MySingleton obj = MySingleton.get Instance();
obj.getMessage( );

//Do this instead
MySingleton.get Instance().getM essage();
}

}



"Saruman" <Anatoliy.Salis tra@gfinet.com> wrote in message
news:dccaaaca.0 310201326.1ee49 662@posting.goo gle.com...[color=blue]
> Unfortunately I am not able to find a direct answer to this question
> neither in documentation nor in newsgroups. I have an application
> biggest part of which is dynamically reloaded in certain circumstances
> - kind of webstart, but dynamic (you have to restart webstart to get a
> new version of an application) and without a GUI.
>
> The problem is that I have some singletons among my objects. Turns out
> that garbage collector has some trouble with affected classes. When I
> load new versions of my classes, the instances of the old versions are
> not finalized. Accordingly their classloader stays in memory, and all
> old classes do too. The classes are therefore never unloaded. The only
> workaround I've found is to explcitly to to every class and set the
> singleton instance inside it to null. Then the famed Java unloading
> kicks in. I understand that class unloading is optional for JVMs and
> depends on implementation (mine is HotSpot), but still - is that how
> it's supposed to work? Is Singleton pattern flawed in these
> circumstances (not that I insist that my usage of it is always
> justified...)?[/color]

[Saruman]

Re: How to unload a singleton

Sorry for misleading, guys and gals. There is NO problem with
unloading classes with singletons, static object references etc. I've
had a problem with ThreadGroup though. We've overridden it to have our
own exception handling (uncaughtExcept ion method). Now, when you
create a new ThreadGroup, it's automatically added to the parent
thread group - 'main' in this case - and stays there until explicitly
removed. Understandably, the class cannot be unloaded - since an
instance of it is still active. Therefore the class loader and all
class definitions stay in memory. The problem is solved by calling
destroy() on an offending ThreadGroup (you have to make sure there are
no threads in it before that). Cheers...

[Saruman]

Re: How to unload a singleton

"SPG" <steve.goodsell @nopoo.blueyond er.co.uk> wrote in message news:<Qc6lb.110 0$GQ2.10559529@ news-text.cableinet. net>...[color=blue]
> My suggestion is never to directly access the object itself, always
> reference it by the getter method, this way you do not hold references to
> the signleton throughout your code, and garbage collection will happen as
> you expect:
>
> public class MySingleton{
> private MySingleton _instance;[/color]
You mean 'static' here and in the method, don't you...
[color=blue]
>
> public MySingleton getInstance(){
> if( _instance == null){
> _instance = new MySingleton();
> }
> return _instance;
> }
>
> public String getMessage(){
> return "Happy Days";
> }
> }
>
>
> public class ReferenceMySing leton{
> public static void main(String args[] ){
> //Do not do this:
> MySingleton obj = MySingleton.get Instance();
> obj.getMessage( );
>
> //Do this instead
> MySingleton.get Instance().getM essage();
> }
>
> }
>
>
>
> "Saruman" <Anatoliy.Salis tra@gfinet.com> wrote in message
> news:dccaaaca.0 310201326.1ee49 662@posting.goo gle.com...[color=green]
> > Unfortunately I am not able to find a direct answer to this question
> > neither in documentation nor in newsgroups. I have an application
> > biggest part of which is dynamically reloaded in certain circumstances
> > - kind of webstart, but dynamic (you have to restart webstart to get a
> > new version of an application) and without a GUI.
> >
> > The problem is that I have some singletons among my objects. Turns out
> > that garbage collector has some trouble with affected classes. When I
> > load new versions of my classes, the instances of the old versions are
> > not finalized. Accordingly their classloader stays in memory, and all
> > old classes do too. The classes are therefore never unloaded. The only
> > workaround I've found is to explcitly to to every class and set the
> > singleton instance inside it to null. Then the famed Java unloading
> > kicks in. I understand that class unloading is optional for JVMs and
> > depends on implementation (mine is HotSpot), but still - is that how
> > it's supposed to work? Is Singleton pattern flawed in these
> > circumstances (not that I insist that my usage of it is always
> > justified...)?[/color][/color]